PDA

View Full Version : Ball screw vs lead screw ... hype?



Farndurk
06-30-2010, 08:22 PM
Well ...... I know it isn't *hype* per se, however is ball type always needed?

I'm pretty new at this, but from what I can tell, ball screw is better for making 400 parts per day (less heat build up on the beds, hence better acuracy?) or if you're mounting heavy parts. I guess the drag of the lead screw type makes for less predictable slide actions and such that affect machining. Backlash is less as well. Got it.

But if you're only making 5 parts per day, no larger than a CD jewelcase and just as light, and the sequence times are around 4 minutes, not to mention backlash compensation in the software .... is ballscrew really needed? Keep the ways/rails/gibs lubed and clean, especially if they are already in excellent and *tuned* condition bla bla and it should be fine, que no?

It just seems like you could use a nicely made plain old ZX45 clone for much more than the *big guys* will tell you (Smithy et al). Of course, no *pro* would ever DARE use a leadscrew with a cnc rig. The shame of it all! ;)

Just curious.

Farmdirt.

wierdscience
06-30-2010, 08:30 PM
Leadscrews properly done are fine,anti-backlash nuts are the key.It is easier and a lot cheaper to get accurate leadscrews than it is ballscrews.IIRC the best I could find for anywhere reasonable money were .009" per foot accuracy,while the mid-range acme leadscrews were .003"per foot.

If you want to see what leadscrews on a CNC mill can do have a look at Evan's mill.Completely homebrew mill featuring Acme leadscrews running in plastic nuts if I recall.Does some very fine work.

dp
06-30-2010, 09:38 PM
Ball screws automate well, as in CNC. Either will work well with hand wheels.

ieezitin
06-30-2010, 10:02 PM
Apples and oranges. You donít have a case.

Friction, think about it the contact point of a ball verses a flat spiral flat. Enough said.

If you were into mass production, balls. Axial or angular balls are king. The next step is magnetism.

psomero
06-30-2010, 10:13 PM
What exactly do you mean by "leadscrew"? That is not a type of screw, but rather a use for such a shaft.

It could be normal 60 degree v threadas or acme... or whitworth? You need to be more specific.

Black_Moons
06-30-2010, 10:16 PM
Acme screws tend to wear, changing the accuracy and backlash over the length of the screw over time.
Ballscrews will wear much less and general allow climb cutting without issue, where as a loose mill might have problems with climb cutting.

beanbag
06-30-2010, 10:57 PM
yes, u need ballscrew for CNC (in general)

RobbieKnobbie
06-30-2010, 11:18 PM
Lead screw is an accepted term for an acme threaded shaft.

You're better off with ballscrews for a cnc application;

* Even though you have compensation in the software, the ammount of comp needed will change more often (relative term there) with a leadscrew
* a properly sized and lubricated ballscrew with preloaded nuts (they are also available without preload - mostly just for moving loads, not positioning though) will give almost infinite life and experience very little wear
* the efficiency of a high lead ballscrew permits the use of smaller drive motors to move larger loads.

For a good comparison between leadscrews and ballscrews try running through the calculations for each type on the Rockford Ballscrew website. That will give you a better feel for which suits your application. They manufacture both and probably don't care which you buy.

Obligatory disclaimer: I'm not connected with RBS except as a satisfied customer.

tumutbound
06-30-2010, 11:32 PM
Definitely ballscrews.
You can get reasonably priced ballscrews from China. These are C7 grade, rolled with less than .002" backlash.

Evan
07-01-2010, 12:51 AM
A lot of points made, some valid and some not.

An acme leadscrew has poor efficiency that ranges from about 90% maximum to as low as around 20 to 30%. The finer the pitch the lower the efficiency because the greater the friction in the nut. The advantages of acme screws are simplicity, low cost, ruggedness, long life and the big one for manual machines: They are self locking if the helix angle is less than about 8 degrees.

The self locking property means that it can't be back driven by machining forces. This is very important when operating a manual machine as it means you won't be fighting varying cutting loads to try and maintain smooth operation of the hand wheels.

Ball screws are not suitable for a manual machine for the reason that they can be back driven and will not stay in place unless they are locked by some sort of friction device. They are much more efficient than acme screws and because of that they allow for smaller motors to drive larger loads in CNC applications. They usually are designed to eliminate backlash which is necessary in CNC machines. Backlash compensation is not an answer for CNC machines as it cannot be used in most operations that use interpolation.

Backlash compensation is relatively easy to implement in acme screw systems by using acetal nuts. The best are made using PTFE filled acetal which has very low friction and virtually no wear. They are designed to be a slight interference fit which also eliminates the requirement for wipers as there is no room for swarf to enter. I have been running an acetal nut on my Y axis on my mill for a couple of years now and it still shows nearly zero backlash.

dp
07-01-2010, 01:01 AM
The self locking property means that it can't be back driven by machining forces. This is very important when operating a manual machine as it means you won't be fighting varying cutting loads to try and maintain smooth operation of the hand wheels.

Ball screws are not suitable for a manual machine for the reason that they can be back driven and will not stay in place unless they are locked by some sort of friction device.

On my little mill there is zero chance of of depending on the lead screw providing much locking. I have the bent end mills as proof! :) I actually allow my gibs to drag some, so my lead screw efficiency is quite low, but it does allow me to make sharp corners. I always lock the unused axis. This mill is hopelessly incapable of even mild climb milling.

Again, a small mill, but I did have the opportunity to use ball screws manually on a machine and it was a pleasure to use. Can't say what it may be like on a larger system. So I'd have to say there's no hard fast rule.

Evan
07-01-2010, 01:47 AM
How well it locks depends on the helix angle of the screw.

Peter.
07-01-2010, 02:03 AM
Backlash compensation is relatively easy to implement in acme screw systems by using acetal nuts. The best are made using PTFE filled acetal which has very low friction and virtually no wear. They are designed to be a slight interference fit which also eliminates the requirement for wipers as there is no room for swarf to enter.

This is very interesting to me Evan because I have a lathe that is metric but converted from an imperial design, and the maker had the great idea of using a 1/2" diameter cross-slide screw with a metric pitch. The pitch is 5mm, which is VERY close to 5tpi. I've been thinking of making an acetal 1/2" x 5tpi nut for it, thinking it might work and serve effectively as anti-backlash too since the pitch is about .004" per turn different. Might try it now.

John Stevenson
07-01-2010, 03:13 AM
Reading the OP original post and addding it to a previous one he's trying to compare the Microproto CNC mill which is a Taig mill underneath with a ballscrew machine or just ball screws.

The Tail uses standard 1/2" x 20 UNF thread as it's leadscrews, not threaded rod, nicely made screws but still vee thread at the end of the day.

They also uses a bronze split nut for backlash adjustment.

I have one of these machines but unfortunately it hasn't done enough work yet to become worn and give a good idea of how servicable it is for engraving over a long period of time.

My concern would be is that the OP wants to do a lot of fine engraving and when backlash does occur how badly will it affect the work.

As Evan has said you can double up with plastic nuts to get a better fit but the use of 1/2" UNF is questionable for an application like this.

.

Farbmeister
07-01-2010, 06:36 AM
Once you put DROs on the axis the issue of acme/ball screw 'accuracy' is moot.

Set it up right and the DROs read position of the table, and really doesn't care how many turns it takes you to get there.

I never understood why servo's encoded off the shaft (well, I know why, but its a poor trade off). If you get your absolute position off the motor shaft, you still have ball screw play and way play. If you encode position of the table who cares about lash... the TABLE is where it should be.

My manual machine has DRO. Its a crutch, but the scales on a hand wheel suck.

John Stevenson
07-01-2010, 07:01 AM
Apples and oranges, we are taking CNC here and with ballscrews you should have no or at least the minimum of backlash present.

Evan
07-01-2010, 08:13 AM
I never understood why servo's encoded off the shaft (well, I know why, but its a poor trade off). If you get your absolute position off the motor shaft, you still have ball screw play and way play. If you encode position of the table who cares about lash... the TABLE is where it should be.


With servos the encoders aren't just about position. They are also how acceleration and deceleration are computed. If the motor is commanded to turn and the encoder doesn't move because it is encoding the table position with lash present then the controller is immediately into it's error band. It will try to increase the power to get things moving since the built in assumption is that the load is very heavy. Lash and back lash must be removed from the system for proper operation with any type of screw. The very high end systems don't use screws at all but instead use linear motors that directly move the load.

tdkkart
07-01-2010, 09:26 AM
Once you put DROs on the axis the issue of acme/ball screw 'accuracy' is moot. Set it up right and the DROs read position of the table, and really doesn't care how many turns it takes you to get there.


It all sounds good, but it doesn't quite work that way. I see it on circuit board drills and mills all the time.
The servo heads to position "X" but when it comes to a stop it lands at X -.001", goes back the other way and lands at X +.001", and so it goes from there. Ends up hunting back and forth and not being able to find what it wants. Ends up just sitting there looking stupid, and then a tiny amount of pressure in the direction of the loose axis and it takes off running again. You'd think it could be programmed out, but apparently not.
A worn screw will make it do it in one area of the travel, a loose thrust bearing will make it do it randomly anywhere in the travel. When it gets bad enough it soon starts causing mislocated holes and/or chatter on the board cuts.

As said, magnetic linear drive is the cat's ass. We've got one machine with it, more coming as soon as the economy picks up a bit.

lazlo
07-01-2010, 10:40 AM
I never understood why servo's encoded off the shaft (well, I know why, but its a poor trade off).

The encoder feedback is used to electronically commutate the motor, so a servo motor has vastly better performance than the same motor running open-loop, especially with changing loads (quick changes of directions with varying forces).

That's why a common induction motor driven by a "sensorless vector" VFD has better performance than an open-loop V/Hz VFD (with the same motor). Put an encoder on that induction motor and you have a full-blown induction servo with even better performance. Many of the high-end CNC machines have induction servos for the spindle.

Farbmeister
07-01-2010, 12:35 PM
When I said I 'don't understand why they encode off the shaft' I meant that its a poor choice.

First off, 'hunting' is a design issue. No good CNC setup is going to hunt. Why? A good setup will be able to calculate acceleration (or Deltas) off the DRO movement.

Most people here evidently are referring to home built or mini-mills with conversions.

Why would you try and derive position of a tool from a motor shaft when you can get EXACT position regardless of wear, lash or any other loss by encoding the table directly?

My mill has a good 1/2 turn lash on XX.. but I don't care because unless the table moves, the DRO wont.

Its that simple.

lazlo
07-01-2010, 12:37 PM
Why would you try and derive position of a tool from a motor shaft when you can get EXACT position regardless of wear, lash or any other loss by encoding the table directly?

Because you still want the high-performance associated with closed-loop motor commutation.

By the way, there are CNC controllers that have dual PID loops: one for the motor encoder, and a secondary PID driven by linear scales on the table. The idea is that the secondary closed-loop will deal with machine flex, wear, ballscrew inaccuracy, while the primary loop squeezes maximal performance out of the axis motors.

But in most cases (high-end VMC's), closed-loop control on the axis motors is more than enough for tenth's accuracy/repeatability.

The CNC Brain project was an attempt to bring dual PID loop control to hobby CNC, but the Devil's in the Details...

Farbmeister
07-01-2010, 01:30 PM
'Closed Loop' simply means a feedback signal used to adjust the primary one. The motor is still a bad location.

Stepper motors can be 'closed loop', they are not the realm of servos, or high end CNC.

And you actually proved my point... the CNC machines with TWO sensors... one on the bed to eliminate the inherit issues of the motors/leadscrews/flex etc.

Think of how much you can save on a CNC mill if you eliminated precision ground ball screws and nuts. DROs are pretty simple in relation to them.

lazlo
07-01-2010, 01:36 PM
Stepper motors can be 'closed loop', they are not the realm of servos, or high end CNC.

Closed-loop stepper controllers are rare, but agreed, you can get similar improvements in a stepper motor performance with closed-loop commutation.

This is Mariss' description of the "Stepper Servo" project he's working (basically, building the mythical closed-loop stepper motor controller):

Stepper Servo (http://www.cnczone.com/forums/showthread.php?t=38763&highlight=Stepper+Servo)

Hi,

This is a progress report on new developments in the "unstallable stepper project".

The goal of the "unstallable stepper project" is very simple. Imagine you have built / bought a 4-foot by 8-foot gantry-style router. You are doing 3D (x,y,z) routing with it when your drunken brother-in-law is over and gets the idea to see what the ride is like if he were sitting on the gantry. The result would be ordinary steppers or servos would immediately stall / fault as soon as the sobriety-challanged one clambered aboard.

Now imagine the motors simply slowed down or even stopped due to the "idiot aboard" overload but they never leave the 3D path in progress. Dislodge the load and the motors pickup to their original speeds and finish the work as if nothing happened. Not a gouge or mark on the finished work afterwards.

So how to get from here to there? A lot of parts that all have to come together is how.

Part 1A, Stepper Servo: You see BLDC [Brushless DC] servos, AC servos, PM DC brush-motor servos. You don't see step motor servos (Vexta Alpha Step and the like, step drives with monitoring encoders don't count because they are half-assed solutions). Why? Because technically it is very hard to servo-tame a step motor. Very, very hard in fact.

Begs the question; why bother what with all the other servo choices? The answer is steppers have a unique speed-torque curve that makes them perfect for 2-mode applications. Lots of torque at low speed work feed-rates and only enough torque for high-speed rapids. Nothing is wasted.

Part 1B, Stepper Servo: Open-loop step motor systems have to be seriously derated. They are running open-loop after all. Closed-loop, a mild-mannered Clark Kent type motor becomes Superman. It literally jumps on the bench from acceleration reaction torque on its way to 15,000 RPM. This from a NEMA-23 2A motor with a 24VDC power supply. It takes 0.03 seconds to go from a standstill to 3,000 RPM. 0.42 seconds later and it's at 15,000 RPM.

The biggest difference? It sets its own optimal rate of acceleration, it cannot resonate, it always gives 100% but cannot cross the line and stall while trying to give 101%. That's built-in.

Mariss

BobWarfield
07-01-2010, 01:56 PM
Think of how much you can save on a CNC mill if you eliminated precision ground ball screws and nuts. DROs are pretty simple in relation to them.

Sounds great, but doesn't work very well in practice for a whole ton of reasons:

- Backlash comp is a hack. All systems have backlash, and the best controllers will apply it even to the small amounts present on a ballscrew. Nevertheless, appreciable amounts of backlash are a bad idea for CNC. It you down compensating for it, creates dwell points on the workpiece that are undesirable and so on. Lots of very common operations on CNC are very unhappy about backlash--interpolating a hole is just one example.

- The better controls often have an "exact position" mode that facilitates approaching the hole from one direction the way a machinist would to take out the backlash.

- CNC'ers love to climb mill because it gives a better finish, reduces tool deflection, yada, yada. That's a bad idea with too much backlash and backlash comp does absolutely nothing to correct the problems of climb milling with backlash. A CNC that can't reliably climb mill is not a happy thing.

- Modern CNC's want to move the axes FAST. The lower friction of the ballscrew is part and parcel of that.

- The PID controllers for servos have to be pretty sophisticated to deal with two feedback sources--encoders + glass scales. That adds cost. The biggest reason to add the scales is not to replace the encoders, but to provide real time feedback for the effects of wear, temperature change, and so forth on what's really happening with the machine. BTW, if you're worried about ballscrew errors, you can attach DRO scales today and map your screws and Mach3 will compensate very nicely for the errors. This is a good use of DRO compensation because unlike backlash, it doesn't have the downsides for climb milling and so forth.

- As an aside, most of the simpler servo drives don't do the commutation Lazlo refers to. They can have the encoder remotely mounted, but there are other problems. Any slip or give between say a timing belt drive from the servo to the ballscrew really messes with the PID tuning of the simple controllers. The dither is much worse there and it becomes harder to tune for a "stiff" axis.

You can reduce the backlash of ACME or v-thread screws as Evan suggests with dual preloaded plastic nuts. But, that mostly works well for smaller machines. As you go up in size, you'll be wishing for ball screws. There's a limit to what forces the plastic nuts can deal with before they deform. I'd be tempted to go there for a project that had light mass and long travels--perhaps a router table of some kind. They're also great for smaller machines.

For a mill of any size, the ballscrews are worth it. Let's say an X2-sized mill is the 50/50 camp. Meaning, bigger mills definitely want ballscrews, smaller mills can be very happy with the plastic nuts, and the X2 you flip a coin.

Given all the goodness they can bring, ballscrews are pretty cheap. Rockfords and Rotons can be had for relatively little, particularly if you're prepared to do the little bit of work to make the dual ballnuts instead of buying theirs. The imports are all over eBay and very cheap too.

Cheers,

BW

lazlo
07-01-2010, 02:26 PM
As an aside, most of the simpler servo drives don't do the commutation Lazlo refers to. They can have the encoder remotely mounted, but there are other problems. Any slip or give between say a timing belt drive from the servo to the ballscrew really messes with the PID tuning of the simple controllers. The dither is much worse there and it becomes harder to tune for a "stiff" axis.

Good point. I was referring to the Yaskawa/Mitsubishi-class servo controllers that you get on a Mazak, Okuma, Mori-Seki...

Evan
07-01-2010, 02:31 PM
The steppers on my mill have built in quadrature encoders. They aren't hooked up since my controllers don't support them but obviously encoded steppers do exist. You can also buy servo drives that are directly compatible with step and direction control signals. The driver will run just about any type of DC motor that is within it's capability and has an encoder attached.

Separating the encoder from the motor shaft introduces non linearity to the feedback loop. There is more than just looseness/slop that must be accounted for. The components in the loop are flexible and will twist under load. They act like springs (torque rods) and that will cause the feedback loop to be much less stable. A feedback loop must be in the negative feedback range to be stable. If there is any lash or twist between the drive and the position sensor it will introduce a phase shift that can push the loop into positive feedback. When that happens the system goes into oscillation and that shows up as hunting. To deal with that the gain of the system must be reduced and that results in greater position error because of increased deadband.

Mounting the encoder to the motor shaft allows maximum possible gain to be applied for maximum position accuracy. The rest of the system is outside the feedback loop so changes to it such as changing the amount of load on the table do not alter the system feedback phase. If table position is used as the encoder source every time anything is changed about the setup the phase shift of the feedback loop also changes, making for a much greater possibility of an unstable system.

John Stevenson
07-01-2010, 02:44 PM
Placing the encoder on the motor is also safer.
If the encoder is on the screw and the helical beam coupling [ that stupid thing with loads of slots in it to weaken it ? ] breaks, the motor won't know where it is and just keeps going looking for a signal.

And on a big laser capable of 37,000 miles per hour it gets exciting, sticky and messy, but not necessarily in that order, fast.

Don't ask, but the repair bill came to more than four pounds seventeen and sixpence.

.

fasto
07-01-2010, 02:55 PM
Good point. I was referring to the Yaskawa/Mitsubishi-class servo controllers that you get on a Mazak, Okuma, Mori-Seki...
I'd add Fanuc to the list - my machine's Fanuc servo drives use an encoder mounted to the motor for commutation as well as position feedback to the CNC.

Farndurk
07-01-2010, 05:38 PM
Wow .. this stuff is like guns. There are about 100 *right* ways to do about any of it. Comes down to the same issue no matter if we're talking Hot Rods, Guns, or machine tools.....

How fa$t do you want to go?

Even's logic re: the acme threads and the angle providing axis resistance makes total sense to me. Some of what has been said here seems like "1 year of experience done 20 times over" instead of 20 years of experience. Reminds me of old timers arguing their case for the 1911 Gub'ment model. Lots of old preconceived notions, zero acceptance for anything not of their world. On the other hand, some of the info here seems more field tested or practical too.

I'm always leary anytime someone says "this is the ONLY way it can be done *properly*" ... or .. the worst of all "that's just common sense" (or the sister phrase of "everybody knows that"). I'm a bit too agnostic to think there is *only* one way to do this or that.

Thanks .. I am very interested in the plastic nut/acme thread stuff. As well as DIY bedrails scraping. Evan .. you've obviously done some out-of-the-box thinking ... you da shiznot, bro. (forgive the pedestrian use of the word "plastic" ... however it is technically correct for those who care about such things).

All proves my point about .. well ... whatever ... and that is *there are 100 right ways to do nearly everything*. Just depends on what is *right* for you.

And yes .. this is all part of me stuck in a decision process that has some great urgency. I have over two dozen pedals that need handstamping that I cannot do anylonger due to an elbow issue. So we are frantically trying to locate a machine/process to get me "off the hammer". I really need to get these devices finished and shipped. Orders come in every day, and the longer I fek around with this project the bigger the pile gets.

So thank guys. I'm nearly there in the decision. The MicroProto seems like it's gonna save our day .... I hope. $3k and free shipping! Such a deal. (geez)

My only other option is a non-electric manual stamping device that indexes over one space after each *stroke* of the stamping lever. $2800. But that is all it will ever do is stamp lettering.

sigh ..... (dang it! this is wearing on me.....)

Brian.

Mcgyver
07-01-2010, 05:59 PM
Sou
- Modern CNC's want to move the axes FAST. The lower friction of the ballscrew is part and parcel of that.


he said you'd do away with precision ground screws using a linear scale for the feedback loop.....he could still be using ball screws just not the expensive ground ones.

I think Farn has a point, linear encoders are more expensive than say Newall scales, but a set up with the accuracy of the newall scales would be a very reliable and accurate system....iirc the the lead accuracy of the newalls is better than a ground ball screw and ground screws are very expensive...but its a likely an academic thing or a solution searching of a problem for a home shop cnc....after all we are talking milling so no one's expecting 10ths accuracy over say a foot.

As a general comment on reading the thread, leadscrews drive a lathe carriage up and down the bed....feedscrews are what you mean for when you're talking mills, crossfeeds etc

Evan
07-01-2010, 06:14 PM
Here is everything you ever wanted to know about leadscrews.

http://www.nookindustries.com/acme/acmeglossary.cfm

sconisbee
07-01-2010, 06:19 PM
Well for what its worth, my CNC lathe has ballscrews and glass scales and is pretty damned accurate, just a little slow as its showing its age a bit. but it seems it has the best of both worlds, although rotary encoders would make any future control upgrades alot easier.

wierdscience
07-01-2010, 08:52 PM
Back to basics,it all depends on what you want.

How often you intend to use it,how fast you intend to push it etc.

If all you want is a bench mill with CNC capability and your happy with +/-.001" and less than 200 inches per minute feed,leadscrews are the answer.

Anything bigger or faster and you need ballcrews.

You could buy a Taiwanese B-port clone new,then buy one of these-
http://www.rockfordballscrew.com/ look under clone kits.

Then buy your motors,software,power supply,do all the wiring etc......

Or you could just buy one of these-

http://www.southwesternindustries.com/swi/prod_kneemill.shtml

rowbare
07-02-2010, 03:45 PM
The Tail uses standard 1/2" x 20 UNF thread as it's leadscrews, not threaded rod, nicely made screws but still vee thread at the end of the day.

They also uses a bronze split nut for backlash adjustment.

I have one of these machines but unfortunately it hasn't done enough work yet to become worn and give a good idea of how servicable it is for engraving over a long period of time.

My concern would be is that the OP wants to do a lot of fine engraving and when backlash does occur how badly will it affect the work.


The backlash can be adjusted out. The trick with something like the Taig is to set a schedule of sorts and tune the machine every so many parts or at least test to see if it should be tuned. Doing the backlash adjustment isn't a long process and replacement backlash nuts and lead screws can be had for about $45 an axis. At that price, you can pretty much treat them as consumables. I imagine he will be able to do quite a few parts between tune ups.

bob