How Can I Tell If Servos Are Tuned?

[lerman]

I'm using Reliance brushed servos and PicoSystems controllers on my Bridgeport Clone.

The system is tuned to the point where the following error is less than half a thou or so.

On some moves, the servos just run over to the selected value. On others, the servo sings or grunts along the way. Milling circles leads to particularly noisy operation. When the position is fixed, the servos seem to sing sometimes. Hitting the manual wheel and displacing it slightly causes the singing to change pitch or to stop occasionally.

Questions:
Is the singing, grunting, etc normal in a reasonably tuned servo? In all cases, the following error seems to be well within what I would consider normal tolerance.

If it is not normal, does anyone have any suggestions on how to better tune the servos. I've followed Jon Elson's procedure.

Thanks,

Ken

[bill south]

Hey Ken;
I'm by no means an expert, in fact, just getting into this cnc business. However, I have read just about everything out there over the past 3 months. Anyway, if you have not checked the geckodrives.com site, they have a bit of information on how to tune the drive system for their geckos. Check the manual for the 320 or 340 servo drives on their site. It may help. Good luck!
billyjack

[ger21]

I'd send Jon an email and ask him.

[lerman]

I'd send Jon an email and ask him.

I've sent so many emails to Jon that I hate to bother him again.

In a sense, this is a more generic question. Does a well tuned servo grunt, grown, and sing at times?

Ken

[Evodyne]

I've sent so many emails to Jon that I hate to bother him again.

In a sense, this is a more generic question. Does a well tuned servo grunt, grown, and sing at times?

Ken

Lerman,

I'll qualify myself as a non-expert, so take what I say in that context. From the things I've read, yes a well tuned servo will sing. In a PID (Proportional, Integral Derivative controller), the integral term accumulates error, no matter how miniscule. The output of the integral section of the controller will grow over time even with the controller in a fixed position at the setpoint. This output is part of the feedback circuit and will drive the motor so as to minimize or eliminate the error. But nothing is perfect (A to D resolution, overshoot, friction, etc.) and the process repeats itself after the move. So, in essence the singing is the motor oscillating around the setpoint in a futile attempt to land right on it. It's not considered something to worry about.

As for the grunting, here is my best guess. The nature of a particular move, dictated by the G-code, is the cause. Straight moves set the setpoint of the move at it's end and the controller sees a substantial proportional error and rockets the axis towards it. No grunting.

On curves or arcs the controller advances the axi (axises?) at different speed relative to each other to affect a curve. Constantly changing, small setpoint changes, lower motor speeds. Grunting.

Like I said, my best guess. I've no practical experiece with servos, but a fair amount with PID loops and controllers. Take it for what it's worth.

If your controlling to half a thousandth and your speed is O.K, your tuning can't be too bad. A loop that's tuned too "tight" will respond quickly, but will overshoot its mark and might, if really far off, oscillate. An loop that's too loose will not overshoot, but will be sluggish.

Evodyne

[ynneb]

My servos also sing and grunt to. At times when they have come to a complete stop there will be a bit of grunting. This is caused by the servo not been in the correct position by a small amount. Lets assume your servo has been told to move your machine from position zero to position 100. The servo will under full strenth do this, but lets assume there is resistance in the movement and instead of moving your machine to 100 it only gets to 99.9. The grunting will now start. The resistance is pushing your servo back to 99.9 and the servo is constantly trying to correct this. It becomes a constant push and pull effort. This is a sign of not gearing down enough, too much resistance in your slide system, badly tuned servos or a combination of all.

Is this a fault of servos? Definately not. The servo is saying to itself I want to do this job well and I want to go to that expected position and hold that position. The machine is saying, Oh no you dont, I will fight you that last poofteenth of a movement.
The servo corrects and turns off and the machine pushes back. The servo detects this and then corrects agian. This causes an ossilation. The sound of this ossilation is a grunt.

The best way to fix this problem is minimise lineal resistance or give the servo greater incremental control via gearing down. If for instsnce you had a worm reduction gearbox, the machine would not be able to push the servo back and there would be no tug of war. Unfortulately with most of our systems the machine can exert pressure back to the servo, thus causing the constant fight.

EDIT: You must note, that while this may sound bad, most of the time we are talking about one/ 2000th of a servo full rotation. After gearing down this translates to maybe a 50th, or less, of a millimetre of push and pull.

[thats_right2]

I've attached a portion of the manual for my servo drives. It may give you a little insight as to what may be happening to your system. I hope this helps you to tune and refine your system.

[webbch]

I'm a little uncertain about servo tuning as well. I envision tuning the PID parameters, most likely while the system is unloaded and not cutting a part. However, I would think those parameters would need to be re-tuned for each and every loading condition experienced in the process of cutting metal.

If this idea is correct, are there any kind of load-compensating algorithms to update PID parameters during a cut, or do you just set them to "work" within an envelope deemed acceptable by your "typical" operations?

Perhaps by definition, "load-compensating" would infer motor current and/or torque feedback?? I'm new to this and am reading many of the posts, but am not sure of an overall "block-diagram" approach to servo control.

Chad

[Evodyne]

I'm a little uncertain about servo tuning as well. I envision tuning the PID parameters, most likely while the system is unloaded and not cutting a part. However, I would think those parameters would need to be re-tuned for each and every loading condition experienced in the process of cutting metal.

If this idea is correct, are there any kind of load-compensating algorithms to update PID parameters during a cut, or do you just set them to "work" within an envelope deemed acceptable by your "typical" operations?

Perhaps by definition, "load-compensating" would infer motor current and/or torque feedback?? I'm new to this and am reading many of the posts, but am not sure of an overall "block-diagram" approach to servo control.

Chad

Thought I was the only one goofy enough to be up surfing at 2:30 AM. Chad, yes you would tune the machine unloaded. But your second thought about needing to retune for every loading condition is incorrect. Don't confuse machine design with machine performance. What do I mean? Well, the ability of you machine to compensate for load is dictated to a great degree by design factors. Motor size (adequate or to small), gearing (which is speed reduction and torque multiplication), and drive electronics/power supply/control software (yeah my machine is capable, but it needs a 48 volt supply and I'm using a 5 volt) can all complement or hurt performance. So design counts. But once that's all sorted out you need to make sure the machine does two things. First, it goes exactly where you tell it to without stopping short, missing the mark, or wandering continuously about your target. Second, it gets there in the minimum time. Tuning takes care of these two things. And, like you guessed, you are setting one envelope that all machine operations will perform under. In other words you are taking care of the worst case, so everything less than that is set be default. A PID loop thats not tuned agressively enough (low PID gain, mainly) will be slower than it can be. If tuned too agressively it can be in such a hurry to get to its target that it rushes by and has to come back-overshoot. If the gain is just right, but the integral term is too low, it may stop very close to the target but never quite zero in on it. Blah blah blah. you get the picture. So...tune it once...get it right...be happy.

Lance

Edit: Congratulations: you are the recipient of my 300th post. Whoopy!

[kong]

Just to confuse you even more, once you think you have your motors just right, put the speed all the way down to around 100mm/min and jog each axis around. If you don't get a smooth movement, just tweak the damping a fraction (usually more damping is required) until it is nice and smooth.
It is easy to tune the servos, but you don't notice the stuttering at slower speeds until you are on the finnishing pass of a piece that just took an hour to cut!

[webbch]

Thought I was the only one goofy enough to be up surfing at 2:30 AM

You were. I went to bed around 12:30-1:00, what's the matter with you?

Edit: Congratulations: you are the recipient of my 300th post. Whoopy!

....And it is in response to my very first post What a ka-winky-dink!

Just to confuse you even more, once you think you have your motors just right, put the speed all the way down to around 100mm/min and jog each axis around. If you don't get a smooth movement, just tweak the damping a fraction (usually more damping is required) until it is nice and smooth.
It is easy to tune the servos, but you don't notice the stuttering at slower speeds until you are on the finnishing pass of a piece that just took an hour to cut!

Thanks for the tip. Sounds like you learned that the hard way . Those lessons are so easy to remember aren't they, and you only make 'em once?

So PID is the way to go eh? My "schooling" was in basic control theory, but my job doesn't use it, so I've never really applied it. I'm interested if there are any other methods of control commonly used in a 3-axis CNC machine controller outside of PID. If so, I'd would be quite interested in the pros and cons of such a method.

Back to reading the older posts.....

Chad