OpenServo.com

[Zeddicus]

How on EARTH does your servo turn the full 180 degrees in 2.2s, with only a 4V supply and a motor winding resistance of 4.3 ohms, considering that you have a reduction of 720 ???? That means a motor speed of over 9800 rpm's, with the weights attached!!!!

I'm having a very hard time trying to find sane parameters that fit this!!!

Do all typical servos have these values, or am I the ignorant one? I still haven't received mine, so I don't really know.

ZZZ

[ginge]

180 degrees = 360 revolutions, and it's 5v

[Zeddicus]

Supply is 5V, but you told me you have the output limited to 80% (204 for the pwm)? This would mean the maximum applied voltage to the motor is 4v.... right? Or did I misunderstand?

At 4V, for a 4.3 ohm winding resistance, you'd get a stall current of 0.93A. If the motor turns 360 times in 2.2s, you get about 9800 rpm's, which is a lot (I think).

I have to get some work done now, but I'll play around with this again later.

Thanks for the info!

Z.

[ginge]

yeah, you got me there 4v it is.

I'm pretty sure my reading are correct... unless I did it wrong.

Something I have noticed, my servos drive the motor using a very small cog, which drives larger ones. This means the motor will have to spin more to move the horn. Surely this doesn't put 9800 rpm out of the question?

You may need to get someone with a standard servo to get some readings for you.

As for what a ""standard servo"" look like, your guess is as good as mine. Although I have some knocking around, they are doing useful things, like moving my security cameras. Maybe I will pull one down and butcher it.

Cheers,

Barry

[Zeddicus]

Barry,
I've played around a lot with my model parameters. I managed to nail another one down, and I got a ratio for another 3. The only one I'm really missing is the motor winding inductance, but I don't think that is critical.

In general, using the parameters that I either deduced or got from you, the model behaves quite well. So much so, that I managed to match your motor current and full speed, as well as the 2.2s movement time for the 180 degrees.

During all this, I had to fool around with the PID controller, and I observed a behaviour that is very similar to what I see in your curves: the pwm output changes wildly for a short time, and some current oscillation happens when reaching the end of movement. I could be wrong, but it looks to me as though your PID is slightly out of tune. According to what I see, your KI should be lowered, though I can't really tell if a little or a lot. Maybe your Kp should be lowered a bit too. At least this is what I did in my model to ease those symptoms quite a bit.

This can also account for what you told me about the servo doing Bad Things when you go to full PWM output instead of 80%.

The parameters I have at the moment are:
Kp=1
Ki=30
Kd=0.1

This will not be directly useful to you, because, unlike the current openservo controller implementation, these parameters are independent of the sampling time.

So this brings up 2 questions:
1.- Are you using PID or IPD?
2.- What controller values are you using?

I can try to check your parameters against my equations, and try to remove the sampling time factor from them.

I also reached one more conclusion: it is advisable to filter the PWM output in some way. It does not make any sense to have the output varying so wildly. A filter of maybe 1Hz-10Hz should be ok (I think). This is similar to what I mentioned at some point about inserting a rate limiter at the output of the controller. A wildly varying output only destabilizes the control loop by making the current change unecessarily, and this will happen even with the motor behaving like a lowpass filter for the applied voltage, because it is not an ideal system.

Well, I hope all this is helpful in some way...

Cheers,
ZZZZZZzzzzZZZZZzzzzzz (almost 3am..... again.....)

[ginge]

Nice.

Regarding my PID being slightly out of tune, you are almost certainly right. They were tuned with a diffrent load, and I didn't see that much oscillation when I changed over.

I will fish out my PID values when I'm in from work.

The PWM strangeness I refer to is more to do with the motor rotating far too fast, and eating all of the current in the PSU, and causing a brownout. 80% pwm was a good tradeoff for that.

Barry

[Guest]

yes, I forgot to mention that your controller was likely tuned either for a different load or a different working point. It works fine as it is though, so the tuning was well done, and if it ain't broken....

I'm going to try and make the model controller more realistic now, by addind the pwm output, and the ADC input. I just have one tiiiiiny problem: I wrote everything in matlab, and it takes about 30 seconds for each simulation, which sucks when you're trying to tweak parameters....

Maybe I should port the whole thing to C/C++.... I could include an opengl simulation of the servo to help visualize what it's doing... and the executable would be available for upload to the wiki..... ummmm.....

ZZZ

[Zeddicus]

stupid cookie double login problem, posted as guest again hate that...

Z

[ginge]

Z, sounds good,

I was going to convert your matlab source to a C++ class myself, but if you are volunteering.

I was hoping to tweak you estimator models to use on the other servos on my arm, It's always good to have a solid model to work with. GL too? You spoil us.

Oh yeah, my PID values:-
Kp = 500
Ki = 4
Kd = 800

Baz

[Zeddicus]

it's pretty much all I can do for the moment: play around with equations and algorithms. I already received my electronic components, and the servos are waiting for me at customs, but I still don't have the pcb's, and in any case, I'll just get home on sunday, after which I have to go on a trip overseas untill mid july, so... I'm just about purple from the frustration

I'll develop the thing under Borland Builder... unless there is another preference?

Cheers,
Z

[ginge]

I get the impression the windows guys use MS Visual Studio (.NET). I am a linux kinda guy, so use gcc tools.

But, you know, whatever gets the job done for you. nothing too proprietary would be wise

Barry

[ginge]

Hey Z, hows the model shaping up?

I'm looking forward to seeing your data.

Barry

[Zeddicus]

hi Barry,

I'm still in Montreal until tomorrow sunday, so I haven't been able to work too much on it. I'm developing the application using CLX, which should (emphasis on the should) compile on Linux as well, though I have little idea how. I have the GUI set up, and I have the motor equations encapsulated as a class. I also have the PID/IPD and other variants encapsulated as a classes.
Since there are so many parameters to handle, I decided to use a configuration file. And since this is the umptieth time I've had to use a configuration file, I got fed up with re-writing the code, so I made a generalized class for that too.
I'm still missing the simulation mechanics as such, and the openGL routines, though that last part is easy.

Also, I'm not sure about leaving everything as classes. Can the AVR compiler handle C++? I think I asked that before somewhere, but I don't remember the answer I got. Sorry about that...

In any case, I'll work on it some on the plane back to Chile, and then I'll have something like 6 days with the afternoons off, so I'll have more time then. I'll let you know as soon as I get it working.

Cheers,
ZZZ

I missed Montreal... lots of memories here for me.....

[mpthompson]

Zedd,

I pulled the following regarding C++ development on the AVR from the WinAVR documentation:

[Zeddicus]

Um... maybe I'll just implement the calls with Ansi C by using structs... I hate not being able to trust the compiler.

Thanks Mike.

Back to drawing board for me...
ZZZ