3.4.0 b3 leaning prints

Fred-Y

@t3p3tony It seems that I got one of my prints leaning on both -X and +Y

On this picture it is leaning on the left and also back.

I spent quite some time today trying to reproduce the issue on simple objects like basic cubes, calibration cubes, ringing test with and without infill (rectilinear, gyroid...) but all seems fine.
I guess I changed something on SuperSclicer ... I even checked I was still on the beta3.

T3P3Tony

@fred-y thanks for spending the time in testing this - it helps!

I will test to see if I can get a lean with the two box prints arranged in different orientations then looking for anything on +/-X

Adrian52

With a 20mm cube and a 20x20x10high cube offset in the y direction printed together, the 20mm cube (on the right) looks like this
undefined
The left cube for comparison printed with no y offset. You can see that the lean reduces after 10mm, but interestingly a smaller lean is still seen. Havent tried a 45deg rotated cube yet.

dc42

Thanks to everyone who has helped us to try to track this down. @chrishamm has managed to reduce one of your leaning prints to a pair of 2 moves that leaves one of the carriages 2 microsteps away from the correct position. I expect to be able to track down the problem now.

dc42

@Adrian52 @CabalSoul @DaveA @Fred-Y @jay_s_uk @SIam I have put new firmware builds with a fix for the issue I was eventually able to reproduce at https://www.dropbox.com/sh/kgah0ylh2zu2f7w/AAD2vQOiTQrFly1p22sob6c7a?dl=0. As with all unofficial builds, use them with care.

CabalSoul

@dc42 thank you exactly what I needed. I am going to test right away.
.
.
.

I've tested it and it seems to work. Thanks again.

jay_s_uk

@dc42 one print completed so far and no leaning!
What was it in the end?

Blacksheep99

@jay_s_uk said in 3.4.0 b3 leaning prints:

@dc42 one print completed so far and no leaning!
What was it in the end?

It was the Pisa effect apparently…….

dc42

@jay_s_uk said in 3.4.0 b3 leaning prints:

@dc42 one print completed so far and no leaning!
What was it in the end?

The new code didn't fully allow for the case of a move ending when a tower that had been moving up was on the point of reversing and moving down, but the required final position was one microstep higher than the theoretical highest point reached, because the stepper motor position had been rounded up to the next microstep to get closer to the commanded position for the end or the move.

jay_s_uk

@dc42 thank you!

DaveA

@dc42 Thanks for the fix. I tested with two of my leaning files and they now print correctly.

Fred-Y

@dc42 I just finished to print my test file and it looks good

I have another print to to today, hopefully it will be good as well.

dc42

All, we have identified that the leaning issue on deltas was not completely fixed by that internal build because there was another less common situation in which a step was taken in the wrong direction. I am currently testing a fix for that.