1746-HSCE2 question

I have an application where I am firing two measuring cylinders (they send out a A/B phase difference pulse) into a part to measure the length and width. The 1746-HSCE2 counts the pulses so I know how far the cylinder extended to contact the part. The part profile is very small, so it will be very difficult to embed a prox in the tooling to detect when it has contacted the part. Right now I am just using a timer to determine when the motion is complete (the retracted switch is off and the extended switch has not been made for 1 second, so the stroke must have contacted the part). This way adds unnecessary time to the cycle and is dependent on the speed of the stroke. If someone changes the flow controls, the system could start taking measurements too soon.

I thought since I am getting the count values from the counter card, there should be a way to monitor these counts to see when they've 'settled' to within a certain tolerance. Since the count usually doesn't stop completely (the values changes slightly up or down even when the cyliner is not visibly moving). I am having a hard time coming up with a method in the logic to do this. Has anyone out there done something like this before? Any suggestions are welcome.

I think you have the right idea

The cylinders have settled when you read two counts that are the same.

If that works then you have a very simple solution, otherwise things get to be a lot more complicated. What throws a wrench in the works is when the objects compress like wood or the feedback is noisy. The changes of getting two counts the same start going down especially if you have high resolution encoders.

We have a stopped bit in the status of our motion controller that does exactly what you want to do. On our latest one we had to add a stopped threshold speed. For you to do the same you would need to calculate at least the differences in counts each time and check if the absolute difference in counts is less than a parameter you can set.

It is funny that expensive motion controllers are often asked to perform this function. Sometimes it is measuring and other times it just pushing an object against a line bar.

It's too bad you're on a SLC. If you had the CLX, with the timestamp from the card, an accurate derivative of the counts would be very easy.

Are you talking about looking for two consecutive scans with the same count (or two counts very close to each other), and calling that the end point? I guess that might work, since the counts are increasing pretty fast compared to the PLC scan time. I might give that a try.

I would determine that the cylinders have reached the part when the peak count value stops incrementing.

One thing that might help is determining if you have creep (count keeps increasing slowly) or oscillation (count jumps up and down). Creep can be a bit of a pain. However, oscillation is a bit easier. Once you start moving toward the target you can reasonably expect the count will not decrease. If it does you can say you have stalled.

Peter's idea is a better all-purpose idea, though. It will handle any case you throw at it. The tough part is getting a decent position derivative on a plc platform that doesn't have good tools to do that. Like CroCop said, this would be nicer on a ControlLogix processor. It might not be any easier but it would be more accurate.

I can't believe I'm helping a Michigan fan. Next thing you know we'll have cats and dogs living together.
Keith