[Al_The_Man]

Here is a usefull paper on the rule-of-thumb for tuning.
Al.

[keebler303]

Al

Do you tune using the step response (instantaneous error), or do you use the actual application velocity profile?

I've recently been fiddling around with an application which must accelerate as quickly as possible, maybe run at max speed for a fraction of a second, and then decelerate as quickly as possible. Throughout this profile position error must be kept extremely small. I think the current servo may be a bit undersize (trying to push the envelope a bit on the speed) so I may stick a bigger unit on it. The biggest issue I am having is the corners of the trapezoidal profile, where error is excessive. Unfortunately the motion profile is generated by an onboard controller so I cannot use S curve acceleration or other tactics to reduce the problem.

Thanks
Matt

[Al_The_Man]

Usually if you cannot attain the F.E. you want it is usually due to undersized motor/low gearing or too high motor-load inertia ratio at the acceleration rate attempted.
Industry standard is usually considered <10:1 motor load inertia ratio.
Any increase in accel/decel cranks this ratio up.
I only use analogue drives, so I don't get involved with step systems. But I find I can tune pretty close with the info on the PDF, I have tuning S/W also but usually there is always tweaking required manually.
Al.

[keebler303]

Thanks Al.

I'm quite sure a bigger motor will make things easier.

Matt

[kevin swan]

Thank You Al

[TomKerekes]

I'm quite sure a bigger motor will make things easier.

I couldn't resist making a comment here. I think in your case you are right but that isn't always the case. I work with high speed servos (laser scanning). We are always looking for methods to go faster and with higher accuracy through tight curves. It is common for people to ask "why just not put on a bigger motor?"

Here is a question I often ask: "When anything doubles in size how much harder is it to spin? (ie. 2 inch cube vs 1 inch cube)"

Rarely does anyone get it right, even engineers. Hint: it is a big number. Any guesses??

That paper is really great and it looks like it has seen a fair amount of use, Thanks Al.

Regards

[mike_Kilroy]

I couldn't resist making a comment here......It is common for people to ask "why just not put on a bigger motor?"

Here is a question I often ask: "When anything doubles in size how much harder is it to spin? (ie. 2 inch cube vs 1 inch cube)"

Can I try to get it right? bigger is better if you get more bang for the buck...

simple math tho:

T=Jw/t

so Torque (#-ft) required to change speed w (rad/sec) in t time (sec) with an inertia J (#-ft-sec2).....

no FM....

So if your bigger motor has too much additional J it will not help you accelerate to speed w in faster time t.

yes, it is often a viscous circle, bigger motor = more inertia, so no help in the end....

this is where us engineers go thru PROPER motor sizing to analyze TOTAL system inertia to design the system so the bigger motor with its bigger inertia DOES help,

[TomKerekes]

Hi Mike,

I agree with that! So what's your answer?

Regards

[mike_Kilroy]

answer? simple: do your homework, do motor sizing with load all facts in the equations and play what if; see if that bigger motor initially picked will improve your process or not or make it worse due to more J. If not improved like wanted, go search for a better motor or change your gearing so your reflected J is large percentage of the total, play what ifs. there are a zillion premade programs to help calculate this if you dont want to do it yourself, for instance, Motioneering (tm) avail fre from Kollmorgen.com

There are als a zillion motors out there with way different ratings. With Kollmorgen there are about 500,000 different choices designed so if a 'better' motor cannot be found, I'd say the person didn't look hard enough.

[TomKerekes]

Here is a question I often ask: "When anything doubles in size how much harder is it to spin? (ie. 2 inch cube vs 1 inch cube)"

Rarely does anyone get it right, even engineers. Hint: it is a big number. Any guesses??

I was looking for a number as an answer.

Nobody even willing to guess?


I like to see it as fundamental principles but this might help.



The bigger problem I run into with a bigger motor is that it tends to lower the first mechanical resonance of the system. Like in Al's Servo tuning technique we tune by cranking up gains until the system goes unstable then back off. The typical reason things go unstable is because we are approaching the system's mechanical resonances where things start becoming crazy and uncontrollable. So:

Bigger Motor=Lower mechanical resonances=
Lower Servo bandwidth=Errors not being corrected quickly

I liked this article:
THE MYTH OF INERTIA MATCHING

Regards

[mike_Kilroy]

J goes up by the fourth power on diam and goes up linearly with length. So if you double the size of cubic inches by making something 2x as long, j only goes up by 2, so it takes 2x the torque to move it (assuming no other j in the system like a load).

But if you make it 2x the cubic inches by making the diameter double, the j goes up by (new diam/old diam)^4 so 16x.

But if you don't mean double the size but mean 1 cubic in block goes to 2 cubic in block like your example (which is NOT 2x the size but rather 8 times the size) then j goes up by 32x.

So you see this is not really a far question for a specific number answer; that's why I gave it back to as T=jw/t.....

Also too much of a generalization to say 'if you double the motor size it will require 8 or 16 or 32x more torque for the same move, because very few systems are ONLY A MOTOR - most include a load that moves also. So if your motor is only 100th of the total system inertia, jw/t shows that doubling the motor size will have little effect on the overall required torque so will indeed produce close to 2x more acceleration. Conversely, if the load is 100 times smaller than the motor, doubling the motor will indeed hurt - unless you change other things like gearing to change this ratio since j changes by the square of the ratio yet speed only changes linearly..... lots to consider so I'd claim unfair question

[mdonovan]

heya Mike, would it be vicious of me to point out that your viscous circle needs to have it's viscosity defined so that we can calculate the additional torque required to compensate?

[mdonovan]

BTW you forgot your analogies about Volkswagens pulling semi trailers...

[Al_The_Man]

I liked this article:
THE MYTH OF INERTIA MATCHING

Regards

Tom, what is your take on motion calculation software such as a touted one like
Electromate Industrial Sales ??
Al.

[mdonovan]

Yes Al that is a very nice article especially in that it highlights the fact that motion system design is complex and difficult. I've been at it for many many years, as has Mike K (who I know personally and have worked with many times btw).

I will say in general that when customers are willing to throw big motors and even bigger drives at small problems our lives are almost always considerably easier!

[Al_The_Man]

I've been at it for many many years,.

Me too, I think as far back as Steam Radio :tired:
Al.

[TomKerekes]

Hi Mike,

I agree 100% with everything you explained. Sorry if I didn’t state the question clearly. 32X was the number I was looking for. I was trying to point out that I’ve always been amazed that as sizes get bigger things tend to slow down real quick (5th power). Where intuitively I would think putting a more powerful thing would help. And there isn’t really any way to get around it without going to exotic techniques.

Here are some Low Inertia Motors from Automation Direct that shows peak torque and moments of inertia:
http://www.automationdirect.com/stat...reservolow.pdf
100W Motor J=3e-6 Kgm^2 Peak Torque=1 Nm
1000W Motor J=260e-6 Kgm^2 Peak Torque=10 Nm
Note the bigger motor has almost 100X more inertia but only 10X more torque!

Check me on this scenario:
Assume we have a 3e-6 Kgm^2 load that we want to accelerate as fast as possible.
Initially we are using the 100W motor direct drive for the best possible acceleration with that motor.
100W Motor J=3e-6 Kgm^2 Peak Torque=1 Nm
Note ideal 1:1 load:motor ratio
Total Inertia at motor = 6e-6 Kgm^2
Max Accel at motor 1/6e-6 = 167,000 rad/sec
Max Accel at load = 167,000 rad/sec

Now change to big 1000W Motor J=260e-6 Kgm^2 Peak Torque=10 Nm
Note “bad” 1:87 load:motor inertia ratio
Total Inertia at motor = 263e-6 Kgm^2
Max Accel at motor 10/263e-6 = 38,000 rad/sec
Max Accel at load = 38,000 rad/sec
System accelerates 4.4X slower

Now load match the big motor by using gear ratio sqrt(87)=9.3
Reflected Inertia is now 260e-6 Kgm^2
Total Inertia at motor = 520e-6 Kgm^2
Max Accel at motor 10/520e-6 = 19,200 rad/sec
Max Accel at load = 179,000 rad/sec
System accelerates only 7% faster

The bigger motor allows crazy speeds (40,000 RPM at the load) , but doesn’t really help with acceleration. And more than likely kills servo bandwidth which is more a function of sqrt(stiffness/inertia).
Did I do that right?

Now if the original system had a much bigger load, and a high load:motor ratio to start with, then increasing the Motor size would help a lot with the acceleration. This is I think what Al is mainly referring to.

But I think it may still be likely to reduce servo bandwidth which might increase Following Errors. It depends on what is causing the following errors. Not enough available acceleration will cause large following errors so a bigger motor will help. But if the following errors are due to disturbances that need to be corrected by the servo loop, then the bigger motor may make things worse.

Regards

[mike_Kilroy]

BTW you forgot your analogies about Volkswagens pulling semi trailers...

M1 tank, not 18 wheeler haha! we could pull it thru quicksand to add the viscous damping you want glad to hear you are well and still enjoying the FL sun!

come on back to dayton to visit and we'll do another servo class at the community college again for ol time sake!

[keebler303]

M1 tank, not 18 wheeler haha! we could pull it thru quicksand to add the viscous damping you want glad to hear you are well and still enjoying the FL sun!

come on back to dayton to visit and we'll do another servo class at the community college again for ol time sake!

Mike

If a local servo class is up for offer, let me know. I know just enough to get myself in trouble.

Matt