Automation Riddle

[JRoss 84]

Very, very carefully. In the factory, we measured the outputs with a tachometer and adjusted them to them to be as close as possible. Then we fed rebar and further tweaked the settings to minimize the drives fighting each other to nothing. Believe it or not, this part of the process works very well, and you'd be hard-pressed to see any fight between the two feed systems. Of course, being a rebar feeding system, there is a lot of inherent vibration that could possibly hide fighting. Even if the drives push/pull each other, the encoder would simply speed up or slow down with the bar. Even if there were slip in the drive system, the encoder is attached to the pressure wheel, not the drive wheel. Since the two wheels turn independently, the pressure wheel should follow the rebar regardless of slip. Even if somehow the rebar was pushed through the pinch wheels and the encoder wheel didn't turn (got jammed, or whatever), the rebar should always go further than it's supposed to. It occasionally stopped rather short of the mark.

[BobLfoot 301]

JR Did I read your O-scope correctly? B- Phase drops to 0V and A-phase went to 4X freq? Any chance the encoder has a malfunction internally? I'll give PM me the answer.

[Steve Bailey 118]

The scope trace looks like the encoder stopped but something was still trying to turn it. It could only move for a fraction of a degree so the A channel signal came on and off. In a properly configured counter module this "chatter" shouldn't affect the count. The rising edge of the pulse will increase the count by one, the falling edge will decrease it. When the B channel signal comes back the count will continue normally. If the rebar is still moving when this sort of anomaly happens, the net effect would be to lose counts and feed too much rebar. Does the same sort of thing happen on the B channel? If it happened on both the A and B channels simultaneously, I can see where that could give extra counts resulting in a short feed. Is something striking the encoder causing it to vibrate? I would expect that to show up on both channels, not just one. Is the DC power supply for the encoder overloaded?

[panic mode 246]

exactly, this looks like normal "in position" situation where position loop is locked and encoder is stopped or actually slightly oscillating around one point which is less than 1/2 pulse. if encoder is wired correctly this will not change count (or it will count one pulse up, one down etc but stay on tatget). if encoder is not wired correctly or if it is used as pulse encoder (single channel), every pulse will be counted and position will drift. problem still could be mechanical if the shaft monitored by encoder get's jammed or seized under load for example.

[JRoss 84]

I think my wave diagram may have confused rather than enlightened people. I was trying to illustrate what I was talking about without having to draw anything myself, so I found an image online that was close. However, as Steve suggesteded, this jumping waveform showed up on both phases in a random pattern, so that the encoder would sometimes lose pulses and sometimes gain them. And yes, this was on a differential, two-phase encoder that was wired and working properly! Steve is close enough that I'll go ahead and give the answer. The problem had nothing to do with noise, although that is what I thought as well. However, the encoder cable is and was shielded properly, and using an oscilloscope proved that there was no EMI noise on the system whatsoever. The problem was a result of mechanical vibration. As I hinted above, the two existing machines are almost identical. The one difference is that on this machine, the feed system (motor and drive wheel) were mounted on an adjustable mount. This was purportedly to allow some up and down movement to control where the rebar was in the feed trough. Not a particularly important adjustment. Over the course of six months of use, the bolts on the mount had loosened enough that the vibration caused by the bar moving through the feed wheels exceeded the vibration specs on the encoder. What results is something called encoder "dither". The vibration caused the optic wheel inside the encoder to skitter, or "dither" back and forth and provide additional pulses to the counter. Because we were using two-phase encoder feedback, the system naturally filtered out some of the problem. But not all of it. This is something that the machine builder has dealt with many times, and as soon as they saw the oscilloscope pattern they could fix the problem. They tightened all the bolts on the mount, and the problem disappeared. After some further confirmation, they welded the mount in place to avoid a repeat of the issue. Encoder dither is not something I had heard of before. Some cursory Google searches don't yield much either, although the term is used in image and sound editing also. Most interesting to me was this article on Wikipedia about how dithering was first used during World War II. So, watch those vibration ratings on your encoders!

[P Daniil 0]

I wish I could have joined this great thread earlier. There are many ways an encoder can give you sleepless nights. ("Eliminating Phantom Movement in Encoder Applications" may be of interest). From what I have seen, the most perplexing and difficult problems to tackle are always the ones involving the index channel.

[panic mode 246]

if it works out but not in machine and cabling etc. didn't change, it is mechanical problem and this was repeated few times. only time chatter as shown on graph can happen normally is in stop position (encoder position oscillates so little that is moving just one pulse) and this is taken care of by input logic of the high speed counter. machine that flexes out of shape under load is not mechanically sound (either design or assembly).

[JRoss 84]

Not sure if I followed all your verbage there, panic! Yes, I suppose you could call this a mechanical problem. But not in the sense that something jammed, bent, broke, or otherwise catastrophically failed. Rather a minor mechanical issue (loose bolts) affected how a portion of the feedback system operated. Check out the specs for the Stegmann encoder in the attached datasheet. There is a line entitled "Resistance to vibration". Once you exceed that amount, the optics inside the encoder will be vibrating enough to created a lot of additional pulses on both phases. The input logic of the counter card can handle that, as long as the "true" pulses are where expected. But what happens when there's enough vibration that a "true" pulse on phase A doesn't line up properly with a "true" pulse on phase B? Or if the additional vibration pulses line up at just the right spot? Lost counts and added counts, no matter the quality of the counter card. The wonder is, that over a 30 foot length of rebar, the positioning was usually only a few inches off, with the Stegmann encoder. The Encoder Products unit (again, my report from the machine builder) caused the rebar to shoot out the end of the machine. I'm guessing the vibration resistance wasn't as high. I would agree about a machine flexing being bad. This wasn't exactly flexing. Just a machine adjustment in a bad spot: right at the epicenter of vibration! That's why it was taken out. It was less important than a working machine. DRS61_User_Programmable.pdf

[panic mode 246]

you are right, it is a matter of definition but machine parts comming loose and vibration of delicate precission device such as encoder in excess of 10-20g/ms or shocks of 50g sounds like quite mechanical problem to me. electrical problem is malfunction of electrical system and that didn't happen (loose connection, trouble with power, noise, ground loops, program and parameters etc.). in fact you could take the whole electrical system and move it to other machine (or swap them out) and it would be just fine... ...because problem is not electrical, it's just that control guys end up on site troubleshooting and fixing everything anyway.

[P Daniil 0]

This great truth should be added to the PLC laws!

[BobLfoot 301]

It is not PLC law #26

[JRoss 84]

A very good point, panic. Put that way, I have to agree entirely. But then, of course it's the controls guys who figure out the solution. [friendly jab at our mechanical counterparts]

[OIDMBTRTECH 0]

All good controls guys are excellent mechanics too. Most mechanics run from that electrical stuff. Since the mount was welded down, I won't ask if the bolts were locktited. Additionally, I won't suggest thay you set the VFD carrier frequency to its lowest setting to help stop the mechanical vibrations that caused the loosening of the parts. Yes, the carrier freq in a VFD can and does cause some machines to vibrate. Not too often but it may have been some of the problem. If you could figure out the freq of the dither and you could compare it to the VFD carrier freq.

[JRoss 84]

It always amuses me when people say they aren't going to say/suggest/recommend something, and then go ahead say whatever it was they weren't going to say. The vibration causing the problem was caused by the ribs on the rebar grinding against the teeth on the pinch wheels, so changing the carrier frequency wouldn't have done much. When we ran the machine with no rebar, there was almost no vibration. As I mentioned, the machine builder has had to deal with dither on previous machines. Most of those previous machines were hydraulic, with no motors except the hydraulic pump, and that usually was on a separate frame.