Inductive linear scale

[zilot]

Hello

I have Ferrary CNC milling machine with SELCA control unit. There are 3 linear scales, and one of them is not working well...Doesn't increment in order when move axis, it generates for i.e 245, 246, 247, then again 245, then 259 etc... I opened it, and seems it is inductive linear encoder.... there are some op amplifiers and one blue 14 pins IC I can not find what is... 14-2-472 and B8004 is what I can read on its top... other two 18 pins ICs are 4051 analog multiplexers... Can someone help with similar symptoms... what this unknown IC could be? It is connected to coils below moving part.

Thanks

 

[SeeFair]

From distant memory, that does look like an inductosyn. Wonderful device, but not very popular... The blue chip may be a resistor pack: multiple 4700 ohm resistors in a single package.

I couldn't read any chip numbers from your photos.

It's been 40 years since I've worked with an inductosyn, but my recollection is that they work as a multi-pole resolver. When I designed mine, you could buy hybrid resolver interface electronics, a big black electronic brick that would do all the hard work. Too big to be on that little pcb. I can't remember the brand...

Again, from dim recollection, the resolver "readout" is part of a feedback loop inside the resolver interface electronics. I can't imagine what would cause it to go unstable, but since your device "sorta-works", perhaps something is awry in that loop.

After writing all this, my guess is swarf on the pickup (the part that moves along the scale). I would drag a clean rag between the pickup and the scale, if clearance allows it. There's probably several cycles of a similar inductive pattern on the pickup (for averaging), and maybe a stray metal chip is buggering it, depending on where it lands.

 

[zilot]

Thanks for the replay, you are right, blue chip is a resistor pack, exactly 4700 ohm resistors....

there are 4051 chips (two of them), and they are analog MUXs 8 to 1, and they drive as I think coils below slider...and there is one 75c914 Hex Schmitt Trigger that is interface between input connector signals and MUXs control inputs....

this inductosyn (as you say it is) is bad, when I connect another one to readout same input it works OK.

I suppose 4051s are bad, and don't generate proper signals for bottom coils...

Have you any hint how does this work at all. I guess electronics from control box generates some pattern for MUXs and they stimulate coils on the bottom, but what then? Is voltage induced on printed traces and electronics on the left measures value, and sends back to the control unit and control unit decides the position?

Does this device absolute reading?

thanks

 

[SeeFair]

I'm having second thoughts. The bottom part of that scale looks like it could be an inductive device, but the top part does not. The top part looks like a grey code or some other binary code to give it some "absolute" indication. Looks like 8 bits on the top part. Is it really possible that there are only 256 cycles on the bottom part? The electronics you describe sound very primitive, so either the guts are somewhere else or this is a technique I am not familiar with. The bottom part would work with either inductive or capacitive pickup, but the top part would only work with capacitive pickup (or contact, but I don't see evidence of sliding contact).

Here is a description of the inductosyn. They mention several methods to decode the device, but their methods to create an absolute encoder do not include a pattern like the top part of your scale. They also mention electrostatic shielding to prevent any electrostatic coupling between the scale and pickup, and you did not mention any of that. So I am going to guess that this is capactive, based on that upper part of the scale. I just didn't think that capacitive devices were well developed until the last 20 years or so.

I didn't quite understand your comment: "this inductosyn (as you say it is) is bad, when I connect another one to readout same input it works OK". Do you mean that the error travels with the readout, or travels with the scale? This scale on another readout works OK, or this scale on another readout doesn't work? I think you mean that if you connect this scale to another axis readout, it works ok; if so, you are have established that the scale and its pickup are good, so you need to look upstream, at the electronics that do the control of your muxes and decode the signals from the pickup.

 

[zilot]

I'm having second thoughts. The bottom part of that scale looks like it could be an inductive device, but the top part does not. The top part looks like a grey code or some other binary code to give it some "absolute" indication. Looks like 8 bits on the top part. Is it really possible that there are only 256 cycles on the bottom part? The electronics you describe sound very primitive, so either the guts are somewhere else or this is a technique I am not familiar with. The bottom part would work with either inductive or capacitive pickup, but the top part would only work with capacitive pickup (or contact, but I don't see evidence of sliding contact).

I don't think it is 8 bits resolution on the top, seems like 12 bits....(just look both photos, where right and left sides are shown). Bottom part is with inductive pickup, tracks are shorted on the right side (it is not shown on photo), and I don't remember, but I think top tracks are also shorted on the right side. I'm currently not there where my machine is...will check for it. There are no contacts below slider, so there must be inductive (or maybe capacitive, but I don't think so).

I didn't quite understand your comment: "this inductosyn (as you say it is) is bad, when I connect another one to readout same input it works OK". Do you mean that the error travels with the readout, or travels with the scale? This scale on another readout works OK, or this scale on another readout doesn't work? I think you mean that if you connect this scale to another axis readout, it works ok; if so, you are have established that the scale and its pickup are good, so you need to look upstream, at the electronics that do the control of your muxes and decode the signals from the pickup.

This scale doesn't work on another readout input (it has 3 inputs for 3 axis), when I connect another scale on this readout input it shows OK. So conclusion is that this scale is bad.

Here is electronic in better resolution... There are 2 OP amplifiers, I suppose they generate AC signal...but have to draw schematics of this...

 

[zilot]

Is there any example of absolute inductosyn, I couldn't find. Have You seen it before?

Do You know, what this z1a104m component is? can't find anywhere? Is it capacitor 100nF?

 

[SeeFair]

Much better pictures! And I agree that that board is the right place to look.

Glad to hear that there are more than 8 bits in the upper part of the scale; that makes more sense.

Just to be clear, anything i say here is speculation, because I'm not familiar with this device. But regardless of whether this device is inductive or capacitive, it probably works as follows: the bottom part of the scale is excited with a sine wave reference. There are two matching sensing elements in quadrature on the pickup, and the amplitude of the detected sinewaves on these two pickups goes as sine and cosine of the position of the pickup with respect to the scale. That sine and cosine gives you fine position within one cycle of the scale. The upper part of the scale gives you coarse position.

I can't see how the top part of the scale would work using induction, but the bottom part could use either inductive or capacitive pickup. Induction might be more straightforward for the bottom part.

The muxes on your pickup are perhaps for the coarse upper part of the scale only, and used to readout those bits one at a time.

I see you have the pickup pcb removed. I don't know how much electronics you are willing to do, but I would replace the electrolytic caps, regardless of what else. If it was me, I would do just that and try it again. If it still fails, I would use an oscilloscope, comparing the behavior of good and bad boards, trying to figure which chip to replace. Or you could just start replacing chips blindly, but that requires a good de-soldering device.

The absolute rotary inductosyn is described in the farrand brochure I attached previously. It requires a second pickup, which looks very similar to the first, but the second pickup has N-1 turns per rev. The details escape me at the moment, but in principle, the two can create a "beat frequency" of once per rev that behaves as a single pole resolver and gives you a "coarse" readout to complement the fine. I don't know how(if) it's done for linear inductosyns.

A little more info on the symptoms might point to the upper or lower part of the scale. How did you notice the problem? How do you know that the scale doesn't count consistently? Is there any periodicity to the error? Does it count properly within one cycle of the scale and have trouble at the transition to the next cycle?

Under the pickup pcb is the actual sensing element. Have you taken that off? (Not that I recommend that you do, but I still would make sure that there's no debris in the gap between pickup and scale).

 

[MachineAmateur]

Replace all those caps and check the resistance of that large resistor in the center of the pcb (if you see any heating on others, check them too). Resistance can be decoded by the color stripes, but if they are burned you have to guestimate. It looks bit crusty. I don't think there could be anything wrong with the scale itself, actually that looks like nice hobbyist project for somebody to use as a basis for cheap linear scale as you can purchase pile of high quality pcbs with pennies nowdays.

But it might be the sensing element crapping out on that board, you have to locate it and measure whether it reacts correctly to the changes.

 

[SeeFair]

I took another look at your scale. The upper part changes at twice the frequency of the lower part. In other words, if the pitch of the fine lower scale is 1mm, the coarse upper scale changes every 0.5mm. At least, that's the way it looks to me.

That's one more indication that it's capacitive. An inductive scale can tell the difference between lined-up and anti-lined-up conditions between the scale and the pickup (0.5mm displacement in the example above); for a capacitive pickup, I can't imagine how you would tell the difference between lined-up and anti-lined-up. So the period of the electrical signal would be half the pitch of the scale if it's capacitive, hence the need for a coarse scale that changes at half the pitch of the fine scale.

By the way, don't most digital calipers nowadays use capacitive scales? I've never taken one apart, but it would be interesting to see if they have a similar scale pattern. I have a crappy HF caliper in my shop that I need a reason to destroy because it is unreliable, but I won't be back there for a week or so. I think the reason it is unreliable is that the gap between pickup and scale is poorly controlled in the HF caliper, so I would pause before removing the actual sensing element in your pickup - you may inadvertently change the gap.

 

[zilot]

Just to be clear, anything i say here is speculation, because I'm not familiar with this device. But regardless of whether this device is inductive or capacitive, it probably works as follows: the bottom part of the scale is excited with a sine wave reference. There are two matching sensing elements in quadrature on the pickup, and the amplitude of the detected sinewaves on these two pickups goes as sine and cosine of the position of the pickup with respect to the scale. That sine and cosine gives you fine position within one cycle of the scale. The upper part of the scale gives you coarse position.

I can't see how the top part of the scale would work using induction, but the bottom part could use either inductive or capacitive pickup. Induction might be more straightforward for the bottom part.

The muxes on your pickup are perhaps for the coarse upper part of the scale only, and used to readout those bits one at a time.

I see you have the pickup pcb removed. I don't know how much electronics you are willing to do, but I would replace the electrolytic caps, regardless of what else. If it was me, I would do just that and try it again. If it still fails, I would use an oscilloscope, comparing the behavior of good and bad boards, trying to figure which chip to replace. Or you could just start replacing chips blindly, but that requires a good de-soldering device.

The absolute rotary inductosyn is described in the farrand brochure I attached previously. It requires a second pickup, which looks very similar to the first, but the second pickup has N-1 turns per rev. The details escape me at the moment, but in principle, the two can create a "beat frequency" of once per rev that behaves as a single pole resolver and gives you a "coarse" readout to complement the fine. I don't know how(if) it's done for linear inductosyns.

I started with schematics reconstruction, and this moving part is sensing part. Muxes are actually demuxes (because they are analog), and their output are in shortcut. Everything that comes from moving part inside is connected to MUXes, and MUXes senses according to some pattern on control inputs. After that there is HF filter, and some amplifiers.

A little more info on the symptoms might point to the upper or lower part of the scale. How did you notice the problem? How do you know that the scale doesn't count consistently? Is there any periodicity to the error? Does it count properly within one cycle of the scale and have trouble at the transition to the next cycle?

Under the pickup pcb is the actual sensing element. Have you taken that off? (Not that I recommend that you do, but I still would make sure that there's no debris in the gap between pickup and scale).

When I move axis slowly, it should count up or down, depends on direction, but it counts in disorder, 2, 3, 4, 5, 6, 7, 5, 6, 10 etc...

No I didn't touch sensing element, I just checked for it, and it is clean, no debris in the gap between pickup and scale.

I'll attach schematic when finish with it.

 

[SeeFair]

The disorder in counting might be a symptom of the HF filter not fully working, leaving more of the carrier frequency left on the signal where it is digitized, resulting in noise. If so, the counts should vary even if the axis is not moving (some positions are likely worse than others because of the sine/cosine thing). Replace those electrolytic capacitors! They have a liquid electrolyte in them that eventually dries out.

When you draw the schematic, please identify which capacitors are electrolytic. It's a bit unusual to use electrolytics for a signal filtering application, but not unheard of.

There's a white component on the lower right side of your circuit board, labelled "1n K 100", which is perhaps a more stable capacitor of the type one would use for filters. It looks crusty, as MachineAmateur would say. Do you know if it's a capacitor? And why does it looks like someone used needle-nosed pliers on it?

 

[zilot]

Еlectrolytic caps are used for DC elimination, and bias. Capacitor labelled "1n K 100" should be common block capacitor and is part of 2nd order RC filter (looks like High Pass filter)...that is what I have reconstructed by drawing schematics until now

Good thing here is that everything can be tested by applying some external test signals (because everything comes from the outside via connectors)... Also this filter section can be tested with function generator, and I'll do it after replacing all capacitors...to check OP amplifiers

Electronics on the left side of scale should be excitation transformers (attached), two transformers for each part, top and bottom tracks (that is maybe why I detected shortcut in the top section, I'm not there to check by visual inspection if segments are connected, but lower segment are connected 100%). Wires to them do not interfere with other parts of inspected electronics.

 

[SeeFair]

Sorry, I understand very little of that. I don't know what a block capacitor is, I don't know why there might be a high pass filter. I see four wires going to the scale and 4 wires going through a spring-type cable wrap to (probably) the pickup. I don't see any transformers.

But it sounds like you are proceeding with confidence. Keep going!

 

[zilot]

This is what I realized until now. Yes 4 wires should be output from that two transformers, and four wires through spring cable are probably supply for transformers (input) and come directly from control unit, they don't interfere with other electronics on our board we discussed in previous posts.. Currently I'm in short on time and can not solve this at once.... will replay as soon as I find more....

 

[zilot]

I finished with schematics. This is what I got. I was wrong, there is no HP filter, there is BP (Band Pass) filter. I replaced electrolytic capacitors, but there was no gain, everything is the same. From the schematics can be seen there is 3 stage amplifier.

I also concluded that upper part is capacitive sensing, and bottom is inductive (You were right). I disconnected connectors from upper and bottom tracks, and upper lines are open, and bottom are shorted.

this is video how it works now (it is Y axis)....

Will put complete electronics from one good scale and see if this works with this, then will know some more...

 

[Hopefuldave]

Hi Zilot,

As the rest of the bits fed to the display are counting correctly (I assume - looks that way in the video) you can assume (I know, ass of u and me...) that the analogue amplifier side is working correctly.

Looking at the counts and the schematic, only one track of the sensor head is either a) not getting any excitation or b) not supplyiing a signal scanned in by the mux (4051) chips. It is one of the more significant "coarser" tracks, which should help identify it.

If you have an o'scope (or can borrow one and get a bit of guidance on how-to), a sensible test sequence might be:

Check the A,B,C and Enable inputs on both 4051 muxes - all should toggle through about 5 volts or more, regularly, being driven from the same buffer/inverter chip. "Confidence check".

*With an O'scope* check the signals on the excitation connector - all should have drive signal, either continuous or (as they're indvidually connected to the display and it may "step" through them remotely - are they definitely separate wires all the way through from display connector to the left excitation connector?) in short bursts.

*With an O'scope* probe the analogue inputs to the muxes while moving the axis enough to get to the bad counting sequence - on "right" all should show varying signal levels, if one doesn't then it points to a fault on the read head - either excitation reaching it or the signal coming back. - on "left" much the same BUT your schematic (as drawn) shows that some inputs are provided wth pull-downs, some are not - expect different signal characteristics, but all sensed tracks should show variations as the scale moves. Again, if one track isn't varying the sensor may have a fault (which could be a dry solder joint, a fatigued conductor, a poor connection in the plug/socket).

If you have a good signal at all the inputs of the muxes, it could be the mux chips themselves - luckily 4051s are a commodity chip and not expensive (a couple of Local Currency Units each), it's worth pulling them and replacng them both if one is suspect! If there's room, it's alway good to install them with IC sockets (I like the Turned Pin ones!) in case they may need future replacement - but see below!

I'm not sure of the age of your DRO, if it was manufactured aroud the intro' of lead-free solders there's a fair chance that one or more (I suspect one, or it'd be really haywire) solder joints has gone dry - and if it's older than that, age will do that anyway! I think it'd be a good idea to take a walk around the board with a good soldering iron, desolder wick/pump then decent quality solder and flux and resoldering all the connections to see whether the fault miraculously disappears (they often have when I've done that on older equipment!).

Hope that helps, rather than confuses!
Dave H. (the other one)

 

[Hopefuldave]

 

[zilot]

I'm not sure of the age of your DRO, if it was manufactured aroud the intro' of lead-free solders there's a fair chance that one or more (I suspect one, or it'd be really haywire) solder joints has gone dry - and if it's older than that, age will do that anyway! I think it'd be a good idea to take a walk around the board with a good soldering iron, desolder wick/pump then decent quality solder and flux and resoldering all the connections to see whether the fault miraculously disappears (they often have when I've done that on older equipment!).

Hope that helps, rather than confuses!
Dave H. (the other one)

Тhis was the first thing I did. And were no improvements... also I checked wires and connectors....everything was fine...

Today I replaced sensing electronics with one from the good scale and everything works fine, so there is no problem with pickup elements below sliding part, no problem with gap between slider and ruler with tracks.... Only problem is in this sensing electronics... And yes I plane to replace all chips, muxes, op amps.... will do just you proposed, first will solder sockets....

I watched some signals with scope, and they are pretty complex.... conclusion is that output from the bad electronics has some greater amplitude spikes on 500Hz that good one... so maybe filter section does not work well. Problem is that LM387 can not be purchased here, and have to wait maybe even month to get it from China.. I thing this OP AMP is discontinued... probably they sell refurbished chips...as I saw there is no pin compatible replacement, and there is no enough room here for some workaround... However will try to replace everything other and see if it works, if not will wait for LM387...

thanks

 

[Hopefuldave]

Тhis was the first thing I did. And were no improvements... also I checked wires and connectors....everything was fine...

Today I replaced sensing electronics with one from the good scale and everything works fine, so there is no problem with pickup elements below sliding part, no problem with gap between slider and ruler with tracks.... Only problem is in this sensing electronics... And yes I plane to replace all chips, muxes, op amps.... will do just you proposed, first will solder sockets....

I watched some signals with scope, and they are pretty complex.... conclusion is that output from the bad electronics has some greater amplitude spikes on 500Hz that good one... so maybe filter section does not work well. Problem is that LM387 can not be purchased here, and have to wait maybe even month to get it from China.. I thing this OP AMP is discontinued... probably they sell refurbished chips...as I saw there is no pin compatible replacement, and there is no enough room here for some workaround... However will try to replace everything other and see if it works, if not will wait for LM387...

thanks

I don't think you need to replace op-amps, if the fault was in those it would affect all the tracks' outputs and it's likely all the digits in the scale would be crazy! If you do, the NE5532 is a good low-noise small-signal dual op-amp if you need to substitute. (assuming 8-pin package it's pretty much a drop-in, same pin assignments etc)

The way it counts through 4 values then back to the start but with a carry into the next digit strongly suggests it's one track of the reader that's not working - perhaps one of the muxes is the culprit? this would give readings like 245, 246, 247, 248 then back to 245, 246, 247, 248 then probably something like 253, 254, 255, 256 and back to 253 - basically one of the bits in the binary absolute position (from the reader electronics back to the display) isn't being flipped as it counts, the next bit is, correctly, so you'd see a jump in the values every few times it went through the measurements, a bit of a complicated pattern to the jumps as the display is converting binary to decimal... looking again at the video, it's definitely one of the high-order bits, unless there are strange values down in the thou" or microns going past too fast for me to read - so it has to be one of the coarse-pitches channels (longer patches on the scale).

Questions: are the spikes a 500Hz signal, or repeating at a rate of 500Hz? Does it match the frequency of one of the A/B/C/Enable lines? Again, that would point to a fault in the excitation or mux sampling side, not the analogue amplifiers.

Hope this helps?
Dave H. (the other one)

 

[zilot]

I don't think you need to replace op-amps, if the fault was in those it would affect all the tracks' outputs and it's likely all the digits in the scale would be crazy! If you do, the NE5532 is a good low-noise small-signal dual op-amp if you need to substitute. (assuming 8-pin package it's pretty much a drop-in, same pin assignments etc)

Problem here is that LM387 has internal DC biasing, so for NE5532 it would be necessary to add some extra component, and pinout is different with NE5532...

The way it counts through 4 values then back to the start but with a carry into the next digit strongly suggests it's one track of the reader that's not working - perhaps one of the muxes is the culprit? this would give readings like 245, 246, 247, 248 then back to 245, 246, 247, 248 then probably something like 253, 254, 255, 256 and back to 253 - basically one of the bits in the binary absolute position (from the reader electronics back to the display) isn't being flipped as it counts, the next bit is, correctly, so you'd see a jump in the values every few times it went through the measurements, a bit of a complicated pattern to the jumps as the display is converting binary to decimal... looking again at the video, it's definitely one of the high-order bits, unless there are strange values down in the thou" or microns going past too fast for me to read - so it has to be one of the coarse-pitches channels (longer patches on the scale).

Unfortunately I didn't trace for control signals (A, B, C), it is quite risky because it is not in good position to handle, and I was afraid to make some shortcut by touching pins...

Questions: are the spikes a 500Hz signal, or repeating at a rate of 500Hz? Does it match the frequency of one of the A/B/C/Enable lines? Again, that would point to a fault in the excitation or mux sampling side, not the analogue amplifiers.

I think it is useful signal, there is 500Hz excitation on capacitive sensing part, and it is modulated with higher sin frequency signal... This OPAMPs probably remove high frequencies and 500Hz should pass (it should be Band Pass filter), but with bad electronics, spikes at 500Hz are much higher... Maybe you are right, maybe just muxs are bad, I'll try to replace them first, LM387 I don't have currently, will try with replaced MUX, then we will know something more..

thanks for analysis