Head pauses after each line of an engrave, causing slowwwness of overall engrave

Scrubbing though the .MOV video in QuickTime player, I see 4 frames at the same ‘left’ position. Scrubbing forward to the ‘right’ position, I see another 4 duplicate frames…

This video is 29.99fps
So:

  • At 1000 speed the full pause is ~4 frames: ~133.3ms (4 frames * 1/29.99ms), and at least >100ms (3 frames * 1/29.99ms)

  • At 600 speed the full pause is ~2 frames: ~66.7ms (2 frames * 1/29.99ms)

Maybe more pause is needed for inertia (let the elastic material of the laser head settle?).
Also, I wonder if the deceleration could be a lot more aggressive to get this thing moving faster…

Certainly too fast for my old brain to see a pause.

2 Likes

@palmercr how do you view the waveform files? Are you snooping the network traffic or something?

At 450LPI, engraving a 3in fret slot (0.023" wide) at 1000 speed, there’ll be at least 100ms * 450lpi * 3in of pausing == 135seconds == (135/60)min == 2.25 extra minutes of pause on top of the actual work being done, and that doesn’t include the time needed for accel/deceleration… :slight_smile:

(3 minutes, if you’re using the 133.3ms measure)

would be interesting to see a video from a pro at 1000… wondering if it has less pause.

The faster the head is moving, the longer it will need to accelerate / de-accelarate.

1 Like

It may just be the last step of the deceleration ramp.

I have nothing to announce at this time regarding how I view the files.

3 Likes

This is what the last few steps of a change of direction looks like when engraving the ticks on the ruler

  20 03 08389 -7193 -10  x-      
  20 03 08388 -7193 -10  x-      
  24 03 08387 -7193 -10  x-      
  26 03 08386 -7193 -10  x-      
  28 06 08385 -7193 -10     y+   
   6 03 08385 -7192 -10  x-      
  42 03 08384 -7192 -10  x-      
  61 03 08383 -7192 -10  x-      
 209 03 08382 -7192 -10  x-      
 199 09 08381 -7192 -10  x+      
 207 09 08382 -7192 -10  x+      
  61 09 08383 -7192 -10  x+      
  42 09 08384 -7192 -10  x+      
  34 09 08385 -7192 -10  x+      
  26 09 08386 -7192 -10  x+      
  24 09 08387 -7192 -10  x+      
  20 09 08388 -7192 -10  x+      

The first column is the delay in waveforms tick, which are 100us. So it is only actually stationary for about 20ms I think. Not really a pause, just the first step of the acceleration ramp.

The second column is the byte in the file in hex, the next three columns are X,Y,Z motor positions in stepper units. I have skipped over the bytes where nothing happens. I.e. when the lines starts with 199 there have been 199 bytes of do nothing. Not the most efficient coding scheme!

2 Likes

@palmercr is this at 1000 speed?
and do you have an idea to explain the difference observed (100-133ms) vs code (20ms)

I think the ruler engraves at 700 but if it uses the same acceleration the last steps of the turnaround should be the same.

Each step is less than 1 thousandth of an inch so you probably perceive it stationary for several steps. The last four steps take 82ms to go both ways.

See the two videos I posted. In one video the pause was 2 frames (at 600 speed) and in the other video the pause was for 4 frames (at 1000 speed). equating to 60ms and 133ms pauses each if we are to trust the sampling rate of a 29.99fps video… crude measure i know. but… should be as accurate as the sampling rate.

If there is any backlash in the system then it will pause for longer.

that would make sense. in another thread I posted about line-by-line engraving error offsets. the observed pauses could easily be due to backlash.

Yes that looks like a faulty machine I think Were the ticks on the founders ruler perfectly straight?

nah, it’s microscopic. In line with what I’d except for backlash. And I see this on other people’s machines too. note, the offsets are only seen in tiny sub 0.5mm text (or thin vertical lines). What the GF needs is a backlash calibration, it would seem (for the offset error… for the pause, idk what it needs).

I found a turnaround with a bigger pause:

  14 03 07496 -7144 -10  x-      
  16 03 07495 -7144 -10  x-      
  15 03 07494 -7144 -10  x-      
  17 03 07493 -7144 -10  x-      
  20 03 07492 -7144 -10  x-      
  20 03 07491 -7144 -10  x-      
  25 03 07490 -7144 -10  x-      
  26 03 07489 -7144 -10  x-      
  34 03 07488 -7144 -10  x-      
  43 03 07487 -7144 -10  x-      
  62 03 07486 -7144 -10  x-      
 607 01 07485 -7144 -10  x+      
  62 01 07486 -7144 -10  x+      
  43 01 07487 -7144 -10  x+      
  34 01 07488 -7144 -10  x+      
  26 01 07489 -7144 -10  x+      
  25 01 07490 -7144 -10  x+      
  20 01 07491 -7144 -10  x+      
  20 01 07492 -7144 -10  x+      
  17 01 07493 -7144 -10  x+      
  15 01 07494 -7144 -10  x+      
  16 01 07495 -7144 -10  x+      
  14 01 07496 -7144 -10  x+      

60.7ms stationary.

It might be to prevent wobble building up as that would affect alignment.

Here is the next line:

  14 01 11459 -7140 -10  x+      
  15 01 11460 -7140 -10  x+      
  16 01 11461 -7140 -10  x+      
  17 01 11462 -7140 -10  x+      
  19 01 11463 -7140 -10  x+      
  20 01 11464 -7140 -10  x+      
  25 01 11465 -7140 -10  x+      
  26 01 11466 -7140 -10  x+      
  33 01 11467 -7140 -10  x+      
  42 01 11468 -7140 -10  x+      
  60 01 11469 -7140 -10  x+      
 179 01 11470 -7140 -10  x+      
 264 03 11471 -7140 -10  x-      
 176 03 11470 -7140 -10  x-      
  60 03 11469 -7140 -10  x-      
  42 03 11468 -7140 -10  x-      
  33 03 11467 -7140 -10  x-      
  26 03 11466 -7140 -10  x-      
  25 03 11465 -7140 -10  x-      
  20 03 11464 -7140 -10  x-      
  20 03 11463 -7140 -10  x-      
  16 03 11462 -7140 -10  x-      
  16 03 11461 -7140 -10  x-      
  15 03 11460 -7140 -10  x-      
  14 03 11459 -7140 -10  x-      

It looks like turn arounds at each end are not symmetrical, which is odd. Also they vary a little from line to line. Perhaps that is due to the quantisation at 10kHz. It makes all the stepper pulses jittery. An engraving section looks like this:

   1 11 08520 -7128 -10  x+        on
   1 10 08521 -7128 -10            on
   1 11 08521 -7128 -10  x+        on
   1 11 08522 -7128 -10  x+        on
   1 01 08523 -7128 -10  x+      
   2 01 08524 -7128 -10  x+      
   1 11 08525 -7128 -10  x+        on
   1 11 08526 -7128 -10  x+        on
   1 11 08527 -7128 -10  x+        on
   1 10 08528 -7128 -10            on
   1 01 08528 -7128 -10  x+      
   1 01 08529 -7128 -10  x+      
   1 01 08530 -7128 -10  x+      
   1 10 08531 -7128 -10            on
   1 11 08531 -7128 -10  x+        on
   1 11 08532 -7128 -10  x+        on
   1 11 08533 -7128 -10  x+        on
   2 01 08534 -7128 -10  x+      
   1 01 08535 -7128 -10  x+      
   1 01 08536 -7128 -10  x+      
   1 01 08537 -7128 -10  x+      
   2 01 08538 -7128 -10  x+      
   1 01 08539 -7128 -10  x+      
   1 01 08540 -7128 -10  x+      
2 Likes

Here is a graph of speed against time during a turnaround. There are only a few discrete speeds you can do at 10kHz so it dithers between the nearest two to get intermediate ones.

The speed is shown here in mm / minute against time in seconds.

5 Likes

Looking further it does seem that most lines have a 60ms pause as each end. The first line seemed different for some reason.

Here is graph that goes right up to full speed where it starts engraving. To get an idea of the real speed I used a one tap IIR filter to average the dithered speed. That is the red line.

It looks simple harmonic motion (like a pendulum, which is the smoothest form of motion) except for the pause.

7 Likes

You know that moment when you are watching a conversation and realise that it is WAY out of your league?
Yeah, getting that now…

Fascinating stuff though

8 Likes

Thanks for the feedback. I’ll share it with the rest of the team.

Please let me know if you have any more questions by posting a new topic or emailing me at support@glowforge.com