Ned, I went ahead and supplied 5 volts to the encoder and followed your instructions for scope use. I could watch the transitions on the screen. Thanks. If the carriage is moved back and forth .0002 the transition will show. Still, the mysterious backlash seems to be present. Small movements seem OK but not large ones. It's hard to count transitions on the screen and watch the indicator too. I got someone to count the transitions on the screen while I watched the indicator. BTW, the indicator is OK, it doesn't have bachlash. Still, I can't see if the shaft is moving or not. So the laser is probably next. However, I think the problem is hysteresis in the rubber belt driving the encoder from the carriage. We'll see. Eric
Hi Eric, I have designed the counters and state machines used to decode quadrature encoders like you are using and there is a good possibility that the problem you are seeing is in the readout head. By putting the scope on the encoder outputs and confirming that they transition every
0.0002 inches including the first, you have confirmed that the encoder is working correctly. The only other test would be to verify that the transitions are on different 0.0002" boundaries for the two channels. One should change on 1,3,5... and the other should change on 2,4,6...
Assuming that the encoder is wired correctly to the encoder, if you don't see the backlash in the scope transitions, it is not in the encoder/mechanical system. Look through the manual for the readout head for configuration stuff about filtering the input data stream. Many of the low cost quadrature encoders are pretty noisy and may generate multiple transitions. It is common to build some kind of filtering into the state machine that turns quadrature transitions into counts. If there is an option to disable filtering on the head, try it.
Another possible source of the problem would be an "index" sensor input. Since quadrature sensors are relative position sensors, a third sensor that triggers once per revolution or at one end of a mechanism can give you a calibrated "home" start point to give an absolute position. If there is any mention of tying the index input to +5V or ground, make sure you have done it. There may be exception states in the quadrature decoding state machine to deal with the index sensor. Or if there is a configuration option to disable it, make sure it is disabled.
Or phosphor bronze... I can see that working for partial turns, but not multiple turns as Eric is looking for. Two opposing straps wound not quite the full circumference, fixed firmly at the ends, will track like a worm and rack.
It would be instructive to disassemble a dial indicator and look inside. Mebbe I would just find a way to mount the encoder to the de-faced indicator, rather than mess with mirrors, lasers, belts, oscilloscopes, quadrature waveforms, ... unless the idea is to mess, to reinvent, rather than measure.
I know how indicators work. And they involve anti-backlash gears. But that misses the point. The encoder should have NO mechanical lash. The disc with the lines is mounted directly to the shaft. A metal ribbon if the friction was high enough. I ordered some fine cable today from McMaster-Carr today. It will be wrapped a few turns around the encoder shaft. I hope that this will provide enough friction while at the same time have low enough tension to keep the friction low in the pulleys the cable will travel over. ERS
The turn-on and turn-off positions do not happen at exactly at the same position when the shaft is reversed. This is where your "backlash" is coming from. If you want better accuracy, use a higher count per rev. encoder.
In article , email@example.com says... I ordered some fine cable today from
What concerns me about using a wrapped cable is that the turns of cable want to walk across the drum as a result of the helical wrap. This will cause the cable to be pulled out of line with it's attachment points, which changes number of pulses per inch of carriage travel. If the stroke is long relative to the drum's circumference the cable may either slip back into line, cross over itself, or walk off the end of the drum.
I've done this, and even with a fairly short stroke, large drum, and low accuracy requirement, it was a real pain to get working reliably. I don't mean to be discouraging, but I haven't heard, or thought of, any idea yet that gives me the warm fuzzies.
Sorry didn't catch the previous post. One way to avoid the cable walking is to set up a parallel shaft nearby and turn a couple "windlasses"- bigger collars on each shaft so the cable can bear more on each. Wouldn't hurt to put notches on the collars to help guide the cable. So the coil goes around the first, then around the second, then back around the first. Not wrapping all the way around each shaft, but looping around the pair of shafts. The aux shaft should be free to spin. Then as the cable drives the shaft (or vice versa), the cable won't walk or slip. Maybe you only need a few loops around the shafts & can keep things fairly compact.
for what it's worth, I have a very nice Trav-a-dial that has a built in encoder - look at what a trav-a-dial is on e-bay or on the web - this one works perfectly (the mechanical part anyway) - I just overhauled it to get a piece of swarf out of it, so now it's all clean inside and is as smooth as new. It's the series 6 unit (one generation old) not the older silver ones. this might be a much better solution to your problem than other mechanical assemblies though I haven't yet gotten any information from the company as to what the encoder specs are.
anyway, I don't really need one of these with an encoder inside, so trade/sale/??? is possible
Greg, I can't quite see waht you are describing. Maybe I'm too tired. The encoder shaft cannot get any bigger. This shaft is 1.000 in circumference. This is so the encoder outputs 10,000 pulses per revolution. Could you explain your idea more clearly so I can understand it? I'm sure everyone else here has already visualized the setup but my brain is fogged this morning. ERS
With a tightly-lashed quadrature encoder, it should be, at most, one count, and usually on the order of 1/2 count. As soon as the "reversing" leading edge is sensed, the quad should pick up both count and direction.
Take your encoder shaft, put a 2nd parallel one "nearby", maybe 1 diameter away. On each shaft put in a series of notches, not deep, just enough to help guide the cable- no more than 1/2 the diameter of the cable. Align the shafts so the notches on one shaft are in line with those on the other. The cable comes in, goes around the outside of one shaft (in the notch), over and around the other shaft (again riding in the notch), then back to the 2nd notch on the first shaft and so on until it has traversed each notch on both shafts- looping around them all the way. With tension on the cable, it will get nice and tight, but it won't walk since its not taking a helical path along the shafts. You should pack in the turns as closely as you can to reduce how far the cable has to shift down the shaft to make the next notch.
I suggested the bigger diameter collars because this approach loses 1/2 of the engagement on each shaft- since its only bearing on the outside surfaces of the shaft pair. Maybe you could cross the cables over as they lead between the shafts & get more engagement that way.
Frictional loss is greater with the 2nd shaft, though maybe you could make the 2nd shaft bigger to reduce the loss a bit.