Extruder motors skip steps with pressure advance enabled

Edgars Batna

@deckingman Don't see M82 anywhere. This video sums up last 3 months of my life: https://www.youtube.com/watch?time_continue=6&v=KjCO0thvlNY

I'll try without filament, but I think I already tried it. I need to make a list...

deckingman

@edgars-batna said in Extruder motors skip steps with pressure advance enabled:

@deckingman Don't see M82 anywhere. ......................

OK - but tell me that you do see an M83. I have this nasty feeling that without either M82 or M83, the board will default to M82 (absolute) extrusion.

Edgars Batna

@deckingman I've posted my config.g from the start with the original post and it's the second line. If there would be a problem such as this, then the whole thing wouldn't work, but it's printing fine with PA disabled.

Just tried without filament - behavior is, I think, marginally worse than with filament. The bigger motor behaves way worse. This would fit the "impossible steps" theory as the rotor inertia is higher and there's no "filament brake" against oscillations now. Gonna post a vid as soon as I get to it, but it's basically the same and it seems that it's the worst during first layer, so it must be speed dependant. The slower the head the more skips they seem to get.

Edgars Batna

@deckingman Video without filament: https://www.youtube.com/watch?v=oXsFLG7AMug&feature=youtu.be

Phaedrux

The video makes me think broken wire.

deckingman

@edgars-batna said in Extruder motors skip steps with pressure advance enabled:

@deckingman Video without filament: https://www.youtube.com/watch?v=oXsFLG7AMug&feature=youtu.be

Well that rules out anything to do with the hot end. We can forget all about jams and blockages. Time to test each individual extruder motor to see if we can narrow it down to one or the other (or both).

Also can you post your complete config.g file. Maybe one of use can spot something that you have missed.

Edgars Batna

@phaedrux All wires have either soldered or fastened connections. These are rather thick 1mm^2 wires with thick insulation. I just tested them and they read exactly motor resistance on the poles on the Duet connector side.

@deckingman The config.g in the original post is the original, I just removed wifi name and password. Motors can be also ruled out, as the nearly 1 year old Nema 17 behave exactly like the brand new Nema 23... Connections can be also ruled out, as these are two separate wires that I tried on 5 separate drivers on 2 separate motors and I use exactly the same type of wiring for the 6 other motors that never have an issue.

Slowly I'm coming to conclusion that there is a bug in the firmware...

EDIT: New findings: Setting extruder microstepping to 256 made it better, but this is not a solution as a single step would have too little torque and they still make knocking sounds, which means they would lose steps under load. I guess the firmware rounds the steps up on lower microstepping. @dc42 This could be, uh, a bug? I summon the firmware gods for clarification! If the resolution is less than required oscillation then it should suppress the oscillation. Does the firmware force 1 microstep movement if required movement is less than microstep for PA? I probably don't know what I'm talking about, but please help.

deckingman

@edgars-batna I've just spent quite a bit of time going through your config files. Sorry to be a pain but just to be clear, you are saying that the config without pressure advance prints fine but the one with pressure advance is crap like in the video? What I mean is, there isn't some other fault crept in coincidentally with you enabling pressure advance? You can simply change M572 D0:1 S0.5 to M572 D0:1 S0.0 and it'll resort to printing just fine yes?

The reason I ask is that there are what look like a couple of anomalies but if it prints fine then it's nothing to worry about.

I'll tell you what looks odd anyway........You have the mixing ratio set to 0.5:0.5. Normally with a push-pull arrangement such as you have you'd set both extruders to 1.0:1.0 otherwise they would run at half speed (as in a mixing hot end). But further up, you set micro stepping to 8X for the extruders, then you set the steps per mm to 88(ish), then you change the micro stepping to 16X. I'm pretty sure that this will effectively change the steps per mm to 176 (88 x 2) which would make the extruders run at double speed and cancel out the 0.5:0.5 mixing ratio thing.

So if the steps per mm should be 88 @16 micro-stepping for those extruders, then ideally you ought to set the micro stepping to 16x in your first M350, take out the second M350 and set mixing ratio to 1.0:1.0.

Maybe - just maybe, there is some interaction between mixing ratio set like that and pressure advance that hasn't yet come to light until now........Long shot - just speculating....

The only other thing I noticed is that all your M569 commands have B, F and Y values which are firmware 2.02 and later - i.e. something new that has recently been added. Again, just speculation on my part but could this be the cause of the problem? Could there be some interaction between these values and the pressure advance algorithm? As this is something new in firmware 2.02 I see no harm in taking those parameters out and simply using M569 Pn Sn.

I'll keep thinking - this is bugging me.

Edgars Batna

@deckingman said in Extruder motors skip steps with pressure advance enabled:

@edgars-batna I've just spent quite a bit of time going through your config files. Sorry to be a pain but just to be clear, you are saying that the config without pressure advance prints fine but the one with pressure advance is crap like in the video? What I mean is, there isn't some other fault crept in coincidentally with you enabling pressure advance? You can simply change M572 D0:1 S0.5 to M572 D0:1 S0.0 and it'll resort to printing just fine yes?

Thanks for going through the file and don't worry about being a pain. Yes, changing PA to 0.5 (or any value, actually) causes step skips to varying degrees depending on the value, 0.5 being that video.

I'll tell you what looks odd anyway........You have the mixing ratio set to 0.5:0.5.

Yes, it's intentional. Basically, if it's set to 1.0:1.0 mix ratio, then the WebControl reports wrong time estimations. This way it extrudes the correct amount, while "fixing" the filament-based estimated time left. I've gone through many iterations with my extruders and I'm 100% sure it extrudes the right amount as long as the steps are not lost.

I'll try setting it to 1.0:1.0 yet again, but I'm quite sure it didn't work. If mixing ratio somehow influences PA like this, then it's a bug. EDIT: Tried, no change.

The only other thing I noticed is that all your M569 commands have B, F and Y values which are firmware 2.02 and later - i.e. something new that has recently been added. Again, just speculation on my part but could this be the cause of the problem? Could there be some interaction between these values and the pressure advance algorithm? As this is something new in firmware 2.02 I see no harm in taking those parameters out and simply using M569 Pn Sn.

I've tried removing them to no avail. The motors are slightly less audible for me with those values. All motors run at those values, which would cause lost steps on X, Y, Z... The accelerations are quite high and any problems would pop up right away. EDIT: Tried, no change.

The thing about microstepping from my last post... Greatly increasing E microstepping has reduced the skips; This must be the first real sign of a parameter affecting the skips. I tried 64 before, but didn't notice much change. At 256 it is no longer marginal, it is totally noticable. That "crazy retract" is no longer happening, but the motors still knock.

deckingman

@edgars-batna I've resorted to comparing your config with mine - not very valid because they are different animals. One thing I noticed is that you have a very high feedrate set for your firmware retraction M207 S3 R0.0125 F20000 Z0.5. So 20,000mm/min which is 333 mm/sec. That's kind of crazy high. Another long shot but could that interact with PA and cause a problem???? Easy to try by using (say) F3000 (50mm/sec).

Clutching at straws here............

Edgars Batna

@deckingman The E speed is limited to F4500. The F20000 is an arbitrary high value. Nevertheless I tried reducing it to F1000 and then reducing max speed to F1000, but it yielded no change.

deckingman

@edgars-batna said in Extruder motors skip steps with pressure advance enabled:

@deckingman The E speed is limited to F4500. The F20000 is an arbitrary high value. Nevertheless I tried reducing it to F1000 and then reducing max speed to F1000, but it yielded no change.

Yeah, I knew the E speed would be limited, but from a conversation I had recently (about something else) with @wilriker, he was looking at the source code and seemed to be under the impression that the extruder jerk value was somehow used to calculate the acceleration for pressure advance. Hence there might have been a link to your problem. But you've tried it so that's another idea shot down........

I'm running out of ideas here. Another long shot - you don't have a config-overide.g file do you? If so, is there anything in it that might be overriding the changes you have made in config.g?

I'm starting to think that maybe there is a hardware problem. Maybe a faulty driver chip or some such. But that wouldn't explain why everything is fine until you enable pressure advance. Nothing makes any sense............

deckingman

OK, so here is something else. I've looked at that gcode you posted which gives you the worse behaviour. Right near the start there is an M204 S10000. Now according to the Wiki, https://duet3d.dozuki.com/Wiki/GCode#Section_M204_Set_printing_and_travel_accelerations

quote - "RepRapFirmware applies the M204 accelerations to the move as a whole",

But according to the wiki, M204 uses P and T parameters - there is no mention of an "S" parameter.

However, if that M204 is being applied, then 10,000 mm^2 is crazy high acceleration.

IMO, the slicer shoudn't be f***ing around with configuration values like that. (I'm really starting to hate PE version of Slic3R now).

So try taking out that M204 from the gcode file. Better still, turn off all "advanced features" of Slic3R PE and slice the file in the most basic manner possible. Even better still, try another slicer such as non-PE Slic3R.

Of course, none of his explains why you can print that file fine without PA but maybe there is some interaction (by design or accident) between a high M204 value and PA.

Just another long shot but worth a try.

Edgars Batna

@deckingman 10000 mm^2 is actually less than half of what the printer can. It has ultra-light head and XY axis from carbon fiber rods. Look at the accelerations in config.g: M201 X24000 Y24000 Z1200 and top speed: M203 X96000 Y96000. I built this to rip people's hands off if needed. If I run one of my test macros that runs 400mm back and forth, XY reaches 1600mm/s. But, remember, this is just travel. Print needs completely different values.

Anyway, I just finished a print with 256 E microstepping and, while not perfect, the results are impressive. Layer consistency went from zero to hero. Still some blobbing, but at least barely any missed steps. This layer inconsistency was the issue I've been fighting all along and it took me 12 months of trial and error and until I was finished with tweaking the hardware...

Now, I find it a bit underwhelming that none of the Duet3D folks have anything to say, because the behavior appears like a bug to me. I can dig into code, too, but I'm so rusty with Linux. Last time I used it was like 2015. Don't get me wrong, Duet3D is darn impressive, but I think this issue might affect more people than just me as I've seen exactly the same inconsistent layer heights here around the threads and I bet it's pressure advance and I bet it goes away by increasing microstepping.

If only someone would like to try reproducing the issue...

I suspect that low E steps per MM and lowish E microstepping breaks PA. I'm talking below 100 steps/mm and 16 or lower E microstepping. I suggest that pressure advance won't work on any setup like this.

deckingman

I hear what you say about the head being capable of acceleration of (double) 10000 mm/s^2. But my point was that, according to the Wiki, M204 applies to the entire move. That includes the extruder pushing filament through the hot end. Which in turn means that the melt rate of the filament has to accelerate at the same speed, which is kinda impossible. So the print head might be capable of ripping people's hands off, but there is no way on God's earth that the filament could be melted and extruded at those kind of speeds and accelerations. Non print moves fine but not extrusion moves.
Maybe I'm wrong but it'd explain the behaviour that's on your video. Or maybe as I said, there is some interaction between M204 and PA (yes, a bug if you like).
Anyway, easy to test to find out...

deckingman

@edgars-batna Oh and don't forget that in your config, you have the steps per mm set to 88 but then you change micro-stepping to 16X from 8X which gives you 176 steps per mm. If you remember, you did that deliberately so that you can set mixing to 0.5:0.5.

I kinda doubt that your E steps per mm and 16x micro stepping are what causes the problem because they are common values for non-geared extruders and AFAIK, nobody else has reported the same problem. But I admit that I'm clutching at straws and have run out of ideas.

wilriker

So, finally reached the end of this thread.

@deckingman Just some info here: M204 actually does understand the Snnn parameter (although it is not documented) and will act as if you provided the value to both Pnnn and Tnnn, i.e. M204 S50 will translate internally to M204 P50 T50. But either way machine limits (i.e. M201) always take precedence.
Simplified example: in case you have set M201 X20 Y20 Z20 E20 and then use M204 S1000000000 it will still limit to the 20mm/s² of M201.

@Edgars-Batna
That one out of the way: I also have ~100 steps/mm@x16 for E and used values for PA between 0.1 and 0.15 without having the issues described as you do. So I rather think it is some kind of issue with mixing extruders. We have seen issues there before. @deckingman should know.

You did ask about what happens if a extruder move is shorter than one microstep. There was a bug fixed with 2.02 regarding this where these extruder moves simply where dropped. Now it accumulates fractions of a microstep and once it has a full microstep accumulated it will issue this. (It should have been like that ever since but there was a bug).

A bit above you said that resetting mixing ratio to 1.0:1.0 did not have any influence. Can you confirm that you also adjusted the steps/mm in this case?

Also can you please once more post the block of settings with

M201 (accel)
M566 (InstantDv in mm/min) (and/or M205 - same but mm/s)
M204 (printing/travel accel)
M92 (steps/mm)
M350 (microstepping)
M567 (mixing-ratio)

as it is now? I lost track whether it is still the same as in config.g in your OP or has changed in between.

deckingman

@wilriker Good to know about M201 taking precedence over M204 - thanks for clearing that up (but Slicers shouldn't mess about with these things IMO).

And yes, I had problems with PA messing with arc moves using a mixing hot end which took about a year to get resolved. But it's fine for me now using around 0.4 - 0.5 PA (but my steps per mm are about 400).

Edgars Batna

@deckingman I suppose wilriker's post clarifies what happens with M204, which I can confirm works as expected.

@wilriker Thanks for looking through it. The thread is as long as the story behind this. I've done like a dozen modifications to various parts of my printer over the last year, but I don't complain. The printer got at least twice better in all aspects now.

Regarding mixing: I've already tried disabling it. Behavior is the same. There's no need to adjust steps/mm for 1.0:1.0 mixing if I set the same value at double microsteps. I'm happy to rerun any tests if required.

To reiterate, if I change this:
M350 X16 Y16 E8:8 Z16 I1 ; Configure microstepping with interpolation
M92 X80 Y80 Z2126.732 E88.208871:88.208871 ; Set steps per mm
M350 X16 Y16 E16:16 Z16 I1

to this:
M350 X16 Y16 E8:8 Z16 I1 ; Configure microstepping with interpolation
M92 X80 Y80 Z2126.732 E88.208871:88.208871 ; Set steps per mm
M350 X16 Y16 E256:256 Z16 I1

Then it skips way less in the worst (third) gcode I originally posted, but it still skips. Trivial prints do not appear to visibly skip, but there are a few minor layer inconsistencies which are with high certainty caused by a skip that is simply inaudible. I repeat, skips are not sporadic and are load-independent; the skips are deterministic.

Regarding mixing,
1.0:1.0:
M350 X16 Y16 E16:16 Z16 I1 ; Configure microstepping with interpolation
M92 X80 Y80 Z2126.732 E88.208871:88.208871 ; Set steps per mm
M350 X16 Y16 E16:16 Z16 I1
M567 P1 E1.0:1.0 ; set tool mix ratios more comments for greedy spam filter: let my posts alone!

0.5:0.5:
M350 X16 Y16 E8:8 Z16 I1 ; Configure microstepping with interpolation
M92 X80 Y80 Z2126.732 E88.208871:88.208871 ; Set steps per mm
M350 X16 Y16 E16:16 Z16 I1
M567 P1 E0.5:0.5 ; set tool mix ratios

M350 recalculates the steps/mm.

@wilriker said in Extruder motors skip steps with pressure advance enabled:

Also can you please once more post the block of settings

Until yesterday evening I ran with the values like in the original config.g. Then I tweaked the values a bit during a print (plus the E256 stepping config above):
M566 X2000 Y2000 Z600 E750:750 ; Set maximum instantaneous speed changes (mm/min)
M203 X96000 Y96000 Z800 E7500:7500 ; Set maximum speeds (mm/min)
M201 X24000 Y24000 Z1200 E1500:1500 ; Set accelerations (mm/s^2)
M906 X2100 Y2100 Z2000 E2400:2400 I40 ; Set motor currents (mA) and motor idle factor in per cent

M593 F20
M204 is per gode file, usually M204 S10000

The rest is same, still at 0.5:0.5.

@wilriker said in Extruder motors skip steps with pressure advance enabled:

You did ask about what happens if a extruder move is shorter than one microstep. There was a bug fixed with 2.02 regarding this where these extruder moves simply where dropped. Now it accumulates fractions of a microstep and once it has a full microstep accumulated it will issue this. (It should have been like that ever since but there was a bug).

But I'm sort of seeing the exact opposite. Instead of accumulating, it sort of forces a single microstep movement. Maybe the firmware got corrupted? Should I reupload?

@wilriker said in Extruder motors skip steps with pressure advance enabled:

That one out of the way: I also have ~100 steps/mm@x16 for E and used values for PA between 0.1 and 0.15 without having the issues described as you do. So I rather think it is some kind of issue with mixing extruders. We have seen issues there before.

I wouldn't move it out that fast. I can print most prints without too obvious problems at lower values, but then some prints just fail. Have you tried the exact gcode as I posted it? But I suppose it won't fit your printer... Maybe this is easier to work with: 0_1548425317168_EccentricGearTest 0 0.gcode

Would be interesting to see layer width consistency on your printer.

wilriker

@edgars-batna I looked into the file you posted. The first thing I noticed is that you sliced using volumetric extrusion. Just for testing could you slice that file again with volumetric extrusion disabled and see if this changes anything? Because that is something I don't use. And maybe you found an issue with volumetric extrusion and PA here.

I am currently in the middle of converting my machine to 24V and thus I cannot print the file you provided anytime soon, sorry.

Ref corrupted firmware: it is quite unprobable that a corrupted firmware will have just this one thing reversed. But it surely cannot harm to reinstall it.