JMC servo "jerk"

[mkuivamaki]

Hello,

I hope I can ask some advices here in English. I recently finished my latest GraniteCNC build where I use Edingcnc CNC720 controller and 200W JMC IHSV60 servos.
Sometimes and randomly there occurs strange "jerks" or "shocks" on motors, at least on Z and Y axis. For example I have a simple straight toolpath but sometimes servos makes a really quick "boom" and jerk what can be seen on traces on workpiece (please see attachment). This isn't repeatable so it occurs randomly. Feels like the servos are trying to compensate lost encoder pulses etc and makes these "jumps". They are not big if looking millimeters, these jumps are only few 0,1mm's at max but "boom" noise is loud and surface quality is not good. Mechanically everything is ok.

Does it sound servo parameter adjustment? Controller issue? If servo parameters, I have already quite stiff setup.

Thanks

[Kjelt]

Is it 100% reproducable in that sense that if you run the G-code again it happens exactly at the same point (delta time since start) ?

Things to already quick check :
- What is your Max. step frequency in the setup field of Eding ?
Your servo's have a max input pulse frequency of 200kHz, to be safe stay at least 10-15% below that number so try 150kHz-175kHz.

- Are your cables to the servo's shielded and at least have the shield hard grounded on one side ?

[mkuivamaki]

I cannot reproduce it at the same points. Same G-code can be better next time etc.
- Max step frequency is set down to 125kHz because I noticed at the very beginning that servos does work more stable with lower frequencies and that speed is still more than enough for me.

- I use CAT6 cables for servo signals BUT the shield is not grounded separately.

[Kjelt]

I cannot reproduce it at the same points.

So to be clear, running the same G-code twice does not produce the same results because sometimes you get these "glitches".
In that case it is not the controller, since it will reproduce almost 100% each time.

Then I suspect an induced error, it could be numerous causes:

- could be the cabling inside the electronic cabinet. If there is a VFD inside and the cables from the VFD are not grounded and shielded properly or too long distance parallel on the servo driver data cables.
- could be in the cabling towards the servo so I would at least connect the shield of the CAT6 cable hard to ground on one side.
- could also be the settings of the acceleration which are too fast, you can also tweak with those.
- could also be an powersupply issue, what diameter DC powercables do you have towards the servos, what is the Wattage of the powersupply(supplies) for the servo. You can test this to run all 4 servo's simultaneously with full speed.

Post some clear pictures of the inside of the electronic cabinet perhaps we can see something there.

[mkuivamaki]

-VFD grounded properly + Line filter (components, spindle cables and wiring instructions from Spinogy)
- Separate line filters on every PSU
- JMC power supplies: 2x600W Mean Well PSP
- Servo power cables are shielded 2x1,5mm^2

I had first major problems with those JMC power supplies, on hard accelerations and decelerations those felt some underpowered. But reduced accelerations and max velocity I got this issue solved. Also needed to add buffer capacitors to PSU terminals. Still a bit weird because I thought that these power supplies would be more than enough.

Here's one picture, a bit old when the cabinet was not 100% ready but almost. My electric cabinet by the way is not fully made of metal (I know Faraday effect would reduce emc issues) but the mounting plate is metal and grounded.

https://photos.app.goo.gl/uARoM8ViNunDTT1e6

Spindle cable goes along Z servo cables in same energy chain etc. But Y-motors are not in touch with spindle cable and they have this same jerking issue sometimes. Feels like X motor might be the only axis without this issue.

Also should I look for JMC tuning again? I heard that less stiff setup would help or adding some more filter. Sorotec also suggested to test another PC since they had also this kind of issues also. My PC is mounted behind the cabinet.

General overview of my machine:
https://photos.app.goo.gl/3f729ygi2at1Di7K9

[P1-Engineering]

Adding a capacitator made a big difference for me https://www.rocketronics.de/shop/de/kon ... r9400.html
If you have something in that range it should be sufficent.

How the VFD-spindle cable comes into your cabinet isn't ideal, you want it as short as possible and the shielding which is open hardmounted to your ground plate. Now you've created a big antenna.
To be sure I've placed the VFD in it's own cabinet, that's the easiest solution. Same for all the pneumatic relays, don't want that stuff near all the delicate electronics.


Personally after 2 years I replaced the JMC's with Delta servos because I'm looking for better performance, was never really happy with the JMC's.
For such a nice machine you built, think I saw it on Insta, I would surely consider the upgrade. With the 760 board and the new differential connection board you're able to run them at the full 400khz.

[Kjelt]

JMC power supplies: 2x600W Mean Well PSP
Also needed to add buffer capacitors to PSU terminals. Still a bit weird because I thought that these power supplies would be more than enough.

Those capacitors should be placed as close to the motor dc terminals as possible, not to the PSU terminals since that can ruin the PSU feedback loop (oscilations which can cause damage to the PSU).
The current step which occurs from a rapid acceleration can perhaps not be delivered quick enough over multiple meters of cable.
But perhaps that is what you already did.

[serum]

Adding a capacitator made a big difference for me

What difference did it make? Did it solve an issue you suffered without? Are you also using the leadshine PSU's? I'm not using leadshine PSU's, yet push it pretty hard without issues. (either with the JMC and the OMC servo's).

In the case of Mikko, I would have replaced the PSU's. We've exchanged some emails/whatsapps in the passed months, and the PSU isn't suited for the task in my opinion. (either that or the cabling is wrong) With maximum acceleration/speed the motors stall and throw a voltage related issue if memory serves well. The added capacitors shouldn't be required. Perhaps some odd resonance occurs at settling speed of the motor tripping voltage of the PSU again causing the motor to reach an over or undervoltage limit, but not enough to trip a real error.

My first guess, based on the previously stated issues, is the meanwell PSU.

If you haven't tried Kjelts tip (place the capacitors as close to the motors as possible) try that first. It could help.

[mkuivamaki]

I can try to put caps near motors (directly to motor DC terminals) and see if there is difference. I monitored Z-axis motor at different speeds and seems that problem occurs more often at slow speeds (~500-1000mm/min). In my test g-code there was just straight lines and Z-position was at constant depth.

https://photos.app.goo.gl/CVVhbkRK1ZcQUSMG7

You can see and hear those spikes on the oscilloscope. Usually it's not that bad all the time, I think I just found the "sweet spot" where this problem occurs.

[Kars-cnc]

This sounds more like a mechanical issue to me... Something with blocking balls in a Ball screw nut or in a linear rail maybe?

Anyway, since it happens while moving steady and without load I think the power supply is unlikely to be the cause.

[mkuivamaki]

It shouldn't be mechanical issue because in that video the monitored motor does not move. If those spikes would occur during travel it would be another thing.
What I think at the moment it could be EMC noise issue. I will also do some other tests:

Running test g-code with:
- Spindle off
- VFD off
- servo signal cable shield grounded (CAT6 cable)

I already tried to put ferrites around the servo signal cables and things got immediately worse. Alarm message was: "Drive position deviation exceeds set value"

EDIT: Tested without spindle and also tested completely without VFD. Grounded signal cable shield. None of these made any improvement.

What I will still test is to buy another CAT6 cable and route it directly from motor to controller (no any other cables nearby) and see does it make any improvement. Still thinking the possibility of faulty motor parameters but not sure which parameter to look first.

[hfjbuis]

What I will still test is to buy another CAT6 cable and route it directly from motor to controller (no any other cables nearby) and see does it make any improvement. Still thinking the possibility of faulty motor parameters but not sure which parameter to look first.

If you are now using a shielded CAT6 cable, you could try an unshielded one to eliminate a ground loop between motor and driver.

[benkr]

I already tried to put ferrites around the servo signal cables and things got immediately worse. Alarm message was: "Drive position deviation exceeds set value"

Fact that this makes this difference tells me that you are not control this signal in a differential way.
The input of the servomotor is normally a LED from an opto-coupler so you should have a signal current going via a wire into the LED and one coming out the LED which goes back to the control board. Those 2 wires should be twisted to each other to eliminate EMC signals from this wire to other devices AND also to eliminate EMC signal coming from other devices to disturb you control signal.
If you put a ferrite around this wire it should have no or minimal effect or gives better results.

If i look to you control cabinet i think i see more places which are not ideal wired with EMC design rules in mind, a very good example is the placement of the mains filters, the input and output wires seems to pass each in a very small distance, this is the most worse way to wire these filters, the capacitance between these wires eliminates almost all of the benefits of using the filter. Also the groundplane on which the filter is mounted should be very very very good connected to the plate on which the other electronic componnts (PSU and VFD) are mounted, not with as single wire but with a wide as possible conducting metal strip.

[mkuivamaki]

Main filters are mounted on the same plate, plate is just bended because I had no room anywhere else.

About the servo control signals, I have just connected them like few instructions have shown (Sorotec and JMC manual). CAT6 cables does have each 8 wires twisted to pairs. Each pair is foil shielded and finally the whole cable has one shield layer.

What is still weird that even for example X axis is moving simple G01 line and Z should be still, the Z motor creates these shocks during X travel. But for example in home position there is no any spikes. So does the other motors generate some emc noise which have effect to Z motor
I will still try to change Z motor parameters to autotune and have a look does it make a difference.