Dual Head Linear Encoder Troubleshooting Assistance

[Pork]

I'm reviving a 3D laser scanner with a bad (impossible to replace apparently) linear encoder.

The encoder was manufactured by Dynamic Research Corp (they sold this part of the company to another company that is now out of business). The encoder has some special requirements at the end of the part number. The datasheet specifies one contact the manufacturer for the meaning of the additional requirements. I've contacted several major manufacturers like Heidenhain and it looks like I'm on my own.

Symptoms:

Bad Read Head

1. The bad read head travels in the X+ direction to the end of travel while homing and crashes.

2. The bad read head channel A+ square wave output is relatively close to 50/50 duty cycle and I believe an acceptable wave form.

3. The bad read head channel B+ square wave output is intermittent and not even close to being 50/50 duty cycle. My o-scope is analog so I'm it's hard to tell if there is a 90 degree shift between signals.

4. The bad read head channel ZR+ does not output an index pulse over the entire length of the scale (I think it should pulse at the middle).
_____________________________________________________________________________________________________________
Good Read Head

5. Good read head outputs decent square waves (50/50 duty cycle) on both A+ and B+ as well as a ZR+ index pulse at the center of the scale.

6. Because of the good head behavior, I think the glass in the scale is in working order.
_____________________________________________________________________________________________________________

What I've done to repair the encoder:

1. Cleaned scale.

2. Very lightly oiled bearings.

3. Removed all Sub-D pins from encoder cable, removed small amount of wire insulation, soldered all pins to wire and reinstalled pins.

4. Checked for continuity at both ends of all read head wires. I also attempted to identify any shorts (none found in the insulated wires)

5. Cleaned optical components under high magnification.

6. Checked for continuity between read head and cable to control card.

7. Did a reflow soldering of all PCB components. The components are very small and I had to add a couple of wires to repair two damaged traces.
____________________________________________________________________________________________________________
Current strategy:

1. After posting this I'll order replacements for the 751580 and LM 393 chips.

______________________________________________________________________________________________________________



Any thoughts?


P.S. I can't attach video but, I have three short oscilloscope reading videos that better convey the intermittent behavior of the B+ channel. If it helps I'm happy to email them.


Thanks in advance

[Pork]

Pictures of the bad read head PCB:

[MadTux]

On Heidenhain scales, there is a 5 photodiode array, along with a lamp.
2x2 photodiode array is used to make sine quadrature current signal, so the machine can detect into which direction it moves
5th photodiode is used to detect reference marks at the end.
Pics of such assembly here:
https://forum.zerspanungsbude.net/download/file.php?id=159815&mode=view
Analog quadrature waveform output:
https://forum.zerspanungsbude.net/download/file.php?id=159817&mode=view
Typical glass scale division on German scales is 0.02mm (new) and 0.04mm (oldish ones), so quadrature period every 0.02/0.04mm. This signal is then interpolated to get like 1µm resolution.

With those LM399s and square with square wave output, there is interpolation happening already on the reading head.
Since the duty cycle isn't 50/50, I'd guess that the analog quadrature output isn't good anymore, probably misalligned optics.
Or the lamp/LED/laser diode whatever used to light the scale, has burned out.
On Heidenhain scales, there are a few screws to adjust the illumination to get good quadrature signal.
https://www.heidenhain.com/fileadmin/pdb/media/img/1078628-22_Interfaces_en.pdf

[Pork]

So you're confident that swapping the LM393s would be a waist of time?

The pic of the diode array you sent was a dead link.

I've attached a pic of the diode array and I don't see any adjustment screws for alignment.

Questions:

1. How was it aligned at the factory?

2. Can the emitter and receiver components be replaced or scavenged from another scale with the same resolution (1250 Lines Per Inch)?

Thanks

[MadTux]

Checking those LM393 in circuit should be quite easy, even with simple DMM, testing SN75158 is equally simple.
If these are fine, I don't see a reason why to swap, a good cleaning the PCB might be more important
https://www.ti.com/lit/ds/symlink/lm339.pdf?ts=1592432366109&ref_url=https%253A%252F%252Fwww.google.com%252F
https://www.ti.com/lit/ds/symlink/sn75158.pdf?ts=1592432187790&ref_url=https%253A%252F%252Fwww.google.com%252F

Adjustment screws are above the lens on the Heidenhain ones.
It goes like: Lamp, beam adjustment screws, lens (collimated beam), stationary grid glass plate, moving glass scale, photodiode assembly.
That stationary grid glass plate is to increase the interference, so to get more light than with just a single slit.
The screws are there to decrease the light, so that the light that every photocell gets, can be adjusted, to optimize quadrature output.

Sure you can probably swap the photocell/electronics sections between scales, but it would probably still require alignment/adjustment and you risk breaking stuff doing so.

Have you checked the light source, is it LED, lamp or laser diode. IR or visible? IR should be visible on camera.

[Pork]

I ran 5 volts to the bad read head and took a picture of the IR LEDs with my camera.

They appear to be working correctly.


Anyone know how to verify the rectangular IR receiving diodes?

I checked them in circuit with a Fluke 177 set to diode mode and I get .55 volts in both directions for all of them. This is not what I would expect from a working diode.

[MadTux]

0.55V is exactly what you get from such a photodiode. Photodiode output is evaluated as variable current, btw.
Put like a 100ohms resistor in parallel to output and measure voltage over diode/resistor pair again. Now you should measure variable voltage, like 0-50mV or so.

Good evaluation electronics apparently use transimpedance amplifiers btw, so there is no voltage over the diode, for  best signal linearity.

[Pork]

So I may have shot myself in the foot then.

When I opened the optical portion of the read head to clean it there was a black tar like substance on portions of the IR receiving LEDs and I cleaned it all off. Was that likely to have effected the output waveform?

[MadTux]

Yeah, dirty optics is probably #1 reason of linear scale failure.
Good cleaning so far fixed them all for me.
#2 and #3 reason is cable breakage and idiots smashing the glass scale by improper handling/carelessness/bad packaging.
Burned out lamp is also quite common.

[Pork]

When I first received the machine and was assessing the damage, I noticed the counter balance cables had snapped at some point.

I believe the head likely made a Z- crash. I'm guessing this was a large contributor to my current situation.

[Pork]

Your comment about cable breakage reminds me. I was having the devil of a time with the small wires from the optics PCB to the quadrature PCB breaking off after reassembling.

I was considering replacing all the wires in the read head because of this.

I noticed that there are two different types of wires used in the read head. The wires running from the optics to the quadrature PCB are very fine compared to the wires that run to the 15pin shielded cable from the quadrature PCB.

Do you know if the wires from the optics board to the quadrature board need to be of a specific requirement or will any small wire of adequate gauge work?

[MadTux]

The short ones inside the head, just normal plain wire of reasonable diameter should be enough.

The long cable outside of head is shielded on the Heidenhein ones, but there they have reason for shielding, since the analog to quadrature interpolation usually happens inside the CNC control/Heidenhain Exe box, far away from the reading head.

If the analog to TTL quadrature happens in the head already, you can probably get away with normal wire, since TTL is much more noise tolerant than small amplitude analog signal.

[Pork]

I made several small alignment adjustments of the optical portion on the bad head and the output waveform improved dramatically.

The machine will now find home without crashing. 

It still isn't generating an index pulse on ZR+ though. 

I will make several more adjustments and plot the movement. Hopefully the missing index pulse is just an alignment issue.

I need to stop for the night but I'll resume tomorrow sometime and post back.

Thanks so much for all the help.

 

[Pork]

Alright, I'm still not getting a ZR+ pulse and could use some help troubleshooting the circuit.

I applied 5 volts to the read head.

Measurements on LM393:

between pin 4 and pin 5 = .160 volts (L)
between pin 4 and pin 6 = 2.272 volts (H)
between pin 4 and pin 7 = .200 volts (L)
________________________________________

between pin 4 and pin 2 = 2.275 volts (H)
between pin 4 and pin 3 = 2.112 volts (L)
between pin 4 and pin 1 = .168 volts (L)

**********************************************************

SN75158 Measurements:

between pin 4 and 5 = .200 volts
between pin 4 and 6 = .063 volts
between pin 4 and 7 = 3.912 volts

between pin 4 and 1 = .061 volts
between pin 4 and 2 = 3.89 volts
between pin 4 and 3 = 1.58 volts

**********************************************************



If I move the read head past the position on the scale where it should trigger ZR+ I get no voltage change on the LM393.

Pin 2 stays high
Pin 3 stays low
Pin 5 stays low
Pin 6 stays high

Cutting the white wire at LM 393 pin 5 results in the following changes:
 
Pin 5 = .324 volts (L)
Pin 6 = 2.28 volts (H)
Pin 7 = 4.95 volts (H)

SN75158 then does the following:

pin 5 = 5 volts
pin 6 = 2.28 volts
pin 7 = 4.97 volts

*********************************************************

Possible reasons the LM393 inverting and non inverting pins 5 & 6 aren't switching?

[Astrodev]

I just came across you posts and at first glance the  components on the board reminded me of a motor positioning system with quadrature encoder output I used to have dealings with, the problem was these would end up hunting because the duty cycle on one of the outputs would go astray, the cause was the hysteresis on one of the comparators, in this case there were 2 trimmer pots for adjusting the hysteresis point on each comparator, so noticing what looks like a pot on the board made me think of this.

I am guessing you would already thought of this but mention it just in case.