Snap Amp Noise affecting USB connection

[TomKerekes]

Hi AZ,

The Kflop has version 1.3 silkscreened on it and I purchased it July of 2019. I also have another k flop that is at least 2 years old with the same silk screen.

ok good those have the Reset filter capacitor added.

The Kflop software is 4.34.

Thanks

I have a hard time seeing anything but the stepper setup causing the problem because if the Stepper DC bus is disconnected the problem goes away. Both servos running will run for hours. The stepper system is so compact and there is nothing there but SnapAmp, Kflop motor and 2 feet of motor wire that is no where near any other wiring. No feedback wiring, no shielding, No grounding. It seems that there is no external path to generate a reset so it must be coming from within the boards.

I'm not sure I agree. No shielding and no earth grounding isn't ideal. You might verify the stepper motor frame is isolated from the coils and then try earth grounding it away from KFLOP. The Motor power supply wiring can be a noise source also.

If I install a filter cap on one of the ribbon cable connectors RESET lines, would this same capacitance / filtering be "seen" on all the other reset lines to the other connectors or would I have to install a filter at each connector??

All the resets are connected together but that may not matter for high frequency noise. KFLOP has a 1000pF ceramic capacitor C37 near JP4 placed in a manner that the Resets from JP4, JP5, and JP7 should pass through.

I will also try running it without the 50 pin breakout. As you can see in the sketch, the only connectors on the K flop are to the Konnect and the SnapAmps. I'm going to completely remove one of the SnapAmps to see what happens if I run one SnapAmp with the stepper.

Good ideas.

I ran a system with two steppers on a Snap Amp about 6 years ago and never saw this problem.

That would point to something being different here. Grounding, Supplies, 5V?

Good luck.

[ZooBDoo]

OK, I found a problem with one of the servo feedback cables, It has two shields, an internal and an external shield. The internal one is supposed to be connected to 5v common, and the external one is chassis ground earth ground. They were shorted together somewhere within the cable. I removed that motor from the equation, removed all those cable connections etc. Same problem with only one servo and one stepper.

I ran a number of tests and I've attached some oscilloscope photos of what I think may be causing the disconnects.
Each photo shows the 5 v noise measured at the 50 pin connector breakout board with 6 inches of ribbon cable with the ground of the scope at the Kflop ground screw where the two SnapAmp ground wires are connected. The scope is "floating". The amplitude gradient is 0.5 v per gradicle.

I changed the ground wires between the SnapAmp and the K flop to 1/4 in wide flat braid and that reduced the amplitude of the ground wire noise between the K flop and the SnapAmps but the resets continued. Ground wire noise was measured with the scope probe ground at Kflop end of the ground wire and the probe tip at the other end.

Each photo is with only one axis enabled and powered.
When both axes are enabled and powered the two photos are superimposed with the noisiest pulse of the servo phasing back and forth, left and right, and increasing / decreasing amplitude as the motor rotates. The stepper pulses stay where they are but the noisiest one is increasing and decreasing in amplitude as the stepper rotates. The other three stepper pulses don't change in amplitude much. I've swapped the stepper motor wiring to see if the noisiest pulse changes but it stays in the same place in the sequence.

When either the stepper or servo runs by itself the noise pulses are still there but they don't reset the KFlop. When I run them together I can see the servo noisy pulse phasing side to side and overlapping the stepper noisy pulse many times but after about 10 to 15 minutes of running, the Kflop resets.

I have a theory that the noisiest stepper and servo pulses are caused by overlap between the firing of the bridge legs. More specifically, where one bridge leg is fired but the other hasn't stopped conducting yet so a high current pulse, and associated noise, is generated. It seems that way to me because the noisy pulses are always the same ones in the firing sequence. None of the other pulses get anywhere near that amplitude.

Is there a way to increase the dead band "wait time" between firing pulses so we give the current in one leg more time to stop conducting before we fire the other leg? IF for no other reason but to prove/disprove the theory.

Or, on another path, is there a way to further filter the Kflop RESET line?

Is the SnapAmp logic optically isolated from the power section?

AZ

[TomKerekes]

Hi AZ,

OK, I found a problem with one of the servo feedback cables, It has two shields, an internal and an external shield. The internal one is supposed to be connected to 5v common, and the external one is chassis ground earth ground. They were shorted together somewhere within the cable. I removed that motor from the equation, removed all those cable connections etc. Same problem with only one servo and one stepper.

I'm somewhat confused. I thought you were going to remove the 50 pin ribbon cable and breakout board? I thought the problem occured with only one Stepper motor. But it seems it needs a Stepper and Servo? I suppose you need the breakout board for the Servo? I thought you didn't have any earth GNDs? Doesn't sound like a good idea to have the two shields tied to Earth and DC GNDs that might couple them together.

Can you run a test without the 50pin ribbon cable? Maybe with the servo motor just homing forever cycling currents open loop?

Or does the problem occur with 2 steppers connected and no 50 pin ribbon cable?

Have you checked for all other combinations of shorts between motor wires, shields, GNDs?

You might scope the Reset# line. The Reset# line is an input and an output. If pulsed (glitched) low it should remain low for a significant time (> 1/8th second). Can you setup the scope to trigger on this?

I have a theory that the noisiest stepper and servo pulses are caused by overlap between the firing of the bridge legs. More specifically, where one bridge leg is fired but the other hasn't stopped conducting yet so a high current pulse, and associated noise, is generated. It seems that way to me because the noisy pulses are always the same ones in the firing sequence. None of the other pulses get anywhere near that amplitude.

Actually even with all the switching the motor currents should be relatively low and constant. The motors have high inductance so the current changes relatively slowly like on the millisecond time scale. The voltage switches at high frequency. Parasitic capacitance in the wiring/motor might cause high frequency currents (micro/nano seconds).

Is there a way to increase the dead band "wait time" between firing pulses so we give the current in one leg more time to stop conducting before we fire the other leg? IF for no other reason but to prove/disprove the theory.

I can't think of a way to do this. The PWM basically varies randomly based on encoder error for the servo and current errors for the stepper.

Is the SnapAmp logic optically isolated from the power section?

No.

Or, on another path, is there a way to further filter the Kflop RESET line?

I can't really think of any. As I stated we have a 1000pF capacitor well placed on the reset line. You might try a larger/smaller. One way to add a capacitor is like this:

[ZooBDoo]

I ran the test with the stepper and servo running open loop with the 50 pin disconnected and had the same result after about 10 minutes.
I installed a 10000pf cap on the K flop 26 pin on the reset line still running open loop on both motors. Same result

I replaced the K flop board with one that was purchased about 3 or 4 years ago and that is still running after 8 hours with the stepper and closed loop servo with the 10000 pf cap on the 26 pin reset line. I didn't run it without the cap yet but will do that.

Can I send you the Kflop in question so you can see if you can tell where the problem with the board is?

AZ

[ZooBDoo]

Update:
I removed the additional external cap from the RESET line and it's still running with both stepper and closed loop servo after 3 hours.
I expect that this will continue running OK.

Now I have to get a good cable for the other servo and get that connected to check the operation with all the motors running.

I couldn't trigger on the RESET line with the scope because I found out my scope probe was faulty.

Thank you for your help with this.

AZ

[TomKerekes]

Hi AZ,

So this is still working now with the different KFLOP board? Indicating the other KFLOP has an issue?

Not exactly sure what we could test here if we can't duplicate the problem.

Are you sure nothing else was changed when swapping the KFLOPs?

[ZooBDoo]

I don't think there is much that could change when swapping out the K flop but I've been surprised before. I've removed and reinstalled the same board a number of times when checking the installation and still had the same problem. Changing to the other board only amounts to umplugging a few ribbon cables, power connector and USB cable.

I suppose the best way to make sure the Kflop board is the issue is to reinstall it and see if the problem comes back. I'll try that later on today.

AZ

[ZooBDoo]

I re-installed the questionable K flop and it RESET in 30 min of running the one stepper and one servo in closed loop.
I re-installed the "good" Kflop and it has been running over 8 hours with no trouble. It does look like the issue is repeatable.
I would like to return this item for an exchange if I could.
I don't know if you'd be able to duplicate the problem there because of the possible wiring differences but it'd be interesting to know if you see any component differences in the board population or assembly.

Also, this forum thread is quite long. Do you think I should edit it to eliminate the "goose chasing" and make it more useable for other viewers?

Thanks again for your insight and suggestions in diagnosing this.

AZ

[TomKerekes]

Hi AZ,

Thanks for taking the time to swap the boards again to further narrow this down.

I understand you contacted our support for an exchange. We will try to do an analysis.

It probably isn't necessary to re-edit the Thread unless you wish to. Maybe add a summary?