Jerky movements, again

Hello, I'm using latest release:
[[language]] Firmware Name: RepRapFirmware for Duet WiFi Firmware Electronics: Duet WiFi 1.0 Firmware Version: 1.16 (20161108) WiFi Server Version: 1.03 (ch fork) Web Interface Version: 1.13
And experiencing issues very close to (but not the same) that described in my old topic on reprap forums: http://forums.reprap.org/read.php?178,546995
Circles are printed with audible clicks (also this clicks sensible then you put your finger on moving tower caret). Few clicks per second (I don't know exact value, about 37 per second). I hear clicks for both first layer brim/skirt and main perimeters.
I'm using Duet WiFi beta board, 12V power supply (LED PSU, 30A), 200 steps/rev steppers and the following movement settings:
[[language]] M350 X16 Y16 Z16 E16 I1 ; Set 16x microstepping with interpolation M92 X100 Y100 Z100 ; Set axis steps/mm M201 X3000 Y3000 Z3000 E3000 ; Accelerations (mm/s^2) M203 X12000 Y12000 Z12000 E4200 ; Maximum speeds (mm/min) M566 X600 Y600 Z600 E600 ; Maximum instant speed changes mm/minute
Maximum jerk speed does not change nothing (I have tried M566 X6000 Y6000 Z6000 E6000 also without success).
Calibration settings at this moment is:
[[language]] M665 R105.83 L214.35 H239.30 B75.00 X0.42 Y0.89 Z0.00 M666 X1.42 Y1.13 Z0.28
I have tried different speeds (3050100 mm/second).
M122 command output after print:
[[language]] MaxReps: 5, StepErrors: 0, MaxWait: 0ms, Underruns: 0, 0
I have tried different slicers (Slic3r, S3D)  also without success.
I have tried to disable heated bed (main power consumer)  without success.
This happens with almost every circular (curved) shapes, test OpenSCAD code for model I printed:
[[language]] cylinder(r1 = 35, r2 = 20, h = 50, $fn = 1024);
Example layer code from S3D:
[[language]] G1 X10.146 Y32.739 Z1.700 E0.0140 G1 X10.947 Y32.480 Z1.701 E0.0140 G1 X11.741 Y32.201 Z1.701 E0.0140 G1 X12.528 Y31.903 Z1.702 E0.0140 G1 X13.307 Y31.587 Z1.702 E0.0140 G1 X14.078 Y31.250 Z1.702 E0.0140 G1 X14.840 Y30.896 Z1.703 E0.0140 G1 X15.594 Y30.522 Z1.703 E0.0140 G1 X16.338 Y30.130 Z1.704 E0.0140 G1 X17.073 Y29.720 Z1.704 E0.0140 G1 X17.797 Y29.292 Z1.704 E0.0140 G1 X18.511 Y28.846 Z1.705 E0.0140 G1 X19.213 Y28.384 Z1.705 E0.0140 G1 X19.904 Y27.904 Z1.705 E0.0140 G1 X20.583 Y27.407 Z1.706 E0.0140 G1 X21.249 Y26.893 Z1.706 E0.0140 G1 X21.902 Y26.364 Z1.707 E0.0140 G1 X22.543 Y25.818 Z1.707 E0.0140 G1 X23.170 Y25.257 Z1.707 E0.0140 G1 X23.783 Y24.681 Z1.708 E0.0140 G1 X24.381 Y24.090 Z1.708 E0.0140 G1 X24.965 Y23.484 Z1.709 E0.0140 G1 X25.534 Y22.865 Z1.709 E0.0140 G1 X26.088 Y22.231 Z1.709 E0.0140 G1 X26.625 Y21.584 Z1.710 E0.0140 G1 X27.147 Y20.924 Z1.710 E0.0140 G1 X27.652 Y20.252 Z1.711 E0.0140
And from Slic3r:
[[language]] G1 X34.040 Y10.097 E0.03229 G1 X33.713 Y11.138 E0.03239 G1 X33.282 Y12.369 E0.03873 G1 X32.886 Y13.384 E0.03235 G1 X32.459 Y14.389 E0.03242 G1 X32.004 Y15.376 E0.03227 G1 X31.516 Y16.352 E0.03239 G1 X30.999 Y17.312 E0.03237 G1 X30.454 Y18.254 E0.03232 G1 X29.880 Y19.179 E0.03233 G1 X29.278 Y20.086 E0.03232 G1 X28.648 Y20.975 E0.03237 G1 X27.990 Y21.845 E0.03236 G1 X27.307 Y22.693 E0.03235 G1 X26.599 Y23.519 E0.03231 G1 X25.864 Y24.325 E0.03237 G1 X25.107 Y25.106 E0.03230 G1 X24.324 Y25.865 E0.03239 G1 X23.519 Y26.599 E0.03233 G1 X22.693 Y27.307 E0.03232 G1 X21.844 Y27.991 E0.03238 G1 X20.974 Y28.648 E0.03236 G1 X19.908 Y29.399 E0.03872 G1 X18.997 Y29.996 E0.03234 G1 X18.066 Y30.566 E0.03241 G1 X17.119 Y31.106 E0.03237 G1 X16.157 Y31.616 E0.03233 G1 X15.181 Y32.097 E0.03231 G1 X14.188 Y32.548 E0.03236 G1 X13.183 Y32.967 E0.03235 G1 X12.167 Y33.356 E0.03228 G1 X11.137 Y33.714 E0.03240
This is happens not only in synthetic tests but also on real models too.
The only difference (compared to previous bug described at http://forums.reprap.org/read.php?178,546995 ) is much less visible patterns on model surface and clicks with higher frequency.
I have seen topics about Duet jerky movements both here and on reprap forums, but not sure that this is the same issue.
Anyone else experiencing such problems?

Usually I'm using this setting:
[[language]] M350 X16 Y16 Z16 E16 I1
I have tried today this setting:
[[language]] M350 X256 Y256 Z256 E256 I0
Print was much quieter and without extraneous sounds and clicks. Printed part surface is also much cleaner.
Also I have tried to change M350 settings during printing (yes, I know that I'm doing not in right way) and printer was completely freezed (I tried two times, both with the same results). DWC shows me something like "no more HTTP sessions available". The only way to restore operation was power off and power on again.
Update:
This setting also gives quiet printing without audible clicks:
[[language]] M350 X128 Y128 Z128 E16 I0
And then I using this setting clicks also very loud:
[[language]] M350 X8 Y8 Z8 E16 I0

Thanks for the feedback I need to try to see if there is a difference between:
M350 X16 I1 and M350 X256 I0 as they both should produce 256 microstepping.
The M350 X8 I0 is expected to be loud as that is 8 microstepping with no interpolation.

1/256 microstepping is very quiet for me, but Duet is usable only with low speeds due to high step rate.
Also documentation isn't very clear about interpolation modes. Seems like only 1/16 microstepping mode is able to use interpolation? Or I'm not right?

I have not tested with interpolation on less than 1/16 microstepping.
What I need to test when I get a chance is the difference between 1/256 microstepping and 1/16 microstepping with interpolation because they should be the same. The interpolation should be dividing each 1/16 microstep up into 16 further microsteps.

No updates? Maybe I can help testing/diagnosing this issue?

It seems that everyone (including David and Tony) ignores me, but I still continue my research.
I just prepared test gcode that you can run without printing something. Code just moves effector around 100 mm circle:
[[language]] T0 G21 ; set units to millimeters G90 ; use absolute coordinates M83 ; use relative distances for extrusion G1 Z10.300 F4800.000 G1 X49.744 Y1.647 F960.000 G1 X49.771 Y0.000 G1 X49.743 Y1.672 G1 X49.659 Y3.342 G1 X49.519 Y5.009 G1 X49.323 Y6.670 G1 X49.071 Y8.323 G1 X48.763 Y9.967 G1 X48.401 Y11.599 G1 X47.984 Y13.219 G1 X47.513 Y14.824 G1 X46.988 Y16.411 G1 X46.410 Y17.981 G1 X45.780 Y19.530 G1 X45.098 Y21.057 G1 X44.365 Y22.560 G1 X43.582 Y24.038 G1 X42.750 Y25.489 G1 X41.870 Y26.910 G1 X40.928 Y28.322 G1 X39.968 Y29.661 G1 X38.949 Y30.987 G1 X37.886 Y32.279 G1 X36.781 Y33.533 G1 X35.633 Y34.750 G1 X34.446 Y35.928 G1 X33.219 Y37.065 G1 X31.956 Y38.160 G1 X30.656 Y39.212 G1 X29.321 Y40.220 G1 X27.953 Y41.182 G1 X26.554 Y42.098 G1 X25.125 Y42.966 G1 X23.667 Y43.786 G1 X22.183 Y44.557 G1 X20.674 Y45.277 G1 X19.141 Y45.946 G1 X17.587 Y46.563 G1 X15.989 Y47.136 G1 X14.420 Y47.639 G1 X12.812 Y48.097 G1 X11.189 Y48.500 G1 X9.553 Y48.849 G1 X7.907 Y49.142 G1 X6.252 Y49.380 G1 X4.589 Y49.563 G1 X2.922 Y49.689 G1 X1.251 Y49.759 G1 X0.447 Y49.773 G1 X2.093 Y49.731 G1 X3.763 Y49.633 F960.000 G1 X5.428 Y49.479 G1 X7.112 Y49.264 G1 X8.738 Y49.003 G1 X10.380 Y48.681 G1 X12.009 Y48.305 G1 X13.625 Y47.875 G1 X15.226 Y47.390 G1 X16.809 Y46.852 G1 X18.374 Y46.261 G1 X19.917 Y45.618 G1 X21.439 Y44.923 G1 X22.936 Y44.178 G1 X24.407 Y43.382 G1 X25.850 Y42.538 G1 X27.265 Y41.646 G1 X28.648 Y40.707 G1 X30.000 Y39.721 G1 X31.317 Y38.691 G1 X32.599 Y37.617 G1 X33.863 Y36.484 G1 X35.052 Y35.344 G1 X36.219 Y34.146 G1 X37.362 Y32.891 G1 X38.430 Y31.637 G1 X39.471 Y30.329 G1 X40.468 Y28.986 G1 X41.419 Y27.610 G1 X42.323 Y26.203 G1 X43.179 Y24.767 G1 X43.987 Y23.302 G1 X44.745 Y21.812 G1 X45.453 Y20.296 G1 X46.118 Y18.735 G1 X46.713 Y17.199 G1 X47.264 Y15.620 G1 X47.762 Y14.023 G1 X48.206 Y12.411 G1 X48.596 Y10.785 G1 X48.931 Y9.146 G1 X49.211 Y7.497 G1 X49.435 Y5.840 G1 X49.603 Y4.176 G1 X49.715 Y2.508 G1 X49.771 Y0.836 G1 X49.771 Y0.836 G1 X49.715 Y2.508 G1 X49.603 Y4.176 G1 X49.435 Y5.840 G1 X49.211 Y7.497 G1 X48.931 Y9.146 G1 X48.596 Y10.785 G1 X48.206 Y12.411 G1 X47.762 Y14.023 G1 X47.264 Y15.620 G1 X46.713 Y17.199 G1 X46.109 Y18.758 G1 X45.453 Y20.296 G1 X44.745 Y21.812 G1 X43.987 Y23.302 G1 X43.179 Y24.767 G1 X42.323 Y26.203 G1 X41.419 Y27.610 G1 X40.468 Y28.986 G1 X39.471 Y30.329 G1 X38.430 Y31.637 G1 X37.346 Y32.911 G1 X36.219 Y34.146 G1 X35.052 Y35.344 G1 X33.845 Y36.501 G1 X32.599 Y37.617 G1 X31.317 Y38.691 G1 X30.000 Y39.721 G1 X28.648 Y40.707 G1 X27.265 Y41.646 G1 X25.850 Y42.538 G1 X24.407 Y43.382 G1 X22.936 Y44.178 G1 X21.439 Y44.923 G1 X19.917 Y45.618 G1 X18.374 Y46.261 G1 X16.809 Y46.852 G1 X15.226 Y47.390 G1 X13.625 Y47.875 G1 X12.009 Y48.305 G1 X10.380 Y48.681 G1 X8.714 Y49.007 G1 X7.087 Y49.268 G1 X5.428 Y49.479 G1 X3.763 Y49.633 G1 X2.093 Y49.731 G1 X0.422 Y49.773 G1 X1.251 Y49.759 G1 X2.922 Y49.689 G1 X4.589 Y49.563 G1 X6.252 Y49.380 G1 X7.907 Y49.142 G1 X9.553 Y48.849 G1 X11.189 Y48.500 G1 X12.812 Y48.097 G1 X14.420 Y47.639 G1 X16.013 Y47.128 G1 X17.587 Y46.563 G1 X19.141 Y45.946 G1 X20.674 Y45.277 G1 X22.205 Y44.545 G1 X23.667 Y43.786 G1 X25.125 Y42.966 G1 X26.575 Y42.084 G1 X27.953 Y41.182 G1 X29.321 Y40.220 G1 X30.656 Y39.212 G1 X31.956 Y38.160 G1 X33.219 Y37.065 G1 X34.464 Y35.910 G1 X35.633 Y34.750 G1 X36.781 Y33.533 G1 X37.886 Y32.279 G1 X38.949 Y30.987 G1 X39.968 Y29.661 G1 X40.942 Y28.302 G1 X41.870 Y26.910 G1 X42.750 Y25.489 G1 X43.594 Y24.016 G1 X44.365 Y22.560 G1 X45.098 Y21.057 G1 X45.780 Y19.530 G1 X46.410 Y17.981 G1 X46.988 Y16.411 G1 X47.513 Y14.824 G1 X47.984 Y13.219 G1 X48.406 Y11.575 G1 X48.763 Y9.967 G1 X49.071 Y8.323 G1 X49.326 Y6.645 G1 X49.519 Y5.009 G1 X49.659 Y3.342 G1 X49.699 Y2.542
This code produces jerky movements with [c]M350 X16 Y16 Z16 E16 I1[/c] command (1/16 microstepping with interpolation) and very smooth movements with [c]M350 X256 Y256 Z256 E16 I1[/c]
I can record video to prove my words.

Thanks for your example. My printers are out of action until my office refurbishment is complete, but I'll try it out when I can. Feel free to remind me on Wednesday if you don't get a response before then.

What happens when you do the exact same gcode with M350 X16 Y16 Z16 E16 I0 ?

One other question: what motor current have you set in your M906 command?

@bot:
What happens when you do the exact same gcode with M350 X16 Y16 Z16 E16 I0 ?
The same thing, but motors are very loud in this mode (relative to interpolated I1), so clicking noise is less audible over steppers noise.
One other question: what motor current have you set in your M906 command?
Thank you, now I'm not alone
I'm using [c]M906 X1500 Y1500 Z1500 E1000 I60[/c] with Wantai 42BYGHW811 motors. 200 steps/rev, 2.5A rated current, 1.8 mH phase inductance according to datasheet. Also I'm using 12V 30A LED power supply.
I have tried also 1.9 A and 1.0 A motor current, nothing changed.
Also I recorded both videos, my phone touches one of delta towers to amplify possible noise. I hope you can listen difference.
UPD: click frequency on first video is about 510 hertz. Clicks appear while effector moves around circle. But there is no clicks while effector goes down or while straight movement (from center to start of circle). 
Yet another video, I put my phone directly on moving caret to record better sound.

Seems like I need to go to Specsavers. What are we supposed to be looking at? Have you actually tried printing anything and if so, is the printed object misshapen?

I'm just want to say that something wrong : ) I'm not experiencing any real problems with printing using Duet WiFi, I'm happy Duet WiFi owner, but this is something abnormal.
Yes, this clicking pattern is slightly visible on printed objects, but I'm unable take good quality photo to show this. And this almost does not affect print quality.

I can't see any difference in the movement between the first two videos but maybe that's because I'm an old guy who wears glasses. You seem to be trying to make videos that record the sound, so are you saying that it sounds different?

I ran roboduet's file on my Rostock Mx V2 and can confirm that there is a different sound as he describes. You can definitely hear a tick which might line up with the interpolation point. Both 128 and 256 microstep without interpolation are smooth and don't have the tick. There is perhaps a discontinuity at some point in the interpolation. I couldnâ€™t hear this when running the same test on my CoreXY, so perhaps it's specific to deltas?
However, I also ran print tests and cannot see any difference in the results! So, to me at least, this appears to be of academic interest only.

I can't see any difference in the movement between the first two videos but maybe that's because I'm an old guy who wears glasses. You seem to be trying to make videos that record the sound, so are you saying that it sounds different?
Yes, that's what I want to say. English is not my native language so sometimes I cannot say what I want in few words so other people can understand me.
Can you hear clicking (510 times per second) sound on last video with black picture? This clicks are felt, if I touch a finger to the carriage. This feels like tower caret stops moving for few milliseconds and then starts moving again.
But there is no any clicks or other sounds then I change microstepping to 1/256. Everything else is the same  maximum jerk settings, accelerations, speed, motor currents and so on.

I ran roboduet's file on my Rostock Mx V2 and can confirm that there is a different sound as he describes.
However, I also ran print tests and cannot see any difference in the results! So, to me at least, this appears to be of academic interest only.Thank you very much! At least this is confirmed now by other people too.
As about differences in print quality  I can see slight differences in external surface, but I'm sure this is absolutely not a problem for me. I can try to find macro photo friendly plastic and right lighting to take photo with differences. Also I'm not sure what is affecting print quality  true 1/256 microstepping instead of interpolated 1/16 or this clicking issue.

The interpolation is only completely accurate when the carriages are moving at constant speed, which is rarely the case on a delta printer. So my guess is that the clicking noises are produced when the drive gets the interpolation wrong, perhaps when a carriage reverses direction in the middle of a move. Dynamically varying microstepping will solve that, when I get round to implementing it.

I hear this clicking sounds with interpolation disabled at 1/16 microstepping.
Also seems like I found bug, printer completely freezes (including ESP8266 web server code) then I executing this commands: [c]G28[/c] to home printer, [c]M350 X16 Y16 Z16 I0[/c] to disable interpolation (interpolation enabled in config) and then start printing file without (!) homing after executing M350 command. This is reproducible every time I enter this commands.