Advanced Search

look latency test
 

Title: Qt eglfs vs Xorg/glamor on Intel J1900 — 1 kHz EtherCAT jitter (not LinuxCNC, GPU/X data point)

Not a LinuxCNC install. Posting because this forum already tracks iGPU/X vs servo-thread latency, and we have before/after numbers on one box.

Hardware: Intel Celeron J1900 (Bay Trail HD 8086:0f31), Debian 13 PREEMPT_RT 6.12, IgH EtherCAT master, 1 kHz, two axes. HMI is a fullscreen Qt Quick app on the same machine. Isolated CPU3 for the RT loop.

Same load, same kernel.

Xorg + glamor + x11vnc:
- P99 |jitter| about 52 us (after removing a measurement bias)
- worst -171 / +58 us
- many outliers; 60-90 points >=80 us per 5 s capture

Qt eglfs / eglfs_kms (no LightDM, no X, no VNC), QT_QPA_EGLFS_NO_LIBINPUT=1:
- P99 |jitter| 14 us
- worst -38 / +45 us
- 0 points >=50 us in ~35k cycles

A separate bug was measuring jitter mid-axis instead of wake-to-wake; that only moved the mean onto 0 us. The spike cleanup is the display path: X/glamor/x11vnc fighting i915 while the RT thread is on the sibling core.

We are not recommending this on Elkhart Lake (J6412/J6413). There, glamor already gave 100-300 us spikes; eglfs would own the GPU even harder.

Happy to hear if anyone running a Qt or other GL HMI beside a 1 ms servo thread has compared eglfs/KMS to a compositing X session.

No worries, for now, but just imagine how that might end up later in the process when we start talking wiring.

That might have been auto correct i swear i wrote servo, im used to 3d printers so maybe i mixed up by habbit. But yes they are servo motors not stepper motors, the servos motors still need a drive board tho, and sorry for the confusion ill double check the spelling

Bunchi wrote:

...maybe im just looking the wrong places tho

No, not the wrong places, probably, but definitely for the wrong thing, 3 times we mentioned those are servo motors and drives, yet you insist on mentioning steppers.

Trying to embed video:

i.imgur.com/mOQMjpb.mp4

i.imgur.com/mOQMjpb.mp4

Some observations:

- When attempting to edit the first post, the editor shows the content only up to the first image. Everything else is gone in the editor. Including in "source" mode. Not sure if the editor has some kind of limit in lines or content. 

- Have not found how to embed videos from imgur. The video button in the editor asks for url, size, height, width, provider (from a fixed list in a drop-down menu), and id. However, the resulting source only showstags, regardless of what is entered. Unclear if something is done in the background.

- When clicking to edit these posts, all of the text shows in blue like a url. Using "source" mode in editor shows the closing url taghas been moved from the end of the first url/link to the end of all text. This happens repeatedly even if in source mode I remove that closing url tag and put it back where it should be and save the post.

This is a test post to duplicate another of my posts in another thread on the forum and troubleshoot why my video links did not work. I may make several edits or reply posts. This thread can be removed by mods eventually. Thanks for having a testing area. And now, on with the test:

==========

Lukasz from Proma-Elektronika kindly sent us a unit to test, and we are quite happy with how it performs on our fiber laser machine.





The height sensor is used to automatically set and/or maintain a fixed nozzle-to-material distance which has been determined appropriate according to the focal point of the laser beam, i.e. "autofocus".

Autofocus is enabled in the following clips, and the readout from the device shows it is maintaining the set height closely, including during/after indexing. The drilling speed was slowed down to make it easier to observe.

The C2V Laser Sense has not had issues with noise from operation or the sparks generated, which eventually became a problem with an optical height sensor used previously with this machine.

Drilling holes:

i.imgur.com/mOQMjpb.mp4
i.imgur.com/CB8KpLx.mp4

Entry holes:


Exit holes:


Straight cuts. Note the visible autofocusing at the very beginning:

i.imgur.com/YGVQZon.mp4

Result:




The testing we did is so far largely qualitative but we can say it comfortably achieves its specified 10 um (0.01mm) resolution, which is important to us. At some point it would also be good to validate the stated sampling frequency (3.65 kHz) and response time (< 2 ms).

Aside from performance, one standout feature of this unit is the information feedback right on the unit. The LED and built-in display are tremendously helpful, both during calibration, and during operation. Really a very useful feature!

Setup was quite simple and the documentation of the device function, modes, wiring diagrams, and calibration routines were helpful. A cat6a cable (shielded, twisted pairs) was perfect for making the data/power cable.

For integrating the C2V Laser Sense to our existing machine using an Aerotech A3200 controller, it was straightforward to just use the analog output from the device. It would be great to see others configure this device with LinuxCNC. Given the many modes of operation offered, it should be possible to get it set up with quite a variety of hardware.

Since precision is our priority, we went with Mode 0 which has a non-linear (exponential decay) relationship between voltage and distance.

Good grounding of the chassis was important to getting good calibration results, and luckily the calibration process is easy. Fast calibration for the reference position is just a button press. The precision calibration requires programming a small routine to get the motion speed and distance required.

The best part is the colors of the LED indicator make it easy to see when calibration should begin, when it is in progress, and most importantly, whether it has failed or completed successfully. And the built-in display even shows the calibration curve; a very nice touch!

In one instance, upon powering up the machine, the device's reading was "frozen" at a fixed voltage regardless of distance, or even touching the nozzle tip. We never figured out what the cause was. Re-initializing the system and checking the grounding did not help. But resetting the unit and recalibrating it resolved the matter. We were not able to reproduce this issue, so I guess chalk it up to machine gremlins.

It goes to show how useful the readout from the built-in display is, as that was the first and most obvious clue that something was not as expected.

Overall, very happy with the C2V Laser Sense. Thanks to Lukasz for sending us this unit. We'll be looking to purchase more units for future machines, hopefully running on LinuxCNC. If others have used this unit with LinuxCNC, please do post your results!

The photos you originally posted appear to be Servo Dynamics/Baldor LD series servo drives. Since they are analog ,they are easy to diagnose.

the fuses are fine we think its something giving too much current since the transformer for the drive boards was smoking when we tried to run it. Ive been trying to figure out if its worth fixing or just finding a replacement that we know works. I know there are a variety of drive boards for small stepper and servos on the small scale but its been hard to find stores that have boards for this size of motor, maybe im just looking the wrong places tho

Currently unused

The part number was in my first post = 74CBT16211CDGGR

Hi

Im based in switzerland and could give it a shot with repair by replacing U3, if i know what chip it is

Hi
I have the 400w version, should be the same. I'll post my comfig as soon as I have access, but it rather looks like an issue of drive side configuration.

-does the manual jog according to the manual work?

-can you readout any error codes? For example when toggling the arrow buttons on the drive

-did you do configure the drive with their software over usb? I did PID tuning there and also had first time power on the motors through this.

-there are some digital inputs that default to overtravel/undertravel with active low according to the manual page 212, make sure these are connected properly or disabled

Again, not steppers, those are permanent magnet brushed or brushless servo motors.
I would not change the drives and motors, that would introduce mechanical issues that might be easy to fix if you have access to a mill.
I would also not change any automation and e-stop wiring, just wire in the required stuff to Mesa and LinuxCNC.
-
Transformers do not blow, usually they just burn in place with some smoke and a distinct smell, but if left like this longer they do blow violently if any wire brakes....makes me ask: Where are the fuses? GFI?

the transformer for the drives is broken, and replacements are old we got one to try and fix it but it blew again so im looking at replacing as much of the old stuff as possible. Im keeping the motors im just not sure where to find new modern drives for steppers this big. I guess what im looking for is would it be easier to find a new drive boards to hook up to a mesa board or should i make a new power delivery for the old drive boards and try and use thoes with the mesa board