step response for force sensor analog feedback

[myo]

Hi all,

I am having a problem in manual PID tunning for servo motor torque control in MAX. I am using PCI-7350 controller, labview 8.2, MAX 4.2. I used Load Cell (50lb) from Futek as force feedback. I try to control force during movement in the application and close the loop with analog feedback from load cell. The cell produce 22 mV (analog value at ADC 1=100) when it is loaded with 10N. Therefore, I used 100 counts (10N) for manual tunning in MAX step response and the results are very unstable. Everytime I step response with identical PID gain, the results come out are not same at all. An example can be seen in attachment file.  

Why does the counts go negative side while fine tunning? I measure with multimeter on AI 1 and AI grd on UMI and the volt come out from load cell  never goes negative. Why is it not same with same PID gain? I set 250 micro sec for feedback update period.

Any comment or suggestion on this is welcome.

Thanks.

Myo

[Jochen]

Myo,

I think the problem might be caused by the low voltage values that you are using. At this level electrical noise and quantization noise might result in a not very stable behavior. I recommend using an external amplifier to better fit the output voltage of your sensor to the input voltage range of the 7350 (e. g. a gain of 100 sounds reasonable to me, but of course this depends on the force range that you need for your application).

Another thing you might try is increasing Td to e. g. 4 or 5 and changing the Kd:Kp ratio to something between 5:1 and 10:1. Please refer to this document that provides good hints for PID tuning in general.
Still I think, the main issue is caused by the low output level of your sensor.

Best regards,

Jochen Klier
National Instruments

[Brian_Beal]

To expand on what Jochen suggested, you will want an amplifier with a high bandwidth.  Some signal conditioners can only do 4 or 5 conversions per second, and you will want much more than that.

[myo]

Hi Jochen and Brian,

I am sorry that I forget to discribe the components used in recieving signal from load cell. I am using SG-04 with SC 2345 and SCB 68. I think it should be enough for amplifying the force signal, right?

Actually, I achieved the resonable step response two weeks ago. Pls refer the attachment. I didn't change anything in hardware setup and components. I really can't find the root cause for this problem now. What should I do?

Thank you for your replies.

Myo

[myo]

Hi,

Addition: My application requires up to 200N. The load cell output is 2mV/V.

Myo

[Jochen]

There are major differences in the PID parameters between your screenshots. Have you tried to use the old PID parameters again? If not, why not?
Even in your older configuration there I can see significant oscillations. Please follow my advise to adjust Kd and Td accordingly. This should result in a better damping.

Jochen

Message Edited by Jochen on 02-14-2008 08:56 AM

[myo]

Hi Jochen,

Now I could do the tuning for step response already. Thank you so much for your advice.

One more thing, now I found that I need to set different PID settings for different ranges of force. In Labview VI, could I set relevant PID setting for desired force range without stopping the motion? Do I need to initialise whenever I changed PID setting? If can, will you give some examples?

Thank you again. 🙂

Myo

[Jochen]

Myo,

you don't have to initialize the board to change PID parameters, but you can't change them while a move is in progress. To change the PID parameters without initializing the board, you should use "Load All PID Parameters.flx" or "Load Single PID Parameter.flx", but again, this will only work, if the axis is in stopped mode.

So depending on the way you have to apply force to your load, this might or might not be helpful for you. If it doesn't help you to change the PID parameters in stopped mode and if you can't find a set of PID parameters that works ok for all force ranges, you probably need a truely adaptive control algorithm. Unfortunately this can't be accomplished with 73xx devices, but if required I can help you selecting another hardware platform that allows this type of control.

I hope this helps,

Jochen