PDA

View Full Version : Electronic-trav-a-dial



Pages : [1] 2

John Stevenson
01-16-2011, 09:03 AM
OK new thread to move away from Black forests thread.

We finished up asking why it was not possible to use a cheap encoder and one of these also cheap Chinese remote displays.

http://www.arceurotrade.co.uk/imagecache/3286dc1a-1abf-4d1d-8f1b-9b4a00bab943_250x125.jpg

Simple cheap encoder with a hardened knurled radiused wheel that can be tilted to get the correct circumference driving into one of these units.

The protocol for the Chines signals is documented on the Shumatech site which is down at the moment but I may be able to get it elsewhere.

Basically it's a reversal of what Scott Schumacher has done.

Any idea's or is anyone up to the job ?

RB211
01-16-2011, 09:31 AM
Shumatech at one point made Quadrature to chinese scale converters you plugged inline. So one could just take a rubber wheel connected to a quadrature encoder, and drive one of those displays with off the shelf components.

John Stevenson
01-16-2011, 09:33 AM
Yes they did, might even have one somewhere.
However what's needed is quadrature to cheap Chinese display.

davidwdyer
01-16-2011, 09:34 AM
I like the idea of this thread. Hope it keep going to a good conclusion. My lathe needs just such a cheap mod.

RB211
01-16-2011, 09:38 AM
Yes they did, might even have one somewhere.
However what's needed is quadrature to cheap Chinese display.
This is a trivial task for a mcu like an Atmel AVR or PIC.

MCS
01-16-2011, 09:42 AM
It's a bit simpler.

Take a microcontroller and feed it with the encoder pulses.

With an update frequency of say 10 times a second, calculate the encoder position to the real position and write it to a simple LCD display.

No special components or conversion required.

The only disadvantage for most people is that it is electronics and software.

BillTodd
01-16-2011, 10:10 AM
John,

The encoder wheel would need a reasonable resolution; e.g. If you used a 5/8" knurl wheel the circumference is approximately 1.96" so a ~1thou" resolution requires at least 2000 pulses per turn.

The next problem is counting speed; Given the speed at which the carriage can be moved it might prove problematic to use a simple PIC interrupt routine, so it may be better to use a device like the Microchip dsPics that have hardware quadrature counters built in.

Of course, once you have that kind of processing power handy, there's no real need to have a mechanical trim when it could be done mathematically. And, there's little point using a cheap display device, one may as well include an LCD driven directly by the PIC. :)

One un-pursued idea I had, was to use a high resolution optical mouse as a sensor. The better mice claim 1600 dot per inch (which is independent of the surface), so quite a useful device could made with just a simple PC application.

Bill

sansbury
01-16-2011, 11:29 AM
I need to have another go at reading Chinese caliper protocol with an Arduino. I tried it a couple years ago but didn't really know what I was doing. Should be more than enough oomph to run one or two scales and an LCD.

RB211
01-16-2011, 12:00 PM
I need to have another go at reading Chinese caliper protocol with an Arduino. I tried it a couple years ago but didn't really know what I was doing. Should be more than enough oomph to run one or two scales and an LCD.

An Arduino is nothing more than an Atmel AVR running an OS. If you want true power of an AVR, use it natively and program it in Gnu C.

lane
01-16-2011, 12:08 PM
Not that the idea may be good.But have you ever tried to stop a moving carriage are mill table on the thousands. With a sweeping dial you can. But with numbers counting up are down. you cant The human mime is not fact enough.With a mechanical Trav A Dial it is easy to watch the dial and throw the feed out at the exact spot. Just some thoughts.

topct
01-16-2011, 12:48 PM
Instead of a digital readout what about using a servo motor to move a needle?

RB211
01-16-2011, 01:01 PM
Not that the idea may be good.But have you ever tried to stop a moving carriage are mill table on the thousands. With a sweeping dial you can. But with numbers counting up are down. you cant The human mime is not fact enough.With a mechanical Trav A Dial it is easy to watch the dial and throw the feed out at the exact spot. Just some thoughts.
What????? Using gears for a mechanical display??? No backlit LCD's, No crystals, No MCU's runnning at 20mhz, no use of A2D, and Interrupts? Nothing to program?? Wheres the fun in that? OOOhh I know, we could use LED's As you approach thousandth, it goes from yellow, yellow, yellow, green. Or we could make the led blink, and the blink rate increases as you get closer to the thousandth!

Yes... One could use PWM to drive an R/C servo to move a needle...
I love over engineering.

sansbury
01-16-2011, 01:05 PM
An Arduino is nothing more than an Atmel AVR running an OS. If you want true power of an AVR, use it natively and program it in Gnu C.

Are you using a mouse and graphical web browser to post here? Do you have a matching skirt and heels to go with that? REAL MEN USE LYNX AND THE COMMAND LINE!!!

For everything I've ever done with an Arduino, the only thing native AVR code has to offer me is more ways to make mistakes.

RB211
01-16-2011, 01:10 PM
Are you using a mouse and graphical web browser to post here? Do you have a matching skirt and heels to go with that? REAL MEN USE LYNX AND THE COMMAND LINE!!!

For everything I've ever done with an Arduino, the only thing native AVR code has to offer me is more ways to make mistakes.

Well, your calling me the one that wears high heals? Your the one defending the use of a "wire" language vs gnu C.
But again, some one is going to come on here, call us both panzies for not using Assembly...
Ahh the ranks of nerdom..

John Stevenson
01-16-2011, 01:36 PM
OK forget it - but I thought it was a good idea and would fill a gap.

Sigh................

tyrone shewlaces
01-16-2011, 01:42 PM
Reading a quadrature rotary encoder and having it send data to a digital display should be a pretty easy microcontroller project. Cheaper ala trav-a-dial than buying long scales for a lathe too (good idea JS!). Also, being micro-controlled, you wouldn't need a radiused wheel to adjust for proper readings too since the micro could adjust it electronically to probably 10 decimal places if you wanted it to.

Whether in Arduino, C or machine language, just pick your preference and from a functional perspective any of them would be more than adequate.

So:
-Rotary encoder attached to wheel that contacts the lathe way.
-AVR microcontroller circuit (Arduino or home-brew) to do the thinking.
-Digital display (I prefer large LED over LCD because it's easier to read, but the circuit will be about the same).
-Enclosure for the micro and display to make it look slick and protect from flying debris.

I'm knee-deep in other projects or I'd jump on it myself. If nobody else gives it a try, I might once things ease up in a few weeks.

p.s. My current project is a PCB machine - essentially just a small CNC milling machine/engraver thingy - so this would be right down that alley. I'm only about 1/3 finished so I expect to be tinkering with it for another 4 to 6 weeks since it's just a hobby thing. It would be cool to make the trav-a-dial a more-than-one-man project just because... then once it's done post the code and stuff for anyone to download and customize if desired.

I'd say it shouldn't be too difficult to make a few permutations of it. One could use the display John posted in the original and convert quadrature to whatever that display needed, one could just stay with quadrature and not need Chinese translation and have it's own built-in board. Just brainstorming here, but maybe it would be a worthwhile thing to figure out.

My guess is that making a circuit from scratch would cost around $20 including an enclosure. How much would an encoder be? There is a place that sells them on mylar for fairly cheap I think.
http://usdigital.com/products/encoders

Interesting idea. I wish I had some free time at the moment.

MCS
01-16-2011, 01:49 PM
Well, your calling me the one that wears high heals? Your the one defending the use of a "wire" language vs gnu C.
But again, some one is going to come on here, call us both panzies for not using Assembly...
Ahh the ranks of nerdom..

Yes, you're both panzies!:D

sansbury
01-16-2011, 01:51 PM
But again, some one is going to come on here, call us both panzies for not using Assembly...
Ahh the ranks of nerdom..

Precisely :)

It's all horses for courses. If you're making 100,000 pieces of a gizmo, the cost difference between a $0.35 PIC and a $2.00 AVR can easily justify an EE spending a week or two fiddling with assembler. For me, if a $30 board lets me get my one gizmo working in an hour or two versus spending a couple days learning to do the same thing with a $5 chip, sign me up for the $30.

Likewise, if I build something useful with an Arduino, anyone here could easily go and buy a board and load up my sketch code with pretty minimal fuss compared to needing an AVR programmer and development environment. AVR is one of the easiest embedded environments out there, but that's still a big hill to climb for people with very limited CS/EE skills.

dp
01-16-2011, 01:53 PM
How is a digital Trav-A-Dial different from a DRO?

sansbury
01-16-2011, 01:57 PM
Back on the OT, it would be very easy to read a quadrature signal and output a measurement on an LCD using the Arduino. I'm doing that plus a bunch more for a CNC pendant with an MPG and 20x4-line display right now. I could probably do the software in a couple hundred lines of code.

The engineering would be mostly in matching the wheel circumference to the encoder count so that one count produces a useful unit of distance. John, when you mentioned tilting the wheel, were you thinking of something as a way to adjust this after the fact?

J Tiers
01-16-2011, 02:04 PM
How is a digital Trav-A-Dial different from a DRO?

Same crap, different sock.

Except lower price, no features. May as well get a DRO, except that will cost about what the lathe cost (typical SB/Atlas/Logan) and do the diameters also, not to mention re-zeroing, better resolution, calibration, and other un-necessary frills;)

But for about the price of a chicom caliper and the blocks of scrap required to clamp it in place, a digital trav-A-Dial might be good.

Thing of it is, I rarely need to do that particular measurement, and when I do, it is usually a short distance, that I may need to a tight tolerance, with many other measurement options. Not sure I'd trust a DRO either.

It would take some doing to get it even within 0.001 over say 18".
Still, for $35 it would be fine.

Black Forest
01-16-2011, 02:04 PM
Don't give up on this idea John. It is a good one. No long scales is a big plus.

macona
01-16-2011, 02:07 PM
Reading a quadrature rotary encoder and having it send data to a digital display should be a pretty easy microcontroller project. Cheaper ala trav-a-dial than buying long scales for a lathe too (good idea JS!). Also, being micro-controlled, you wouldn't need a radiused wheel to adjust for proper readings too since the micro could adjust it electronically to probably 10 decimal places if you wanted it to.

Whether in Arduino, C or machine language, just pick your preference and from a functional perspective any of them would be more than adequate.

So:
-Rotary encoder attached to wheel that contacts the lathe way.
-AVR microcontroller circuit (Arduino or home-brew) to do the thinking.
-Digital display (I prefer large LED over LCD because it's easier to read, but the circuit will be about the same).
-Enclosure for the micro and display to make it look slick and protect from flying debris.


Problem is dumping it into the goofy serial format that the display wants.

And then building a housing that will hold and encoder, bearing, and drive wheel, and the spring mount to hold it to the ways (They use quite a bit of pressure) you are near the cost of a regular DRO for a small lathe.

The wheel on a trav-a-dial is about 1.5" diameter. I am guessing that large for the larger contact area. You would need a 2500 line encoder to give you .0005 resolution. You are looking at a $75 encoder.

tyrone shewlaces
01-16-2011, 02:31 PM
Problem is dumping it into the goofy serial format that the display wants.

???
Sending data to a digital display is one of the first basic things that microcontroller hobbyists always do. Most of the LCD displays have a built-in driver that makes it dead easy.


And then building a housing that will hold and encoder, bearing, and drive wheel, and the spring mount to hold it to the ways (They use quite a bit of pressure) you are near the cost of a regular DRO for a small lathe.

Come on man! That's the fun part ;)


The wheel on a trav-a-dial is about 1.5" diameter. I am guessing that large for the larger contact area. You would need a 2500 line encoder to give you .0005 resolution. You are looking at a $75 encoder.

Mostly true at first inspection. But I'd bet money that there is a fairly simple solution to skinning that cat and bringing it down closer to $25 or so. For instance, I do a fair amount of precision work but .001" accuracy would do for me most of the time. I don't think I'd trust a genuine trav-a-dial any more than that. I wouldn't trust a .0002" DRO to actually face two surfaces to a tight tolerance for me either. I have to rely on the machinist to take care of that. Displays only tell you where the scale is and don't factor in the other dozen variables.

The Artful Bodger
01-16-2011, 03:39 PM
You could always try something likes this (which I have posted some time ago):-
http://bbs.homeshopmachinist.net/showthread.php?t=44306
http://farm4.static.flickr.com/3486/3294903470_996ce149f4.jpg

Accuracy for carriage travel would be easier to achieve than for slide travel as the later requires flexible cable.

MCS
01-16-2011, 04:09 PM
You could always try something likes this (which I have posted some time ago):-
http://bbs.homeshopmachinist.net/showthread.php?t=44306

Accuracy for carriage travel would be easier to achieve than for slide travel as the later requires flexible cable.

My main objection toward the "Chinese protocol" is the slow display update frequency.

I made LED displays with an update frequency of 100Hz, which is over the limit of the human eye.

The result is a dynamic display, where with the "Chinese protocol" the response is later, it's more static, so the human reaction is to wait until it's stabilised.

macona
01-16-2011, 04:20 PM
???
Sending data to a digital display is one of the first basic things that microcontroller hobbyists always do. Most of the LCD displays have a built-in driver that makes it dead easy.


Not a generic LCD, the one John posted. That takes the serial signal from a digital caliper or scale and displays it elsewhere. The protocol is documented but if I were to do it I would just make a new display. and VFD, LED, or OLED. I like displays that emit light.

Ooo.. Even better. Nixies!!

macona
01-16-2011, 04:26 PM
My main objection toward the "Chinese protocol" is the slow display update frequency.

I made LED displays with an update frequency of 100Hz, which is over the limit of the human eye.

The result is a dynamic display, where with the "Chinese protocol" the response is later, it's more static, so the human reaction is to wait until it's stabilised.


That has been my gripe with capacitive scale dros. Slow. My anilam continuously updates the display as far as my eyes can tell.

John Stevenson
01-16-2011, 04:44 PM
You could always try something likes this (which I have posted some time ago):-
http://bbs.homeshopmachinist.net/showthread.php?t=44306


UK until you want 40" scales.

The Artful Bodger
01-16-2011, 05:08 PM
UK until you want 40" scales.

How frequently do you make a cut that is of that length? You will notice that my slides are way less than the possible carriage travel and thats why the slide respositions on the cable via the slip joint.

John Stevenson
01-16-2011, 06:46 PM
I very rarely make a cut of that length but I don't want to be connecting and disconnecting cables just to reach the end of the bed.

Whilst your idea is suitable for you with two calipers mounted on the rear of the machine it would not suit me as my guard travels with the toolpost and would obscure it and there is no other place to mount these that is easily visible.

Not being an EE I though what I thought was a sensible idea into the ring, it must be sensible as it's commercially made at high $$$, but translated to cheap encoder and readout that has no length limitation and easy to mount.

tyrone shewlaces
01-16-2011, 07:26 PM
... and readout that has no length limitation and easy to mount.

Bingo! One size fits all. That's the big appeal for me. If this thing could be made to work, and I don't see any big reason why it can't, then the biggest hurdle for most folks will be just mounting it on their saddle somewhere.

And Macona - yes I think nixies would be the thing. That would decorate my vintage lathe just about right I think:)

As far as streaming to the display John showed, I think that's been done several times already too, so one less thing to figure out. I'll steal others' work any day if it saves me some time. Plus I'm not that smart anyway so lucky for me that solves a dead end.

I was talking to a buddy about it earlier today and he mentioned the idea of using a mouse wheel for the encoder, which is obviously not nearly high enough resolution. But that got us talking about using lasers for measuring without the need for contact nor encoders at all. Now that would be exciting if it was doable. If they can shoot the moon with a laser and measure the distance to within an inch 30 years ago, could we do it on a lathe these days to within .001" or better? Hmmmm.

macona
01-16-2011, 07:39 PM
Lasers can measure down to the sub-micron range. They use them for mapping silicon wafers. I have a few of them, they will measure down to .125u but their range is only 500u. Lasers are also used to map ball screws on cnc machines. But they are not something you want in a dirty shop environment.

I think there are such things as a digital trav-a-dial. Problem with them is accumulated error. The further the travel the more its going to be off.

Another option is a spring loaded cable reel at one end. There are a couple companies that make DROs for large machines based on this.

John Stevenson
01-16-2011, 08:01 PM
Another option is a spring loaded cable reel at one end. There are a couple companies that make DROs for large machines based on this.

This is the type of DRO I bought many years ago, I still have them on the cross slide of two lathes but to be honest on laths cross slides they are useless because of the doubling up of errors caused by diameter machining, DRO is 1 thou out and that's 2 thou on the work.
Given that glass scale DRO's have come down in price, these 'cheap' ones have gone up but their resolution has come down.

http://www.bwelectronics.co.uk/systems/sysintro.html

J Tiers
01-16-2011, 09:19 PM
The big killer on the cheap Trav-A-Dial is the wheel and encoder....

The wheel must be quite accurate, although the round edge idea means it need not be 'absolute".... And the encoder resolution needs to be good , meaning expensive.

But.... if the encoder were of lesser resolution, and the wheel were geared UP in one step to the encoder, one could use a much lower resolution, cheaper, encoder. A single anti-backlash gear setup would not cost a lot in china. Electronics would be cheap, uPs are basically priced on the basis of program development cost, not part price per se.....

What is the usual DRO overall end-to-end repeatability/accuracy? You'd not want much less, I suspect about 0.003 over the whole bed length would be considered lousy..... even for a cheap device.

KEJR
01-16-2011, 09:55 PM
I seem to have come by late, but how would you get the mechanics of this right? Is the knurled wheel going to dig into your lathe bed underside? Won't this cause more accuracy problems than the encoder itself? I have to admit I've never used a mechanical travedial, so some of these ideas might be lost on me.

As far as the electronics, I agree with most of the others. A simple project with AVR or Arduino wouldn't take much. They sell quadrature chips to offload the micro if needed. All you "really" need is a micro, LCD (or 7 segment) and a few buttons. This can be packed into a box pretty easily, especially if your display can be bezel mounted. I thought of doing a readout at one point with quadrature inputs, but couldn't think of a good sensor unit to hook up to. Maybe if you had precision ground ballscrews on a mill, for instance.

I can't help but think that by the time you are done you will end up with one of those cable driven systems and then ask yourself "why didn't I jsut buy one of those $500 DROs?? with glass scales" Don't get me wrong, if this can be done mechanically very simply it would be a cool project.

I do think you could want something better on your cross slide(if used ont he lathe) , but alot of folks seem to be OK with large dials and careful measurement as you approach your dimension (or an indicator mounted on a moveable slide, for instance). Maybe you could do an encoder on your cross slide which should be as good as your screw if you take out backlash.

Keep the ideas coming, this makes for good reading...

KEJR

KEJR
01-16-2011, 10:29 PM
I haven't done and exhaustive search, but US digital has an encoder that is 512 lines (2048 pulses post quadrature) for $37


http://usdigital.com/products/encoders/incremental/rotary/kit/e8p/


We'd have to double check the price and all that, but with a 1/2" diameter contact wheel that gets you under 0.001" on your resolution (but not down to 0.0005")

This one is more money but offers 2880 cts/rev at $47 qty 1. $34 qty 10.

http://usdigital.com/products/encoders/incremental/rotary/kit/e7p/

You will need a 4x quadrature decoder to get these resolutions, but this is readily available from $5-$10 or so.

KEJR

dp
01-16-2011, 11:14 PM
Any friction device will get a buildup, bump into chips, pick up cat hair or bellybutton lint, what ever. If you need to know more about what can go wrong with rollers in a non-clean working environment, shake the ball out of your old non-optical mouse. bleh! Then check inside at the rollers. Quadrature bleh!

J Tiers
01-16-2011, 11:55 PM
The mechanical mouse is crazy simple, relative, cheap, but functional.

I notice that the optical type, which I had more than 20 years ago, are now totally mainstream. The old ones needed a special pad, but for 15 years or so, that has not been required.

Anyone know the sensitivity of the present optical mouse in actual movement? it seems somewhat speed dependent, and is probably around 20 or 30 thou. Direction is obviously detected.

I wonder if that system could be used..... probably with a mechanical magnifier of some type... simplest would be different sized disks.

if it were possible, that could be pretty darn simple, and cheap. But I have no clue if there is any way to calibrate it. I don't know how it works, probably somewhere in the ******pedia it is explained.

Edit:
Ah, a short search produces the info... it is an image comparison camera system, with DSP etc.... makes sense, and is quite likely to be usable, but not in the form used in the mouse.... some different programming would be required.
http://www.pantherproducts.co.uk/Articles/Mice/Optical_Mice.shtml

dp
01-17-2011, 12:13 AM
The best mouse ever invented is the new Mac mouse except for the carapace there are no moving parts. And no wire. http://www.apple.com/magicmouse/

That is what an encoder should be like - Bluetooth enabled, too.

Edit: Ewww - just watched the video on that page. It's a bit dated...

The Artful Bodger
01-17-2011, 12:36 AM
I very rarely make a cut of that length but I don't want to be connecting and disconnecting cables just to reach the end of the bed.

What happens is the caliper hits its end stop and the cable slips, there is no need to do anything but the device stops working.



Whilst your idea is suitable for you with two calipers mounted on the rear of the machine it would not suit me as my guard travels with the toolpost and would obscure it and there is no other place to mount these that is easily visible.
The lecturn is attached to the carriage.

form_change
01-17-2011, 04:20 AM
If only there was a gear rack that ran the full length of the bed - then you could mount a dial with an anti-backlash gear on it and read directly.

Oh wait...

Michael

mike4
01-17-2011, 04:39 AM
What happens is the caliper hits its end stop and the cable slips, there is no need to do anything but the device stops working.



The lecturn is attached to the carriage.
Have a look at many of the latest all terrain cranes they use a cable attached to the boom which drives an electronic encoder to measure how much the boom is extended , this is fairly accurate and could be adapted to a lathe or mill if the accuracy could be matched to machining requirements .
These work in often very dirty enviroments without problem.

My 02c .

Michael

BillTodd
01-17-2011, 08:20 AM
Anyone know the sensitivity of the present optical mouse in actual movement?



The accuracy of the mouse is down to the number of pixels in the linear arrays on the chip. It is these that detect the light and shade from the reflected surface, as long as the surface has enough variation to 'cover' the chips surface then the sensitivity should be that of the chip.

One snag, is that the mice chips tend to switch off at regular intervals thus potentially losing steps (I also suspect they may average steps to prevent glitches, since the claimed resolution is not actually required by the PC).

It would be interesting to see a data sheet for the mouse chip, I wonder if they have a diagnostic mode that would allow the pulses to be seen and counted?



it seems somewhat speed dependent,
That's a function applied by the driver within the PC.

Bill

ikdor
01-17-2011, 08:43 AM
As you can get optical mice that have over 5000 dpi of resolution and are wireless, they would make a really nice dro.
You could use a netbook as the DRO display.

Igor

J Tiers
01-17-2011, 09:18 AM
Can you relate the pixels to units on the surface in some way to get a consistent cal? I don't see a good way to do that directly, unless I am not seeing the system correctly.

You would have to use focal length to relate pixels of observed movement, and distance, it seems.

BillTodd
01-17-2011, 01:35 PM
Can you relate the pixels to units on the surface in some way to get a consistent cal? I don't see a good way to do that directly, unless I am not seeing the system correctly.

The actual distance moved per pulse is not that critical, as long as it is less than, or at least of the same order, as the required resolution. Calibration is a simple mathematical process.


You would have to use focal length to relate pixels of observed movement, and distance, it seems.

Sensing distance does not effect the accuracy too much (try lifting your mouse 1/8" off the mouse mat and see what little difference it makes), but it would make sense to track the flattest possible surface at a fix height.

Bill

BTW anyone wahting to use those cheap chinese scales with a PC might be interested in my pic based DRO here (http://www.cnczone.com/forums/pic_programing_design/62704-pic_based_dro.html)

ikdor
01-17-2011, 01:48 PM
Ah, got an example sensor here:
http://www.avagotech.com/pages/en/navigation_interface_devices/navigation_sensors/laserstream/adns-9500/
5000cpi resolution, 150ips max speed

And it costs a whopping $6 at Future and $0.30 for the lens.
Not too shabby considering the complexity of the sensor.

It will probably drift when you go back and forth as it doesn't use absolute references, but you could put a homing sensor fairly close to where you normally work. Not as good as a real DRO, but could be functional for a fraction of the cost.

Now, which of you retirees has some spare time.....:D

Igor

The Artful Bodger
01-17-2011, 01:54 PM
I am not confident a mouse is the way to go...

Just do this simple test, open Paintbrush and move the mouse over the drawing area and you will see the coordinates displayed at the lower right.

Now put something solid on your desk top and move the mouse up to it, note the coordinates, move the mouse away and bring it back. With my optical mouse they are never the same. Bad news I think.:rolleyes:

BillTodd
01-17-2011, 01:55 PM
Ah, got an example sensor here:
http://www.avagotech.com/pages/en/na...eam/adns-9500/
5000cpi resolution, 150ips max speed

Looks interesting :)


I found this older Agilent mouse sensor (http://www.ian-nic.pwp.blueyonder.co.uk/mouse/ADNS2051.PDF) which has a 800 cpi and four useful quadrature outputs

ikdor
01-17-2011, 02:03 PM
I am not confident a mouse is the way to go...

Just do this simple test, open Paintbrush and move the mouse over the drawing area and you will see the coordinates displayed at the lower right.

Now put something solid on your desk top and move the mouse up to it, note the coordinates, move the mouse away and bring it back. With my optical mouse they are never the same. Bad news I think.:rolleyes:

Ah, but this is due to the fact that windows uses a non-linear gain for the mouse pointer. The faster you move the mouse, the higher the ratio between screen vs. mouse inches.
Your point is still valid though, it will still not read the same going back and forth. The question is, how bad is it exactly......

Igor

BillTodd
01-17-2011, 02:31 PM
I am not confident a mouse is the way to go...

Just do this simple test, open Paintbrush and move the mouse over the drawing area and you will see the coordinates displayed at the lower right.

Now put something solid on your desk top and move the mouse up to it, note the coordinates, move the mouse away and bring it back. With my optical mouse they are never the same. Bad news I think.:rolleyes:

This is mainly caused by the acceleration function in the mouse driver (and the fact that Windows only works at an 18fps rate)

Bill

oldtiffie
01-18-2011, 01:03 AM
I will see if I can post some pics later of my very effective, easy and cheap digital caliper as an electronic/digital "Trav-a-Dial" on my lathe - later.

John Stevenson
01-18-2011, 04:40 AM
I will see if I can post some pics later of my very effective, easy and cheap digital caliper as an electronic/digital "Trav-a-Dial" on my lathe - later.

Will it travel 40" to 84" and have a remote display where I want it?
If not don't bother as this post is not about that.

This post is aimed at the various clever people on this board who can interface a cheap encoder with a cheap remote display.

Weston Bye
01-18-2011, 06:07 AM
Some 16-18 years ago or so, I built a number of interfaces for gathering, reformatting and transferring data from up to 24 Mitutoyo indicators to a PC via RS232 serial. I did this with the venerable Z80, and I built 25 of the systems for a Ford plant. This was in the days when Coordinate Measuring Machines were just coming into service there and they needed a temporary solution. This plant was using fixtures with multiple measuring points - multiple Mitutoyo digital indicators and a technician with a clipboard (and resulting transposition errors). They needed a more reliable data gathering method. There were other such systems available, but not packaged for industrial (ab)use, and not with such a large number of indicator inputs.

I presume that the Japanese Mitutoyo data format is what you guys are referring to as Chinese. While completely satisfactory for static measurement, the data transfer/translation speed will place a limit on practical feedrates. It seems to be the long way around to translate quadrature to Chinese to Cheap Display, unless... the parts are available for pennies.

The "proper" way, from an engineering point of view, would be quadrature to a PIC microcontroller to a LCD display. I have done such things with TTL logic chips, with the Z80, and with firmware resident microcontrollers, but alas, not with a naked PIC - the cheapest way.

Another thing for my bucket list - but surely someone will get to it before I do.

J Tiers
01-18-2011, 08:51 AM
The actual distance moved per pulse is not that critical, as long as it is less than, or at least of the same order, as the required resolution. Calibration is a simple mathematical process.




Not so fast..........

If you use a mouse sensor to read a scale for an encoder, you need to make sure you get readings from the marks, and not from some other "feature". And you need to make sure you read every mark sequentially, or you will drop counts and accumulate errors.

That's where I am heading with that question..............

That sort of issue makes or breaks the encoder solution.

BillTodd
01-18-2011, 01:09 PM
If you use a mouse sensor to read a scale for an encoder,

Keep up at the back... ;)

No, the mouse sensor would be use to 'read' a surface (e.g. the side or underside of one of the slide-ways) on the machine directly. All the surface needs is sufficient granularity to keep the sensor happy, then the resolution is determined by the sensor's image array.

For those not familiar with how an optical mouse works, this is from the Avago Technologies sensor datasheet;


The sensor is based on LaserStream technology, which
measures changes in position by optically acquiring
sequential surface images (frames) and mathematically
determining the direction and magnitude of movement.
It contains an Image Acquisition System (IAS), a Digital
Signal Processor (DSP), and a four wire serial port. The
IAS acquires microscopic surface images via the lens and
illumination system. These images are processed by the
DSP to determine the direction and distance of motion.
The DSP calculates the delta x and delta y relative displacement
values. An external micro-controller reads the delta x and delta y
information from the sensor serial port.



you will drop counts and accumulate errors
That's a possibility with most types of DRO sensor.

How repeatable are Trav-a-dials? (John?)

Bill

The Artful Bodger
01-18-2011, 01:45 PM
Bill, if I understand correctly the optical mouse reads displacement between frames. Nothing is perfect and the displacement value cannot have an infinite resolution so there is always some error between frames. With a large number of frames these errors add up.

It seems to me that the sensor must actually 'read' a fixed scale to get the results we desire.

Bmyers
01-18-2011, 02:02 PM
How repeatable are Trav-a-dials? (John?)

Bill

From trav-a-dials website:
The result is a mechanical system that is capable of accurate measurement (after calibration) of .0003” and near absolute repeatability.

MCS
01-18-2011, 02:08 PM
From trav-a-dials website:
The result is a mechanical system that is capable of accurate measurement (after calibration) of .0003” and near absolute repeatability.

The trick lies in the use of the word "near".

The exact moment it has to be "absolute" it is "near". Which will, according to Murphy, result in the removal of too much metal, instead of too little.

Form "near" to "absolute" can be a long journey.

BillTodd
01-18-2011, 02:29 PM
Bill, if I understand correctly the optical mouse reads displacement between frames. Nothing is perfect and the displacement value cannot have an infinite resolution so there is always some error between frames. With a large number of frames these errors add up.

All the device does is store a count of how many pixels is has to shift one frame onto the previous to get the best correlation between the two. The camera's pixel spacing determines the resolution.

Think of it this way:

In a typical glass scale there's a regularly spaced grill (the scale) and a sensor (with a somewhat fuzzy light source), (actually there's a pair of sensors but that's not important now). As the two move past each other, the grill regularly casts a shadow over the sensor, the changes in light level are counted.

In the mouse sensor, there's a regularly spaced set of sensors and the 'shadow' (light-level variation) is cast by the obliquely illuminated surface. As the sensor moves over the surface, the DSP calculates how many pixels the 'shadow' has moved between frames. The pixel count is accumulated.

BillTodd
01-18-2011, 02:57 PM
Some 16-18 years ago or so, I built a number of interfaces for gathering, reformatting and transferring data from up to 24 Mitutoyo indicators to a PC via RS232 serial. I did this with the venerable Z80, and I built 25 of the systems for a Ford plant. This was in the days when Coordinate Measuring Machines were just coming into service there and they needed a temporary solution. This plant was using fixtures with multiple measuring points - multiple Mitutoyo digital indicators and a technician with a clipboard (and resulting transposition errors). They needed a more reliable data gathering method. There were other such systems available, but not packaged for industrial (ab)use, and not with such a large number of indicator inputs.

:) I cut my teeth on a Z80 too - my favourite microprocessor.


I presume that the Japanese Mitutoyo data format is what you guys are referring to as Chinese. While completely satisfactory for static measurement, the data transfer/translation speed will place a limit on practical feedrates.

The Chinese protocol is simpler than the Mitotoyo one, just a string of bits in either Binary or BCD depending on the type of scale. The data update rate is usually only 300mS per reading, but can often be increased to 20mS by toggling the data and clock lines. At 20ms the display is just about usable as a DRO


The "proper" way, from an engineering point of view, would be quadrature to a PIC microcontroller to a LCD display. I have done such things with TTL logic chips, with the Z80, and with firmware resident microcontrollers, but alas, not with a naked PIC - the cheapest way.


I've no doubt you could do it :)

Bill

Peter.
01-18-2011, 03:27 PM
For a simple test I just hot-glued my PC mouse to the head of my Mitutoyo digital caliper and clamped the depth probe to a small piece of wood on my desk to anchor the caliper. I glued a couple of guides to the desk to constrain the mouse movement to x-axis only.

Fired up Draftsight and ran the mouse to the caliper end-stop and back a few times. Four times running the same co-ordinates came up in draftsight when it hit the stop, then it hit a totally different value and repeated that a few times. I think this is the windows accelleration feature sensing different mouse moving speed.

Not much of a test and probably full of flaws but it does show that the sensor is somewhat repeatable. It also showed that any slight lift of the mouse changed the x co-ordinates so any sysem using this would have to have the sensor orientation very tightly constrained.

Perhaps a u-shaped track with the sensor mounted in a low-friction slider, running in the track? More like a DRO scale than a trav-a-dial I guess.

The Artful Bodger
01-18-2011, 03:33 PM
It is not too difficult (trivial really) to disable mouse acceleration.

http://www.nextlevelgamer.com/tweaks/disabling-mouse-acceleration-in-windows-xp

Peter.
01-18-2011, 03:35 PM
I had already set all the reg values to 0 but accelleration remains. Easy to check just by watching the screen and moving the mouse at different speeds.

BillTodd
01-18-2011, 04:00 PM
I had already set all the reg values to 0 but accelleration remains. Easy to check just by watching the screen and moving the mouse at different speeds.

Peter the acceleration could be built into the microcontroller within the mouse (it would make sense for a mouse to have this feature as the polling rates of USB or PS2 aren't great).

I've found a mouse with a 800cpi Agilent sensor (for which I have a data sheet), I may have a tinker if I can find the time. :)

Bill

The Artful Bodger
01-18-2011, 05:06 PM
I tried mouse acceleration off and it is no good.

Even moving the mouse a short distance slowly between two fixed stoppers on the desk the numbers slowly creep away from what was started with.:confused:

Mouse is no good, even if we took polling rates and acceleration right out of the picture the mouse can never get (in navigation terms) a 'fix' and is only as good as dead reckoning.:(

oldtiffie
01-18-2011, 06:10 PM
It seems to me that the thread is wandering away from the requirements of the OP.


OK new thread to move away from Black forests thread.

We finished up asking why it was not possible to use a cheap encoder and one of these also cheap Chinese remote displays.

http://www.arceurotrade.co.uk/imagecache/3286dc1a-1abf-4d1d-8f1b-9b4a00bab943_250x125.jpg

Simple cheap encoder with a hardened knurled radiused wheel that can be tilted to get the correct circumference driving into one of these units.

The protocol for the Chines signals is documented on the Shumatech site which is down at the moment but I may be able to get it elsewhere.

Basically it's a reversal of what Scott Schumacher has done.

Any idea's or is anyone up to the job ?

Or is the OP requirement neither practical or/nor possible?

tyrone shewlaces
01-18-2011, 06:23 PM
It seems to me that the thread is wandering away from the requirements of the OP.
Or is the OP requirement neither practical or/nor possible?

No it's not. And it's definitely practical and possible. It's just a matter of working the problem.
The reason it's not straying away from the OP is that we've simply wandered down the path of having a non-contact reading method rather than a friction wheel like a trav-a-dial which, being mechanical, requires a mechanical mechanism to do its thing. If we're going down the electronics path, why not explore a non-contact reader since a micro could convert that easily to digital readout - IF the non-contact method is repeatable and accurate. If it is, a non-contact lathe longitudinal readout would be dead easy to make and attach. If it's not, then the rotary encoder and friction wheel will work and then it's a matter of working that problem. It would work as well as a trav-a-dial anyway.

oldtiffie
01-18-2011, 06:32 PM
Thanks TS.

I'd have thought that a non-contact method would be subject to some adverse conditions if it were operating in the same environment as a Trav-a-Dial (or many lead screws).

The real Trav-a-Dial seems to do pretty well "down there".

Would it be possible or practical to attach an encoder to a real Trav-a-Dial or at least the wheel from a Trav-a-Dial?

I'd imagine that similar accuracy to that of a Trav-a-Dial is required.

tyrone shewlaces
01-18-2011, 07:29 PM
I'd have thought that a non-contact method would be subject to some adverse conditions if it were operating in the same environment as a Trav-a-Dial.

True, but a friction wheel is exposed to some less than ideal conditions too. I've scrapped parts before due to crud with a trav-a-dial and my own lack of lathe hygiene. The thing with non-contact, non-encoder sensor is that it would read dirt the same as a clean surface and not gather up the dirt along the way (generally speaking).
I don't know if an optical surface thing like is being discussed would actually work either because I don't know that kind of stuff. But if it did work, it would solve a lot of problems.


Would it be possible or practical to attach an encoder to a real Trav-a-Dial or at least the wheel from a Trav-a-Dial?

I'd imagine that similar accuracy to that of a Trav-a-Dial is required.

Possible and probably practical, but the obvious problem there is the requisite scrapped out trav-a-dial for hacking. It's just some sort of wheel and thus probably a simple lathe project.
The mechanical accuracy would come from just having the wheel be concentric, and achieving non-slippage. Any adjustment for scaling the length readout to display accurate travel would be a simple correction factor entered into the microcontroller, thus the friction wheel could vary quite a bit from a "perfect" diameter and be easily calibrated to suit.
The trav-a-dial wheel has a convex circumference which you can adjust mechanically to calibrate. To do it electronically would likely be simpler and much faster and not require a convex surface unless otherwise desired for some reason.

Peter.
01-18-2011, 07:38 PM
I see no reason why both avenues could not be investigated.

I for one would love to be able to fit basic displays from either contact or optical senders especially if the cost was far less than the £1000 or so it would cost me to fit up both my lathe and mill. The electronics and programming is beyond me but I can follow instructions in that respect and build the hard bits.

tyrone shewlaces
01-18-2011, 07:56 PM
Since I'm sitting here thinking, I figured I'd throw out my knee-jerk idea for solving the friction wheel/encoder resolution problem.

I'd say the trav-a-dial wheel diameter is possibly about what you want for reliable contact. Say you want .0005" display resolution. That means 2000 steps per inch on an encoder. If the best you can do with a $30 encoder is 2000 per revolution, then you're talking about a 5/16" diameter. That's pretty small, so why not use a larger wheel to contact the lathe bed, then the opposite side of the diameter contacts a 5/16 (or whatever) wheel connected to the encoder. So the friction wheel is kind of an idler. This accomplishes the resolution problem and moves the assembly a little further out from the lathe bed to make room for the body of the encoder.

You probably want some kind of springy type of mount so consistent pressure is applied between surfaces. This is what the trav-a-dial does too so no need to re-invent that wheel.

Anyway, I thought I'd just throw that out there for brainstorming purposes. Personally, I hope an optical mouse type of sensor will end up being the solution though. That would just make things very easy.

The Artful Bodger
01-18-2011, 08:23 PM
Tyrone, having an idler would double the slip, if any.

tyrone shewlaces
01-18-2011, 08:27 PM
Tyrone, having an idler would double the slip, if any.

So don't have any, hehe:p

Mike Nash
01-18-2011, 08:58 PM
Here's the link to the no longer available quadrature to chinese scale format solution. The links on the left side of the page still contain the BOM, schematic and code needed. This makes a quadrature device look like a chinese scale. But be aware that many/most of the newer scales and calipers have moved to various new formats.

There are only 4 electronic components.

http://www.shumatech.com/web/products/qcc-100

J Tiers
01-18-2011, 09:18 PM
Keep up at the back... ;)

No, the mouse sensor would be use to 'read' a surface (e.g. the side or underside of one of the slide-ways) on the machine directly. All the surface needs is sufficient granularity to keep the sensor happy, then the resolution is determined by the sensor's image array.



I would think that was a tough one..... all kinds of crap accumulates, but an image is an image....... maybe it would be worse clean.

And, yes, the wheel shares an issue there.....

but the resolution probably does not approach 0.001" repeatably with typical blurry surfaces and poor edges on features. Oil, which typically drips down on the side of the ways, would not bother the trav-a-dial, but would probably mess up the mouse images.

I was thinking of the mouse as a very cheap encoder component, allowing a more coarse and cheap encoder wheel to perform like a very dense, fine one. Same thing as reading the surface, but "at one remove", and operating in more controlled conditions, with an optimum, clean, "target".

Robin R
01-18-2011, 10:19 PM
Would it be possible to use something like GPS technology, something like 3 radio transmiters around the shop and a reciever that identifies the direction the signals come from. Obviously this is just a concept and I have no idea how to put it into practice. If it was possible, it would use the same transmiters for all the machines in a shop and it might even be possible to move the reciever from one machine to the next.

oldtiffie
01-18-2011, 10:32 PM
What about laser technology? Use an emitter (remote) and a reflector (on the machine)? Can laser read to a resolution 0.001" or better over say 1 metre? 2? 3? 5?

Provided the emitter and the reflector are adequately clean and the "line of sight" not impeded - would it work?

If it did it would take all components out of the crap and crud environment of a machine.

The results could be "radio-ed" (wireless) transmitted to any convenient LCD/DRO or other type of read-out or display.

Could it be interferred with/by "stray" or other radiation or surges etc.?

Having something electronically sophisticated in a clean environment is one thing. Having it on a moving (or any?) part of a machine may be quite another entirely.

J Tiers
01-18-2011, 10:46 PM
What about laser technology? Use an emitter (remote) and a reflector (on the machine)? Can laser read to a resolution 0.001" or better over say 1 metre? 2? 3? 5?

Having something electronically sophisticated in a clean environment is one thing. Having it on a moving (or any?) part of a machine may be quite another entirely.

laser interferometry can do a lot better than we need..... at a high cost. Simpler methods might work too, but if the bits are not all on one machine, there are the issues of vibration and things coming between them to deal with. Probably better to all be on the machine itself.

Lots of electronic stuff is on, in, and around machines now, it isn't a problem with proper design. That's what us EE types get paid to do.... make it work without problems. Sometimes we even succeed in doing that.

The Artful Bodger
01-18-2011, 11:04 PM
Radio waves? Maybe the saddle position could be determined using Lecher Wires at UHF?

form_change
01-19-2011, 02:21 AM
My suggestion a way back about using the rack on the machine was bit tongue in cheek, but really, what is wrong with mounting an encoder on the carriage with an anti-backlash gear?
This would give an output continuously along the bed; because we are using an anti backlash gear would not have any (appreciable) hysteresis and cuts out all the stuffing around with lasers, GPS, wires, scales and aliens pacing the carriage in their flying saucers. All that would be needed is a readout that would take the pulses and turn them into distance. Depending on the functionality required and the components used, that might be as simple as a counter (for example 1 pulse = 1 thou)

Michael

topct
01-19-2011, 07:58 AM
There's one system I have seen, Spherosyn, that uses steel balls inside of a tube. The reader head wraps around it and slides along it's length. I'm going to guess that the head has a series of coils? or maybe hall effect transistors? that would either generate a voltage or a signal to drive the display as they passed over the balls.

Or maybe there is a way to read off of the teeth of the rack that's already on the lathe?

Mark K
01-19-2011, 10:50 AM
Questions for those who've installed Trav-a-Dials.

What sort of procedure is used initially to impress the 'rack teeth' into the lathe bed? Is there a feature in the mount that provides for this force?

And does this work with a hardened bed?

Curiously, Mark

Spin Doctor
01-19-2011, 11:07 AM
As neat as this idea is I've had a couple of ideas in the back of my head for a "Mechanical Readout" Perhaps cable driven with a counterweight at one end to maintain constant tension. maybe someting like a stopwatch that you could reset to zero

macona
01-19-2011, 12:17 PM
There's one system I have seen, Spherosyn, that uses steel balls inside of a tube. The reader head wraps around it and slides along it's length. I'm going to guess that the head has a series of coils? or maybe hall effect transistors? that would either generate a voltage or a signal to drive the display as they passed over the balls.

Or maybe there is a way to read off of the teeth of the rack that's already on the lathe?

The newall works inductively similar to a LVDT transducer or a resolver.

Problem with gear teeth on a lathe is the racks are not what one would call precision. They were never intended to be.

topct
01-19-2011, 01:48 PM
The newall works inductively similar to a LVDT transducer or a resolver.

Problem with gear teeth on a lathe is the racks are not what one would call precision. They were never intended to be.

How hard would it be to build a transducer like that? And to stuff some steel balls in a tube.

I will be sitting down for any replies. :D

The Artful Bodger
01-19-2011, 01:52 PM
As neat as this idea is I've had a couple of ideas in the back of my head for a "Mechanical Readout" Perhaps cable driven with a counterweight at one end to maintain constant tension. maybe someting like a stopwatch that you could reset to zero

My cable system rarely gets a positive response, but I like it!:p

The Artful Bodger
01-19-2011, 02:03 PM
I am wondering how accurate are the graduations on an engraved metal scale (a.k.a. ruler)?

Fix one of those to the bed and aim your USB microscope camera at it, maybe engrave a vernier scale to assist with interpolation.

oldtiffie
01-19-2011, 02:09 PM
Again - here is the OT.


OK new thread to move away from Black forests thread.

We finished up asking why it was not possible to use a cheap encoder and one of these also cheap Chinese remote displays.

http://www.arceurotrade.co.uk/imagecache/3286dc1a-1abf-4d1d-8f1b-9b4a00bab943_250x125.jpg

Simple cheap encoder with a hardened knurled radiused wheel that can be tilted to get the correct circumference driving into one of these units.

The protocol for the Chines signals is documented on the Shumatech site which is down at the moment but I may be able to get it elsewhere.

Basically it's a reversal of what Scott Schumacher has done.

Any idea's or is anyone up to the job ?

So far it seems that the (analogue) Trav-a-Dial with an accuracy of 0.0003" and reading to/of 0.001" over the entire travel distance is in no immediate or foreseeable danger of being replaced and/or being made obsolescent or obsolete (redundant?) by a digital equivalent (with remote display).

It seems that the Trav-a-Dial and John's unfulfilled requirement for a digital replacement for it are going to be around for a while yet.

The Artful Bodger
01-19-2011, 02:29 PM
OK new thread to move away from Black forests thread.

We finished up asking why it was not possible to use a cheap encoder and one of these also cheap Chinese remote displays.

http://www.arceurotrade.co.uk/imagecache/3286dc1a-1abf-4d1d-8f1b-9b4a00bab943_250x125.jpg

Simple cheap encoder with a hardened knurled radiused wheel that can be tilted to get the correct circumference driving into one of these units.

The protocol for the Chines signals is documented on the Shumatech site which is down at the moment but I may be able to get it elsewhere.

Basically it's a reversal of what Scott Schumacher has done.

Any idea's or is anyone up to the job ?

OK, addressing the opening post specifically:-

For a cheap "hardened knurled radiused wheel" take a look inside your coffee grinder!:)

The Artful Bodger
01-19-2011, 02:46 PM
Further ideas regarding the opening post:-

For a hardened wheel for the encoder I suggest using a knurling wheel.

For combined computing and display system drag that old laptop out from under your bed, you only threw it under there because the battery was stuft and you do not need the battery on your lathe! All the better if it can run in DOS, (not a virtual machine), real honest to goodness early 80's DOS! Then you can use the comm port in interrupt mode to read 4 inputs at least.....and it will not take an age to boot.:) I would need to check but I think even your latest laptops can be booted in DOS but it might not be very practical unless it has a floppy drive.

Precis, look for a laptop with a floppy drive and dont care about the battery.

The laptop will have a bigger display, be more flexible and be even cheaper than the CCRD (cheap Chinese remote display).

Starting with a knurling wheel of arbitary size means there will be some number crunching required to get the desired scale but even a lethargic 8086 processor can do that provided it is not bogged down with a GUI operating system (for example Windoze).

The hardest part will be making the encoder as it needs to signal both distance travelled and direction, it might even require some variation of Gray Encoding.

Even harder, impossible for many, would be the software development but fortunately on a forum like this we only need one person to do that bit!

Thats how I would do it.

Mark McGrath
01-19-2011, 02:48 PM
Questions for those who've installed Trav-a-Dials.

What sort of procedure is used initially to impress the 'rack teeth' into the lathe bed? Is there a feature in the mount that provides for this force?

And does this work with a hardened bed?

Curiously, Mark

The Travadial slides on it`s mount.There is an adjustment screw pushing on a flat spring leaf which pushes against the Travadial.The adjustment screw is tightened until the flat spring has deflected so much out of straight.Visible through a hole in the side of the bracket.
I would prefer to say the wheel had a very fine straight knurl rather than teeth,they were not all that prominent.
As for accuracy I don`t know.We used to calibrate over a set distance,that distance being whatever length standard we had available,usually 2-3ft.
Doing it that way only guaranteed it was accurate over that distance and did not take any account of accuracy at intermediate distances.
They`ve been on the go a long time,I fitted my first Travadials about 1970 and the electronic ones in the eighties.
My view on this project is that there seems to be a fair bit of reinventing the wheel involved and at the end of the day,accuracy,cheap and longevity do not normally go together.

form_change
01-19-2011, 03:07 PM
I don't agree that the rack is not accurate enough to measure from.
CVA had a dial on the traverse wheel as factory standard for lengths along the bed. I have a photo that Tim took of his apron that clearly shows this but I can't find it at the moment. It's on my list to make one up for my lathe.
Look at any set of dial calipers and it is a rack driving the dial.
I would have far greater confidence in a gear/ rack arrangement than a knurled wheel held on a hardened smooth metal surface by a spring.
One thing that I haven't noticed being mentioned is resolution - is the aim to be able to position accurately to 1 thou, 5 thou, 1/10 of a thou? The trav-a-dials I've seen are mechanically driven off lead screws so with backlash and wear how accurate are they? Depending on the age of the lathe I'm guessing that they could easily vary up to 5 thou anyway.

Michael

BillTodd
01-19-2011, 03:22 PM
This is an interesting video:

Here four optical mice are being used to sense position and orientation.

http://www.youtube.com/watch?v=nB1SjhCiXMQ&feature=related

topct
01-19-2011, 03:31 PM
Further ideas regarding the opening post:-

For a hardened wheel for the encoder I suggest using a knurling wheel.

For combined computing and display system drag that old laptop out from under your bed, you only threw it under there because the battery was stuft and you do not need the battery on your lathe! All the better if it can run in DOS, (not a virtual machine), real honest to goodness early 80's DOS! Then you can use the comm port in interrupt mode to read 4 inputs at least.....and it will not take an age to boot.:) I would need to check but I think even your latest laptops can be booted in DOS but it might not be very practical unless it has a floppy drive.

Precis, look for a laptop with a floppy drive and dont care about the battery.

The laptop will have a bigger display, be more flexible and be even cheaper than the CCRD (cheap Chinese remote display).

Starting with a knurling wheel of arbitary size means there will be some number crunching required to get the desired scale but even a lethargic 8086 processor can do that provided it is not bogged down with a GUI operating system (for example Windoze).

The hardest part will be making the encoder as it needs to signal both distance travelled and direction, it might even require some variation of Gray Encoding.

Even harder, impossible for many, would be the software development but fortunately on a forum like this we only need one person to do that bit!

Thats how I would do it.

Something like this,

http://www.lindsayengraving.com/other_interests/dro.html

He uses a cable.:D

The Artful Bodger
01-19-2011, 03:46 PM
Something like this,

http://www.lindsayengraving.com/other_interests/dro.html

He uses a cable.:D

I thought cables were verboten!:D

Bmyers
01-19-2011, 04:36 PM
googling "hacking optical mouse" shows we arent the only ones interested in position from a mouse

for example:
http://tinyurl.com/4rayuz6

Mark McGrath
01-19-2011, 04:42 PM
I don't agree that the rack is not accurate enough to measure from.
CVA had a dial on the traverse wheel as factory standard for lengths along the bed.
Michael

I think that dial is calibrated at 0.020" per division.

macona
01-19-2011, 06:19 PM
How hard would it be to build a transducer like that? And to stuff some steel balls in a tube.

I will be sitting down for any replies. :D

Possibly you could use two variable reluctance gear tooth sensors set 90 degrees out of phase to each other in reference to the gear teeth. Then you should have a quadrature sinusoidal voltage output. You could interpolate that and create a standard ttl quadrature signal out.

oldtiffie
01-19-2011, 07:42 PM
John Stevenson's OP (again):



OK new thread to move away from Black forests thread.

We finished up asking why it was not possible to use a cheap encoder and one of these also cheap Chinese remote displays.

http://www.arceurotrade.co.uk/imagecache/3286dc1a-1abf-4d1d-8f1b-9b4a00bab943_250x125.jpg

Simple cheap encoder with a hardened knurled radiused wheel that can be tilted to get the correct circumference driving into one of these units.

The protocol for the Chines signals is documented on the Shumatech site which is down at the moment but I may be able to get it elsewhere.

Basically it's a reversal of what Scott Schumacher has done.

Any idea's or is anyone up to the job ?

And to respond Michael(form_change):

I don't agree that the rack is not accurate enough to measure from.
CVA had a dial on the traverse wheel as factory standard for lengths along the bed. I have a photo that Tim took of his apron that clearly shows this but I can't find it at the moment. It's on my list to make one up for my lathe.
Look at any set of dial calipers and it is a rack driving the dial.
I would have far greater confidence in a gear/ rack arrangement than a knurled wheel held on a hardened smooth metal surface by a spring.
One thing that I haven't noticed being mentioned is resolution - is the aim to be able to position accurately to 1 thou, 5 thou, 1/10 of a thou? The trav-a-dials I've seen are mechanically driven off lead screws so with backlash and wear how accurate are they? Depending on the age of the lathe I'm guessing that they could easily vary up to 5 thou anyway.

Michael

Racks are not all that accurate and consistent even when new. They are pinned /doweled to the lathe bed. With localised (ie non-consistent/universal actual and on-going) wear rate the result from any wheel driven by the wheel will be (a) variable.

Driving from the lathe rack will be no more accurate than the threading dial which is driven by the lead-screw which should be at least as and probably more accurate than the rack.

As distance on the required (new) pseudo (digital) Trav-a-dial is only measured in one direction at a time, back-lash should not be an issue any more than it is with a digital DRO on a mill or lathe slides where the lead-screw back-lash and end-play is canceled out - as it if when screw-cutting with a lead-screw that has considerable gearing back-lash between the head-stock and the tool-post/carriage.

John Stevenson has it right - the core component is a very accurate "cutter" like the real Trav-a-dial wheel where the accuracy over-all is 0.0003" and the dials are calibrated in 0.001". (That's no mean effort for a "knurl(ed) wheel").

In the interim, and until such time as the gurus here find a purely electronic/digital solution with a remote display, the most feasible answer seems to be a standard DRO (accuracy = +/- 0.0002") on the back of the lathe seems to be the way to go.

Perhaps John was "having us on" as I'd bet he knew that the solution to his OP was and is the Holy Grail of many people and many BBS's where the discussion had petered-out (fizzled) and died with no viable solution as seems to be (or is/being) the case here.

Perhaps its a John Stevenson "wind up" as I am sure that if there was a solution John would know about it and either have posted it instead of his OP here or else he will let every one dash hither and yon and exhaust their efforts - and then (and only then) tell them/us.

Perhaps a USB camera over a real Trav-a-Dial will do the job pending a solution here.

The Artful Bodger
01-19-2011, 09:18 PM
The rack could be used with perfect confidence if there was enough electronic smarts to map the errors of the rack. Run the carriage back and forth using the lead screw while the electronics measures the error relating to each tooth.


(Then go and borrow ten more lead screws of the same pitch and run them through too and average the result.)

oldtiffie
01-19-2011, 10:27 PM
AB.

You seem to be making the assumption that the lead-screw - even when new - is accurate to the degree necessary to equate it to slip guages or similar. Unless you are very fortunate your lead-screw may not be that accurate even when new - the more so if it is worn.

"Mapping" a lead-screw, or a very accurate ball-screw or a DRO really does require very accurate (pre-set) "master rods" between the lathe head-stock or bed or temporary "stop" and the carriage.

Many lead-screws are just rolled (like a screw or bolt) on a continuous process, some (few) are screw-cut on a master thread and others are hardened and precision-ground for pitch and consistency.

The Trav-a-Dial has an accuracy over-all of 0.0003" which would require an extra-ordinarily good lead screw to match let alone better it if that lead-screw was to calibrate an electronic pseudo "Trav-a-Dial" or a common lathe rack.

oldtiffie
01-20-2011, 12:17 AM
As an interim or final solution, can anyone better this - with "digital caliper" accuracy"?

https://www.machineryhouse.com.au/viewImage.php?id=4564&imageNumber=0

https://www.machineryhouse.com.au/Products?stockCode=D665#

PDF manual is here:
https://www.machineryhouse.com.au/images/99691.pdf

The Artful Bodger
01-20-2011, 12:31 AM
AB.

You seem to be making the assumption that the lead-screw - even when new - is accurate to the degree necessary to equate it to slip guages or similar. Unless you are very fortunate your lead-screw may not be that accurate even when new - the more so if it is worn.

Thats why I said to get ten feed screws and take the average errors.

I would think that for a home shop the lead screw is about as accurate as we are likely to get.

oldtiffie
01-20-2011, 01:40 AM
Perhaps.

The electronic/digital Trav-a-Dial must have an accuracy of at least +/- 0.0005" (half a thou = 5 tenths) everywhere over its length which suggest an accuracy approaching +/- 0.0001".

Even a perfect lead-screw is "elastic" and even if the load (carriage) is quite constant and will be constant throughout the calibration and use (which I doubt) the extension of the lead-screw will vary as Young's Modulus (aka "E") (the ratio unit stress to unit strain within the elastic limit) suggests as the effective length of the lead-screw reduces as it pulls the carriage forward during calibration. I have neglected torque which even if constant will cause a variation in total "twist" as the effective lead screw varies with length.

The Trav-a-Dial "knurled wheel" works at its optimum just as and where it is - on the front of the carriage on the tail-stock side and bearing directly on the lathe bed.

The Artful Bodger
01-20-2011, 03:33 AM
Well Tiffie, I forgot who asked about using the rack for reference but if the lead screw is inadequate for calibrating the rack then I assume the only alternative would be to borrow a Trav-a-dial for the calibration.

oldtiffie
01-20-2011, 04:29 AM
Thanks AB.

I've been waiting for that answer.

I am not at all sure that many lathes just will be able to mount an analogue (the "real (analogue) deal") Trav-a-Dial as it is a fair sized unit and will require a mounting bracket to be made as well as some drilling and perhaps modification of or to the lathe.

As recalibration will be needed periodically or at short notice that "borrowed" Trav-a-Dial will need to be readily available over time. It would not be hard to really strain the relationship with who ever owns it.

I'd expect an accuracy of the order available from a good digital caliper over a distance of say 150mm/6" for the
https://www.machineryhouse.com.au/viewImage.php?id=4564&imageNumber=0

https://www.machineryhouse.com.au/Products?stockCode=D665#

that I posted earlier and all for about OZ$90 all ready set up.

Here is the accuracy and repeatability of one of my calipers:
http://i200.photobucket.com/albums/aa294/oldtiffie/measuring/Digital_caliper2-1.jpg

http://i200.photobucket.com/albums/aa294/oldtiffie/measuring/Digital_caliper4.jpg

I will post a few pics later of how I just use a pair of good shop digital calipers on the lathe - costs nothing, nothing to make (just stuff from the shop - and takes 5 minutes to set up and less to move or remove.

Later.

John Stevenson
01-20-2011, 04:53 AM
T

I will post a few pics later of how I just use a pair of good shop digital calipers on the lathe - costs nothing, nothing to make (just stuff from the shop - and takes 5 minutes to set up and less to move or remove.

Later.

Will they measure 40 " or more,????????? if not don't bother as this thread isn't about using fixed length calipers.

oldtiffie
01-20-2011, 05:13 AM
Try this one - 900mm effective measuring range.
https://www.machineryhouse.com.au/Products?stockCode=D667

I would not be surprised if longer items were available on order complete or made to order.

Only needs some-one to ask instead of sitting on their freckle.

As it is quickly demountable it may suit others.

My 6" caliper does all that I need in that regard.

BillTodd
01-20-2011, 07:25 AM
If you want a long scale solution (i.e. up to 25m/80' ) then this magnetic tape looks quite good:

http://www.machine-dro.co.uk/index.php?target=products&product_id=347

http://www.machine-dro.co.uk/index.php?target=categories&category_id=60

It would take some careful fitting to ensure absolute accuracy and would certainly require addition environmental protection, but at <£60 a metre (plus reading head etc.) it is cheap.

Bill

oldtiffie
01-20-2011, 07:45 AM
Thanks Bill - sincerely.

I could not really see anything beating that both for accuracy, any length, consistency as well as ease of installation.

I can't see cost being an issue given what it can do and where you can fit/install it.

It seems to meet every criteria (other than cost) in/for/by the OP.

J Tiers
01-20-2011, 09:06 AM
The problem with caliper-based measurement devices is that calipers aren't really that accurate anyway. Quite a bit of that is "mechanical", due to the really poor design of the item for precision measurement purposes, but the scale likely does not justify more than the general accuracy expected.

Now, before the screaming starts, lemme say I am distinguishing between "resolution" and actual accuracy...... where 'accuracy" means that when it says, for instance 15.173", the distance really IS that.

The way the scale is made is fine for 6" where the expectation of accuracy is most below 1", but perhaps not so much above that. One could easily argue that the trav-a-dial is unlikely to beat it, and that's a consideration that is worth a good deal of thought.

As for calibration......

The OBVIOUS method involves comparison of movement to either gage blocks , or jig bore rods. Both are made very accurate, gage blocks clearly more so. But jig bore rods usually have more possible distance span, especially if you use both sets. I reckon you can get over 26" with the set I have. You can cal the rods against blocks.

There are many other possible methods of cal, but the direct length comparison is simple and direct.

oldtiffie
01-20-2011, 09:26 AM
The problem with caliper-based measurement devices is that calipers aren't really that accurate anyway. Quite a bit of that is "mechanical", due to the really poor design of the item for precision measurement purposes, but the scale likely does not justify more than the general accuracy expected.

Now, before the screaming starts, lemme say I am distinguishing between "resolution" and actual accuracy...... where 'accuracy" means that when it says, for instance 15.173", the distance really IS that.

The way the scale is made is fine for 6" where the expectation of accuracy is most below 1", but perhaps not so much above that. One could easily argue that the trav-a-dial is unlikely to beat it, and that's a consideration that is worth a good deal of thought.

As for calibration......

The OBVIOUS method involves comparison of movement to either gage blocks , or jig bore rods. Both are made very accurate, gage blocks clearly more so. But jig bore rods usually have more possible distance span, especially if you use both sets. I reckon you can get over 26" with the set I have. You can cal the rods against blocks.

There are many other possible methods of cal, but the direct length comparison is simple and direct.

JT.

Here are the accuracy figures for one of my digital calipers.
http://i200.photobucket.com/albums/aa294/oldtiffie/measuring/Digital_caliper4.jpg

What you say is quite correct as accuracy degrades over distance from a digital "zeroing".

If I had a 40" (~1 metre) scale and if I were to zero it at any point and not exceed more than 6" from that zero, then the scale should maintain its "6" accuracy from the electronic/digital zero even when the digital is not physically co-incident the mechanical zero.

John Stevenson
01-20-2011, 11:35 AM
You have already posted the calliper lie sheet only six posts earlier - yawn.

The Artful Bodger
01-20-2011, 02:24 PM
I am not really sure how I could determine that something over 40" long is 'out' by a tenth. Thats one part in 400,000 and I suspect a mosquito farting in the shop would raise the tempurature enough to cause that error?

BillTodd
01-20-2011, 02:45 PM
Just had a reply from the Masaaki Kumagai, the guy who's been experimenting with mouse sensors for robotic uses.

He says the few lost counts in the video are due to the error from the velocity calculation and that otherwise the repeatability of the devices is good with few or no lost counts :)

Bill



This is an interesting video:

Here four optical mice are being used to sense position and orientation.

http://www.youtube.com/watch?v=nB1SjhCiXMQ&feature=related

(I wouldn't normally quote my own post but the original was some way back)

MCS
01-20-2011, 02:54 PM
I like the magnetic tape idea.

Sony has 2 types: SL331 (resolution 2 mu) and SL130 (resolution 0.01 mm)

Pricey, but maybe there is something similar to be found here.

Absolute distance is a function of temperature among other things, but close to absolute repeatability at a certain point is good enough for me. Sony has also AB-output (quadrature), which makes the rest of the electronics also easy. I have already the 24-some lines of AVR code for that.

BillTodd
01-20-2011, 03:17 PM
Sony has also AB-output (quadrature), which makes the rest of the electronics also easy. I have already the 24-some lines of AVR code for that.

I estimated the maximum speed I can move my carriage is about 200mm/s; with a 2um tape that's 100,000 pulses per second. Make sure your counters can keep up or you'll find it'll suddenly lose its position. (i.e. be wary of interrupt driven counters ;) )

Bill

Robin R
01-20-2011, 03:29 PM
I don't think anyone has posted this yet, it seems to fit the bill, other than perhaps on price. http://www.warrenmeasurement.co.uk/traksensor.htm

MCS
01-20-2011, 03:36 PM
I estimated the maximum speed I can move my carriage is about 200mm/s; with a 2um tape that's 100,000 pulses per second. Make sure your counters can keep up or you'll find it'll suddenly lose its position. (i.e. be wary of interrupt driven counters ;) )

Bill

I think you are slightly overestimating. 200mm / sec is 12 meter/minute.

I worked with an Gildemeister CNC lathe. It's maximum speed was 5 meter/minute, which was quite scary. Normally I set it at 1-2 meter a minute, still scary enough, because it's not the speed, but the approach which gets on your nerves.

For an AVR @16Mhz 100000 pulses a second is 160 instructions between pulses. That's a holiday.

BillTodd
01-20-2011, 03:45 PM
I think you are slightly overestimating. 200mm / sec is 12 meter/minute.

OK :) slight over estimate (Just tried it - best I managed on the lathe was ~60mm/s before my arm started to hurt :LOL: ) Maybe peak speed could reach twice this.

Bill

MCS
01-20-2011, 04:00 PM
OK :) slight over estimate (Just tried it - best I managed on the lathe was ~60mm/s before my arm started to hurt :LOL: ) Maybe peak speed could reach twice this.

Bill

Somewhere, somewhere on this world there will be the manual carriage speed championchips. Sprint and marathon.

It must be.

It's the eternal question: "What else can you do with it?"

BillTodd
01-20-2011, 04:59 PM
championchips

World class pun :thumbs-up: :)

J Tiers
01-20-2011, 10:19 PM
So.........

Do we REALLY have a "prototype" in the form of a "Trav-A-Dial", or does that device not quite come up to snuff either?

The caliper "lie sheet", which I paid no attention to the first time, does show that even they will only claim "optimistic" numbers on the order of 0.001 at 300mm, or about 12". I don't know how that would work out for 4 foot, it might be out several thou at least. One would assume the caliper-based measurement device would be similar. The scale probably isn't any more accurate, it is just a strip of plastic with the metal pattern on it, subject to stretch, etc.

A wheel against the bed might be no better, possibly a good deal worse, depending on a lot of factors, many of which can be calibrated out.

If the "Trav-A Dial" is capable of "X" accuracy, "we" are unlikely to do much if any better.

So the question is whether there is a reasonable shot at actually getting an accurate result from any wheel-based device. No matter if the "Dial" does, OR does NOT do it to 0.02 mm over a metre, for instance, why do "we" think that a wheel and "mouse guts" is going to do that well? Is that a realistic expectation in a perfect world, let alone in the environment it will exist in?

The design idea needs a good hard look at first principles, never mind details of how to do it, how much an encoder costs, etc. The design idea may simply be not feasible.

A "proof of principle" test is needed to show that it is realistically possible

The Artful Bodger
01-20-2011, 10:44 PM
The way I see it is any shop competant member of this forum could likely make a wheel that would wear itself a 'rack' in the bed and be as reliably repeatable as the Trav-a-dial.

There is no reason why we could not devise some sort of chopper disk or other encoder system to drive the type of Chinese display Sir John mentioned in the opening post.

So it is more a matter of will than anything else required.

Alternatives, and it has been entertaining to consider some of them, show it is really hard to come up with something that betters the principle of a rack and pinion where the smaller the pitch of the rack the greater the potential accuracy and the Trav-a-dial is in essence a rack and pinion system of very fine pitch.

J Tiers
01-20-2011, 10:55 PM
The way I see it is any shop competant member of this forum could likely make a wheel that would wear itself a 'rack' in the bed and be as reliably repeatable as the Trav-a-dial.



Ah, but how repeatable IS that? Have we seen the actual specs? Has anyone who has one tried to confirm the specs on their unit?

And.... the wheel may be less easy than you think..... let's see your shop-made version! Presumably after you are re-settled..........

SVS
01-20-2011, 10:57 PM
IMHO the Trav-A-Dial properly calibrated and maintained is good enough for semi precision work over long spans and anything but the highest precision over short distances. (rigid criteria, but you get the idea)

Three failure modes that I see, (calibration, slip, major mechanical failure) regular cal checks, wiper maintenance and cleanliness take care of first two, you would notice the third, so no harm done.

If the proposed substitute can match those criteria PLUS give a readout in a useful location AND retail for say 30% of a decent DRO I'd buy one. Otherwise I'll stick with the original and just suffer the nerve wracking feeling of looking low and to the right at the dial while getting pelted with chips from high left.

I really like the concept and would willingly try one or build the hardware if an EE type handles the software. So far though I'd rather take a chance on a slipping Trav-A-Dial than the constant dithering and failure to home I see in my cheap digital calipers.

tyrone shewlaces
01-20-2011, 11:00 PM
Sigh...

Trav-a-dial does NOT "wear" a rack into the bed to make it repeat or not slip. It is simply friction and they've done a pretty good job of making it work for a couple decades or so (prototype enough???). They aren't perfect, but they do work pretty well. If it DID wear a rack into the bed way, you can count on it not being very accurate from one area to another. I'm not going to go into why, but you could count on it varying over the length.

It's pretty amazing how even the simplest ideas are so willingly shot down by folks who have a "feeling" it won't work. Those folks might as well just give up and quit trying anything new. I know folks like that and they seem to get by. Not much fun IMO.

Anyways, somebody take a scrapped out trav-a-dial and attach an encoder to see if it works I guess. Keep it cheap and use the display in John's original post & translate the output to display properly (as previously mentioned).

I'll go away now.

Robin R
01-20-2011, 11:26 PM
I'll try again as no one seemed to have read my post #120. This is a rotary encoder with Quadrature square wave output, depending on the model the resolution is 0.0001" to 0.0005". It seems to be available to suit most major displays. There isn't a price posted on the website, but I would expect it to not be cheap. http://www.warrenmeasurement.co.uk/traksensor.htm

SVS
01-20-2011, 11:41 PM
Ruins the plan if it's not cheap Robin.

Just dashed out to the shop. Then came back.

My 7a Trav-A-Dial has a smooth satin finish on the crown of the wheel. No Knurls. Diameter is roughly 2" by .185" thick. Inch hand reads 10" per rev. and .100" per rev. for the fast hand.

J Tiers
01-20-2011, 11:46 PM
It's pretty amazing how even the simplest ideas are so willingly shot down by folks who have a "feeling" it won't work. Those folks might as well just give up and quit trying anything new. I know folks like that and they seem to get by. Not much fun IMO.

Anyways, somebody take a scrapped out trav-a-dial and attach an encoder to see if it works I guess. Keep it cheap and use the display in John's original post & translate the output to display properly (as previously mentioned).

I'll go away now.

Hey, I'm not "shooting it down"....... I like the idea.

But I try to be realistic, it's one of those "engineer things". I like a challenge, but I like it a lot better when I have a pretty good idea that it is "do-able", or even that it "should be" do-able.

If the best and most experienced practitioners, which would arguably be the Trava-a-dial folks, could only do "X" repeatability etc, when they obviously had a very good reason to try to do better, that is a really good argument against being able to beat them at their own game for 50 bucks.

You would need to have a good idea why they were hung up at "X" accuracy etc, and a good reason why your idea is better.

However, I do NOT KNOW what accuracy they got. I don't even know what they *claimed*, which is probably as good as it ever got.

This is one of those things where less accuracy is not worth just proportionately less, probably in reality it is worth exponentially less. And at a certain point it isn't worth anything, really.

So the question of what is realistically attainable is rather important.

J Tiers
01-20-2011, 11:52 PM
Ruins the plan if it's not cheap Robin.

Just dashed out to the shop. Then came back.

My 7a Trav-A-Dial has a smooth satin finish on the crown of the wheel. No Knurls. Diameter is roughly 2" by .185" thick. Inch hand reads 10" per rev. and .100" per rev. for the fast hand.

Southwestern Industries provides essentially NO specs..... other than a general "resolution "0.001", no distance quoted, no repeatability number quoted. A claim of "near absolute repeatability" is made, again no spec.

I'd like to see what the result of running the thing up and back over the whole bed length, between fixed stops is. Does it always result in the same pointer location at each end?

Another thing..... it really seems only to work over 10". That's all the dial shows. So anything more would be multiple dial revolutions, and accuracy etc specs (if they existed) are probably not intended to apply over such a distance

However, for those who claim that there is no pattern and nothing "worn" into the bed..... Southwestern Industries say the following:

"After a few traverses, the special pattern scratched into the wheel creates the mirror pattern on the running surface, forming a micro rack-and-pinion system that cannot slip."

Hopefully that puts ONE issue to bed..................

The Artful Bodger
01-20-2011, 11:59 PM
Ah, but how repeatable IS that? Have we seen the actual specs? Has anyone who has one tried to confirm the specs on their unit?

I did mention that I had no practical way of measuring 1 part in 400,000 (one tenth in 40").


And.... the wheel may be less easy than you think..... let's see your shop-made version! Presumably after you are re-settled..........

If I was doing it I would start with a wheel from a knurling attachment and I would run it on a aluminium strip and I would figure out some way of biasing the pressure on the wheel to reduce backlash.

But I wont be doing anything as my cable system seems to be quite good enough for the scale of work I do.

J Tiers
01-21-2011, 12:04 AM
I expect tha measuring to a couple thou would be good enough to show what's up....;)

SVS
01-21-2011, 12:18 AM
Playing with the math a little, a wheel with a diameter near 1.59" would turn twice in 10". If my 2" guestimate without dissasembling for accuracy is close they are scaling the 10" in roughly 1.59 turns. Freaky!

oldtiffie
01-21-2011, 03:51 AM
As the pace has slowed to a halt here, I thought I'd post how I use the "depth" function of a good digital caliper as a DRO. It is easy to set-up and remove and costs nothing and has a maximum error of 0.001" in any/either direction from the digital zero as distinct and separate from the mechanic zero/stops on the caliper.

I just use "Kant-Twist" clamps to hold the caliper to the back or front of my lathe bed. The "jaw" end is packed up with two 3/8" washers and the "bottom" (where the depth "needle" comes out) is supported on a bit of scrap hot rolled 5/8" x 1/8" flat bar about 1 1/2" long. The packing is to ensure that the caliper reading head and depth needle slide freely.

I have to slide the needle by hand but that's OK. The needle end contacts the carriage.

I can zero at any point as or for a reference. I can also zero off any "spacer" (made from scrap) anywhere and any time.

If I am not using caliper/DRO I either cover it in clear plastic sheet or remove it and remount it as and when required.


Mounted on the back of the lathe bed:
http://i200.photobucket.com/albums/aa294/oldtiffie/Lathe_misc/Caliperguage2.jpg

http://i200.photobucket.com/albums/aa294/oldtiffie/Lathe_misc/Caliperguage4.jpg


A carriage stop mounted on the back of the lathe bed between the saddle and the head-stock"
http://i200.photobucket.com/albums/aa294/oldtiffie/Lathe_misc/Caliperguage5.jpg


Caliper fitted to the front of the lathe bed:
http://i200.photobucket.com/albums/aa294/oldtiffie/Lathe_misc/Caliperguage6.jpg

It does have a minor problem with interference to or with the tail-stock but that's OK as it takes no time to remove and set up again.

John Stevenson
01-21-2011, 04:09 AM
To answer a few relevant posts.

At the moment i use the carriage hand wheel for the Z measurement on the lathe I use most, the CVA doesn't have one although later ones did.

The TOS reads to 1/2mm or 20 thou which is reasonable for me given MY type of work, diameters need to be very accurate, shoulders not so.

The TOS moves 80mm for one complete turn and it's the accrued lengths that sometimes catch me out, a digital readout will save the counting and addition pluse an easy way to swap between imperial and metric.

On this lathe I can get out of trouble easily by fitting a 40" glass scale and connecting it to the unconnected Z readout on the DRO, simple chequebook engineering but that only cures one machine for me, it doesn't do the rest or anyone elses.

The Warren Industries ones are about $800 for the encoder and you need to DRO so to be honest unless you have a lathe with a 100" plus bed it's cheaper to buy a glass scale.

I threw this into the ring to get feedback as there are some clever people on this board and it did turn some interesting features up as regards the mice etc.

macona
01-21-2011, 05:26 AM
However, for those who claim that there is no pattern and nothing "worn" into the bed..... Southwestern Industries say the following:

"After a few traverses, the special pattern scratched into the wheel creates the mirror pattern on the running surface, forming a micro rack-and-pinion system that cannot slip."

Hopefully that puts ONE issue to bed..................

I have had three or four of these units around and none of them have any sort of special pattern on the wheels. I call the above marketing BS. It's friction and works well. There is quite a bit of pressure pushing that thing against the ways. When I installed my DRO to replace the trav-a-dial the agreed exactly.

oldtiffie
01-21-2011, 05:31 AM
For measuring steps and depths, this attachment on a caliper is hard to beat. I use it in conjunction with the caliper "step" function.

Being digital, I can zero it anywhere.

http://i200.photobucket.com/albums/aa294/oldtiffie/Calipers/Dig_Caliper1.jpg

It is much more convenient than a depth micrometer (which gets saved for the more accurate jobs).

MCS
01-21-2011, 05:53 AM
Tiffie, if you have to post photos from tied-up calipers, please align them first.

My eyes hurt from the induced sine errors.

oldtiffie
01-21-2011, 06:06 AM
They are aligned alright - "photo problem" - mostly - although your are correct as regards the last pic. I should have done better there but that aside any error is negligible.

http://i200.photobucket.com/albums/aa294/oldtiffie/Lathe_misc/Caliperguage6.jpg

The cosine error is negligible.

For the caliper to be off-set such that 150mm travel on the caliper scale was actually 149mm on the lathe axis the off-set angle would be 6.62*

Acos 149/150 = 0.99333 = 6.62* (ie 6.62 degrees).

Weston Bye
01-21-2011, 06:13 AM
I worked out the least cost electronic hardware for using a quadrature encoder, a PIC microcontroller, and a parallel LCD display. Nothing magic here. The basic "system" would display travel and have a Zero button. Cost for purchased items, exclusive of encoder would be $50 or less, and could be easily wired up on perfboard. You would have to provide your own enclosure.

I haven't developed the program, that would take me a while for the PIC, but I could have a working system over the weekend using a more costly (under $100) firmware-equipped microcontroller

http://www.vestatech.com/docs/PDFDataSheets/MC077Data.pdf

similar to the Arduino. Indeed, the Arduino would also do the job. These last options would open the door for feature creep like the addition of a keypad and presets, and is overkill for the basic DRO operation. I have used the Vesta controller for CNC operation with an X-Y table - coming to Digital Machinist in Spring 2011.

BillTodd
01-21-2011, 07:40 AM
Here are a few bits I've sourced that would be useful to make a digital encoder wheel

The disk is a 3" (75mm) 1800 line encoder wheel from a HP ink-jet printer (dead after just a year!) with quadrature output reader PCB. It's a bit on the large side, but its high resolution (7200 cpt) means that the contact wheel could also be ~3" diameter while still giving a good resolution.

The cylinders are head bearing assemblies from hard-disk drives (the large ones are from older half height drives, the small one from a typical 1/3 height 3 1/2" you'd find in a modern PC). They are super precision devices with preloaded bearings and would be ideal to couple the encoder wheel to the contact wheel


http://wktodd.webspace.virginmedia.com/test/trava-dial parts.jpg

Bill

J Tiers
01-21-2011, 08:25 AM
I have had three or four of these units around and none of them have any sort of special pattern on the wheels. I call the above marketing BS. It's friction and works well. There is quite a bit of pressure pushing that thing against the ways. When I installed my DRO to replace the trav-a-dial the agreed exactly.

I'm calling BS on the claim of BS..... At least it seems there has to be something, and "court testimony" here confirms it.

The only person who has described the wheel in any "positive" detail (I can't find the post at the moment) said his has a matte finish, which sounds like an etch, but NOT an etched pattern like knurls. The fine tooth of an etched surface, possibly a specific "Rawal" pattern, would have very high friction against a softer surface, for low/no slip. Yes, it could self-align, although I would think that a knurl tooth would be better.

A smooth polished surface would, if hardened, be very prone to slip, as is regularly pointed out here with regard to holding end mills in a chuck. obviously the dial has much less drag torque, but...... we are talking tenths, that's not much to slip on what has to be a very accurate 1.0000:1.0000 relationship.

KEJR
01-21-2011, 08:40 AM
I think you guys have very successfully talked me into glass scales. :)

BillTodd
01-21-2011, 08:46 AM
I'd guess that the trava-dial requires very little force to spin it slowly, with the only significant force required as the carriage is traversed at speed due to the internal gearing (something a digital version would avoid).

I suspect the high pressure contact is as necessary to avoid problems with soft contaminants (oils grease etc.) as it is for traction.


Bill

MCS
01-21-2011, 08:53 AM
I suspect the high pressure contact is as necessary to avoid problems with soft contaminants (oils grease etc.) as it is for traction.


Bill

A measuring system and high pressure contact is a contradictio in terminus.

A measuring system is observative, not a player in the game.

BillTodd
01-21-2011, 09:02 AM
A measuring system and high pressure contact is a contradictio in terminus.

A measuring system is observative, not a player in the game.

That's philosophy, not engineering ;)

topct
01-21-2011, 09:13 AM
Here's an interesting pdf, shows some details that would be important to include in a design.

http://www.southwesternindustries.com/swi/Support/Documents/DRO/Technology.pdf

MCS
01-21-2011, 10:49 AM
That's philosophy, not engineering ;)

Engineering has it's philosophical qualities, as it has its legal qualities.

Here we are talking about a system, where the manufacturer claims that it is "near absolute".

Yet you own already half of the measuring system, while the manufacturer doesn't know the material nor the geometric properties of it. That's not measuring fundamentals.

The device needs to do destructive activities while running in and has an abundance of mechanical and electronical adjustabilities.

A marketeers dream: it never has to work, every problem can be denied by the supplier. Claims on one side, unsolvable problems on the other side, just because every application is unique and therefore unproven.

The Artful Bodger
01-21-2011, 02:17 PM
I have had three or four of these units around and none of them have any sort of special pattern on the wheels. I call the above marketing BS. It's friction and works well. There is quite a bit of pressure pushing that thing against the ways. When I installed my DRO to replace the trav-a-dial the agreed exactly.

If the wheels are not polished then there is a pattern, of some sort.

J Tiers
01-21-2011, 10:00 PM
Here we are talking about a system, where the manufacturer claims that it is "near absolute".

Yet you own already half of the measuring system, while the manufacturer doesn't know the material nor the geometric properties of it. That's not measuring fundamentals.

The device needs to do destructive activities while running in and has an abundance of mechanical and electronical adjustabilities.

A marketeers dream: it never has to work, every problem can be denied by the supplier. Claims on one side, unsolvable problems on the other side, just because every application is unique and therefore unproven.

Right, then.....

Your problem, however, should you choose to admit it, is that the device really DOES work. And it works well, as indicated by at least one poster who indicates close to perfect agreement between the DRO and the Trav-A-Dial.

We would like a bit more quantification, to assess how likely it is that "we" might succeed in making a device to attain that accuracy with different innards. But none of us (other that you) are questioning that the product is successful and useful.

John Stevenson
01-22-2011, 04:01 AM
There is only one answer.

Someone has to get off their arse and build one to prove the concept.

I can do the wheel arrangement standing on my head, after that I'm dead in the water.

macona
01-22-2011, 04:20 AM
I'm calling BS on the claim of BS..... At least it seems there has to be something, and "court testimony" here confirms it.

The only person who has described the wheel in any "positive" detail (I can't find the post at the moment) said his has a matte finish, which sounds like an etch, but NOT an etched pattern like knurls. The fine tooth of an etched surface, possibly a specific "Rawal" pattern, would have very high friction against a softer surface, for low/no slip. Yes, it could self-align, although I would think that a knurl tooth would be better.

A smooth polished surface would, if hardened, be very prone to slip, as is regularly pointed out here with regard to holding end mills in a chuck. obviously the dial has much less drag torque, but...... we are talking tenths, that's not much to slip on what has to be a very accurate 1.0000:1.0000 relationship.

Like I said I have had 3 or 4 different travadials in hand. All of them had smooth surfaces on the wheels, no apparent pattern, most likely ground. And unless the wheels are glass hard they are not going imprint anything on a hardened lathe bed.

I have even had the things apart for cleaning. Nothing terribly complicated, spring anit-backlash gears.

MCS
01-22-2011, 05:31 AM
Right, then.....

Your problem, however, should you choose to admit it, is that the device really DOES work. And it works well, as indicated by at least one poster who indicates close to perfect agreement between the DRO and the Trav-A-Dial.

We would like a bit more quantification, to assess how likely it is that "we" might succeed in making a device to attain that accuracy with different innards. But none of us (other that you) are questioning that the product is successful and useful.

First: I don't have a problem.

Second: I'm not alone:


I think you guys have very successfully talked me into glass scales. :)

Third: I have a calculator.

Weston Bye
01-22-2011, 05:49 AM
From the horse's mouth:


The Trav-A-Dial is a mechanical marvel. It detects machine tool motion by the contact of the gage wheel with a running surface. The wheel is made of tool steel and is very hard – 63C on the Rockwell scale. The Trav-A-Dial's wheel is pushed up against the surface by a special mounting base. After a few traverses, the special pattern scratched into the wheel creates the mirror pattern on the running surface, forming a micro rack-and-pinion system that cannot slip.

Wheel motion is translated through the Trav-A-Dial through a set of very precise gears. In fact, Southwestern Industries’ gear rolling technology enables us to make some of the world’s most precise instrument gears. In order to eliminate backlash, these gears are preloaded in the axial and radial directions – no mean trick.

The result is a mechanical system that is capable of accurate measurement (after calibration) of .0003” and near absolute repeatability.

The Trav-A-Dial gives you precision measurement in an analog display preferred by many machinists.

Advertising hyperbole? Perhaps, but a grain of truth somewhere. Having no personal experience with the product I can only engage in idle speculation.

This system puts me in mind of a railroad locomotive and the tractive effort conveyed from the driver (wheel) to the rail. Theoretically, the wheel is in point or line contact with the rail. Not much contact area. Certainly the mass of the locomotive contributed to the tractive effort by forcing the driver into the best achievable engagement and possibly causing some localized distortion of the driver and rail. In the steam era, locomotives were equipped with sand dispensers to improve "engagement" and boost traction, but these were used sparingly to prevent undue wear.

The systems are similar but with different goals. The locomotive is concerned with conveying enough one-time engagement to move the freight. The Trav-a-Dial is concerned with repeatable engagement to ensure repeatability.

If the advertising is to be believed, then the wheel must be impressing a repeatable pattern on what ever surface it is running on, and best theoretical practice would be never to remove or otherwise disengage the wheel from the pattern that it had established. Though remote, even "tooth-to-tooth" variation could introduce error.




There is only one answer.

Someone has to get off their arse and build one to prove the concept.

I can do the wheel arrangement standing on my head, after that I'm dead in the water.


I am doing some playing with the electronics, but getting articles out takes priority at the moment. Would your wheel arrangement include a quadrature encoder?

A thought about the mouse discussion above: Incorporate the "mouse guts", optical or mechanical, within the travadial enclosure to provide a controlled athmosphere or environment. Although this might be combining more than one less-than-perfect technologies in hopes of arriving at a perfect outcome.

Weston Bye
01-22-2011, 06:17 AM
I think you guys have very successfully talked me into glass scales. :)

I think the quest here in this thread is for a perfect shop-made position detector. Making glass scales is beyond most of our capabilities.

John Stevenson
01-22-2011, 08:01 AM
I am doing some playing with the electronics, but getting articles out takes priority at the moment. Would your wheel arrangement include a quadrature encoder?


It would if I knew which one to connect to or had access to the same.
I can get 1000 count ones like this.

http://www.cnc4pc.com/Store/osc/product_info.php?products_id=301?osCsid=9c980d5c27 1a0bebd0c7d33c37deb425


Data sheet here.

http://cnc4pc.com/images/hk30_DataSheet.jpg

John S.

BillTodd
01-22-2011, 08:31 AM
John,

If you need something quick albeit at a price...

This Quarature reader display (http://www.machine-dro.co.uk/index.php?target=products&product_id=331) should work OK with any of these encoders (http://www.machine-dro.co.uk/index.php?target=categories&category_id=51) (check with the firm to be sure).

http://www.machine-dro.co.uk/images/product_images/thumbnail_dro-mg-mdu-1x_ad04.jpg

http://www.machine-dro.co.uk/images/category_images/thumbnail_rotary_encoders_web.gif

[edit]


It would if I knew which one to connect to or had access to the same.
I can get 1000 count ones like this.
I'm not sure, but I think that is 250 lines with 1000 counts in quadrature (i.e. counting the up and down edges of both channels) to get your 1 thou" resolution you'd need a wheel 0.318" (1/pi) in diameter which seems a bit tiddly to me.

J Tiers
01-22-2011, 09:12 AM
Myself...... I tend to think that the wheel-based system, as a home shop project, is not likely to succeed for most people. I think it can be done, and could work fine, but several issues must be overcome first.

The question then is whether the result is worth the effort.... whether this problem is interesting enough to create a solution even if it isn't very useful in a practical sense.

And, I find, upon consideration, that I would likely rather have the OTHER axis first....... While there is no scale on the long axis, it tends to be used less, and for me is "worth less" (but not entirely "worthless").


First: I don't have a problem.

Second: I'm not alone:

Third: I have a calculator.

Your problem is that you are a "Trav-A-Dial denier"....... you apparently claim (in the face of contrary evidence) that the original device, which has been sold and successfully used for many, many years, doesn't work, can't work, etc.

That isn't the point here...... we all agree it DOES work, and we'd like a decent version cheaper.... since the original (single axis) item is competitive with a two axis DRO in price.

There is more than one member of the "flat earth society" who has not quite realized that the earth is not flat. That has no bearing or effect on the true situation.

MCS
01-22-2011, 09:35 AM
There is more than one member of the "flat earth society" who has not quite realized that the earth is not flat. That has no bearing or effect on the true situation.

Ahh, that way!

If one put questionmarks to feasibility, ability, cost and such matters, one is a member off the "flat earth society".

Well, it surely must simplify life.

John Stevenson
01-22-2011, 09:50 AM
Ahh, that way!

If one put questionmarks to feasibility, ability, cost and such matters, one is a member off the "flat earth society".

Well, it surely must simplify life.

So EVERYTHING you do has to be for profit ? I'll bet you are one barrel of laughs.

MCS
01-22-2011, 10:44 AM
So EVERYTHING you do has to be for profit ? I'll bet you are one barrel of laughs.

I rather make parts than tinker with gizmos, if you twist it long enough you can call that a longing for profit.

Deja Vu
01-22-2011, 12:56 PM
Heh! I'm seeing a dramatic drift seeping in here! :D

The Artful Bodger
01-22-2011, 02:23 PM
Myself...... I tend to think that the wheel-based system, as a home shop project, is not likely to succeed for most people. I think it can be done, and could work fine, but several issues must be overcome first.


I think most folks here could manage to make a wheel. The principal requirement as I see it is that the wheel must be hard enough to mark the lathe bed (or whatever) so that a positive locking track is created.

The finer the pitch of the 'rack' the better and a visually matt surface would be more accurate that a big chunky rack, and much easier to form.

The wheel must be of small diameter to get high angular movement to drive the encoder with accuracy. An alternative would be to use a gear to drive the encoder faster than the contact wheel, but then backlash has to be taken care of of which there are various methods available the easiest being the system where two gears have a spring between them to exert constant pressure against a pinion.

J Tiers
01-22-2011, 04:07 PM
I think most folks here could manage to make a wheel. The principal requirement as I see it is that the wheel must be hard enough to mark the lathe bed (or whatever) so that a positive locking track is created.



No doubt.....

THIS wheel needs to be very round, crowned just right, with a very good surface finish, hardened, then etched or vapor blasted, etc to get a fine tooth pattern, random or regular. The "spindle" it is mounted on must be *exactly* on-center, to avoid periodic errors of over 0.0001 or so.

It's not done in every shop, you know, even in the commercial world.

This does not mean it cannot be done. It simply means that it is not 'easy"..... unless you are quite satisfied with accuracy of perhaps 1/10th of the Trav-A-Dial. I expect most anyone could hit that target.

Sheesh.........

We got one guy saying it is so hard to do that even actual Trav-A-dials don't really work, and another guy saying that it's a walk in the park......

That sounds like a challenge to me....... but I have a machine to scrape in, which means I have several precision fixtures to make. And I don't really want a T-A-D that bad right now.....

Who's up?

oldtiffie
01-22-2011, 08:35 PM
The OP:


OK new thread to move away from Black forests thread.

We finished up asking why it was not possible to use a cheap encoder and one of these also cheap Chinese remote displays.

http://www.arceurotrade.co.uk/imagecache/3286dc1a-1abf-4d1d-8f1b-9b4a00bab943_250x125.jpg

Simple cheap encoder with a hardened knurled radiused wheel that can be tilted to get the correct circumference driving into one of these units.

The protocol for the Chines signals is documented on the Shumatech site which is down at the moment but I may be able to get it elsewhere.

Basically it's a reversal of what Scott Schumacher has done.

Any idea's or is anyone up to the job ?

Has this proved to be too hard or has the thread - and concept and the answer to the OP - died? or is it on its last legs (about to die)?

It it going to be resuscitated or resurrected or just quietly laid to rest after going out without a whimper let alone a bang?

Started off with such a flourish and lots of promise too.

SVS
01-22-2011, 09:01 PM
Tiffie, this is apparently way over your head. Takes time and your drivel helps nothing. Particularly disgusted by the posting of your caliper bodge after two requests from OP not to do so.

Anyway,

Wasn't a 7200 count/turn encoder mentioned earlier? If so a 1.14" diameter wheel would count every .ooo5" or a 2.29" wheel every .001"

Mechanically a bigger wheel is better....less rolling resistance, less likely to skid, and avoiding the precision gear issue is a plus if a high count encoder allows the necessary resolution.

Deal with concentricity by turning the wheel and axle in one setup. And MAYBE you could calibrate in the encoder count to display phase rather than crowning the wheel and rocking it.

J Tiers
01-22-2011, 09:21 PM
Deal with concentricity by turning the wheel and axle in one setup. And MAYBE you could calibrate in the encoder count to display phase rather than crowning the wheel and rocking it.

That's better, of course, and then there are the bearings..... let's not go THERE......;)

But "calibrate in the encoder count to display phase rather than crowning the wheel "???

The crowning is for two reasons, as I see it....

First, it assures that any position is effectively "flat against the surface", as opposed to having just one position that has the wheel in full contact, and any other one contacting at the edge only.

Second, as there is a small change of effective radius it changes the length cal.... in a rather direct manner

I'm pretty sure you see that latter point.... but to cal for a circumference error, one would have to actually slip phase, or effectively "drop 'n' fractional counts per count" * to do a cal off of the direct count. Is that what you mean?

Is that not effectively an interpolation to a higher count resolution, and can you "count on" it being as accurate? It seems rather similar to interpolating on a dial caliper to a greater resolution than the gears etc actually justify. If it would work well enough, it is a good idea.

What I see though is that you will do this interpolation many many times in one measurement, concatenating the results to get a total, while the questionable caliper interpolation occurs once only, and is unlikely to cause an error exceeding the spec resolution. The error in the result of a few hundred interpolations could exceed the basic resolution.

Alternately you could either get an encoder wheel with very good "jitter" spec, or simply go to a higher resolution wheel to begin with.

* If you got to the point of dropping one count every "N" counts, you would have a periodic error.

SVS
01-22-2011, 09:35 PM
Yes, I know why Trav a Dial crowned the wheel. The suggestion to calibrate in the electronics was based on laziness regarding building a proper rocking mount for calibration.

The electronics is way beyond me, but seemed possible that the programming could be just as happy calling one count .0004999999" or.000500001 instead of .0005"

I can defintely see that an iterative rounding error is problematic. (I think that is your objection, and conceed the point.)

oldtiffie
01-22-2011, 10:03 PM
Why are so many just emulating or duplicating the Trav-a-Dial principle by having a hardened wheel running on the side of the lathe bed?

All that is required, as I understand it, is to measure relative movement between the saddle and the lathe bed with a reasonable degree of accuracy over the traveled length. John Stevenson did say that while he required diameter accuracy (cross-slide) he was not all that fussy about 0.010">0.020" (more or less) for saddle travel ("step" work).

If a taught wire were to be stretched and located parallel to the lathe spindle axis in the vertical and horizontal planes, with a wheeled "follower" rolling along the wire and rotating the wheel/s to drive an encoder the output of which was passed for electric sub/post processing and the output sent to a remote display, I would imagine that most if not all requirements would be met.

My MIG welder wire drive principal seems to be a viable concept if run in reverse - ie the wire driving the wedge-groove and flat wheels (which are geared together to prevent/reduce "slip") and one of the gears is direct-driven by a variable-speed motor. The tension of the rollers on the wire is very nicely adjustable.

If that drive were connected to the lathe carriage with its wheels gripping the wire and the rollers driving an encoder it might work.

It would not be difficult to measure the lathe saddle travel in terms of distance per revolution of the rollers.

Here is the drive:

http://i200.photobucket.com/albums/aa294/oldtiffie/MIG_welder/MIG-Welder4.jpg

http://i200.photobucket.com/albums/aa294/oldtiffie/MIG_welder/MIG-Welder6.jpg

http://i200.photobucket.com/albums/aa294/oldtiffie/MIG_welder/MIG-Welder7.jpg

http://i200.photobucket.com/albums/aa294/oldtiffie/MIG_welder/MIG-Welder2.jpg

J Tiers
01-22-2011, 10:29 PM
Another way is to bag the rollers, use a very accurately ground supply reel, with only one layer of wire, and read the revolutions with an encoder.

However, one may get into trouble with wire sag, which can pull off extra 'tenths" of wire and cause an error of measured length at long distance. That COULD be corrected for in software.

SVS
01-22-2011, 10:43 PM
Four points on wire system.

It's been done

Doesn't fix the slip issue, just moves it to a different place on the machine and adds stretch and sag issues.

Less compact

Hardware is really the minor issue. The electronics will determine success.

On edit:See JT covered some of my thunder.

oldtiffie
01-22-2011, 11:01 PM
Tiffie, this is apparently way over your head. Takes time and your drivel helps nothing. Particularly disgusted by the posting of your caliper bodge after two requests from OP not to do so.

.................................................. ............



The electronics certainly is but I don't think that the mechanical aspects are.

http://i200.photobucket.com/albums/aa294/oldtiffie/Calipers/Caliperguage4.jpg

I put the "caliper" up as so far it is the only feasible solution to a stepping problem on a step or steps not exceeding 150mm/6" long. It is out of the way - as much as the Trav-a-Dial is, costs nothing, is positive reading with and accuracy of +/- 0.001", is already available in most shops and takes just minutes to mount and dismount. If it gets ruined just get or buy another.

I can set it with or without gauges or spacers and the carriage only requires a simple stop.

http://i200.photobucket.com/albums/aa294/oldtiffie/Calipers/Caliperguage5.jpg

The OP all but suggested it:


Originally Posted by John Stevenson
OK new thread to move away from Black forests thread.

We finished up asking why it was not possible to use a cheap encoder and one of these also cheap Chinese remote displays.

http://www.arceurotrade.co.uk/imagecache/3286dc1a-1abf-4d1d-8f1b-9b4a00bab943_250x125.jpg

Simple cheap encoder with a hardened knurled radiused wheel that can be tilted to get the correct circumference driving into one of these units.

The protocol for the Chines signals is documented on the Shumatech site which is down at the moment but I may be able to get it elsewhere.

Basically it's a reversal of what Scott Schumacher has done.

Any idea's or is anyone up to the job ?

I did suggest a similar "caliper-like" solution with the effective 900mm working range here (or order in a longer one):
https://www.machineryhouse.com.au/Products?stockCode=D667

If the working range does not exceed 600mm this "caliper" with a remote display works very well:
https://www.machineryhouse.com.au/Products?stockCode=Q656

with its remote display:
https://www.machineryhouse.com.au/Products?stockCode=Q650

Not cheap but very effective and can be mounted and running in very little time.

If it were me and pending an answer here, these options are very viable either as a permanent stand-alone response or while waiting for a complete "bolt on" "ready to go" simulation of a Trav-a-Dial or an electronic alternative to it.

dp
01-22-2011, 11:07 PM
The avagotech site has an interesting laser navigation sensor:

http://www.avagotech.com/pages/about/archive/laserstream_0228_embedded/

And I've had an idea. The idea is to use such a sensor and a position detector to provide navigation and positional information in one dimension.

For a position sensor - imagine a laser beam passed through a diffuser such that a line is generated at a target. Laser levels use this idea to paint a line on a wall to align objects such as picture hangers.

Now imagine that target is a light sensor and further, it is behind a tapered mask so that if the beam were at the vertex of the mask, little output would be generated while the same beam at the wide end of the taper would produce orders of magnitude greater output from the sensor. Next imagine a second sensor is next to the first and is behind a rectangular mask so that regardless of beam position, the output remains fixed.

Now imagine the laser beam, which itself is a planar wedge, is shone at an angle off parallel to the plane of the the laser movement and strikes the sensor which is positioned so to intercept the beam. As the distance between the source and target changes the position on the sensor that is illuminated changes. In one sensor the output is a function of distance, beam energy, and temperature while in the other, the output is a function of beam energy and temperature. The difference is the range between the light source and target.

You now have an analog position sensor and temperature/beam energy detection system that is looking for an A/D device to digitize the output.

SVS
01-22-2011, 11:15 PM
You are hopeless.

Twice when you offered to post your bodge JS said words to the clear effect of No thanks, Off topic, and don't bother if it won't cover full axis travel.

Now you've managed to post the crap once again, plus roughly 5 reposts of JS's OP with your chiding reminders that we aren't working hard enough at the project.

Point is: ITS A HOBBY AND IT TAKES TIME AND IT WILL BE DONE WHEN IT IS DONE AND YOUR BULL**** DOES NOT HELP THE PROCESS

oldtiffie
01-22-2011, 11:41 PM
Well,

some other than John Stevenson have similar needs and may be interested in alternatives - which I've provided.

But it it doesn't fit within "hobby" budgets it might not "fly" for many here - when-ever it appears.

John Stevenson made his request in the context of a commercial shop.

I take your point about the "no rush" and "all in good time" and " 'cos its a hobby" as that may have been my response if I were working on it for other than myself. I am quicker (less slow?) when a project is for my shop.

SVS
01-22-2011, 11:53 PM
Allrighty then.

You got your 2cents in. Sit back with me and wait for the EE types, and by all thats holy if you repost the caliper bodge in this thread I'm gonna have a stroke. If I survive it I'm buying a plane ticket to OZ and bringing a club.

oldtiffie
01-23-2011, 12:03 AM
I'm waiting - patiently.

J Tiers
01-23-2011, 01:15 AM
Well, as far as I can see:

* the mouse option is sorta open...... questionable as to reliable resolution and avoidance of accel issues, but that may depend on the chip.... Also questionable as to the surface and reliability of reading it, although that might be fixed by reading a clean prepared surface inside a box, with a wheel actually riding the way and moving the read surface......
Clearly needs some detailed study of teh mouse chip data.

it would help if the mouse chips were actually available in small quantities.... many of these things made in bazillions for commercial products are not available unless you need at least fractional bazillions per month.

Mouse chip seems to be the probable cheapest, and potentially has the least hardware issues if the chips are suitable, but also has the most questions all the way from basic accuracy through reliable pickup without "slip", to the possible ways to cal it.

* The wheel and encoder wheel approach is still very open, with questions as to the wheel shape and resolution of encoder. Also questions as to the means of cal, whether software or a hardware solution such as tilting a crowned wheel.

Very few questions as to the ability of the concept to "work", but some questions as far as the capability of making the wheel. It might be pretty cool if a suitable wheel were found, perhaps one not necessarily intended for this. That would take care of the "how to make a whee;l" problem.

* "other" means..... basically other means of picking up wheel position, really, as really "other" means starts to look like a DRO scale... Which brings up a point..... if the DRO scale can be made cheap enough and good enough to work directly, then it seems possible to wrap that idea into a circle to pick up wheel position.

This is pretty open, hasn't been addressed significantly

...the usual scale, similar to an encoder wheel...has been done.

...the capacitive system such as used on a caliper, but circular

...ball bearings as used in one kind of DRO, in a circular tube instead of linear tube....

... other unspecified means usually used in a linear DRO, but wrapped in a circle instead.

...entirely other??????

oldtiffie
01-23-2011, 01:30 AM
Another way is to bag the rollers, use a very accurately ground supply reel, with only one layer of wire, and read the revolutions with an encoder.

However, one may get into trouble with wire sag, which can pull off extra 'tenths" of wire and cause an error of measured length at long distance. That COULD be corrected for in software.

JT.

As an EE you will be aware that "wire sag" as in any cable or wire under tension is "catenary sag" (as in power line cables etc.).

As a mere non-EE, I'd like to suggest that as the working distance/length will be the maximum travel of a lathe, that the nearer the suspension/tension points for the wire are to those end points and the larger the tension the better. I doubt that temperature variation in a shop would be too significant.

Most stuff on the web that I could find for catenaries is pure Math and little of no applied math - with this exception:
http://www.spaceagecontrol.com/calccabl.htm?F=0.674427074&a=12&q=0.00044091&g=32.18503937&Submit+Button=Calculate

If using the MIG welder feed roller system you will be limited to the maximum wire size that the rollers can handle - in the case of my welder the maximum wire size is 1.2mm ~ 0.048".

J Tiers
01-23-2011, 01:48 AM
normal sag does look to be less of a problem over short distances under 1m, I hadn't run any numbers...... but dirt, oil, etc getting on cable might well throw that off, as weight changes.

Frankly, the cable would probably work, but looks too darn "exposed"....

Then also, vibration, if it resonated with the cable or wire at the given tension, could really throw off the measurement, as the wire accelerates up and down, with a LOT more "displacement" than gravity accounts for.

Think about a guitar string......any "note" or harmonic that is in tune with it will cause it to sound.

oldtiffie
01-23-2011, 02:03 AM
JT.

I am sure that any "nuisance-level" resonance can be damped.

Keeping the wire "clean" is a shop "house-keeping" matter the same as cleaning the Trav-a-Dial track on the lathe bed is.

I'd have thought that piano wire would be OK.

That wire feed drive on my welder is very good as it adjusts from "just slipping" to no slip at all at which level, if the wire jambs it just makes one big birds nest that I have to pick apart and pull out.

I will post a clarifying pic of the gear drive as the driving gear drives the idler gear and both drive the wire. The drive roller has "V" in which the wire engages and the plain/non-vee idler keeps the pressure on the wire to force it into the groove - not unlike a round belt in a vee pulley - and just as effective.

oldtiffie
01-23-2011, 03:49 AM
Clarifying pic of MIG welder drive here:

http://i200.photobucket.com/albums/aa294/oldtiffie/MIG_welder/MIG-Welder10-1.jpg

The drive roller drives idler via the two gears so the idler cannot slip relative to the driver. The driver drives the vee-d roller via a pin. The idler roller is attached to the idler gear and uses spring tension to squeeze the wire into the driver roller vee.

Another thought is a warping drum or a capstan or winch drum with say 3 or 4 turns of the wire wrapped around the drum. They rarely slip. You possibly seen them as (coffee) "grinders" on yachts etc. The capstan/winch can be used to drive an encoder.

Weston Bye
01-23-2011, 05:46 AM
The biggest drawback with the reverse-wire-feeder is that the wire is deformed and trades diameter for length as it passes between ther rolls. Them that has a wire feeder and wants to experiment should try passing the same length of wire repeatedly through the rolls. Soon the rolls will loose traction, and the wire will be longer than before.

Similarly, if a suitable surface is not available on the lathe bed, travadial recommends adding a piece of bar stock for a running surface. Too thin a piece of bar stock will change in dimension also, probably bowing out toward the wheel. displaced metal has to go somewhere.

John Stevenson
01-23-2011, 06:53 AM
I'm waiting - patiently.
Well while you are waiting patiently why don't you go into that well equipped shop you keep posting endless pictures of and


BLOODY USE IT

to make something you can post on here instead of all the pretty posed pictures you keep regurgitating.


[EDIT] Instead of regurgitating a wire feed system go back and read post #34 and look at the link.
Been done. Got one, you want pretty posed pictures? sorry can't do it's being used.

luthor
01-23-2011, 07:18 AM
Well while you are waiting patiently why don't you go into that well equipped shop you keep posting endless pictures of and


BLOODY USE IT

to make something you can post on here instead of all the pretty posed pictures you keep regurgitating.


[EDIT] Instead of regurgitating a wire feed system go back and read post #34 and look at the link.
Been done. Got one, you want pretty posed pictures? sorry can't do it's being used.


In Tiffies defence, I did see some swarfe on his lathe back in post 137, which indicates that it has been used at least once.

John Stevenson
01-23-2011, 07:31 AM
In Tiffies defence, I did see some swarfe on his lathe back in post 137, which indicates that it has been used at least once.

I think the "at least" is not needed in that sentence . :)

MCS
01-23-2011, 09:33 AM
In Tiffies defence, I did see some swarfe on his lathe back in post 137, which indicates that it has been used at least once.

Well, it may be the case that, in reaction to the complaint of "posed" photos, we see here the introduction of "posed" chips.

It is not an indicator, chips can be get, lend, stolen or bought.

J Tiers
01-23-2011, 11:29 AM
The biggest drawback with the reverse-wire-feeder is that the wire is deformed and trades diameter for length as it passes between ther rolls. Them that has a wire feeder and wants to experiment should try passing the same length of wire repeatedly through the rolls. Soon the rolls will loose traction, and the wire will be longer than before.


Which is a really good argument for reading turns of the wire drum, with a single layer on it.... Any off-angle as wire unwinds is likely to be minimal.... same like the sag if a decent tension is applied, and the drum is of decent size so that not too many turns are needed.

Alternately, the wire could simply be deflected around a "reader wheel".

I think the squashing effect would be fairly minimal, if the basic nuisance factor of having a wire is handled, AND the issue of wire vibration is dealt with.

yes, Tiffie, the piano wire is exactly the item I had in mind... guitar strin was mentioned as an example of "sympathetic vibration". Although it seems the "vibrations" here are turning distinctly "unsympathetic".........

sansbury
01-23-2011, 11:58 AM
If anyone wants to try this but scared of the silicon part, I would be happy to supply Arduino code for this. I'm working on a project which entails an LCD and an encoder so it would be an easy hack. I have cnc on my machines so not much interest in this but happy to lend a hand. You can buy an Arduino for $30 and program it with any pc and a USB cable. Easy peasy.

Fwiw, it will be much cleaner if one edge count in the encoder equates to one increment of measurement. At this scale floating-point math can become annoying and either compounds the error or just smushes it around. Or be off by a clean factor of ten like .0009, .0008 per step so you can just accumulate the error and correct it every N counts.

dp
01-23-2011, 01:20 PM
...entirely other??????

An entirely other was proposed at #176.

Here's another test - nicely done, by the way, using mice.

http://www.youtube.com/watch?v=nB1SjhCiXMQ

And an application of a mouse navigator:

http://www.youtube.com/watch?v=RDuq6y9siOE

BillTodd
01-23-2011, 02:39 PM
the issue of wire vibration is dealt with.

Inside each of the Sony Magnascale devices on my lathe is a tort wire, approximately 2.5mm diameter. The long scale is 1m, Sony make them up to 2.2m.

The wire is magnetised with a sine wave signal (just like a tape recording). As it passes through read head it generates a signal in a pair of magneto-resistive sensors.

The absolute accuracy is claimed as +/- 0.0035mm (+/- 0.005mm x effective length in metres). Resolution is 0.5um.

Vibration is not really an issue.

BillTodd
01-23-2011, 02:40 PM
An entirely other was proposed at #176.

Here's another test - nicely done, by the way, using mice.

http://www.youtube.com/watch?v=nB1SjhCiXMQ

...

That's the one I posted earlier :)

dp
01-23-2011, 02:46 PM
That's the one I posted earlier :)

Missed that. BTW, found that while looking for images for your coax indicator. Not to hijack this thread, but what does it look like inside?

Edit: Oops - never mind. You've been busy! :)

J Tiers
01-23-2011, 03:45 PM
Vibration is not really an issue.

evidently not in their case, but they are , if I "get " your description, not reading a "loose" wire, but one pre-tensioned, where the error would be "stretch" of a stretched wire.

In the case of an un-reeling wire, care would be needed to ensure that vibration would not lead to large excursions of the wire, which , depending on frequency and amplitude, might lead to reading the wrong amount of unreeled wire.

The vibrations, if large, could make the "average tension" higher, and unreel a bit more wire. We ARE talking tenths as resolution in order to get 0.001 accuracy.

Anything significantly worse than that would mean the distance could be just as well read off a long scale by a traveling magnifier and pointer, like the first jig bores.

dp... in # 176, the avago link basically says :"ours is really good"...

The idea presented is interesting, I wonder what resolution could be obtained? How long a distance do you envision being able to quantify?

I missed that post as it was in the middle of the "Tiffie tiff"

BillTodd
01-23-2011, 04:01 PM
evidently not in their case, but they are , if I "get " your description, not reading a "loose" wire, but one pre-tensioned, where the error would be "stretch" of a stretched wire.
Yes that's correct. The wire is not piano string tight, but it is tensioned to keep it straight.


In the case of an un-reeling wire, care would be needed to ensure that vibration would not lead to large excursions of the wire, which , depending on frequency and amplitude, might lead to reading the wrong amount of unreeled wire.

Ah yes, I can see vibration in that case would be a big problem :)

Some years ago, I tired to make a cable driven pen-plotter, it had this exact problem - At times the cable would shake the whole machine!

You might be able to control the vibration problem by using a thin tape , rather than a wire (vibration in the tape is damped down by its air resistance). Or, by laying the cable onto a surface as its unwound (friction damping), this would also prevent any 'sag' effects.


Bill

SVS
01-23-2011, 07:40 PM
Sansbury,
I'm scared of the silicone part.

I'm curious what price and and specs are on the encoder and LCD you are working with.

If I build one of these contraptions it would go on an 80" centers lathe, and possibly a 3 axis system for my Cincinatti Vercipower vertical mill. Need travels of roughly 48", 36", and 26"

I really think the mill needs an actual DRO interface with hole circle functions, or a faked up substitute through a pc.

I like the idea of feeding a mill DRO with the encoder T-A-D because the mill does not have great access for glass scale mounting and protection and I believe the E. T-A-D could be a rugged and unobtrusive little chunk.

Weston Bye
01-23-2011, 08:05 PM
If anyone wants to try this but scared of the silicon part, I would be happy to supply Arduino code for this. I'm working on a project which entails an LCD and an encoder so it would be an easy hack. I have cnc on my machines so not much interest in this but happy to lend a hand. You can buy an Arduino for $30 and program it with any pc and a USB cable. Easy peasy.

Fwiw, it will be much cleaner if one edge count in the encoder equates to one increment of measurement. At this scale floating-point math can become annoying and either compounds the error or just smushes it around. Or be off by a clean factor of ten like .0009, .0008 per step so you can just accumulate the error and correct it every N counts.

Sansbury has proposed probably the most practical solution with a widely available controller and display that can be duplicated by other hobbyists. All that is needed is to add the encoder and mechanics - within the capabilities of a reasonably accomplished machinist or hobbyist. The mechanical design should try to arrive at at least one pulse per .001" or .01 mm of travel depending on your preference.

Bmyers
01-24-2011, 12:00 PM
I ordered an Arduino controller and LCD display to at least play with the electronics end of this thing until it warms up enough in the shop to make the mechanical portion. I dont see why a wheel with a O ring tire wouldn't provide sufficient traction to ride on the bed way.

dp
01-24-2011, 12:08 PM
It is a friction drive and can't reliably provide varying positions or location. That requires something like a gear and rack, or embossed wheel that imprints its track if you go with mechanical devices, or a photo mapping tool that records and compares position using imagery (like finger print comparators optical mouse sensors).

Bmyers
01-24-2011, 12:29 PM
It is a friction drive and can't reliably provide varying positions or location. That requires something like a gear and rack, or embossed wheel that imprints its track if you go with mechanical devices, or a photo mapping tool that records and compares position using imagery (like finger print comparators optical mouse sensors).

i think this will be a first experiment:
http://www.martijnthe.nl/2009/07/interfacing-an-optical-mouse-sensor-to-your-arduino/

J Tiers
01-24-2011, 10:11 PM
The first problem with a rubber tire is the varying radius... depends on pressure and the amount of "squish" on the rubber. If radius varies, the distance traveled per turn varies, and any remnant of accuracy below the nearest inch is probably out the window.

Actually, that is the last problem also..... after that one you don't need any more anyway ;)

dp
01-24-2011, 10:18 PM
It is possible to manage the radius with squish in the mix by rigidity. The next problem is tyre slippage over time. That kind of squish will cause an elastic tyre that is not firmly attached to the wheel to walk around the wheel over a period of time. I rather like the swipe sensor on the new Apple Magic mouse. No moving parts and yet rubbing (gesturing) the carapace in any direction will scroll the current pane left/right (for example) just as does a track ball.

http://www.patentlyapple.com/patently-apple/2010/05/two-of-apples-magic-mouse-patents-finally-surface.html

J Tiers
01-24-2011, 11:20 PM
It is possible to manage the radius with squish in the mix by rigidity.

'Ow so?

Rigidity seems to mean either a rigid hardened wheel, or rigidly holding at a fixed distance.

I'm betti ng on the hardened wheel, because you can jam in in hard and still hold exact radius. Holding the wheel at a constant distance "may" work... or may slip.

dp
01-24-2011, 11:31 PM
'Ow so?

Rigidity seems to mean either a rigid hardened wheel, or rigidly holding at a fixed distance.

I'm betti ng on the hardened wheel, because you can jam in in hard and still hold exact radius. Holding the wheel at a constant distance "may" work... or may slip.

If you can hold the dimension from the axle to the rack/track, what ever it becomes known as, you control the turns/distance traveled. It is the same regardless of the degree of compression in the elastic component. It is the same with dual wheels. Regardless of the relative inflation of each tire, both tires rotate at the same speed and cover the same distance over the ground.

J Tiers
01-25-2011, 12:08 AM
If you can hold the dimension from the axle to the rack/track, what ever it becomes known as, you control the turns/distance traveled. It is the same regardless of the degree of compression in the elastic component. It is the same with dual wheels. Regardless of the relative inflation of each tire, both tires rotate at the same speed and cover the same distance over the ground.

Yas.... hold at fixed distance as I suggested..... But that makes amount of friction (read "amount of squish") and dial calibration (also read "amount of squish") interactive with each other, which is maybe not the best situation.

oldtiffie
01-25-2011, 05:21 AM
Once again, I go back to the OP for statements of purpose and intent as regards requirement in the environment in which it is intended to operate.

OK new thread to move away from Black forests thread.

We finished up asking why it was not possible to use a cheap encoder and one of these also cheap Chinese remote displays.

http://www.arceurotrade.co.uk/imagecache/3286dc1a-1abf-4d1d-8f1b-9b4a00bab943_250x125.jpg

Simple cheap encoder with a hardened knurled radiused wheel that can be tilted to get the correct circumference driving into one of these units.

The protocol for the Chines signals is documented on the Shumatech site which is down at the moment but I may be able to get it elsewhere.

Basically it's a reversal of what Scott Schumacher has done.

Any idea's or is anyone up to the job ?

As I see it, it is a type of "Trav-a-Dial" with just the "wheel" and without the analogue dials but perhaps driving an encoder instead. The Trav-a-Dial accuracy is 0.001". John said later that he was not too concerned about accuracy along the lathe bed/spindle axis as his "steps" were not of a high order of accuracy.

Here are some of the details from the Trav-a-Dial site:
http://www.southwesternindustries.com/swi/prod_measurement1.shtml


Trav-A-Dial
Precision measurement of machine tool travel.
The Trav-A-Dial is a mechanical marvel. It detects machine tool motion by the contact of the gage wheel with a running surface. The wheel is made of tool steel and is very hard – 63C on the Rockwell scale. The Trav-A-Dial's wheel is pushed up against the surface by a special mounting base. After a few traverses, the special pattern scratched into the wheel creates the mirror pattern on the running surface, forming a micro rack-and-pinion system that cannot slip.

Wheel motion is translated through the Trav-A-Dial through a set of very precise gears. In fact, Southwestern Industries’ gear rolling technology enables us to make some of the world’s most precise instrument gears. In order to eliminate backlash, these gears are preloaded in the axial and radial directions – no mean trick.

http://www.southwesternindustries.com/swi/Images/product_images/TRACDIAL_97a.jpg

The result is a mechanical system that is capable of accurate measurement (after calibration) of .0003” and near absolute repeatability.

The Trav-A-Dial gives you precision measurement in an analog display preferred by many machinists.

http://www.southwesternindustries.com/swi/Images/product_images/TRAVDIAL_97.jpg


Model 7A
Inch reading only
001” resolution
10” memory

http://www.southwesternindustries.com/swi/Images/product_images/TRAVDIAL_95.jpg



So the analogue wheel is radiused to adjust to a correct effective diameter to correctly calibrate the rotation the wheel (and the indicator dials) in terms of unit turns of the wheel in relationship to unit axial travel of the lathe carriage relative to the lathe bed and via the saddle to the cutting tool.

The wheel radius also makes it easy to ensure that an adequate surface of the wheel and its track on the lathe bed is sufficient to over-come static and dynamic friction which could cause "wheel-slip" and loss of accuracy and the calibration reference zero.

The relationship of turns to distance relies on the effective diameter and circumference of the "wheel"

The circumference "C" = pi x wheel effective diameter "D"

ie C= 3.1416 x D

3.1496 is a constant but C ans D are variables and a rise or %-age rise of fall in one will cause a similar %-age rise or fall in the other.

As the analogue Trav-a-Dial "C" must be very accurate than so must its effective diameter "D" as other than tilting the wheel, there is no other way of compensating for any error/s in effective diameter "D".

This, as I understand it, is not the case in a digital "Trav-a-Dial" as an error can be corrected by insertion of a constant.

For example, for 4" travel per revolution of the wheel, the wheel circumference needs to be 4":
C = 4 = 3.1416 x D

D = 4/3.1416 = 1.2732"

For C to be 6":

D = 6/3.1416 = 1.9099"

For C to be 8"

D = 8/3.1416 = 2.5465

But if say we were using the 1.9099" diameter wheel and it was only 1.9080" the error in diameter 1.9099 - 1.9080 = 0.0019" is an error of:

((1.9099 - 1.9080)/1.9099) = 0.0019/1.9099 = 0.0009948 which is a constant and can be applied digitally as a correction: 1.9080 x 1.0009948 = 1.9099.

The diameter of the wheel will need to take into account the resolution of the encoder so that effective circumference (= effective travel) must be such that the number of "steps" on the encoder = effective circumference = effective travel (inches)/1,000 for an effective saddle travel to have an effective accuracy of 0.001" to match the 0.001" accuracy of the Trav-a-Dial.

That the math and the "mechanical" aka "nuts and bolts of it as I see it.

I will leave the electronics and the remote display to our collective intrepid EE's as directed by


Allrighty then.

You got your 2cents in. Sit back with me and wait for the EE types, and by all thats holy if you repost the caliper bodge in this thread I'm gonna have a stroke. If I survive it I'm buying a plane ticket to OZ and bringing a club.

So the issue is restricted to EE's only by some power vested in or invoked by SVS which by extension excludes all non-EE's.

Wasn't it nice of him to offer to bring over some of his (HSM board??) friends and club-mates (his club?) to see me in OZ - if he doesn't have a stroke first.

I am quite sure he will have good cause to remember his trip to OZ.

I am still waiting patiently for SVS and his EE compatriots.

Weston Bye
01-25-2011, 06:18 AM
I am still waiting patiently for SVS and his EE compatriots.

It seems that an electronic solution has been put forward. Certainly not as Athena - sprang fully formed from the forehead of Zeus, but a workable plan nonetheless.

http://en.wikipedia.org/wiki/Athena


If anyone wants to try this but scared of the silicon part, I would be happy to supply Arduino code for this. I'm working on a project which entails an LCD and an encoder so it would be an easy hack. I have cnc on my machines so not much interest in this but happy to lend a hand. You can buy an Arduino for $30 and program it with any pc and a USB cable. Easy peasy.

Fwiw, it will be much cleaner if one edge count in the encoder equates to one increment of measurement. At this scale floating-point math can become annoying and either compounds the error or just smushes it around. Or be off by a clean factor of ten like .0009, .0008 per step so you can just accumulate the error and correct it every N counts.

Bmyers
01-25-2011, 07:01 AM
OK new thread to move away from Black forests thread.

We finished up asking why it was not possible to use a cheap encoder and one of these also cheap Chinese remote displays.

http://www.arceurotrade.co.uk/imagecache/3286dc1a-1abf-4d1d-8f1b-9b4a00bab943_250x125.jpg

Any idea's or is anyone up to the job ?

Here is how I see it. Those little displays are about $69.00 on eBay. You still need some sort of interface to go from quad encoder to the signal the DRO reads. An Arduino is $30.00 and LCD another $12.00, this provides the ability to modify code and to read any input you can wire to it. I still want to pursue a digital travel dial, but I have abandoned the Chinese display. I think it just complicates the electronics. As soon as I get my Arduino and an optical mouse I am going to start tinkering.
I think anyone who can machine and read a drawing could build a Arduino based travel dial. If I get this working in a reasonable amount of time I will share code and drawings.
I am also looking for a cheap laptop for programming. Any help would be appreciated.

SVS
01-25-2011, 08:29 AM
Tiffie, that was a fine summation of the issues. On topic even.

BUT.

Broke no new ground, and I fail to understand your interest in the subject.

Doubtful you'll actually build one.

NOTHING is stopping YOU from forging ahead. Please show your work here. We are dying to see it.

I have no fear of the mechanical part. I have a lot to learn regarding the electronics....don't even know what I don't know. I'm studying, when I have an estimate of price and shop time I'll make my decision. Might happen next week, next year, next decade, who knows, so don't let lil' ol' me slow ya down.

oldtiffie
01-25-2011, 08:57 AM
Well SVS,

it seems that we are both in the non-EE camp and unlikely to change anytime soon - so I leave it to the experts.

I certainly did not break any new ground as from a mechanical aspect it is hard to improve on what Trav-a-Dial has done as it had already broken any ground that needs breaking in that regard.

All that I did was to define the maths and mechanics including a T-a-D type wheel but driving an encoder but with provision for calibration of the encoder points against actual physical distance by the lathe carriage in relation/relative to the lathe bed to achieve an accuracy of 0.001" as is the case with the T-a-D.

How the electronics are processed and displayed on a remote screen is of no interest to me as electrons is outside my bailiwick.

The unit must not only be accurate but realistically rugged and reliable enough for the environments it is intended to work in.

It must be cheap enough for the market that it is intended to serve - small and large shops as well as HSM-ers. The real/original T-a-D is priced beyond most small shops and many HSM-ers.

Further, the T-a-D is in "inch" units only - no metric - and many HSM-ers are entirely or mostly metric or change from one to the other - digitally. This is a standard feature in the caliper set-ups I showed as well as the unit shown in John Stevenson's OP.

Other than a variation/s to or adaption/s of a T-a-D wheel set-up or principal, many are but variations of the DRO principles.

I can but wait patiently for the EE's to come up with either proof of concept, a working prototype or a product ready for market.

In the meantime I will keep using my trusty caliper or DRO adaptions.

J Tiers
01-25-2011, 08:57 AM
Perhaps Tiffie is FORGETTING an important point......

As I believe he has often mentioned, to get an "accuracy" of 0.001, you must generally have an actual instrument accuracy in the area of 0.0001 to 0.0003.

The rule is either 10x or 3x, depending on whose numbers you quote, or, if a scientific basis is wanted, how you assume the errors add. Linear addition argues for the 0.0001, RMS addition argues for the 0.0003.

In any case, that adds at least a "astronomical order of magnitude" to the required true accuracy to obtain the apparently desired 0.001".

The concept of "merely adding a constant" for calibration is attractive, but needs consideration as to the math.

Adding to what?..... if the basic count is just barely enough to do the job, then it is necessary to do interpolation, for which you may have no numeric justification, in order to avoid the buildup of errors and their sudden correction. The result could be on the order of the movement of Passover around the calendar, where there is drift and then a jump as the corrective month is employed.

Alternately, the original resolution needs to be larger, in which case you add expense which might as well have been used towards a better adjustment method, or better basic accuracy.

it may be necessary to do math to a better precision, possibly requiring a faster or mor capable and expensive processor.

The crowned wheel begins to look good, as the crown has other good features as well.

As for "inch", I own several dial calipers which do both inch and metric, the 'technology" to do that even as a mechanical solution is available, the T-A-D folks simply didn't do it...... Electronically the count "value" as displayed can be adjusted to suit metric, inch, old French inches, or any other unit as desired.

oldtiffie
01-25-2011, 09:09 AM
JT.

You have done a fine job of defining the limitations and practicalities of such things.

The are big hurdles to jump.

If there was a big enough demand for it to be a commercial and practical proposition, I can but wonder why it hasn't progressed beyond the expensive "inches only" analogue T-a-D.

Perhaps we are finding out why that may be so.

But perhaps we will have a "Eureka" moment here.

I hope so.

http://en.wikipedia.org/wiki/Eureka_(word)

http://en.wikipedia.org/wiki/Eureka_effect

oldtiffie
01-25-2011, 09:39 AM
Perhaps Tiffie is FORGETTING an important point......

As I believe he has often mentioned, to get an "accuracy" of 0.001, you must generally have an actual instrument accuracy in the area of 0.0001 to 0.0003.

The rule is either 10x or 3x, depending on whose numbers you quote, or, if a scientific basis is wanted, how you assume the errors add. Linear addition argues for the 0.0001, RMS addition argues for the 0.0003.

In any case, that adds at least a "astronomical order of magnitude" to the required true accuracy to obtain the apparently desired 0.001".

The concept of "merely adding a constant" for calibration is attractive, but needs consideration as to the math.

Adding to what?..... if the basic count is just barely enough to do the job, then it is necessary to do interpolation, for which you may have no numeric justification, in order to avoid the buildup of errors and their sudden correction. The result could be on the order of the movement of Passover around the calendar, where there is drift and then a jump as the corrective month is employed.

Alternately, the original resolution needs to be larger, in which case you add expense which might as well have been used towards a better adjustment method, or better basic accuracy.

it may be necessary to do math to a better precision, possibly requiring a faster or mor capable and expensive processor.

The crowned wheel begins to look good, as the crown has other good features as well.

As for "inch", I own several dial calipers which do both inch and metric, the 'technology" to do that even as a mechanical solution is available, the T-A-D folks simply didn't do it...... Electronically the count "value" as displayed can be adjusted to suit metric, inch, old French inches, or any other unit as desired.

JT.

I sort of made mention of the say 2" travel wheel similar to a T-a-D wheel having 2,000 or multiples of 2,000 encoder step. Given the T-a-D has more than adequately solve the "no slip" problem and providing that no encoder steps are lost, I'd have thought that with a calibration error compensation factor that the accuracy problem had already been adequately addressed.


As I believe he has often mentioned, to get an "accuracy" of 0.001, you must generally have an actual instrument accuracy in the area of 0.0001 to 0.0003.

The rule is either 10x or 3x, depending on whose numbers you quote, or, if a scientific basis is wanted, how you assume the errors add. Linear addition argues for the 0.0001, RMS addition argues for the 0.0003.

That's quite true. I did say that several/many times but it was in the context of checking a physical measuring or sizing device against another else instead of becoming a validated measuring device it operates as a comparator.

ikdor
01-25-2011, 02:21 PM
The concept of the mouse is flawed by it reporting movement in a 30 pixel wide window. That means there is at least a 3% error for each step. The error might average out a bit and make a very crude digital TAD but you would need to reset it regularly. I'll give it a try if I ever find the time.

I think the best bang for the buck would be to make a small piece of electronics reading a magnetic tape and showing it on an on board small display.
So not as compact as a TAD, as you'd need to stick the tape on your lathe, but you could run it for a long time of an AA battery so it doesn't need wires.

Total cost would be around 100 euro including the tape and you can have a resolution down to 0.5um (0.00002") so you're basically stuck with the accuracy of the tape itself. I'll try to build it and post some photo's, but don't hold your breath as my fatherhood has seriously reduced my hobby time.

Igor

MCS
01-25-2011, 03:01 PM
I think the best bang for the buck would be to make a small piece of electronics reading a magnetic tape and showing it on an on board small display.
So not as compact as a TAD, as you'd need to stick the tape on your lathe, but you could run it for a long time of an AA battery so it doesn't need wires.


Igor

That's the spirit! The gettofication off the linear scale.

I think the TAD is a great 1960's design. I can see in my imagination man in white laboratory coats explaining the use to man in blue overalls.

If you replace the clock with an encoder, you have an 1960's design with an 1980 digital appendix. Direct connecting the encoder to the contact wheel is not possible because of the applied force, so you need a backlash free geartrain too. Progress: little, only that the position of the readout is somewhere else.

Encoder plus wheel plus geartrain can be considered equal in cost to an commercial reading head plus magnetic tape.

So your idea is simpler, more 2000's, repeatable and cheap.

Unfortunely my analog electronics are not that good, but I'm going to look into it. And as everybody is getting the same idea at the same time, it must be somewhere on the internet.

macona
01-25-2011, 03:39 PM
Unfortunely my analog electronics are not that good, but I'm going to look into it. And as everybody is getting the same idea at the same time, it must be somewhere on the internet.

Nothing analog about it. He is referring to the magnetic strip encoder, I think. Still, they are not cheap though. I have one in the machine I am turning into a laser cutter.

Ill stick with glass scales.

John Stevenson
01-25-2011, 03:43 PM
The tape is avahere for £5.63 per 100mm

http://www.machine-dro.co.uk/index.php?target=categories&category_id=60

For UK users and US Digital in the US has 500 count strip for £42 per 10"

http://usdigital.com/products/encoders/incremental/linear/lin/

The Artful Bodger
01-25-2011, 03:58 PM
It would still be interesting to see the accuracy of a pinion engaging the lathe rack, backlash accommodated of course, if nothing else it would at least be highly repeatable which is the greatest uncertainty in the contact wheel approach.

Hmmmm, having thought about it for a moment.... you could get one 'good' screw the length of the carriage travel, mount that in the spindle with a nut on the tool post and calibrate against that, then you could move the nut a bit using the compound parallel to the spindle axis and repeat the calibration. Doing that a few times and taking the average would get a fairly accurate calibration, I would think.

IMHO you only need custom electronics for the display and processing if you are wanting to make and sell the devices, otherwise and as per an earlier suggestion by me, use an otherwise obsolete and unused laptop computer.

BillTodd
01-25-2011, 04:13 PM
It would still be interesting to see the accuracy of a pinion engaging the lathe rack, backlash accommodated of course,
My Hardinge's carriage wheel is calibrated in 10 thou" divisions, they are quite wide and could easily have been 2 thou".

However, a lathe's carriage rack and pinion is often subjected to considerable wear (a notable weak spot on HLVs) which may make it unsuitable to use as a scale.

Bill

MCS
01-25-2011, 04:34 PM
Nothing analog about it. He is referring to the magnetic strip encoder, I think. Still, they are not cheap though. I have one in the machine I am turning into a laser cutter.

Ill stick with glass scales.

Generally translated from the Sony catalog:

"Magnescale is a magnetic measuring system usin a multiple flux sensitive reading head detecting the movement.

The scale is placed with a period of 0.2 mm on magnetic materiaal tape or bar shaped."

A period makes the scale analog, because the readout is brought to higher resolution using two reading heads, the signals sinus/cosinus interpolated.

These interpolators can be in the DRO housing, or in the reading head, they translate the signal to AB phase.

My country fellow man Ikdor proposed to create this from the ground up, which can be done, where later the (cassette) tape can be replaced by the cheap magnetic tape as in John's link.

The "out of stock" of the 0.5 mu reading head is a stimulus.

The advantage is the low resolution of the tape, making for example mounting not critical, followed by the disadvantage of the need of the converter.

I would start with the official tape and take it from there.

macona
01-25-2011, 05:02 PM
The problem with the stuff from US Digital is it is only available up to 500 counts per inch which would give you an ultimate resolution of .0005. Also the longest length you can get is 34". There was some talk of longer ones a few years ago for things like CNC routers but they never showed up.

The pricing on the magnetic encoders is pretty good. Really good for a long machines with the max single length of 25 meters.

SVS
01-25-2011, 05:33 PM
Direct connecting the encoder to the contact wheel is not possible because of the applied force.......

Why Not????

~2" OD wheel with stub axles(one piece, turned in one setup)-A coupling can be incorporated in one axle to drive the encoder. So long as the encoder is less than 4 or 5" OD most lathes will have a suitable mounting location and thanks to the remote display the back of the saddle is a good as the front.

For myself, I'm bailing out if encoder gearing is required. The backlash and precision hassles tip the scale on a marginal situation.

Also, earlier in the thread J Tiers questioned if the 10" range of the dial might be the T A D's max travel for quoted accuracy.....S.W. Industries calls it a 10" Memory....Makes sense but find it odd to hear "memory" regarding a mechanical device.

oldtiffie
01-25-2011, 05:38 PM
That's the spirit! The gettofication off the linear scale.

I think the TAD is a great 1960's design. I can see in my imagination man in white laboratory coats explaining the use to man in blue overalls.

If you replace the clock with an encoder, you have an 1960's design with an 1980 digital appendix. Direct connecting the encoder to the contact wheel is not possible because of the applied force, so you need a backlash free geartrain too. Progress: little, only that the position of the readout is somewhere else.

Encoder plus wheel plus geartrain can be considered equal in cost to an commercial reading head plus magnetic tape.

So your idea is simpler, more 2000's, repeatable and cheap.

Unfortunely my analog electronics are not that good, but I'm going to look into it. And as everybody is getting the same idea at the same time, it must be somewhere on the internet.


Direct connecting the encoder to the contact wheel is not possible because of the applied force, so you need a backlash free geartrain too.

I don't think so.

If the encoder is close and direct coupled to and axially aligned with the "T-a-D" (type) wheel there is no side/lateral thrust on the encoder and there is no back-lash or "play" between the wheel and the encoder.

With its etched "rack and pinion" drive, the TAD has effectively solved the back-lash problem there too.

Back-lash is really not a problem at all as this is not CNC machinery with high precision ball-screws and zero-adjusted end-play. It is only a non-CNC-ed shop lathe. As all "slack" and "back-lash" is automatically taken up/out just after the start of uni-directional saddle movement for cutting, it is the same as taking up back-lash when power-feeding or screw-cutting.

Similarly, as the T-a-D wheel has solved the lateral thrust problem, with the encoder coupled to the wheel, there is no angular back-lash either.
http://en.wikipedia.org/wiki/Oldham_coupler

http://www.mekanizmalar.com/oldham.html

http://www.google.com.au/search?hl=en&rlz=1T4IRFC_enAU360AU360&q=oldham+coupling&aq=0&aqi=g10&aql=&oq=oldham+cou

I have an Oldham's coupler between the 180 step and 1/2-step stepper motor on my "Divisionmaster" digital indexer which is connected directly with an Oldhams coupler to the worm shaft on my 6" "Vertex" rotary table. There is no back-lash between the stepper and the worm shaft nor is there any end thrust as the coupler eliminates both.

http://divisionmaster.co.uk/divisionmaster.html

http://divisionmaster.co.uk/examples.html

So far as I can see, all of the maths, mechanical and "nuts and bolts" issues (real and imagined) have been addressed and solved previously.

It seems that all that is required is to get the electrical and digital stuff done.

oldtiffie
01-25-2011, 06:05 PM
Why Not????

~2" OD wheel with stub axles(one piece, turned in one setup)-A coupling can be incorporated in one axle to drive the encoder. So long as the encoder is less than 4 or 5" OD most lathes will have a suitable mounting location and thanks to the remote display the back of the saddle is a good as the front.

For myself, I'm bailing out if encoder gearing is required. The backlash and precision hassles tip the scale on a marginal situation.Also, earlier in the thread J Tiers questioned if the 10" range of the dial might be the T A D's max travel for quoted accuracy.....S.W. Industries calls it a 10" Memory....Makes sense but find it odd to hear "memory" regarding a mechanical device.

I agree with you SVS as regards the coupling of the encoder spindle to the wheel spindle.

If direct coupled (Oldhams coupler) there is no back-lash or end-thrust/play either.

Gears are only required to drive mechanical indicator dials as on the T-a-D, but as the wheel is only driving the encoder, there is no need for gears at all. The "distance" will be shown digitally on the remote display.

The Artful Bodger
01-25-2011, 06:56 PM
My Hardinge's carriage wheel is calibrated in 10 thou" divisions, they are quite wide and could easily have been 2 thou".

However, a lathe's carriage rack and pinion is often subjected to considerable wear (a notable weak spot on HLVs) which may make it unsuitable to use as a scale.

Bill


Bill, thats why any rack system would need a method of calibrating it and if we have electronics in the mix we might as well make full use of them and 'map' the errors of the rack which is what I was suggesting with my reference to a calibration screw.

As far as I know nothing in the real world is infinitely accurate, nothing is 'flat', nothing is 'straight', nothing is 'round'. The most we can do is measure and calibrate against something that we believe is at least of the accuracy that we require. If we cant do that we can take the average of lesser references and hopefully achieve the accuracy we require.

Electronics make this averaging and the creation of a error profile achievable.

The more I think about it the more I become inclined to believe the lathe rack is well worth further consideration.


[Now I am trying to figure out a way of getting rid of backlash between the rack and the handwheel spindle.]

Weston Bye
01-25-2011, 08:12 PM
... The more I think about it the more I become inclined to believe the lathe rack is well worth further consideration.


[Now I am trying to figure out a way of getting rid of backlash between the rack and the handwheel spindle.]

An encoder engaging the rack could be equipped with a clockspring to keep the encoder pinion biased always in one direction against the face on one side of the rack teeth. The clockspring would have to have enough "wind up" to travel the whole length of the rack.

oldtiffie
01-25-2011, 08:36 PM
No need for a spring - a counter-weight on a cord over a roller will do.

I am not all sure that a lathe rack is all that accurate. If any sort of accuracy and avoidance of cyclic error is required that rack and pinion meshed set will need to be very good.

Weston Bye
01-25-2011, 08:41 PM
Agree on both points, Tiffie.

The Artful Bodger
01-25-2011, 09:41 PM
...... but I think they would have to be rather strong as they would have to carry the working pressures of the saddle travel...............

Ignore that bit, I was having a brain fade....

macona
01-25-2011, 10:18 PM
Thats what these gears are for:

http://www.antibacklashgear.com/

oldtiffie
01-25-2011, 10:38 PM
Those gears are 24>120 pitch (DP?).


Nordex Anti-Backlash gears are available in two designs - Interchangeable and compact. Anti Backlash gears are available in 24 Pitch to 120 Pitch in stainless steel and aluminum, pin hub and split hub. Nordex also offers Anti-Backlash Miter Gear sets. Please consult Nordex customer service department for unlisted sizes. Custom machining available on request.

at:http://www.antibacklashgear.com/

I am not sure that they would suit too many lathe racks.

J Tiers
01-25-2011, 10:50 PM
Well, I sure don't agree with whoever said that the encoder must have gears, or that there is no advantage in reading the wheel rotations via an encoder.

1) The encoder need not be exactly the same size... it could be bigger, as there is usually some extra room over the ways without hitting the T/S. The whole thing can tilt, I believe the TAD does. You don't need to just tilt the wheel and have some sort of coupling.

2) There is a HUGE advantage in bringing the encoder inside, as it then is protected and kept clean. if the wheel is right, and the path kept reasonably clean and clear of chips, it should work well.

oldtiffie
01-25-2011, 11:42 PM
Thanks JT.

Taking that at face value - as a non-EE - it seems that all we have to do is get all the electric string and stuff set up to have the (corrected/modified/adjusted??) encoder out-put converted to characters on a remote digital display.

As long as the effective circumference of the wheel (which is close-coupled to the encoder) in thousandths of an inch (0.001") has the same number of 0.001's as there are pulses per revolution of the integral encoder the problem should be getting close to resolution - as you say.

This solution seems to accord with the OP in every way but at the accuracy of Trav-a-Dial (ie 0.001").

The OP is (still):


OK new thread to move away from Black forests thread.

We finished up asking why it was not possible to use a cheap encoder and one of these also cheap Chinese remote displays.

http://www.arceurotrade.co.uk/imagecache/3286dc1a-1abf-4d1d-8f1b-9b4a00bab943_250x125.jpg

Simple cheap encoder with a hardened knurled radiused wheel that can be tilted to get the correct circumference driving into one of these units.

The protocol for the Chines signals is documented on the Shumatech site which is down at the moment but I may be able to get it elsewhere.

Basically it's a reversal of what Scott Schumacher has done.

Any idea's or is anyone up to the job ?

J Tiers
01-25-2011, 11:56 PM
Thanks JT.

Taking that at face value - as a non-EE - it seems that all we have to do is get all the electric string and stuff set up to have the (corrected/modified/adjusted??) encoder out-put converted to characters on a remote digital display.

As long as the effective circumference of the wheel (which is close-coupled to the encoder) in thousandths of an inch (0.001") has the same number of 0.001's as there are pulses per revolution of the integral encoder the problem should be getting close to resolution - as you say.

This solution seems to accord with the OP in every way but at the accuracy of Trav-a-Dial (ie 0.001").


In principle, yes, as a marketing-level description.

Making suitable bits and getting it working to desired accuracy might be more of an issue, there are always details. I expect that you want a bit better than teh minimum resolution on the encoder, but it would work that way, to +- 1 count overlaid on top of the wheel-to-distance conversion accuracy.

That's actually a bit worse than +- 0.001", depending on the wheel cal.

oldtiffie
01-26-2011, 01:21 AM
That's so JT.

But wheel diameter (and therefore circumference) error - within reason - can be corrected by a software correction factor (see previous) that can be accurate to four or more decimal places.

Its all a matter of comparing the actual saddle travel in 0.001's of an inch compared to the number of encoder pulses per circumference of the wheel.

Say 10 x revolutions of a 4" circumference wheel (diameter = 1.2732") = 40" = 10 x 4,000 = 40,000 encoder pulses.

If say the actual travel was 39.988" the error would be 40.000 - 39.988 = 0.012".

The error as a ratio to 40.000" is (40.000 - 39.988)/40.000" = 0.012/40.000 = 0.0003.

The correction factor is 1.0003

When applied it becomes 40.000/1.0003 = 39.988" which will be the out-put to the remote so that the remote display shows the actual distance traveled.

Its all relatively simple maths that just needs to be incorporated into the electronics. I would imagine that changing the correction factor from time to time as the units is re-calibrated should not be a problem to key in either internally or via a USB port

MCS
01-26-2011, 05:32 AM
Ok, forgot the Oldham coupling, but found the chip.

http://www.austriamicrosystems.com/eng/Support/Design-Resources/Demo-Eval.-Progr.-Boards/Linear-Encoder/AS5311-DB-Demoboard

So, a magnetic scale reading head can be bought as a component, reducing the problem to digital.

ikdor
01-26-2011, 06:49 AM
http://www.austriamicrosystems.com/eng/Support/Design-Resources/Demo-Eval.-Progr.-Boards/Linear-Encoder/AS5311-DB-Demoboard

So, a magnetic scale reading head can be bought as a component, reducing the problem to digital.

That encoder head was exactly the one I had in mind. They cost about 10euro each, so combined with the tape from machine-dro you'll end up near 100 euro for the total.

Igor

MCS
01-26-2011, 07:07 AM
That encoder head was exactly the one I had in mind. They cost about 10euro each, so combined with the tape from machine-dro you'll end up near 100 euro for the total.

Igor

Just ordered 3 samples for free.:D

J Tiers
01-26-2011, 08:44 AM
That's so JT.

But wheel diameter (and therefore circumference) error - within reason - can be corrected by a software correction factor (see previous) that can be accurate to four or more decimal places.

Its all a matter of comparing the actual saddle travel in 0.001's of an inch compared to the number of encoder pulses per circumference of the wheel.

Its all relatively simple maths that just needs to be incorporated into the electronics. I would imagine that changing the correction factor from time to time as the units is re-calibrated should not be a problem to key in either internally or via a USB port


Tiffie, you may need to study..........

math in the encoder and digital world does not involve an arbitrary number of decimal places as desired.

On paper, one can simply write another number. In a "microcomputer", the number must be represented as a certain number of "bits".

One does not "invent" the value of these bits, they are data. But the data has to exist. AND the smallest number that can be represented is ONE "bit".

if you are satisfied to count 1:1 by 0.001", then in say a 2" distance, one might manage to encode 2000 "counts".

However, now the numbers you can actually represent are LIMITED..... the only numbers NEAR 2.000 inches that you can represent with the data you have, is 1.9999 or 2.001 inches.

You may have the resolution in 2 "bytes" to represent 65,535 different numbers, but you have no reason to pick one over another, because all you know is that you counted the 1,999th count. You don't have any idea, with that data, whether you are really at 1.9991 or 1.9994, 1.9997, etc.

You can't use time to interpolate, because you don't know that the motion has not stopped, slowed, etc.

As a result, calibration changes are rather "stepped"..... in 0.001 steps.

Applying a *constant* correction per 2000 counts, you can, over 10", choose the full count, 10.000, or you can choose 9.995, or 10.005, by dropping or adding one count over 2". But you can't get to 9.999 or 10.001, because each of those would require knowing the measured distance to 0.0005 instead of 0.001 only.

if you decide to drop ONE 0.001 per 10 inches, now your number looks good, but you have a "drift " error, which is corrected periodically, and your real accuracy is not better, although resolution of the correction is improved. You take 5 steps forward and one step back.

Nonetheless, you can't do better than a "window" of 0.002" plus the basic error of distance measurement. It is inherent in measuring only to 0.001. In fact, your error is probably at best plus or minus 2 counts, for a 4 thou "window".

You must measure to a better resolution in the beginning in order to have the data. That also allows you to apply simple corrections, since you can correct easily by as little as 0.0001 per rev of the wheel.

Any fancy scheme to read phase, etc, is in reality, improving the resolution of the encoder, and the encoder wheel accuracy needs to justify that added resolution. You can't cheat.

Otherwise you you can't justify the number you get. Your "resolution" is better than your "accuracy".

oldtiffie
01-26-2011, 11:17 AM
Thanks JT.

That was one of the "holes" in my "logic" that I could not "see" (didn't know about).

If the "one encoder pulse per 1/1,000" were increased x 10 to say 10 pulses per 0.001" of travel, are there encoders with that sort of numbers of pulses per revolution?

In short, an encoder with say 2,000 pulses per revolution would have an equivalent circumference of 2,000/10 x 0.001" = 0.200" of travel.

Is that "10" factor that I arbitrarily selected OK or are there others, and if so, what are they?

Can you tell me what an ideal factor is for a resolution of 0.001" and also what encoders are there that are not too physically large but which have high counts of numbers of pulses per revolution?

This is going to determine the limits of wheel effective diameter and travel per revolution of the close-coupled wheel and encoder sizes as well as the available distance-measuring resolution.

I anticipate that there may or will be limits and/or compromises that may need to be made or applied.

My WAG is that my error was that I assumed that the electronics would not "lose" pulses where-as that may not be the case. I assumed that as my stepper motor on my DivisionMaster did not lose pulses provided that I keep the speed and load (read: torque) within acceptable limits that all would be well.

I think that I can see where my errors in logic are (at least I hope/think I can).

I would appreciate it if you would put me on the right track so that I can get the mechanical stuff sorted out as best as is reasonable in the circumstances.

I guess that what I need to be able to do is to solve for the best "distance per revolution" of an encoder by using it to solve for the best affective wheel diameter.

If the electronics limit the numbers of pulses per minute that it can process without "losing" any pulses (0.001" steps) then that may limit or be affected by the maximum saddle speed that can be used since the last electronic "zero-ing".

I am limited to the physical mechanical stuff outside the magic box that has the electronics in it as electronics are pretty well a complete mystery to me, hence I am compelled to leave it to others such as yourself.

I don't know what sort of torque is required to turn an encoder either. There is a limit that the wheel/lathe bed friction interface has before the encoder torque causes the "measuring wheel" to physically "slip" and so potentially "lose pulses" due to the physical limitations of a friction drive.

But that may well be solved at least in part if the driving wheel has a sort or hardened "knurled" or embossed contact surface to "etch" itself into the disk travel path on the lathe bed as is the case with the original analogue all-mechanical wheel driving/contact peripheral surface.

Many thanks for having the courage to "get into it" instead of just "standing back".

I realise that this may all end up in a heap but I'd like to give it my best shot to see how close we can get to 0.001" travel resolution.

BillTodd
01-26-2011, 12:11 PM
math in the encoder and digital world does not involve an arbitrary number of decimal places as desired.

While true to a point, there is a way around this...

Say, I can only measure my 2.291831..." *wheel diameter to +/-0.001".

Suppose I make the wheel and it is actually oversize by about a couple of thou, the peripheral error would be PI times this, so my imaginary wheel is now really 7.207.." instead of 7.2" around.

Each pulse from the counter is now actually 7.207/ 7200 = 0.0010009722.." long.

If I counted the pulses assuming them to be 1 thou each after a couple thousand pulses (~2") the display would show 2.000 but would actually be 2.00194.." nearly 2 thou" out.

However, it is a relatively trivial matter to accumulate 0.0010009722 per pulse while only displaying the first three decimal places. Now the display would show 2.002 (rounding up the to the least significant digit) less than a tenth out.

Of course, measuring the error is the biggest problem, but with a suitable gauge (say 10" long) it would be possible to measure the error to a suitable number of digits.

Bill

*[0.001" per pulse with a 7200 ct encoder]

The Artful Bodger
01-26-2011, 02:19 PM
One of the challenges for a home shopper such as I would be to get useful calibration. Can anyone comment on my suggestions to calibrate against a series of gauges of unknown accuracy and averaging the results?

MCS
01-26-2011, 03:51 PM
Just a simple example of how a calculation can be brought into comprehendible numbers. With it you can also calculate the feasability of a rotary encoder.

The linear sensor given to me has a 12-bit resolution in a cycle of 2 mm.

This equates to 4096 pulses per 2 mm.

So to get millimeters I have to divide it by 2048.

To get hundreths of millimeters I have to divide it by 20.48. Which is not practical. The point on a display is just a LED which is on, the number on the display is an integer number. So I multiply the readout by 100.

For example: the real position is 1.75 mm, equating to an encoder count of 3584.

3584 * 100 / 2048 = 175.

These integer multiply divide conversions can solve within reason most problems, if you have enough resolution to start with.

oldtiffie
01-26-2011, 06:18 PM
Bill.

Thanks.

If I get a consistent number of pulses per unit length using an accurate say 10" guage, then as you say, calculating the unit length per unit pulse is an easy maths problem and a calculated correction factor could be inserted to correct any error.

Once a constant distance traveled per encoder pulse was known the actual wheel diameter becomes irrelevant as its only function is to provide its circumference (= 3.1416 x D) which, assuming no "slip" at the friction interface defines how far the saddle will move per revolution of the encoder.

The main problem as I see it from JT's advise is getting the electronics to out-put that actual physical movement of the saddle as say 10.000" at the remote display.

If I understand JT correctly, there is a minimum number of pulses that are required to accurately show any given length accurately in the electronics.

If there is a known error it may be able to be corrected electronically but that way is out of my field and I have to leave it to others.

If there is a maximum number of pulses per second that the electronics can process, then depending on the pulses per revolution of the encoder produces, that will limit the revs/minute of the encoder and the maximum travel speed of the saddle.

I have no idea of what torque is required to accelerate or retard the encoder nor what steady speed of it takes either as that may exceed the limit of the friction interface between the wheel and the lathe bed to transmit it without physically slipping and causing an unknown number of encoder pulses to be lost. The Trav-a-Dial positive drive is far superior to a plain friction drive and would require a lot less thrust than a plain roller. The "etched" T-a-D will require much lighter bearings and structure. The T-a-D dial with its raised surface and its etching of the lathe bed is very close to an acts as a very accurate rack and pinion.

I keep getting the sense that this way of measuring saddle travel is really best described as a real T-a-D with the gears stripped out and an encoder close-coupled with an Oldhams coupler to the T-a-D embossed/"knurled"?? wheel fitted to it instead.

I keep reverting to a complete T-a-D with a web-cam focused on the dials (calibrated to 0.001") with the web-cam image shown on a remote monitor.

The T-a-D is in the optimum position on the lathe as it is relatively easily see/read and install and is not obstructed nor does it obstruct anything else on the lathe as a DRO or scale may. It is not as likely to be in the swarf and cutting fluid envelope as DRO's and scales may be and which don't always take kindly to that sort of operating or working environment.

It was to avoid that aggressive environment that I used my "clamp on caliper" methods as they are easily mounted and dismounted and as they cannot move further than 6"/150mm will be within 6"/150mm of the previous caliper zero and will have an error of +/- 0.001". Many can be connected to a remote display for output.

The T-a-D wheel as required/suggested in the OP is getting to be less likely or affordable.

Pending a solution to the encoder/electronics matter, it seems that a scale with a reading head ("caliper" or DRO) seems to be the optimum affordable solution that is readily available at many traders/outlets and are "bolt on out and go out of the box" very accurate solutions.

Having said that, I am quite happy to progress as far as I/we can until it works or is proved not to be possible or feasible.

oldtiffie
01-26-2011, 06:23 PM
Thanks MCS.

Appreciated.

That un-muddies the waters more than somewhat.

If nothing else this thread/project has really made me appreciate the concept and development of the real Trav-a-Dial to being the excellent mechanism that it is.

J Tiers
01-26-2011, 10:38 PM
One of the issues is that the math to add little imaginary bits to the counts requires several things.

You may have to do the math in more complicated ways, since it may take several bytes to accumulate the counts.

For a 36" travel, to nearest 0.001, you already require 36000 counts. That brings you to 2 byte math, which can count to 65535.

If you then need to add small amounts to that, such as the 0.0000009722 added per 0.001 in the prior example, you need resolution to another 1/1028, which requires another 2 bytes, since one byte gets you 255 only.

So you are accumulating and adding on 4 bytes. Not a problem, in general, as there are C libraries for most chips (not all of them correct, however) to do that. And that particular problem is actually easily solved even if you write in assembler.

if any conversion is required, to metric or from metric, etc, then it may be handy to use a library routine to save trouble, as math other than addition becomes tedious to write. The calculations get longer, and results etc take more space.

As the error becomes smaller, you may get clever, and wait to add corrections until the number becomes more significant...because the number of bits needed to express the result increases. But a correction strategy dependent on the magnitude of the error is cumbersome.

However, you STILL don't get away from the resolution issue.... counting 1000 per inch gets you a basic uncertainty of plus or minus 0.001 pretty much independent of other errors. Its just a question of the decision as to whether it is 0.001 or 0.002..... The precise point of the change is not known closely, or the encoder would be better than claimed.

you will do well to be within 0.005 over a few inches, if you include other errors of 0.001 or so.

You need probably to count to an additional bit or two resolution so that the result can be reasonably correct. That does tend to increase costs.

Then also, how is the cal INPUT? You need some means to put in a probably installation-dependent cal factor. That means a human interface of some sort, at added cost. The old tilted crowned wheel looks pretty good.

It's all do-able, but fundamentally it is a problem which has been already solved, by the DRO. With the DRO existing as a possible solution, one really need not consider any alternate solution which is not accurate to 0.001 max error.

How cheap would a display need to be to make it attractive if it had an error of +- 0.002? How cheap if the accuracy was within 0.005?

There you have it.... at some point it simply isn't worth ANYTHING, as you can do just as well with a rule stuck on the lathe bed.

dp
01-26-2011, 11:33 PM
You could always have 10" indexes that trip a secondary counter for 10's of inches and have 65535 counts per 10" rather than over the entirety of the travel? Or 1" indexes, for that matter.