More strange pressure advance behaviour

deckingman

David, I've found something really significant. Please read this post fully.

In summary, and contrary to what we've both been thinking, increasing extruder micro stepping from 16x to 256x cures the problem (or at least it's a usable work-around)!

That is to say, printing the version of the tapered cylinder mentioned above, which has 3 perimeters but no infill, with 256x micro-stepping for the extruders, pressure advance happens just once at the start/end of a circle, as I would expect. Printing the spiral vase version of the same object, after the first 3 base layers, pressure advance doesn't appear to be applied ever (that's also fine by me as print speed is slow under those conditions).

However, if I set micro-stepping to 16x, (which is what I have always used until a short time ago) both gcode files print appallingly badly. That is to say the observable behaviour is as if pressure advance is being applied multiple times throughout any given circle and the print head seems to pause and wait for that to complete before it moves on the next segment. This applies to the spiral vase version of the file, as well as the version with 3 perimeters. Running M122 while this appallingly bad behaviour is happening, doesn't report anything untoward that I can see.

Printing either of those files at 16x micro-stepping but no pressure advance also results in good prints with no pauses or jerky motion. It's a combination of 16x micro-stepping plus 0.5 pressure advance that causes the problems.

Both files show significantly unequal segment sizes. I created the stl from an OpenScad file using $fn set to 300. This causes Slic3R to "do it's own thing" and change the number of facets to something else (64 IIRC). That may or may not be the cause of the unequal length segments.

Possibly the unequal size segments cause pressure advance to trigger but only with low micro-stepping. Is that possible? Does it make any sense to you?

I did some further testing with extruder micro-stepping set to 32x and 64x. In each case , the observable problem was less severe. i.e 32x was better than 16x and 64x was better than 32x but still not good.

Finally, I did an extreme test using 3 extruders (all the above prints were done with a single extruder) with 256x micro-steppping and pushed the speed up to 300%. The objects were sliced at 60mm/sec so I was printing at somewhere up to 180mm/sec. All with pressure advance set to 0.5. No problem! Although under those conditions with 0.5 pressure advance, the extruders do crazy things but there was no hesitation or jerky behaviour of the print head.

I've created a folder within that shared folder that I have already linked to called "AsOf140718" and put both the gcode files in there, as well as my latest config.g.

Finally, I'm still on older firmware -
Firmware Electronics: Duet Ethernet 1.0 + DueX5
Firmware Version: 1.20.1RC2 (2018-01-01)
Web Interface Version: 1.20

Hope this all helps shed some light on what the heck is happening.

Ian

dc42

Thanks Ian, that really helps. I have a couple of theories that I need to follow up about why low microstepping and high pressure advance together might cause the problem.

deckingman

@dc42

David, I've just tried printing one of those files again with lower pressure advance. The problem is present but not as severe even when using a value of 0.1 but with a value of 0.05 there is no observable issue (might just be too mild to be noticeable). To be clear, that's using a single extruder at 16x micro-stepping. So it's not necessarily high pressure advance but the problem is more sever the higher the pressure advance value.

What's bugging me is that it isn't being widely reported. It's not like it's difficult to spot - it's really bad and ruins the print. Quite noticeable even with pressure advance set to 0.1.

dc42

When you say it's with a single extruder, do you really mean that, or do you have the other extruders set to a very low percentage in the mix? I think it may make a big difference.

deckingman

@dc42 said in More strange pressure advance behaviour:

When you say it's with a single extruder, do you really mean that, or do you have the other extruders set to a very low percentage in the mix? I think it may make a big difference.

Dam. My bad. I thought I had mixing ratios set to unity but I've just checked and somewhere along the line I'd uploaded a config with mixing ratios of 0.96:0.02:0.02. That's what I was using. When I get time, I'll test again with unity mixing ratio. Sorry....

deckingman

OK. Tested with mixing ratio of 1.00:0.00:0.00 and the result was the same as with a mixing ratio of 0.96:0.02:0.02. That is to say very jerky head movement and by observing the extruder, the appearance that pressure advance is being applied multiple times around the circle. It is observable (if you watch the extruder gear) at lower pressure advance values like 0.1 but much less severe that at higher values. That's with 16x extruder microstepping. I can toggle the fault on and off by changing the extruder micro-stepping "on the fly" from the gcode console. So entering this M350 E16:16:16:16:16 will cause the fault to happen, and entering this M350 E256:256:256:256:256 will cure the fault.

deckingman

Trying to understand why this problem isn't being more widely reported. It's probably a unique combination of various factors but in an attempt to eliminate one of them, I defined Tool 0 simple as M563 P0 D0 H1 and commented out all the lines to do with mixing etc. Printing the tapered cylinder with 16x micro-stepping for the extruder and 0.5 pressure advance exhibits the same bad behaviour as having the tool defined to use multiple extruders but with the mixing ratio set to 1.00:0.00:0.00, so it isn't that. Oh well.....

wilriker

@deckingman I guess that a value of 0.5 for pressure advance is even high for a long bowden setup as far as I can tell. So maybe that's the reason why it is rarely ever reported.

deckingman

@wilriker said in More strange pressure advance behaviour:

@deckingman I guess that a value of 0.5 for pressure advance is even high for a long bowden setup as far as I can tell. So maybe that's the reason why it is rarely ever reported.

The thing is, I can see the same behaviour with lower values although at 0.1 or less, most people probably wouldn't notice unless they looked really closely. 0.3 or more is quite alarming though so I'm sure there must be some other reason why I see this behaviour. There are a lot of other things that are a bit unique to my printer. Two A and two B motors, 5 colour Diamond hot end, 0.5mm nozzle, big build volume, the use of Slic3R which generates strange looking segments sizes for arcs, - the list goes on. It's strange why upping the extruder steps per mm to 256x from 16x cures the problem though.

deckingman

Following on from this thread https://forum.duet3d.com/topic/6090/problem-printing-circles/26 I tried re-slicing the object with "detect thin walls" disabled and "resolution" set to 0.1. The problem persists with extruder micro-stepping set to 16x, so no change. The gcode file still shows large variations in segment size which may be acting as trigger for whatever is causing the issue?

OBELIKS

Check this one also, at the end: https://forum.duet3d.com/topic/6078/core-xy-movement-issues/57
He swapped M92 and M350 in config.g

deckingman

@obeliks said in More strange pressure advance behaviour:

Check this one also, at the end: https://forum.duet3d.com/topic/6078/core-xy-movement-issues/57
He swapped M92 and M350 in config.g

Thanks for trying but yes I've been trying to help that OP also so I'm well aware.

Edit. But just for the hell of it, I made those same changes (even thought it's not correct) and it didn't help my issue.

dc42

Hi Ian, I tried printing your tapered cylinder on my delta last week and again today with various pressure and advance extruder microstepping values. Even at pressure advance set to 0.4 and extruder microstepping set to 1 the XY movement is still smooth, although I can feel and hear the extruder vibrating.

[EDIT: I only tried the spiral vase version.]
[EDIT 2: I also tried with a mixing extruder configured, mix ratio 1:0.02:0.02, microstepping as low as 1:1:16. XY movement still smooth.]

So the next step is for me to replicate your configuration on my bench setup and see if I can see any jerkiness in the movement of the XY motors. But I have a feeling that you are not running very recent firmware. So please can you upgrade to firmware 2.01beta2 if you are running an older version, and verify that the problem is still there and it goes away at higher microstepping. Stay with DWC 1.20 (the latest DWC beta doesn't work properly with CoreXY) and add M564 H0 to config.g if you don't have it in there already.

PS - did the jerkiness happen at all heights of the tapered cylinder, or only in certain regions?

deckingman

@dc42

Hi David. As per post dated 14th July, I am indeed still on 1.20.1RC2 so I'll update and report back.

I can't say if the jerkiness happens at all heights. It does for the skirt and the first few layers, after which I've always aborted the print (didn't seem any point in continuing).

deckingman

Well now, it seems to be fixed

So I updated the firmware to 2.01beta2 and put extruder micro-stepping down to 16x and I cannot provoke the problem. Tried single tool, mixing tool, ratios from 0.96:0.020.02 to 0.34:0.33:0.33. Tried the tapered vases - both vase mode and 3 perimeter model. Tried the original test cylinders. These were sliced at mostly 90mm /sec and I changed the speed from 30% (30mm/sec) all the way up to 300% (theoretically 270 mm/sec but in reality limited by acceleration to somewhat less). Tried pressure advance from 0.1 to 0.5. Tried all combinations of those sometimes crazy settings and behaviour is fine and normal.

That is to say, no matter what I do, pressure advance gets applied once only at the end of each circle, instead of multiple times during the circle and there are no pauses or jerky behaviour of the head movement.

So something that changed in firmware between 1.20.1RC2 and 2.01beta has fixed it although I haven't seen anything in the release notes that would explain it. Maybe it was the change to RTOS??? Who knows? Who cares? I'm just happy it's fixed.

PS, the top layer surface finish on those test cylinders at some speed way in excess of 90mm/sec with highish pressure advance was as good as 30mm/sec - almost flawless.

dc42

Great! I'll cross that one off my list. I made some changes to the pressure advance calculations in release 1.21, so those may have fixed it.