Universal table woes (long) - Page 9

# Thread: Universal table woes (long)

1. Hi Ross,

Originally Posted by AlfaGTA
Good call on the centering scope....wonder if that would have enough magnification to be able to really see the encoder marks...will have to check that out...
Of course you will need to try this to see if it works. But I think it should work. Here is my logic:

Your encoder has 9000 lines, and the approximate diameter of the disk is 120mm. Hence the circumferance of the disk is pi x 120mm ~ 380mm. Hence the distance between the lines will be about 380mm/9000 = 0.040mm = 40 microns. I expect that the lines themselves are between 1/4 and 1/2 of this width, so 10 to 20 microns.

That should be easily visible in a 30x centering scope. These scopes typically permit you to center objects to 3-5 micron accuracy.

(If it's not visible with a centering scope, then use a "real" lab microscope with x100 or x200 magnification, clamped or fixed to the headstock/ram.)

Cheers,
Bruce

2. Originally Posted by ballen
Of course you will need to try this to see if it works. But I think it should work. Here is my logic:

Your encoder has 9000 lines, and the approximate diameter of the disk is 120mm. Hence the circumferance of the disk is pi x 120mm ~ 380mm. Hence the distance between the lines will be about 380mm/9000 = 0.040mm = 40 microns. I expect that the lines themselves are between 1/4 and 1/2 of this width, so 10 to 20 microns.

That should be easily visible in a 30x centering scope. These scopes typically permit you to center objects to 3-5 micron accuracy.
Do you know for sure that rotary encoder etched line circles are not necessarily perfectly concentric with the disc ID upon which they are etched...or is that more a theory of what might be the case sometimes ?

3. Hi Don,

Originally Posted by Milacron
Do you know for sure that rotary encoder etched line circles are not necessarily perfectly concentric with the disc ID upon which they are etched...or is that more a theory of what might be the case sometimes ?
This is purely theory: I have no special knowledge or training in these matters (though I do understand how the encoders work).

Cheers,
Bruce

4. Aluminum
Join Date
Oct 2016
Country
SWITZERLAND
Posts
55
Post Thanks / Like
Likes (Given)
16
16
Just had a closer look at the 9000 lines encoder of my FP2NC 2038 table and I think I just realized why it has 2 reading heads.

One is for quadrature encoder and and the other is for reference marks.
Outer ring has 9000 lines and inner ring is mostly dark, with a few reference marks, 180° on each side.

Similar to Heidenhein linear scales, but there they get along with 1 lamp for quadrature and reference marks. On rotary encoder, they probably needed more space with line encoder, so they had to add another lamp for reference marks. So with a dead lamp, you either loose encoder or reference marks.

Also means that for alignment, you want to get correct rotary alignment of glass encoder ring with reference marks, so like that 0° reference mark means your T-slots are parallel to X-axis, along with good concentricity of glass encoder ring on the worm ring.

BTW table and encoder was all filthy with old coolant as some coolant leaked into the worm gear assembly (probably by 6x allen screws that hold table to worm gear). Oil was rusty and nasty, so I stripped everything down. Got it cleaned and mosty assembled by now, fortunately no real damage done by coolant leak. Lamp assembly/encoder ring is currently having a bath in isopropyl alcohol.

Worm gear adustment is kind of funny on that table, because on most rotary tables, worm gear backlash adjustment is done by tilting worm in/out of worm gear ring. On 2038 table, worm is pulled/pushed in/out of worm ring by 2x2 screws on the back of table. Worm screw thread is narrow on 1 end and gets thicker towards the front of table for that. So you have to remove the locking nut (below cover) on the worm shaft (shaft rides on 2x tapered roller bearing) before pulling the bearing assembly out.

5. Diamond
Join Date
Dec 2002
Location
Benicia California USA
Posts
8,061
Post Thanks / Like
Likes (Given)
1801
2410
Think your evaluation of your table might be different on different tables.....
Have seen a few different setups on these tables and their related encoder setups....
Some have two lamps, have seen ones that have only one lamp , and even a setup that had two lamps, but only one reading head.....

Will tell you that the 2038-1100 table on my FP4NC has two lamp assemblies and two reading heads.
That encoder has a PC board with a set of DIP switches and you can select via the switches either or both the lamps to be powered....and i will tell you that the angle readings
are produced with either or both the lamps powered..Signal read from either read head via "O"-scope was essentially the same ...So at least on that setup, it does not seem that one lamp is used for reference marks (a feature that at least the Dialog controls do not use on the rotary)

Based on my exposure to these tables and their encoders the only safe conclusion i could make is that "your results may vary."
Seems there are lots of differences on these.

As to the gunk...yes the bolts can be a source of leakage. I use a metric "Dowty" style washer under the head of my table bolts.
Don't forget the "O" ring that seals the neck extending from the center of the table.
On my 2038 the neck was pitted from corrosion (think coolant condenses there) and mine required fitting a "Speedi Sleeve" to negate the pitting in order to get a good seal with the new "O" ring.

Cheers Ross

6. Aluminum
Join Date
Oct 2016
Country
SWITZERLAND
Posts
55
Post Thanks / Like
Likes (Given)
16
16
You're correct, Ross.

7. Dear DMD,

IMO it would be poor design if one head was used for reference marks and the other one for the quadrature scales. But you can easily test your theory. Just cut out the power to the lamp in what you think is the reference mark head (disconnect either of the white wires). If you then can measure angle but lose your reference mark, then you have proved your theory.

I don't think this would work because if you look at Heidenhain specs for these scales, the reference mark has to be timed with respect to the quadrature signals and also have an amplitude which matches those. This is tuned from onehead to the next. So not likely that two different heads would match correctly in phasing and amplitude for this.

Cheers,
Bruce

8. Aluminum
Join Date
Oct 2016
Country
SWITZERLAND
Posts
55
Post Thanks / Like
Likes (Given)
16
16
Bathing the encoder optics in isopropyl alcohol probably wasn't the smartest idea, since it dissolved the glue that was holding together the grating windows above the photocells and I ended with an rotary encoder kit inside my IPA dish.
The good thing was that I could clean off all the rust ;-)

Fortunately got everything back together without smashing any unobtainium Heidenhein glass stuff

Notes on assembly, the metalized side of the grating windows goes down, the metalized side of the encoder ring up, so that the lines/metalized sides on both are the closest together. Won't work otherwise
Both also have a small "9000" on it, which has to be normally readable and not mirrored when mounting.

Used cheap transparent chinese silicone adhesive for assembly (Kafuter 705 from ebay). That stuff isn't too strong so that it's possible to separate things even after it has cured, in case you do stupid things, like me .

Metal piece that holds the encoder ring was then mounted to table and centered with dial indicator. After that, 705 silicone on top of that ring and carefully put encoder ring on top, gets also centered with dial indicator and checked a couple of times while curing.

For windows, I aligned them by feel, most important thing is that the windows are later parallel to the encoder ring and centered on the optics.

Then put everything back together and solder the wires back on. Use something to keep the worm wheel in place (as mentioned above), so that it doesn't get pulled down, I used a couple of pieces of scrap metal for that.

9. Aluminum
Join Date
Oct 2016
Country
SWITZERLAND
Posts
55
Post Thanks / Like
Likes (Given)
16
16
Next is alignment of the optics. I didn't care for the miniature allen screws below the lamps, I moved them out of the optical path.

Disable the other lamp with DIP switch and use 2 differential amplifiers and a scope in X/Y mode (I'm a classic Tekronix freak, so I used 7854 with 2x 7A22, adjustable filters in 7A22 are great to get a cleaner signal) to look at one set of photocell signals (the one you want to align). Soldered small wire loops into the connection box to make probing more easy.

With DIP switch 1/2 disabled (reference signals), you get 90° quadrature signal between grey/blue and yellow/pink. red/green is 45° to both other signals.

Now align window distance and tilt, the lower photocell PCB and the lamp on top until you get the best circle possible on scope (90° Lissajous pattern). Damn fiddle job, btw.

Next thing is optical encoder ring runout. Outside of glass ring isn't concentric with optical lines, so the scope is needed again. Enable both lamps with DIP switch. Both photocell assembly signals are added/averaged together. If the encoder ring has any runout, the velocity of the encoder lines is different on each photocell assembly and you'll get interference between both photocell assemblies. No runout means that the movement of lines below both photocell sensors is synchronized/the same. This drawing should make it more understandable:

Interference is visible as Lissajous ring that gets big and small to a point, or as dual 90° quadrature sine wave that varies in amplitude. Dialog4 axis counter looses track of lines in that situation => wrong reading with both lamps enabled.

Keep a dial indicator for sanity check and try to get a circular pattern on scope while turning the table. Slightly loosen the screws that hold the metal piece with encoder ring. I used a screwdriver to slightly pry the metal piece that holds the encoder ring into concentricity. Basically, if you have runout and get little little interference , the center of runout is in center between both photocell assemblies. Lots of interference means that the center of runout is to the left or right.

I used a metal punch to make 4 small marks on the top of the worm wheel piece as reference for alignment and then wrote down the runout of the dial indicator on paper once I found an good solution for runout. And then used the dial indicator to get back and improve upon that solution. With perfect runout, you should have a circle or a point that doesn't change in size for an entire table rotation.

Final thing is to use the small excentric slot screw on top right of the encoder assembly to align both photocell assemblies for max signal, i.e so that both photocell are synchronized on the encoder lines and see a line/slit at the same time. Finished!!!!

Check encoder with probes removed and put table back on, don't forget to put the big O-ring seal and some good grease in between.

10. I am impressed that you were able to get this working correctly. Kudos!

I was surprised to learn that the two heads have their signals added together. Is that also the case for the reference marks?

The screws that you removed (0.7mm hex head, as I recall) are there to balance the circle and set the amplitude.

I am curious, how did you set up the scope to act as a transimpedance amplifier (current to voltage conversion)?

Cheers,
Bruce

11. Aluminum
Join Date
Oct 2016
Country
SWITZERLAND
Posts
55
Post Thanks / Like
Likes (Given)
16
16
No idea whether reference signals are also added together.

DIP switch 1/2 seems to enable/disable reference signals. Could measure the reference spike between some cables, when enabled. But since there are only 6 signal cables (9 cables total, 2 cables for lamp power, 1 for shielding and 3x2 differential signal pairs), some other signal gets deleted with reference signal enabled. Dialog4 didn't like reference signals enabled (counted weirdly), so I disabled them again and did no further investigation.

Most HH EXE interpolators seem to simply use a resistor to convert photo cell current into sine voltage, page 9:
http://stanki-used.com.ua/image/PASS...enhain_EXE.pdf
So you can just power up Dialog4 and measure quadrature voltages at the connection box.

Of course, if HH had used a real transimpedance amplifier input on their EXE boxes, you couldn't simply measure any voltages, since always 0V and only variable current present on cables.

Then it probably would be easiest to disconnect big connector at servo drive and supply 5V for lamps with your favourite power supply and use some suitable resistors to convert photocell current to voltage yourself.

Open circuit voltage of photocells is something like 600mV (you want to stay much below that, for good linearity) and output current should be about 11microamps. So something like a 10kOhm resistor is needed to get an about +/-100mV (200mVpp) signal from photocell current.

12. I originally tried to monitor the current from the photocells as you suggest, monitoring the voltage across a ~10k resistor. But it didn't work because the impedance of the photocells changes drastically as a function of the current.

The best documentation for the 11uA signals that I found was pages 115-120 of the Heidenhain Users manual for the PWM9 Diagnostics Kit. Here is their recommended input circuit:

This requires an op-amp that makes the circuit (a transimpedance amplifier) insensitive to the impedance of the source.

The nine wires that you describe are indeed a shield, power +- for the lamp, and then three signal pairs. One pair is for the zero-phase sinusoidal signal, one pair is for the 90-degree phase sinusoidal signal, and the last pair is for the reference mark signal.

13. Aluminum
Join Date
Oct 2016
Country
SWITZERLAND
Posts
55
Post Thanks / Like
Likes (Given)
16
16
Disabled one lamp.

With 2 lamps enabled, you align it, mount the table and wait a couple of hours. Then temperature in the shed changes a couple of degrees or something and that thing is out of alignment again. One of the sensors only needs to shift like 0.01mm or the bearing inside to have a similar inaccuracy to get that thing to count wrongly.

I also now believe that Deckel cheap'd out on signal evaluation electronics for runout compensation. Both sensor signals need to be evaluated independently, like feed both sensors to an independent EXE610 and then digitally calculate the average count value of both EXE610 output signals for display. Like (EXE610[1] + EXE610[2]) / 2

Found that info on page 22:

What Deckel did, is connect the photocells of both reading heads together and directly feed them into one EXE610.

Might be fine as long as the table is new, but as soon as it wears and the bearing runout or something gets worse and the sensor signals are like (1+(-1))/2, the EXE box has 0 signal and looses count.

14. Fantastic work of digging out that reference. Finally this explains what Heidenhain was thinking! Well done! But is an EIB1500 the same type of interpolation device as an EXE610??

15. Aluminum
Join Date
Oct 2016
Country
SWITZERLAND
Posts
55
Post Thanks / Like
Likes (Given)
16
16
EIB1500 unfortunately has a completely different interface, 1Vss input and some machine specific (digital) serial output.

2x EXE610 and some DIY count averaging circuit for 2x EXE610 TTL output signals (CPLD / small FPGA, or the dirty way with microcontroller) probably would be the easiest way to go.

But since there is only 1 set of signal wires (would require lots of rewiring) and I'm not that good with digital electronics (I'm more an analog guy), I think I just leave it as is, with 1 lamp switched off.

Interestingly, the big connector for table DRO, below X servo drive and the wiring box in the photo above has 14pins, enough for a complete set of signals of both reading heads.

So it might very well be that Deckel originally had something like that installed and later cheap'd out on it.

#### Posting Permissions

• You may not post new threads
• You may not post replies
• You may not post attachments
• You may not edit your posts
•