[durundal]

Hi Tom,

I am trying to resolve an issue with encoder motion to the same count not returning to the same physical location. I have a DMM DYN4 servo drive connected to the Kanalog quadrature differential inputs, with an individual twisted shielded pair for each A/B/Z pair, with the shields tied to Kanalog ground. When I move the axis against a dial indicator (the motor is connected to a zero backlash ballscrew) and drive the axis via software jog buttons in KmotionCNC + and - several times and then command it back to the starting location, the dial indicator shows a few thousandths of an inch off, at seemingly random amounts (something like a max of 0.004” error over a travel distance of 1-2”, but it is cumulative with additional distance traveled). Cutting down the accelleration and velocity in the trajectory planner did not seem to help.


The DYN4 motor has an absolute encoder that sends digital data to the drive, which then converts it to quadrature incremental signals that get sent to the Kanalog. If I note the absolute encoder position via it’s USB interface to the DYN4 it appears to return to the the same count it started at once I manually adjust the servo position until the dial indicator is zeroed. This makes me think that either the DYN is not sending out the correct number of quadrature pulses to the kanalog, or that they are being attenuated or not being counted somehow. Any ideas on how to troubleshoot this? The floating DYN4 encoder ground is tied to the Kanalog ground, which has a single line to protective earth through the PC power supply I am using (Kflop USB has an isolator on it and the 24V power supply is floating output, single ground verified with a multimeter and unplugging the power supply cable), and the quadrature channels are as below (with the Z axis used for this testing)

Thanks,
Andy

//// JP1 Differential inputs
#define YAXISMOTOR_QA 0 // y axis motor A
#define YAXISMOTOR_QB 1 // y axis motor B InputChan0
#define YAXISINDEX_Q 2 // y index
#define XAXIS1INDEX_Q 3 // x index slaved InputChan1
#define XAXIS1MOTOR_QA 4 // x axis motor slaved A
#define XAXIS1MOTOR_QB 5 // x axis motor slaved B InputChan2
#define XAXIS2MOTOR_QA 6 // x axis motor A
#define XAXIS2MOTOR_QB 7 // x axis motor B InputChan3
//// JP2 differential inputs
#define XAXIS2INDEX_Q 36 // x index
#define ZAXISINDEX_Q 37 // z index InputChan4
#define ZAXISMOTOR_QA 38 // z axis motor A
#define ZAXISMOTOR_QB 39 // z axis motor B InputChan5
// #define unused 40
// #define unused 41 InputChan6
#define MPG_QA 42 // mpg pendant A quadrature bit
#define MPG_QB 43 // mpg pendant B quadrature bit InputChan7

[TomKerekes]

Hi Andy,

I can't see anything you are doing incorrectly.

You might try moving manually with all motor supplies off to see if it still accumulates counts. That would provide clues whether it is noise or something about the way the Drive outputs signals.

You might try lots of short direction reversals to try to determine if it could be a lost count on each or most reversals.

You might check the A+ A- B+ B- voltage levels at Kanalog. They should all toggle between about 0.5V and 3V. Sometimes they operate marginally correct with a broken wire and floating signal.

Kanalog has only moderate termination on the encoder inputs of 470 ohms. Ideally the Termination should be ~120 ohms. So you might consider paralleling 160~200 ohms across each + to - input.

KFLOP has digital filtering to reject noise glitches. KFLOP samples the A B signals at 16.67 MHz and requires them to be stable for 7 samples (by default) before considering the signal to have a valid new changed state. 7 samples = 0.42us. You might try increasing the filter value. Numbers higher than 14 will start reducing the max count rate below 1Mhz. The setting is global to all Encoders and can be changed by executing the statement below with the number 7 changed to a larger value (255 max):

FPGAW(ENC_NOISE_FILTER_ADD)=7;

Regards

[durundal]

Hi Tom,

Replies in red below.

Thanks for the help,
Andy

You might try moving manually with all motor supplies off to see if it still accumulates counts. That would provide clues whether it is noise or something about the way the Drive outputs signals.
The drives need to be powered to read the absolute encoder and convert it to the incremental quadrature outputs unfortunately

You might try lots of short direction reversals to try to determine if it could be a lost count on each or most reversals
Appears to be a function of distance traveled, dithering 0.001” still results in offsets but takes a long time


You might check the A+ A- B+ B- voltage levels at Kanalog. They should all toggle between about 0.5V and 3V. Sometimes they operate marginally correct with a broken wire and floating signal.
Measuring across the differential with a multimeter seems to show +2.6 or -2.6V as it moves, but not sure if I am measuring pulses or averages or what. I have ordered an oscilloscope to be able to get insight into this sort of thing...

Kanalog has only moderate termination on the encoder inputs of 470 ohms. Ideally the Termination should be ~120 ohms. So you might consider paralleling 160~200 ohms across each + to - input.
Added 200 ohm resistors, no apparent effect.

KFLOP has digital filtering to reject noise glitches. KFLOP samples the A B signals at 16.67 MHz and requires them to be stable for 7 samples (by default) before considering the signal to have a valid new changed state. 7 samples = 0.42us. You might try increasing the filter value. Numbers higher than 14 will start reducing the max count rate below 1Mhz. The setting is global to all Encoders and can be changed by executing the statement below with the number 7 changed to a larger value (255 max):

FPGAW(ENC_NOISE_FILTER_ADD)=7;
Success! Changing this to a value of 12 seemed to do the trick, a value of 10 reduced the drift significantly but did not eliminate it. If my motor max RPM is 5000 and I get 8192 pulses per rev = 682kHz does this mean I am fine just leaving this filter value higher and declaring success? Similarly if I raise it to 14 (or theoretically up to 24 if it was right on the edge) does that get me additional margin against noise?

[TomKerekes]

Hi Andy,

Regarding voltages: with a multi-meter you would want to move some random distance and stop. The voltage should then be approximately +2.6V or -2.6V. It seems they are connected correctly.

Regarding digital filter samples: it’s hard to say what would be the optimal value without knowing the noise signature. Too large of a value might cause a valid state change to be missed with certain glitch spacing. Maybe 14. I assume you meant to say counts/rev?

Regards

[durundal]

Hi Tom,

I reworked the wiring to move the AC lines further away physically from the Kanalog and DC lines, this seems to help with the overall noise. I discovered that the quadrature input for channel 4 (that was losing pulses) doesn’t seem to function with any inputs after reworking (tested by moving a working cable set from another axis to that channel), I have unused quadrature channels and moved those lines for the time being but am concerned that it might have fried the TI 26LS32AC chip that is reading it? I am not sure what I would have done to fry it when reconfiguring the layout, I am pretty sure I didn't do anything like put direct voltages on JP1/JP2 or something like that. The oscilloscope shows an odd change in voltage (the ~1V changes in the middle of a pulse) when probing these channels, in the middle of each differential pair pulse it drops slightly (ungrounded probes in yellow and cyan, math function of A-B in purple). The clean signal (blue and magenta) is from the new channel that doesn't lose counts. I also noticed that another axis has the same lost count problem to a lesser degree and also shows the same ~1V level shift on the encoder and index lines. These readings are with the main servo drive power disconnected, only the servo logic power lines connected and manually turning the shafts.

Thanks,
Andy

[TomKerekes]

Hi Andy,

I would guess that the Step is occurring when the other A or B channel is switching. Some sort of cross connection between A and B. It might not even be relevant without a proper GND reference.

I can’t really think of anything to add other than the obvious to check for miss wiring, shorts, bad connections, etc. including the connections between KFLOP and Kanalog.

Make sure you aren’t configuring the KFLOP encoder Inputs as outputs. Configured as Step/Dir pins will also drive as Outputs.

It should be easy to test Kanalog. It just the differential receiver with 470 Ohm termination.

Regards

[durundal]

Hi Tom,

Measuring the inputs (cables all unplugged):

0A 470 ohms
0B 469
1A 468
1B 470
2A 415 <---
2B 470
3A 470
3B 470
4A 157 <---
4B 467
5A 470
5B 300 <---
6A 467
6B 470
7A 470
7B 467

It doesn't look like there is any debris on the Kanalog board to account for the lower resistances, do these lower values imply the silicon has broken down? Are channels 0 & 1 on the top receiver, 2 & 3 middle upper, 4 & 5 on middle lower, and 6 & 7 on the bottom? Are the unpopulated encoder connectors on the PCB worth interrogating in some way?

Thanks,
Andy

[TomKerekes]

Hi Andy,

do these lower values imply the silicon has broken down?

I suppose it might, but hard to say how the connected receiver would behave without power.

Are channels 0 & 1 on the top receiver, 2 & 3 middle upper, 4 & 5 on middle lower, and 6 & 7 on the bottom?

Yes the Receiver ICs are in order:

U1 handles channels 0,1
U2 handles channels 2,3
U3 handles channels 4,5
U4 handles channels 6,7

Are the unpopulated encoder connectors on the PCB worth interrogating in some way?

No.

To test in detail you might test every one of Kanalog's 16 differential receivers which are connected to 16 KFLOP IO pins (IO 0-7 and 36-43). Use something like a 1.5V battery and 470ohm series resistor to apply +0.75V to each differential + - input pair. The pin to KFLOP should go high. Reversing the battery should apply -0.75V and cause the pin to KFLOP to go low.

Regards

[durundal]

Hi Tom,

Performing the battery test, all cables disconnected from servos and servos unpowered (only 24V and 5V power supplies powered on in the cabinet)

definitions:
ok - moves to on when matching + - pairs, returns to off when battery removed
latches - stays high or low depending on which battery polarity was applied
no response - does not change state regardless of applied polarity

0A bit 0 - ok
0B bit 1 - ok
1A bit 2 - ok
1B bit 3- no response
2A bit 4- ok
2B bit 5 - latches
3A bit 6 - bit flips autonomously while battery held in the "on" polarity (no flips in negative or when battery removed)
3B bit 7 - latches
4A bit 36 - ok
4B bit 37 - no response
5A bit 38 - no response
5B bit 39 - no response
6A bit 40 - ok
6B bit 41 - ok
7A bit 42 - latches
7B bit 43 - ok

I don't know if this is expected or not but the pins default to this after powering on the Kflop but without running any user programs:

0A on
0B on
1A off
1B off
2A on
2B on
3A off
3B on
4A on
4B off
5A off
5B off
6A on
6B on
7A off
7B off

Thanks,
Andy

[TomKerekes]

Hi Andy,

It would seem several of the IC's have been damaged. "latches" are probably ok and normal as the IC's have hysteresis. The initial power up state with floating inputs is not likely to be important.

I assume you were just looking at the KFLOP IO on the Digital IO screen and not actually measuring the voltage at the KFLOP pins.

You should be aware that the Kanalog Terminal Screw Heads can't be used as a reliable connection or test point when the screws are loose.

Those IC's are normally very robust and can handle +/-25V and significant ESD without damage. You might check your grounding. Maybe something like earth ground has large voltage spikes and that is somehow coupled into the encoder circuitry.

Regards

[durundal]

Hi Tom,

Yes, I was reading the digital IO screen for the status, and the screw terminals were screwed shut before taking readings. Earth ground gets some noise on it at 40kHz when the servos are enabled which I assume is induced currents from switching? I have had the board for a while now but only recently got to the point of cutting stuff which is where I noticed the lost counts, it’s possible that before the grounding was reconfigured something was not right. The servo signal ground (isolated from protective earth ground on the drives I believe) was not connected to the Kanalog ground until the cabling rework, might that have caused it? Some of the 240V lines were running in proximity to 24V lines as well.

I am not very sure of my surface mount skills, do you offer a repair service?

Thanks,
Andy

[TomKerekes]

Hi Andy,

I suppose missing grounds with things floating to high voltages might have caused it. I wouldn't expect cable cross talk would be sufficient to cause damage.

It usually isn't cost effective to repair these boards. Where are you located?

If you (or you know someone) reasonably skilled with a soldering iron you should be able to replace those ICs as they are not very fine pitched. If you don't have hot air soldering equipment you might cut each lead individually flush with the IC body with a razor blade or exact o-knife. A magnifying glass/lamp helps. Then you can remove each pin separately with a soldering iron. Clean off excess solder from pads with solder wick. Then place and solder in the new chip.

Regards

[durundal]

Hi Tom,

I have ordered some replacement chips and soldering accessories to give repair a shot. I am in Washington state.

For preventing future problems, I am trying to make my grounding as correct as possible. Currently I have the following:

- Isolated 5V power supply to Kflop JR1 5V and ground, no other connections
- Isolated 24V power supply to buttons and proximity sensors returning to opto isolated inputs and outputs, with its return tied to protective earth/chassis ground at a single point, and also to one of the Kanalog GND lug holes
- 2x SSRs controlled with the Kanalog relay drivers, sourced from the 24V supply (the SSRs control big 24V Allen Bradley contactors for main power/e-stop and a stand alone [switched loads not connected inside the cabinet] auxiliary circuit for vacuum/water pump)
- Drain lines from shielded cables running to Kanalog grounds for +/-10V ins & outs, and to the 24V return bus for others (including differential lines from the servos), shield drains connected on the Kanalog/24V return end only.
- Ground connection from Kanalog ground to isolated signal ground on DMM servo drives

The wiki says:
Kanalog has a single GND common to KFLOP, control logic, relay drivers, and Analog signals. The GND should normally be isolated from Earth GND to avoid ground loops and injecting earth GND noise into Kanalog/KFLOP DC GND. Standoff mounting holes are completely isolated with the exception of the GND lug holes marked "GND".

Questions:
Does this imply I should move the SSRs to 5V power for control (I had been thinking the higher voltage would be less likely to pick up noise but since the SSR can change at 3V maybe that doesn't matter) and have the Kanalog ground floating from chassis ground? I don't think I want to have the 24V ground floating as the 24V limit switch and e-stop circuits are fairly extensive and roaming around the machine and could be cut accidentally potentially.

Should all shield drains be moved to chassis ground?

Thanks for all your help on this,
Andy
Kanalog has a single GND common to KFLOP, control logic, relay drivers, and Analog signals. The GND should normally be isolated from Earth GND to avoid ground loops and injecting earth GND noise into Kanalog/KFLOP DC GND. Standoff mounting holes are completely isolated with the exception of the GND lug holes marked "GND".

[TomKerekes]

Hi Andy,

Grounding is complex. I'm not an expert but here are some thoughts:

- Isolated 5V power supply to Kflop JR1 5V and ground, no other connections

I think this is good. However this may not be what you actually have. You might verify with an ohm meter there is no connection to any of your other GNDs. Its usually a good idea to remove all connections between GNDs, verify the circuits are really isolated, then add any connections that you wish and where you wish them to be. Otherwise it is hard to know for sure there aren't any inadvertent connections.

Isolated 24V power supply to buttons and proximity sensors returning to opto isolated inputs and outputs, with its return tied to protective earth/chassis ground at a single point, and also to one of the Kanalog GND lug holes

Connecting to Kanalog GND lug is probably a bad idea. This defeats the opto isolation between the 24V circuitry running all over the machine and KFLOP/Kanalog GND. Also the connection from Earth GND to Kanalog GND may inject Earth GND noise directly into the KFLOP/Kanalog GND as well as possibly forming a ground loop.

2x SSRs controlled with the Kanalog relay drivers, sourced from the 24V supply (the SSRs control big 24V Allen Bradley contactors for main power/e-stop and a stand alone [switched loads not connected inside the cabinet] auxiliary circuit for vacuum/water pump)

Again if you use the same 24V power supply for the Opto isolated circuits as the KFLOP/Kanalog GND circuitry then the opto isolation is defeated. If your SSR could run from 5V instead of 24V you could use the Kanalog +5V instead.

Drain lines from shielded cables running to Kanalog grounds for +/-10V ins & outs

Probably good. Although normally recommended to connect signal shields to the receiver's GND.

and to the 24V return bus for others (including differential lines from the servos), shield drains connected on the Kanalog/24V return end only.

Probably not good. Especially for the 5V encoder signal shields. My thinking is the 24V proximity sensor circuity and shields is running all over the machine and has picked up all manner of noise. That is why all this circuitry should be completely isolated from KFLOP/Kanalog 5V signals and GND. But then all that noise is injected into the encoder shields and in turn into the encoder signals.

Questions:
Does this imply I should move the SSRs to 5V power for control (I had been thinking the higher voltage would be less likely to pick up noise but since the SSR can change at 3V maybe that doesn't matter) and have the Kanalog ground floating from chassis ground?

Yes for the above reasons

I don't think I want to have the 24V ground floating as the 24V limit switch and e-stop circuits are fairly extensive and roaming around the machine and could be cut accidentally potentially.

Not entirely clear what you mean. That might be best with a single connection to earth GND.

Should all shield drains be moved to chassis ground?

No. My thinking is to look at the purpose of the shield. Usually one of two purposes: #1 to prevent noise from being radiated. An example here might be shields around motor/spindle/solenoid wiring. These I would connect to earth/chassis GND to attempt to contain and drain the noise away from other signals. This however may make earth GND noisy. #2 to prevent noise from being received and getting into sensitive signals. An example might be encoder signals. In this case the shield should be connected to a quiet GND near the receiver.

HTH

[mmurray70]

Did you get this sorted out after? Im having some similar issues with losing position between a DMM drive and Kanalog.

[durundal]

Did you get this sorted out after? Im having some similar issues with losing position between a DMM drive and Kanalog.

Hi mmurray,

Yes, and sorry for the delayed reply, it seems to work again after replacing the chips. I ended up using a product called "ChipQuick" as recommended by the EEVBlog guy to desolder the chips. It is basically a low melting point alloy (similar to Woods Metal or cerro alloys) that you melt onto the leads and it stores enough heat to allow everything to be liquid on all points, at which point the chip can be popped off. It is expensive stuff and if you have to do more than one contiguous chip you can slide the molten blob around to the next one without re-melting more fresh material. I was careful to not overheat the pads or scrape against them as this can lift the copper off the PCB apparently. Using some braid to get any excess off before putting down the new chips and cleanup with IPA seemed to work nicely. The alternatives I found are trying to clip each lead and desolder them one by one, or use a hot air system with chip-specific blower guides to melt the leads simultaneously if you're doing this a lot or can't tolerate contamination.

I think I also realized where the fault was - I was using the relay FET switches to enable main power to the servo drives via one of the SSRs driving a big contactor, the SSR was sharing the same 24V ground that was tied to the Kanalog ground plane elsewhere, and in doing so that was creating the ground loops but only when the FET switches were enabled. I didn't notice this when probing the grounds as I was purposely avoiding enabling the main 240V relay while probing as I didn't want to accidentally short anything or have anything take off uncommanded, and could separately enable power to the drive's logic supplies. Now only SSRs running on the Kflop 5V are on those lines and everything else is moved to the opto ins and outs on a 24V supply, a dedicated isolated 5V is running the Kflop/Kanalog, and a second 5V supply tied to the 24V return and protective earth for 5V accessories (MPG, water flow meter, etc), which are being sent to isolated inputs on a Konnect board. Hopefully this avoids any future grief... many thanks to Tom for his support in troubleshooting!

Thanks,
Andy

[mmurray70]

Delayed reply? Thats pretty fast, I only posted like 20 mins ago. So did you notice any other problems with inputs and outputs when that chip was bad or was it simply random loss of motion on on axis? Where any of the other axes affected?

Everything else seems to be working fine on my machine, and I dont have any position error when spindle is stopped. Only with spindle running.

[durundal]

Delayed reply? Thats pretty fast, I only posted like 20 mins ago. So did you notice any other problems with inputs and outputs when that chip was bad or was it simply random loss of motion on on axis? Where any of the other axes affected?

Everything else seems to be working fine on my machine, and I dont have any position error when spindle is stopped. Only with spindle running.

I had meant to say what fixed it when I replaced the chips but lost it in my pile of things to do. The chip started intermittently losing counts but one channel ended up dying entirely. I tested the inputs with a AA battery and resistor to toggle them high or low to make sure they switched and some didn’t switch at all on the failed channels. I have a WJ-200 Hitachi VFD and didn’t notice any problems with the spindle running, just when the motors were moving, and simultaneous axis motions were more likely to cause it, FWIW.

[mactec54]

Delayed reply? Thats pretty fast, I only posted like 20 mins ago. So did you notice any other problems with inputs and outputs when that chip was bad or was it simply random loss of motion on on axis? Where any of the other axes affected?

Everything else seems to be working fine on my machine, and I dont have any position error when spindle is stopped. Only with spindle running.

Yours problem sounds like a noise problem, which is mostly caused by wiring, and shields not correctly terminated