What's new
What's new

Gear Hobbing: Servo or Stepper motor

drcoelho

Stainless
Joined
Feb 19, 2017
Location
Los Altos
I'm early in developing a gear hobbing add-on to my Deckel manual mill along the lines of what Andy Pugh (YouTube and YouTube) and John Stevenson (Electronic hobber - test run) did but with more professional components. The approach is to have an encoder on the spindle, a PLC controller, a horizontal rotary axis with support and the work piece on an arbor, and quality gear hob. I'm hoping to create very high quality gears. My question goes like this:

- should I use a stepper motor or a servo motor on the rotary?

My inclination is to go with a very high quality servo motor such that there is absolutely no discrete movement in the axis, just really smooth movement. The thought is that the stepper motor might introduce discrete errors in the cutting process.

What are your thoughts on this?

P.S. I would be programming the PLC myself with tight closed-loop processing to keep the spindle and rotary motor in perfect sync, and would use very high sampling rates on the spindle encoder. The Deckel offers a lot of flexibility in terms of setting the helix angle, the ability to use horizontal spindle,etc
 
I would agree you want high stiffness and accuracy in the fixturing and control of motion, so I'd use a good servo and perhaps a brake (not just relying on the servo for stationary holding), but you've lost me a bit on using a manual, presumably box way machine for the cutting.

Wouldn't you have some clearance in the ways, therefore lose some control of position and motion due to the mill itself, no matter how good your added gear fixturing is?
 
I'll be using an FPS 300 M (Deckel Clone) brand new from germany, it's quite a stiff machine and I'm doing relatively small gears...it's quite an accurate machine and the horizontal spindle is SK40 tooling so quite stiff as well with overarm support. The Deckel has powered x and y and z, so as long as I'm sensitive to the backlash (e.g. drive it only one direction during cuts) it seems to me that I shouldn't have issues with the ways. I've also got a Brother Speedio, but it's unclear to me how I would wedge the gear hob into that machine. Not production use here, just small prototyping, but want the ability to create high quality gears on demand and not have to outsource.
 
Give it a shot, see how it works out. Do you have a high-power toolmaker's microscope? You'll want it (or even better, one of the metrology imaging camera setups) to confirm accuracy of the tooth form and uniformity of pitch.

I'd also try to get some of the real gearmakers here to give feedback, at best I'm just a tyro.
 
Servos are your best bet, gives you position feedback to the control.

I'd strongly suggest you look into Galil motion control cards galilmc.com They are all set up for electronic gearing.
 
Servos are your best bet, gives you position feedback to the control.

I'd strongly suggest you look into Galil motion control cards galilmc.com They are all set up for electronic gearing.

Just checked Galil out, very helpful lead, thanks!
 
closed loop feedback is normally better. that is a separate digital readout or scale device or encoder that sends actual position info back and if for any reason its out of programmed position it moves it back to position.
.
for example on a Prototrak mill in programmed position drill mode if cutting vibration moves position or even i force handle to move X axis out of position as soon as i let go it magnetically moves back to position. usually .0005 off it moves slow and .001 off it moves fast. only time i had a problem is position sensor using fine tooth wheel in contact with side of slide was jammed with a metal chip from turning. it read slide did not move as programmed and gave X full power til it jumped a inch and finally encoder wheel turned and it sensed movement. obviously it moving a 1" extra til it registered movement can break tooling and or scrap a part
.
but by sensing actual position and continually moving or applying force to programmed position if its not on position it can have its advantages when it is working correctly. i ended up replacing fine tooth encoder wheel with different sensor type that has ball bearings in a SS tube and sensor reads the ball bearings in the tube. it is a much more robust less sensitive to dirt or chips design. wasnt around 30 years ago. new technology thing
.
many cnc with a rotary table if it senses it moved out of position it at least stops with a alarm to let operator know it moved out of position. otherwise you might not realize it moved out of rotary position. backlash can be a problem and often even rotary tables have backlash. some horizontal cnc B axis rotary position at B0. is checked with a .0001" indicator everyday. checking calibration and ability to hold calibration. ideally it should maintain calibration for weeks, months, years, etc. how often you check ability to hold tolerance or calibration depends on what you are making of course and size of parts
.001 degree error about .001" per 57" error or .0001" per 5.7" part, if you got .002 degree backlash obviously that can be a problem
 
Hi drcoelho:
You wrote:
"I'm hoping to create very high quality gears."

How are you defining a very high quality gear?
Are your criteria set by performance expectations or by conformance to an AGMA standard?

Since you describe wanting gears for prototyping, I presume you want them to work, but you don't care so much that they can be certified to AGMA standards.

So what does an acceptable gear look like to you?
Will it transmit high torque?
At high speed?

Does it need to be particularly smooth to transmit motion?
Particularly accurate to transmit motion?

Does it need to last forever?
Under unforgiving conditions like poor lubrication or shock loads?

Obviously, an understanding of what you must have will inform what equipment you must build (or buy) to get to your goal.

This subject and similar subjects have come up on this forum quite often.
One stands out in my mind in which the goal was to prototype miniature gears for a small planetary gearbox and the hope was to make AGMA class 10 gears using a hobbing attachment on a Swiss lathe.
The knowledgeable gearmakers among us gave a whole bunch of very useful cautionary notes; I recall particularly that a custom gear maker named Zahnrad Kopf related his severe disappointment with the quality of gears cut with hobbing attachments on CNC lathes, and as I recall he had the measurement data to back him up too.

So your success will depend a lot on what you actually want.
I've successfully cut prototyping gears a bunch of different ways and they have met my goals for those prototypes, but the performance criteria for those drivetrains were pretty modest, and whenever I needed a GOOD gearset, I always bought or had custom made by gearmakers whose equipment and processes I could rely on.

Please realize I'm not trying to discourage you from enjoying this build and if it works and you can make good enough gears to prove your designs you have a winner.
But I caution you to be realistic about getting "very high quality" gears from a setup like you linked to.

With regard to your question about steppers or servos, sadly I have no expertise to offer.

Cheers

Marcus
Implant Mechanix • Design & Innovation > HOME
Vancouver Wire EDM -- Wire EDM Machining



 
Servos are your best bet, gives you position feedback to the control.

I'd strongly suggest you look into Galil motion control cards galilmc.com They are all set up for electronic gearing.

It was a long time ago and they certainly have made improvements, but my experience with Galil was dismal. I used one of their cards in a PC to replace the punch tape programmed system on my Boston Digital mill. The motion control card from them worked all right but the G&M code program they charged me $500 for was a joke. It turned out that it was a beta test version that was unuseable. I wound up working with the kid who programmed it who seemed to understand programming but knew nothing about machining. We had difficulties over seemingly simple things like absolute vs relative modes. He would say that it was sort of absolute, meaning that it was most of the time but not always. I had to get a mold made and he finally sent a program that I could use with a lot of workarounds and I got the job done, then bought an Aerotech card, which I am still using.

Interestingly, it took three tries to buy the Aerotech one. The order would go in, then evaporate. The salesman finally tracked it down. Aerotech's main business seemed to be selling complete to be selling complete systems, motors, driver amplifiers, and controllers, not CNC retrofits. Engineering said that the last time they sold one to a machine shop, the customer could not make it work and they had to send a crew to another city to settle a $2,700 sale. I managed to talk to an engineer and explained that my primary business is electronics and I have several patents on motor controls and servos, which satisfied them. When I installed the card and tested it, the mill would make the programmed move, then jump 1.8" at max speed. The glitch is still there; they had me change the parameters to move it into a decimal place beyond .0001" where it is never invoked. At least there I was able to talk with someone who knew what he was doing.

No doubt both Galil and Aerotech have moved far past the cards I got, so this may be moot. I still have the Galil card if you want it. It works fine, it is just the CNC program that sucks.

BTW, my 4th axis uses a Baldor servo motor connected to a gear by a timing belt and the encoder is on the motor so it depends on the belt for accuracy. It reads to .001 degree and claims .002 accuracy. It seems to work all right. Someone gave me the 4th axis because it was made for a particular Tree mill and no one knew how to drive it. I made a simple driver fed by the Aerotech card with good results.

Bill
 
I remember that thread, wonder what the results were. Frustrating when so often you don't know what worked and what didsn't...

Pretty sure he came back and said that basically Marcus and Zahnrad were right... I'll see if I can find it.
 
I've been using Galil cards since the very first ones that plugged into a PC buss.
They do not understand G-code. You have to buy software or write your own.
The sync function can be difficult to "rehit" or lock onto a fast moving axis.
Rigid tap on a mill is like falling off a log, multi pass threading on a lathe not so easy.
I really like them but am biased as I have a in-house written cnc side that I have over 50.000 lines of code invested in.
There is also another mode where you throw axis points at the card and it follows. Used for robots and handy for multi axis grinding.

Servos are (sometimes) better at smooth movements at very low speeds.
Steppers do not jump instantly from one to the next and micro drives less.
Welcome to the world of mechanical and electrical time constants. Get it right and life is good, get it wrong and you get flats.

Any gearing, belts or boxes introduces errors. Your infeed leadscrew will also have cyclical errors per turn.

The real world in these motions is not as easy as it seems on the computer.
Servos are "softer" they all have a lag which you need to match on the other axis working with them and all that tuning time.
Steppers are cheaper and may do the job.

Actually checking a gear or a set mesh is not a easy task if you don't have the rather expensive measuring machines.
Bob
 
I'm not going to use g-code, I'll be coding in native PLC language whatever that ends up being. Quite honestly, all I need is a robust PLC with good I/O that has fast processing. My plan is to pickup encoder position from the spindle, and pickup position from the servo, and implement a phase locked loop (my code) with averaging features for smoothness to keep the two locked in sync.

I just want my gears to be very high quality for my own purposes, which will be light load, light use, but very delicate operations and tolerances. I'm not selling these, for my own use, don't really care about certification, just end use results. The quality of my results will be driven I suspect mostly by the quality of the gear hobbers I use, which I'll buy commercially. I'll likely do deburring manually.
 
I was wrong. He made gears. He even said they measured good. But somebody (Marcus) asked if they actually worked, and he never came back with an answer.

Thread is here: Manufacturing of micro gears

This is a great thread to read. I'm not planning on making gears that small nor that accurate. My thoughts are along the lines of adapting a Deckel manual milling machine to be comparable in quality to an old school hobbing machine.
 
.... My plan is to pickup encoder position from the spindle, and pickup position from the servo, and implement a phase locked loop (my code) with averaging features for smoothness to keep the two locked in sync.

That all sounds so easy if you say it fast.
What could go wrong?
Any signal you get from the spindle has come and gone and you are late to the party.
Bo
 
That all sounds so easy if you say it fast.
What could go wrong?
Any signal you get from the spindle has come and gone and you are late to the party.
Bo

Hm. I'm no expert with gear hobbing but I do know a thing or two about programming, and given sufficient money applied to the problem it seems to me that I can find a fast enough PLC with some predictive software programming (my code) that should get the spindle and servo synchronized. After all, CNC machines do this all the time. Not saying it is trivial, just saying I'm pretty sure I can write the code.....

Also, I had already envisioned having the programming doing some averaging of spindle speed values such that when the spindle settles down to a more or less fixed speed, I would then get the rotary spinning in a perfectly matched speed. The idea to get away from the latencies associated with reading from spindle then feeding speed to the rotary...if the spindle is at constant speed, one could just tune the rotary servo to be the perfect matching speed.
 
What are you doing with gearing to the 4 axis? direct, timing belts, worm gear or ?

I'm thinking a servo powered rotary such as a Sankyo RCD, with a servomotor matched to a PLC from Allen Bradley,Yasaka,Mitsubishi....the nice thing about the Sankyo is that it has no backlash. The rotary/servo motor would be controlled by the PLC which is in turn getting samples from a high resolution encoder attached to the spindle.
 








 
Back
Top