Shutdown: Timer too close when running TEST_RESONANCES/SHAPER_CALIBRATE

General Discussion
1

I get the infamous error “Timer too close” when running TEST_RESONANCES or SHAPER_CALIBRATE.
Environment:

  • Delta Kinematic
  • Stepper settings:
[stepper_?]
step_pulse_duration: 0.000000190
rotation_distance: 32
microsteps: 16
full_steps_per_rotation: 400
homing_speed: 120
  • Klipper host:
Intel(R) Core(TM) i7-4710MQ CPU @ 2.50GHz
64bit x86_64 with 8 cores

The error occures at approx. 70Hz.
Here’s the sar output when the error occures

17:10:37        CPU     %user     %nice   %system   %iowait    %steal     %idle
17:10:38        all      0,87      0,00      0,50      0,00      0,00     98,63
17:10:39        all      0,88      0,00      0,38      0,00      0,00     98,74
17:10:40        all      0,75      0,00      0,63      0,00      0,00     98,62
17:10:41        all      1,00      0,00      0,38      0,00      0,00     98,62
17:10:42        all      0,88      0,00      0,63      0,00      0,00     98,50

I only got a working shaper by decreasing the µStep to 8…

2

Due to the special Delta kinematic the MCU needs to create a lot more steps than for a Cartesian one. Looks like your board cannot create enough steps without lagging behind.
Post a klippy.log showing the issue.

3

Last year (july) I successfully run those tests with 1.8° Stepper and 32µSteps with the same hardware (Printer & MCU).

I also tried with and without STEP_ON_BOTH_EDGES. With “off” the shutdown should occure much earlier because of doubling the MCU load. But - interestingly enough - the error was in the same frequency range.

klippy.zip (27.9 KB)

4

Can you post a full unmodified log?

Just spitballing, but I think the acceleration changed from 7k to 10k during the test but I’m not sure if that would explain what you are seeing.

5

No, because I get also failures when testing Y axes - and here it fails already at ~70Hz.
Attached yesterdays and todays log. Today I had to go down to 4uszep to run a successfull calibrate.
klipper_all.zip (3.4 MB)

6

Now you apparently upgraded to 0.9 steppers. I’m by no means a Delta expert, but to my knowledge it needs all 3 axes to either move in X or Y direction. With doubling the steps on one axes, this means 6-folding it, when all 3 axes are used, right?

7

Yes, but I also reduced the uSteps to the half - compared to last year. So 0.9 * 16uSteps compared to 1.8*32 uSteps. That’s no difference in steps/rotation.
Addendum: last year I used a 2-core, 1.6GHz host. Now I used a 8-core with 3.5GHz each.
And - according to the doc - Timer too close means the host can’t handle the load…

8

My understanding is that you still have a factor of 3:
3 steppers x 2 (.9 stepper) / 0.5 (uSteps) = 3

Virtualized?

9

No - plain hardware.
And no - steps/rotation is the same on each axis. So the sum is the same…

10

So, finally it showed that the mcu (MKS Robin Mini) was simply too weak.
Switched over to a Supernova board - now it runs fine with 128µSteps…