[nicad]

I don't have a scope available to verify, but I have a couple of smaller NEMA23 servo motors With HEDS 9100 500cpt encoders that would lose position while running if the motor's power cable and shielded encoder cable were next to each other (4' long cables). I added the 1uF capacitor across the power pins on the 9100, and no more position loss. Thanks for the recommendation Mariss!

[Mariss Freimanis]

1) The cable length doesn't make the middle picture look the way it does. Noise on the encoder +5V pin makes it do that. I intentionally coupled 20kHz 100nS noise pulses to the +5V pin; when the noise amplitude reached 500mV the encoder went berserk.

2) The 'notch' in the yellow trace is caused by the blue trace signal going from 5V to 0V. The wires in the 25' cable have capacitance between them. The longer the cable the greater the capacitance. The signal change on the blue trace gets capacitively coupled to the yellow trace producing a notch that is R * C seconds long. R is 1K at a logic 1 and 20 Ohms at a logic 0. That is why you don't see a notch on the blue trace when the yellow signal goes from 5V to 0V.

3) Ceramic capacitors have extremely low ESR (equivalent series resistance) and minimal series inductance. These are the qualities needed to bypass noise and render it harmless. If a 1uF MLCC (multi layer ceramic capacitor) is used then it is sufficient by itself. Otherwise a it has to be paralleled with an aluminum electrolytic if the MLCC is much smaller (say 100nF like the 104 you mention).

Mariss

[Mariss Freimanis]

nicad,

Noise pulses have a distribution of amplitude and width. The drive is the source of noise pulses and they increase in amplitude as load on the motor increases. That is why some people erroneously said the G320 isn't good above 4A. When you have something that is on the verge of not working, it is the occasional rare high-amplitude noise pulse that injects an error count into the system.

A 1uF MLCC completely eliminates the encoder's sensitivity to noise. It is an absolute necessity because the HEDS-9100 doesn't have any bypass capacitors at all. The third picture in the series shows the effectiveness of bypassing. Note the 100mv 'fuzz' riding on the waveform. That is noise that has been rendered harmless in both amplitude and rise/fall time. The capacitor integrates (in the calculus meaning of the word 'integral') the noise.

Mariss

[BobWarfield]

This last weekend I decided to install the bypass caps Mariss recommends. I'm running Homeshopcnc servos, which have the HEDS encoders.

There have been people using the same setup who say they're having no problem whatsoever, but it seemed like cheap insurance, so I went ahead and did it.

3 allen head bolts will remove the access cover on the servos:



You can see there is a little wiring harness in there. It's dead easy to just bend the leads of the tiny little capacitor to piggyback into place:



I bypassed pins 1 (Gnd) and 4 (+5V). With the cap in place, just put the harness back onto the encoder. You don't need to remove the assembly as I did to get a better picture, just unplug, insert cap leads, replug, and button it up.

Total cost was about $2 as Mariss suggested. Digikey was only too willing even for the small order to send me the parts right away. Took about 10 minutes of my time to plug them in.

What was the result?

The servos were immediately quieter at idle and they needed a little retuning, not much, and I just touched them up by ear. However, I got a couple of immediate benefits. First, some odd little glitches I used to get when jogging (especially small jogs, such as when measuring backlash) just went away. The system was much smoother. Second, I've been fiddling with a part trying to tune up my accuracy and this made a big difference there too.

It was pretty clear that I had been getting noise on these encoders, though I never did look at them via o-scope or anything.

This mod is so easy, I highly recommend it just as noise "insurance".

Cheers,

BW

[Hellpit]

I had the same issue until last night.

After two nights of bashing my head into a wall, and swapping out everything, the last thing that I could account was the position of the G320 in relation to the others.

In my installation, I have 3 G320's installed right next to each other, and the fourth 1" away. The drive that was sandwiched between the other 2 was constantly jumping and loosing position. All the encoder wires were segregated from the drive wires. It seems that the other 2 drives on either side of it puke loads of EMF into the center drive glitching it out. (My guess is the main power wires are then too close to the encoder wires on the next drive)

I would pull the center drive away from the other 2, and there was no more issue.

I installed the cap today, and no more problems. Thanks!

As a note to others: Mount your drives at least 1" apart to avoid this issue. I hope this helps somone.

[Larken]

This last weekend I decided to install the bypass caps Mariss recommends. I'm running Homeshopcnc servos, which have the HEDS encoders.

There have been people using the same setup who say they're having no problem whatsoever, but it seemed like cheap insurance, so I went ahead and did it.

3 allen head bolts will remove the access cover on the servos:



You can see there is a little wiring harness in there. It's dead easy to just bend the leads of the tiny little capacitor to piggyback into place:

Cheers,
BW

Bob , did you make that Encoder housing for the motor? Does it hold on by the 2 4-40 screw or is the other screws ?
Nice design ! I need one for our wood routers to keep the dust out.

Larry K

[BobWarfield]

Bob , did you make that Encoder housing for the motor? Does it hold on by the 2 4-40 screw or is the other screws ?
Nice design ! I need one for our wood routers to keep the dust out.

Larry K

Larry, homeshopcnc makes those housings.

Cheers,

BW

[LUCKY13]

I just wanted to follow up on this with my setup. All though I had to let things set for a long time before I got to finish this project I now have my system ready to bolt onto the mill.


I keep having a intermiting problem and after adding the caps per this thread it totally cured the problem. If you have these encoders dont even try and run without this mod, I could see the cure on the scope and recreate the problem by removing the caps. What was really strange if one day the system would be stable and the next it would start faulting out of no where. I guess resistance of electronics can be effected by many things we dont think of and these differences from one day to the next would allow this system to work some times but not others.

ANyway its a 100% fix, and even if your system is working this will make it work better/smoother, and probably even allow a retune to a better level of performance.


Jess

[robhrzic]

The U.S. Digital single ended encoders (HEDS technology) I purchased simply could not drive even my shortest cable which is less than 4-feet in length.

One option for those with this type of encoder is to purchase the line driver, cable and receiver. See attachment.

Randy

So, is it better to use this line driver setup, get the EM1 module to replace the original HEDS unit, or both?

Both of these will work with the G320 drivers, correct?

[robhrzic]

Or, just scrap the US Digital encoser and go with the AMT102-V encoder as Mariss had suggested previously. I know that this is the cheaper $$ option, but which would be preferred for longevity and reliability?

[fatal-exception]

I just thought I would chime in too and say that I did the same to the Z axis encoder and it totally fixed the problems I was having with it losing position. I think that it's happening on the z and not the X/Y because the encoder cable is quite a bit longer.

Before this fix, I was losing up to .001 per z move, which renders the machine pretty much useless for the stuff I do. Today I countersunk 250 holes in a sheet of plexiglass and they are all the same depth! Woohoo!

I think I will put the caps on the x/y and rotary just as a precautionary measure.

Thanks to all here for the information.

Paul

[LUCKY13]

So, is it better to use this line driver setup, get the EM1 module to replace the original HEDS unit, or both?

Both of these will work with the G320 drivers, correct?

No just use the HEDS and add the capasitor to the cable as shown in Bob Warfields post. No line driver needed, no need to relace encoders.

Now he just plugged his in to the connector, I soldied mine into the cable. It was a 100% fix and I am running cables longer than I should be.


You can see the little hump in the wiring end where I heatsrinked over the soldered in capasitor to protect it and keep it from moving around or shorting. Real simple, real cheap. Solder the capasitor with one leg on the red/hot wire, and the other on the ground/black wire. Done deal, runs very smooth with no more problems.

FatalException,
yes I would add them to axis that where not even having problems. Like Mariss said the caps should have been inistalled in the encoder from the factory. IDK, maybe they where hoping to sale some line drivers


It wasnt a maybe I hope it helped thing, I could create the problem, hook the cap up and fix it, unhook it and the problem show right back up.



Jess

[goldenfab]

I just finished converting a knee mill to CNC controls using a KFLOP motion control board, G320X drives, KL34-180-72 servos, 72V power supply, and HEDS-5640#A06 encoders. Everything seemed find until I ran my first program and noticed I am loosing position on all 3 axis. I did the 1uf capacitor and 1k resistors mod however I used electrolytic capacitor instead of ceramic because that is all I had in the shop at the time. I'm going to get some ceramic ones on order. Otherwise everything seems to operate ok. I tuned the drives by ear, some buzzing but overall seem ok (except for loosing position). I could not figure out how to hook up my scope. None of the signals look rlike Gecko says. Either something is wrong with my old scope or I don't know what I am doing...

I used shielded cable (computer serial cable) for the encoder wires. I don't have the cable shields connected to ground because I did not get around to it and it seemed to work ok without. After seeing that I'm now missing steps I connected one end of the shield to earth ground but surprisingly now the servo will not engage, it just faults out immediately and won't stay engaged. I'm completely puzzled. Ideas to try are welcomed.

The servo (ground wire not DC "-") and all my AC devices are earth grounded to a central point. Should the power supply "-" and encoder "-" be earth grounded as well?

[Larken]

Those encoders are fine (but differential would be better, with twisted pair wire. ). Don't go with those AMT's
Since your loosing on all 3 axis...Maybe your loosing steps on the incoming Step/dir. Whats the pulse width that the Kflop is outputing ? Is the Dir setup time within spec ?
( On some servo drives you could read the internal counters and diagnose this easily } .

Also how close are the encoder cable to the power cables ? keep them apart at least 2"

Larry K

[goldenfab]

Thanks for your input Larry. I think you are right, I was loosing steps on the incoming step to the drive.

From the gecko website:

When using COMMON connected to GND:
The minimum specified STEP logic '1' time is 1 uS.
The minimum specified STEP logic '0' time is 2.5 uS.
Use narrow logic '1' step pulses.
The STEP input cannot be driven with an open-collector driver.
3.3V and 5V logic level drivers can be used.

When using COMMON connected to +5VDC:
The minimum specified STEP logic '1' time is 2.5 uS.
The minimum specified STEP logic '0' time is 1 uS.
Use narrow logic '0' step pulses.
The STEP input can be driven with an open-collector driver.
The STEP input cannot be driven with 3.3V logic unless COMMON is connected to 3.3VDC.

The pulse time I am outputting with the KFLOP is 2.5us. I'm assuming this corresponds to the STEP logic '1' for the gecko drive. I had COMMON on the gecko drive connected to +5VDC when I was having trouble. I just now switched COMMON to GND and am not loosing steps as far as I can tell. I should be up and running now. Grounding the encoder cable shielding still keeps the servo from engaging. I'll just keep it un-grounded for now. There is a voltage potential between the cable shield and earth ground of a few mV. I think this is normal but how grounding it makes things go bad has me completely stumped at the moment.

[Larken]

only ground the shield on 1 end... to chassis gnd, not logic gnd.