[pstockley]

A lot of more modern CNC equipment seems to use direct drive AC servos with high encoder counts (2500 and up). I assume they are doing this to eliminate potential backlash and get fast rapids. It would seem like the downside is you loose mechanical advantage of gearing so need bigger more powerful servos.

For my lathe conversion I have the option of going direct or indirect and I have pretty big servos (750w for x and 1000w for z). I am really keen on getting the X axis as accurate as possible and with the minimum amount of backlash. Even .0002" backlash can result in a part being over/under size by nearly .0005".

Anyone got any opinions?

[unterhaus]

it's really easy to get oversized servos nowadays. Might be worth it. I'm pretty sure that when I'm done my mill will go way faster than I want being geared down 2:1

The only thing with my mill is that mounting the motors directly would be a big problem, so that option is out for me.

[NC Cams]

If you mount the encoder to the servo, you'll multiply your error by any slop in the drive.

However, if you mount enconder to ball screw and drive it via belt drive or such, you'll be positioning the screw, not screw via drive via motor.

Yes, you still have slop in screw and/or gibb but it is as direct as you can get.

A direct drive motor on the Y axis of a Bridgeport mill puts the motor smack dab against the family jewels. IT is bad enough dodging the fold away handle but now a motor too???

Consider the packaging - both yours and the machines as part of the direct vs indirect drive selection process.

We got into this situation with our cam measuring machine and moving the encoder to the spindle (corresponds to screw in CNC). By doing so, we TOTALLY got rid of postion error problems. And we use 3600 count/rev encoders and the machine does not tolerate hiccups while the axis is rotating while reading a cam.

Drive motor is little 12vdc Dayton gear motor with 150:1 gearing and another 4:1 between gear motor and spindle that rotates the cam. Lift encoder measures lobe lift to 0.0000001" and the rotary and linear have to jibe or else you get junk cam measurement data....

I'd contend the accuracy of the concept is "satisficient" (satisfactory and sufficient) for CNC work. If SNL can come up with "strategery" and turn it into an accpetable lexicon, I can give "satisficient" a shot at infamy.

[pstockley]

I did think about mounting the encoder on the ballscrew. However, I hear this is a bad idea if there is any appreciable backlash. In this case, the servos continuously hunt which can lead to premature ware of the brushes.

[NC Cams]

Backlash causes problems no matter what:

a. if encoder is screw mounted, motor brush issue may occure after long term use due to hunting (which will cause bad parts ni the mean time any) Our system was 10 years old and only screw and support bearings were causing problems. We redid motors anyway but this was redundant and preventative rather than remedial fix.

b. If encoder is motor mounted, motor will be in proper position BUT parts will be bogus due to "god knows where status" caused by table/screw being some place you THINK it is but really isn't. Been there, done that and found solution for it as it became INTOLERABLE.

Unless you direct drive screw with servo (not always possible/desireable like if you have servo in your crotch at Y axis if it is a mill), you have to take remedial action accordingly with positioning motor and/or encoder - life is filled with compromises.

I lived with backlash (screw/bearing/motor hunt induced) until I got tired of junk parts. Now that we have NO appreciable backlash (motor mounted encoder due to packaging constraint and super precise screws and bearings), we have parts that are deadly accurate and borderline jewelry.

My prior post re: mounting endoder direct to "screw" ala what we did with our cam measuring system shows that it can be done (driven member encoder mounting) to accuracy levels that are QUITE precise.

[Geof]

I did think about mounting the encoder on the ballscrew. However, I hear this is a bad idea if there is any appreciable backlash. In this case, the servos continuously hunt which can lead to premature ware of the brushes.

The encoder should be mounted to the motor because backlash between the motor and encoder can cause problems.

Regarding your question about direct drive and servo size I think it is instructive to look at what commercial machines have. The Haas GT20 has a 400 watt 1500 rpm max servo direct driving the X axis ball screw which has a lead of about 3/8". The Z axis servo is larger in physical size and is also direct drive but I cannot see the rating. The Haas TL1 has a 450 watt 1500 rpm servo on the Z axis. The rapids on the GT20 are 710 ipm and the maximum thrust is above 1000 lb for both axes.

I think your servoes should be plenty big enough for direct drive.

[pstockley]

Thanks everyone for the info. I am somewhat surprised at the size of the servos on the Haas. I thought that the Z would be bigger.

I will go direct drive. Packaging is actually not a problem as the x axis servo will be mounted at the back of the carriage. I am putting ground ballscrews with 5mm lead on X (class 3 precision) and Z 10mm lead (class 5 precision). I will also buy some decent bearings for the ballscrews. I don't like the idea of spending $800 per set for ballscrew bearings so I will look for a slightly lower end alternative. I don't mind spending $200 to $400 per set. I will have to do some research into this.

[Al_The_Man]

How its done with higher end systems including motion cards like Galil where the problem of motor hunting can be eliminated when using say linear scales on the final positioner and some backlash exists.
This is done by using an encoder on the motor as well as the scale or ball screw encoder on the final motion.
In this manner the PID tuning is separated between the two feedback items and is done in two parts.
The control obviously has to accept both feedback sources.
There is a comprehensive tutorial on the Galil web site on this issue.
Al.

[NC Cams]

Direct drive versus remote drive will always be debated and compromised.

Ditto that for encoder mounting (motor vs screw).

We went thru the issue in excruciating detail when we were trying to update/stabilize our OEM built Bridgeport Eztrak mill. Ditto that for our cam machine which we bought from a vendor and then redid as previously described.

We had 0.003" direction change errors in X and Y with the factory system (motor mounted encoder) for out BPT.

First we attacked ball screw bearing slop. ALthough OEM bearings were like new (preloaded 6204'S mounted DF), they didn't have adequate rigidiy. SO, we shimmed them to increase rigidigy. Got it to 0.001" but you couldn't hardly turn the handles.

So we then scored some super trick surplus ball screw bearings. These ended up with NO slop but this uncovered "looseness/backlash" in the screws themselves. These were checked and pronounced "like new, why do you want to screw with them??" Never the less, we had them preloaded with O/S balls - the guy looked at us like we were crazy but did it and gave us the "you bought it, its yours" sermon.

Still had slop BUt now under 0.001 and not repeating. So we then spent time adjusting the slop out of the gibs. Got unbelieveable 1;1 movement of table BUT now the servos' (which were tuned for OEM acceptable slop) were now doing wierd things to compensate for slop that wasn't there anymore.

At this point, we'd done all we could using our expertise and called in the calvary (Brian at BPT Machine in Carol Stream, IL) He tuned the servos and now we're milling parts to accuracies that approach that of CNC ground - on a Bridgeport mill. Even a who knew of our history with doing undoable things thought this one would be our demise. At this point and to his surprise, were still undefeated in the "they can't do that dept"...

Point is this: the generalities are true IF you use general wisdom and approaches in dealing and/or tolerating machine performance. If you systematically ELIMINATE the machine vagaries, you find that things don't quite work the same way. Machines respnod to "blueprinting" just like engine do.

Did we spend WAY too much money "fixing" our BPT??? Yes.

But the few thousand we spend is pale in comparison to what it would have taken to buy/lease/support the Haas tool room mill that would have had to be bought to match what our 12 year old, paid for Extrak can now do. Besides, the "how the hill did they do that???" is as priceless a compliment as one could ever imagine.

Humming birds and bumble bees can't fly but they do.

10 year old refurbed Bridgeports can't mill as good as ours does but even a factory trained expert was impressed with what he saw ours do. Speed and accuracy cost money. How fast do you want to go? Likewise, how accurate do you want it to be???

Systematic deviation elimination will create machines that perform in an amazing fashion...

In my opinion, the prior "...encoder should be mounted to the motor..."

should be rephrased to

"...encoder is typically mounted to the motor but it can be mounted to the driven elemment if drive slop and/or clearance is/are properly and systematically eliminated..."

We did it and it can be repeated as time and budgets permit... yes one can compromise - we chose NOT to.

[NC Cams]

Ball screw bearing alternatives

The high buck NSK ball screw bearings we used (20TAC47's) are designed for the thrust spoken of in post #6.

Less costly and lower thrust in 20TAC47 size should be 7204A5TYDUHP4 or PA7 mounted DF

A bit less costly and less thrust would be 7204CTYDUHP4 or PA7 mounted DF

Least costly but would need preloading by outside firm would be 7204BYG. Have them preloaded to 150lb DF by KAF Mfg in Stamford Ct.

YOu have the P/N's, you can now simply price shop

Keep in mind that we're dealing with a mill and the TAC bearings we used were for a huge VMC - like major overkill.

NOTE: change size as needed to basic part numbers listed - suffixes should stay the same. ALways check for availability when it comes to machine tool bearngs BEFORE you cut any metal...

[pstockley]

Thanks for the info NC that has saved me some legwork. I will do some shopping once I get the ballscrews.

[Ron22]

I agree with NC first try and remove as much mechanical backlash/slop in the system as possible.
Like Al said most higher end cards allow for dual feedback. Commonly an encoder on the motor for velocity control and a linear scale on the table for positional control. This will eliminate any backlash in the system because position is control by the scale by passing the ballscrew and gear box.