Unable to complete quad gantry levelling

I have a stock SV08 that I bought two weeks ago. I am unable to get a quad gantry level to pass. I keep seeing the following errors:

  • Probe sample tolerance exceeded

  • Too many retries

  • Retries aborting: Probed points range is increasing. Possibly Z motor numbering is wrong

I was told by some folks on the CN3DP Discord server that the issue is the stock probe. I am planning on replacing it but are looking for some troubleshooting tips on how to diagnose this problem and potentially fix it.

Open up printer.cfg and find the “probe” section:

[probe]
pin: extra_mcu:PB6    
x_offset: -17                  
y_offset: 10             
#z_offset : 0
speed: 15.0
speed: 5.0
samples: 2
sample_retract_dist: 2.0
lift_speed: 50
samples_result: average
samples_tolerance: 0.016
samples_tolerance_retries: 2

You can open up the tollerance or get crearive.  My SV07 sensor is a bit flakey I overcame it by:

[probe]

speed: 15.0
speed: 3.0 #lower slower
samples: 5 #more samples
sample_retract_dist: 2.0
lift_speed: 50
samples_result: median #use the sample in the middle of the group
samples_tolerance: 0.036 #allow more scatter
samples_tolerance_retries: 2

Using the median keeps outliers from skewing the average.

The SV07 only runs a 2 point gantry level but my machine was making (on average) 8 or 9 passes and erroring 30% of the time.

After the changes it makes 4 or 5 passes and (so far) never errors. Each pass is slower but the complete process is faster.

I suggest a bed soak of 30min (to be sure and eliminate others possibles issue), then perform the QGL (twice if the first fail), then the bed leveling process. If that continue to fail during the QGL, “Possibly Z motor numbering is wrong” (the Z motors can be plugged in with the wrong driver, check the sticker on each). If that continue after that, verify if the QGL works properly :

  • That should try one time and report values + corrections (corrections are applied automaticly)
  • Then it’s retry : if the “range is increasing” (so that is worst that the previous one), that is an issue with Z leveler (motors, belts, one screw badly fixed…).

You can ask a technical support from sovol by email (they does not answer here directly, only by email).

Thanks to both of you for your suggestions.

After making changes to samples_tolerance and a nice long heat soak, I am down to the front-right (Z4) probing failing with complaints about sampling exceeding tolerances.

I have an email in to Sovol support so we shall see what happens.

I wonder if I should just wait until I install my BTT Eddy Duo + eddy-ng to give this any more attempts…

2 Edged sword

On one hand getting rid if the sensor that is most likely causing the problem is likely going to solve the issue.

On the other hand, launching into a major upgrade with a non-functional starting condition makes debugging upgrade issues harder.

Klipper has PROBE_ACCURACY built in. Try 5 or 6 sets near each corner. How bad is it really?

1 Like

Here are the results from running PROBE_ACCURACY in all four corners:

Front right at probe at 317.500,37.500

probe accuracy results: maximum -0.275625, minimum -0.288750, range 0.013125, average -0.285750, median -0.286875, standard deviation 0.003769

probe accuracy results: maximum -0.271875, minimum -0.286875, range 0.015000, average -0.282562, median -0.284062, standard deviation 0.004112

probe accuracy results: maximum -0.275625, minimum -0.286875, range 0.011250, average -0.284437, median -0.285937, standard deviation 0.003462

probe accuracy results: maximum -0.266250, minimum -0.281250, range 0.015000, average -0.277500, median -0.279375, standard deviation 0.004743

probe accuracy results: maximum -0.264375, minimum -0.283125, range 0.018750, average -0.277875, median -0.279375, standard deviation 0.005675

Front Left

PROBE_ACCURACY at X:57.500 Y:37.500 Z:1.715 (samples=10 retract=2.000 speed=5.0 lift_speed=50.0)

probe accuracy results: maximum 0.037500, minimum 0.011250, range 0.026250, average 0.019875, median 0.017813, standard deviation 0.007556

probe accuracy results: maximum 0.026250, minimum 0.007500, range 0.018750, average 0.012375, median 0.011250, standard deviation 0.005447

probe accuracy results: maximum 0.020625, minimum 0.003750, range 0.016875, average 0.008813, median 0.007500, standard deviation 0.005104

probe accuracy results: maximum 0.018750, minimum -0.001875, range 0.020625, average 0.006000, median 0.003750, standard deviation 0.005675

probe accuracy results: maximum 0.202500, minimum 0.178125, range 0.024375, average 0.184875, median 0.181875, standard deviation 0.007075

Back Left

probe accuracy results: maximum 1.066875, minimum 1.050000, range 0.016875, average 1.053750, median 1.052813, standard deviation 0.004961
probe accuracy results: maximum 1.066875, minimum 1.050000, range 0.016875, average 1.053375, median 1.050938, standard deviation 0.005155
probe accuracy results: maximum 1.068750, minimum 1.048125, range 0.020625, average 1.053000, median 1.050938, standard deviation 0.006173
probe accuracy results: maximum 1.080000, minimum 1.051875, range 0.028125, average 1.062563, median 1.061250, standard deviation 0.007642

Back Right
probe accuracy results: maximum 0.127500, minimum 0.114375, range 0.013125, average 0.119625, median 0.119063, standard deviation 0.003437

probe accuracy results: maximum 0.163125, minimum 0.148125, range 0.015000, average 0.150938, median 0.150000, standard deviation 0.004296
probe accuracy results: maximum 0.193125, minimum 0.176250, range 0.016875, average 0.188625, median 0.191250, standard deviation 0.005447
probe accuracy results: maximum 0.226875, minimum 0.202500, range 0.024375, average 0.208313, median 0.206250, standard deviation 0.006887
probe accuracy results: maximum 0.243750, minimum 0.217500, range 0.026250, average 0.223500, median 0.219375, standard deviation 0.007676

All 4 corners are very similar and consistent but about 2x the range I’d expect… Based solely on my experience with my SV07.

Either something is binding in your gantry or the Prox switch is no good.

That all 4 corners read so similar points to the probe not the gantry.