MDC2460 realistic position accuracy expectations

Hi, me again with a different question. I mostly scrapped the analog position sensor in favor of a quadrature encoder. I finally have something working. Now, I need to get some realistic expectations around the performance of this controller.

I can now make the controller function properly in all three different position modes. They all perform differently as one would expect. PID tuning seems to be an art I'm going to have to invest time in mastering. I have had varying levels of success at tuning the PID and I've tried a couple of different strategies including the one in the manual.

Some of the position modes give me much snappier response and begin moving the actuator (via attached geared motor) promptly but have other undesirable behavior. For example, very high frequency oscillation from the motor when changing directions but no real overshoot on arriving at a destination position, especially if direction changes are very rapid. The "position relative" behaves best in that regard but is really sloppy/sloshy with the accuracy. On a 4096/rev count encoder with a total count range of 12800 over the 6" the actuator moves, it's accurate to within +/- 1.95%, i.e. about 250 counts. I can improve upon that somewhat with additional P gain but then it begins to behave closer to the other modes where direction changes result in high frequency and momentary oscillations again. I can compensate for that to some extent with differential gain but of course that brings about other issues and when I try and compensate for those with integral gain, the slop/slosh comes back.

Now 1.95% accuracy isn't all that bad, it just slightly changes where the actuator stops depending on which direction it was moving on the way to the new position and for this application that could be a problem. I might be able to compensate for that in my control logic by tweaking the destination position to add a couple hundred counts worth of normalized position value to the desired destination but for small position changes, that could be a real problem.

Now don't get me wrong, I'm not necessarily complaining, I'm just hoping someone with the voice of experience with these controllers can help manage my expectations. I typically have to move the slider in Roborun a little ways before it will even begin moving the motor and once it does, I'm already 250+ counts away from the previous position. I'd love to get the position accuracy down to within 0.5% if possible but I still need maximum speed when changing positions from one end of the linear actuator to the other.

Thanks in advance for your thoughts :)

You seem to have found reasonable, but different, PID gains, and/or closed loop algorithms, for two different conditions - changing direction vs. approaching destination. Might it be worth thinking about a script that changed the gains or algorithm to match what the actuator is doing?

Whoa, help me wrap my head around that. Is that even possible/practical/possible?

How much inertia and gearing are we talking here? The controller should be able to hit this with proper PID tuning.

Anyway, reading through your post again, I think this may have less to do with PID, and more to do with the limited resolution of Relative Position Mode. Have you tired count position mode?

Not a lot of inertia and gearing is quite high as would be the case with a worm gear. Based on the construction of the actuator, I'm pretty sure it has to be a worm gear. I have tried the position count mode but I couldn't get the PID tuning dialed in.

In position count mode, the performance would either suffer from overshoot or be too sloshy. I'd start at 0, then increase p-gain until it started to over-shoot. Then increase d-gain until the overshoot was tempered but on rapid direction reversals, the motor would kind of rapidly whipsaw a number of times before accelerating toward the destination.

position relative mode was overall the most well behaved. I can try position count again.

Worm gears mean everything is going to be geared down by quite a lot. But at the end of the day, you should just make your P gain stiffer, and get better position feedback.