PDA

View Full Version : home made linear encoders??



Mcgyver
09-07-2005, 06:17 PM
reading David's efforts on hooking up encoders got me thinking....do the electronic geniuses here think that a home made version of a Newall encoder is possible?

If you don't know what a Newall encoder is, check out http://www.newall.com/LEDs/operation.htm

They are a very neat linear encoder technolgoy, y are accurate to like 1 micron over a meter and have no moving parts.

Unlike glass scales, maybe, just maybe, that head with the coils could be a diy project - I don't know enough about electronics to know if thats nonsence or not, but if possible it would open up cheap dro and closed loop servo bases cnc systems for the home shop, which imo would be very cool.

what do you guys think - would it be feasible and you have ideas on how to build the encoder unit??

topct
09-07-2005, 07:05 PM
Aren't these very expensive? There is a reason why. And does an HSM need this kind of accuracy?



------------------
Gene

IOWOLF
09-07-2005, 07:13 PM
If you can do it just as accurate and still be cheaper than a store bought DRO,Great, they are sure easy on the eyes.

------------------
The tame Wolf !

Leigh
09-07-2005, 07:39 PM
<font face="Verdana, Arial" size="2">Originally posted by Mcgyver:
reading David's efforts on hooking up encoders got me thinking....do the electronic geniuses here think that a home made version of a Newall encoder is possible?

If you don't know what a Newall encoder is, check out http://www.newall.com/LEDs/operation.htm

They are a very neat linear encoder technolgoy, y are accurate to like 1 micron over a meter and have no moving parts.

Unlike glass scales, maybe, just maybe, that head with the coils could be a diy project - I don't know enough about electronics to know if thats nonsence or not, but if possible it would open up cheap dro and closed loop servo bases cnc systems for the home shop, which imo would be very cool.
what do you guys think - would it be feasible and you have ideas on how to build the encoder unit??</font>

Feasibility varies directly with the depth of your pockets http://bbs.homeshopmachinist.net//biggrin.gif

The system is quite simple, and would not be too difficult to implement (obvious patent questions aside). The problem is in the balls. If you want accuracy, your balls have to be very uniform. McMaster-Carr sells a 1/16" stainless steel ball with 10 micron sphericity and 700 micron diametric tolerance for $6.26/100.

Then you have to design the circuitry, have a PC board designed and fabricated, write the DSP code, design the housing, design the ball carrier, design the mounts, design the cabling and define the connector.

Then you have to design the controller, the display, the user interface, the housing, write the microprocessor software, define the power supply. Layout the PC board and get it fabricated.

Then you must buy the parts, stuff and solder the PC boards, assemble the various components, make up the interconnecting cables.

And test it http://bbs.homeshopmachinist.net//eek.gif

A piece o' cake http://bbs.homeshopmachinist.net//biggrin.gif

[This message has been edited by Leigh (edited 09-07-2005).]

Peter Sanders
09-07-2005, 08:20 PM
Hi

There is a 2 part (I think) article in MEW Model Engineers Workshop, on using a computer mouse as such a device. It does seem to be remarkably accurate and is very simple to make. I don't have the issue numbers.

hammerhead74000
09-07-2005, 08:27 PM
Don't forget calibrating it against a known standard, too...

Davis In SC
09-07-2005, 08:36 PM
I was under the impression, that the Newall unit must be "Taught" at the factory.... I saw a cutaway of the scale at a Travers Tool Customer Day. Seems like they had a utit operating under water...

Mcgyver
09-07-2005, 09:03 PM
Leigh, your points are good, but basically brush everything away except the linear encoder.

The reason I say this is if one thought through a dro or closed loop cnc system for a diy perspective that didn’t' rely on screw accuracy (i.e. base on linear encoders), everything else is fairly accessible. I'm in the process of making most of the things you mentioned for a pcb mill viand have learned there is lots of open source designs and the components are not expensive.

The exception is the linear encoders, which as far as I can tell, meant glass up until these puppies. The reference is accessible to all of us, ball bearings + the electronics, while beyond my ability to design, seemed to have diy potential.

700 micron diametric sounds awful, I’m taking a micron to mean 1/1,000,000 meter or about 3/100,000 thou? I don't know specs on high class ball bearings (the ball, not TIR) but that’s what Newall says they use.

So assuming ball bearings are reasonably accurate and priced, a suitable non magnetic tube can be found, and whatever you are doing with the ttl output exists or is assumed, any thoughts on how to build one?

I willing to accept that it may be impossible, but not because of what they retail for...there has to be a technical constraint, i.e. we build lots of $500 or $1000 items in our shops for $50. Now, it may still be impossible, but if I can I'd like to find out why before giving up.

Leigh
09-07-2005, 09:29 PM
<font face="Verdana, Arial" size="2">Originally posted by Mcgyver:
...So assuming ball bearings are reasonably accurate and priced, a suitable non magnetic tube can be found, and whatever you are doing with the ttl output exists or is assumed, any thoughts on how to build one?</font>

The problem is: The TTL output that you mention is the result of the processing done by the DSP. This is definitely non-trivial. I've designed many similar systems, and I can tell you that it takes some very good engineers and programmers to make something like this work, and considerable $$$.

ibewgypsie
09-07-2005, 11:03 PM
Yeah,, guys...

A pair of magnetic pickups like under your car on the balancer? put two of them out of phase reading the "rack" on your lathe cross slide it'll do the same thing..

OR? I did years ago a 35mm roll film encoder. It rolled up and across, back and forth. was black as pitch and blocked the light.

One chip in it, scratches the emulsion and it is dead. So.. that ideal is kinda dead unless you remote mounted it on the wall.


ThE rolling cdrom is the way to go. CUt slots in it, using .50 worth of emitter-detector pairs.. if you have to buy interuppter modules *emitter-detector pairs you might as well buy the clarostat 128 ppr rotary encoders since they are $50.


[This message has been edited by ibewgypsie (edited 09-07-2005).]

Mcgyver
09-08-2005, 06:04 AM
looks like there is more to these units than I thought. http://bbs.homeshopmachinist.net//frown.gif might have to, choke, buy them

David, my hope was for a high quality linear encoder instead of rotary - the rotary on the back of the servo is ok, but it really doesn't know where the thing is, just how many revolutions. If I were able to deploy linear encoders close loop, it opens new possibilities like controlling to less than a tenth for a wire edm, bolting the system onto any old machine regardless of leadscrew condition and other weird and wonderfull creations plus of course the regular dro function.

Anyone have a price list for them? In true industrial supply fashion, I found it hard to get pricing info

ibewgypsie
09-08-2005, 10:06 AM
Yeah. Lemme look at this.

Cable mount is the way to go. Remote the encoder away from the chips.

You do make chips don't you?

Using a sliding glass encoder is the way most companies done this. It uses a window, a mask and led emitter/detectors.

David

Mcgyver
09-08-2005, 10:29 AM
<font face="Verdana, Arial" size="2">Originally posted by ibewgypsie:
You do make chips don't you?
</font>

if you are talking about the metal kind, the answer is YES, backed up with pics! if you are talking about the silicon kind, the answer is not yet http://bbs.homeshopmachinist.net//biggrin.gif

Wirecutter
09-08-2005, 12:52 PM
I think I agree with Leigh. The Newall folks seem like they're explaining how these things work, but I doubt they would do that if they thought just anyone could do it.

It might be a neat experiment, though. My little "How to Run a Lathe" book talks about winding coils on a lathe, too. Hmmm.


<font face="Verdana, Arial" size="2">Originally posted by Peter Sanders:
Hi

There is a 2 part (I think) article in MEW Model Engineers Workshop, on using a computer mouse as such a device. It does seem to be remarkably accurate and is very simple to make. I don't have the issue numbers.</font>

Now a combination of that and what ibew is suggesting - that sounds easier.

[This message has been edited by Wirecutter (edited 09-08-2005).]

SGW
09-08-2005, 02:24 PM
On the Newall encoders, I imagine there are several pickups, reading several ball bearings at various spacing. It seems to me the rate of change in the signal that occurs when the pickup traverses across the point where two balls touch is going to be very different than the rate of change that occurs when the pickup is traversing the "fat" part of a ball.
By spacing out the pickups, one could select among them and use the pickup that was giving the best (however one might define "best") signal at that point in the travel. But sorting all that out might be pretty interesting.

Mcgyver
09-08-2005, 02:38 PM
<font face="Verdana, Arial" size="2"> The Newall folks seem like they're explaining how these things work, but I doubt they would do that if they thought just anyone could do it.
(edited 09-08-2005).][/B]</font>

you may be right, on the other hand, there is a big difference between what industry would find worthwhile doing for themselves and what a home shop type would, plus there are likely patents that would protect them in commerce


<font face="Verdana, Arial" size="2">I imagine there are several pickups, reading several ball bearings at various spacing. It seems to me the rate of change in the signal that occurs when the pickup traverses across the point where two balls touch is going to be very different than the rate of change that occurs when the pickup is traversing the "fat" part of a ball.
</font>

I think you are right, the change in voltage/linear distance at the appex is going to approach zero. must be reading more than one ball at a time, with each reader 90 degrees apart. when one reader hits a voltage delta limited, logic starts reading the second.


<font face="Verdana, Arial" size="2">Now a combination of that and what ibew is suggesting - that sounds easier</font>

maybe, but they're two different things, aren't they rotary encoders? the newall's are linear. if the newalls are impossible to make, a rotary solution is great, but if there was a way to make a linear encoder it can do a lot of things....

so you have a coil buzzing away and a second one picking it up, the delta in voltage between pulses is some curve relationship (sine?) to the distance. AD converter and based on the voltage delta, look up or calculate the linear equivilant. Not sure how to tell what direction its going though. granted the above is gross simplification


[This message has been edited by Mcgyver (edited 09-08-2005).]

ibewgypsie
09-08-2005, 02:56 PM
I like the ideal of .... CHEAP..

A cdrom, with slots cut in it. Remember it's not only high resolution you are after it is also pulse width. If it pulses too quickly as a small diameter encoder does (while BillG's software is out on a walk) it misses the count.

Not good.

BUT, the dos version is kicking butt. I may put the scope on the pins to see just how fast.

The clarostat 128ppr encoders are about .75 in diameter, kinda short pulse duration. I kept having to speed the loop timer up over and over and over to keep catching them. Having a larger diameter wheel is one way to solve this. You know the kind we can machine on our machines. OR a metal strip photo-sentisized adn etched holes? That'd work too.

Paul Alciatore
09-08-2005, 04:08 PM
Some key points I saw in their web site:

The balls are preloaded. This is likely a means of calibration by compressing the balls by the amount needed to reach a standard distance. So absolute diameter is not as important as having all the balls the same size.

They do say they use four pickups. But they also say they have additional items that are located without disturbing the spaceing of the balls. They are for determining the absolute position on power on. With only the balls, you will not get absolute positioning at power up.

Not on the web page, but my slant: The balls are perhaps 1/4" to 3/4" in diameter. It is going to be important for the ID of the tube to be very uniform. There will almost certainly be some allowance between the balls' OD and the tube's ID and this can be OK if it is not too much. But, the balls will stack slightly askew and this could throw the pickups off if it is too much. It would be better if the offsets are all in one plane and alternate (up/down or left/right), but how to do that? Perhaps a slightly oval tube?

The real problem is if the tube's ID varies. Then the balls will have different offsets and that will really throw the sensors off.

A final observation. They say that microprocessors are used. Well, a digital signal processor (a term probably suggested by the lawyers). I suspect this is to correct for a very non linear signal that comes off the sensors. Probably a table lookup thing that translates a set of sensor signals to a linear value. These tables may be individually created for each tube. These tables may also compensate for any offset in the balls' positions. Be prepaired for something like this. It is probably a big part of the expense.

In my opinion, it may be better to use a "rack" cut in a piece of round stock (drill rod?) and covered with a piece of tubing to keep chips out. The rack's teeth need not be involute form, rectangular slots would work just fine. Or how about just precision, threaded rod with a tube covering it? Any errors can be corrected in the "digital signal processor". Either of these would eliminate the offset problem you have with the balls. And it would seem that an Acme or square thread could provide a more linear output from the sensors.

Paul A.