4 Attachment(s)
Tuning Servos trying to determine what info is relevant
I think I'm on the right track. This is the highest I've managed to get my p (2.4) value and when I make a move of 4000 the graph looks good. When I make a move of up to 40,000 it looks like the position lags substantially. When the move is 400 it looks like the position overshoots. I have to graph for 3 seconds before the output line settles down. I don't know how to evaluate the quality of information I get from these graphs. I either look as though i'm way off or doing well based on the size of the move.
I am hoping someone here could give me some insight about what info to let guide my next move...
Re: Tuning Servos trying to determine what info is relevant
Hi gbman7a,
It seems you understand the basic idea is to get the red plot to closely follow the blue plot. When well tuned you may not be able to see any difference so plot error instead.
The green plot is what is being commanded to the amplifier. Note your green plots are saturating (flat top). Longer moves are worse as you try to go faster. Please read this post/thread and let us know how much you understand.
Please don't take photos of your screens. Instead use Alt-Print Screen, paste to Paint.exe, crop, and save to *.png file. Always include the IIR Filters and Configuration Screens so we know all your parameters.
It would also be helpful to know what kind of machine this is and what the encoder resolution is so we can relate to real speeds and distances.
2 Attachment(s)
Re: Tuning Servos trying to determine what info is relevant
Hi Tom,
Thank you for the quick response. The machine is a 1992 MotionMaster 3 axis router. It has Glentek GA370 servo amplifiers. The servos are Baldor MTE 4090 BLBCE. The machine has Tachs running back to the servo amps. The encoders are BEI R15C which are 1000 pulses/rev. Sorry about the the pictures! I will take screen shots in the future.
Re: Tuning Servos trying to determine what info is relevant
Hi gbman7a,
Thanks. We'd also need to know the mechanics (lead screw pitch or whatever). The idea is to know how many encoder counts per linear inch or mm of travel.
The plot is still saturating. You might also read this.
The screen shot shows Filter2 has some coefficients but not the filter that created them. See this article.
Re: Tuning Servos trying to determine what info is relevant
I read the post/ thread that you suggested. I have a fairly good understanding of what p i and d do. I don't have a full grasp on how the output settings and error settings impact p but i will reread that thread a few times that'll take a while. I can see that limiting the output would make you unable to achieve adequate acceleration. I increased my output to 2000 and did a very rough fast tune on p and was able to achieve much better acceleration at a much lower p value.
I was reading that thread and writing my post, i had not seen your reply #4. I will read that information.
Re: Tuning Servos trying to determine what info is relevant
HI Tom, we determined that the machine operates at 5081 counts per inch traveled on the x axis.
We took an image of the filter screen but no filters have been applied.