Servo Tuning Advice

For a torque loop I would add P and D until it cannot be made stable
(more D will allow more P up to a point) then back off on both proportionally until its stable again. (do this with no I term)

Then as I said before, add I until its unstable and then use 1/2 to 1/3 of that amount of I
I suspect you need 10 to 1000 times more I

PCW wrote:

Yes, heres a static error (250 uInch) in the middle that the Integral term should pull in
I would make the integral term larger until it gets unstable and then use about 1/2 to 1/3 of that amount. Seems like the deadband could be smaller as well just so it does not interfere with interpretation of the plots

When I make "I" large the whole thing starts shifting up (like adding more bias). Is it normal for this to happen? Then should I use the bias to pull it back to 0?

No that makes no sense. Something is wrong if 'I' term behaves this way. it should at the least start to pull in the static error to 0

A plot would help

But most of the tuning threads on the forum that I've looked at said that a lot of I was bad.

After about 5000 or so I stopped moving the error away like the bias, and started to bring the steady state error in line.

You _may_ be able to reduce the width of the error spikes at the
beginning and ending of motion by increasing 'I' further

I am a bit frustrated again. Made some different length moves at different speeds and the steady state error came back with a vengence. I am not sure where the solution lies. But right now I'm to buisy with other things to mess with it.

I've finally had a little time to play with it again. I was not able to able to add enough I to cause instability. However having an I of 50K did result in a very interestingly odd shaped f-error plots in hal-scope. I'll post a pictuer of it later. Also I don't understand why adding "I" is messing with the bias number so much.

Here is the hal scope for I @50000

Are there any simple tricks to make the ouput scale positive? I wonder if I have a positive/negative problem screwing with the PID math. I ran it backwards a little the other day and I thought things seemed a little better, so I switched 2 encoder wires to reverse that, but I haven't found a simple way to reverse the motor, (I don't want to deal with swiching a power wire and re figuring out the hall sensors (that was much to painful the last time). I will have a look at the drive software I might be able to do it there. I know it can for step/dir, but I don't know about analog torque mode.