See also https://forum.duet3d.com/topic/18412/rrf-2-03-pressure-advance-causes-20-overextrusion/21

Steps as counted by logic analyzer connected to step and direction pins:

Without pressure advance:
fwd_count 1942964
rev_count 961423
net_count 981541
mm 2484.913924

Same file with pressure advance set to 1.0:
fwd_count 6334445
rev_count 5157566
net_count 1176879
mm 2979.440506

Logic analyzer output, no pressure advance
Logic analyzer output, pressure advance 1.0
Python script to count steps

With no pressure advance: https://drive.google.com/file/d/1oFvrJnxZc6fvLlGZqTdCqO7GB4Zfzdy_/view?usp=sharing

Python script output:
fwd_count 1942964
rev_count 961423
net_count 981541
mm 2484.913924

2.48m, dead on what the slicer expected.

Proof positive that PA causes overextrusion and that this is a firmware bug, NOT a mechanical or configuration issue.

For now this is confirmed only on 2.03, will do an upgrade to at least 2.05 and probably 3.1.1 soon.

@bot tried reducing the resolution of the STL to about 0.5mm and it still totally overextrudes, but turning it down in the slicer seemed to work better... weird.

I am out for the weekend now, so no more experiments for a while.

Need to grok the code. Sounds like it needs some low level improvements on top of any improvements to pressure advance, based on what @Edgars-Batna is saying.

Funny, I just quit my software job, was hoping it was forever. Didn't even last a day.

@deckingman So why the hell is bowden so popular? It isn't simpler, it isn't cheaper, it isn't better.

@bot said in bug: pressure advance causes overextrusion:

@jschall It seems like you've lost interest, based on the other thread. What model of logic analyzer were you using? Do you have any tips on how to get quickly up and running measuring step pulses in the way you did?

I used a Saleae.

I will still test 3.1.1...

I'm really surprised there's no counter anywhere for total steps taken on each drive. Maybe I should try to add that (would have to work out how to build).

Alternatively I can hook up a logic analyzer and have it count the steps...

Logic analyzer capturing, wrote a python script to analyze the results, running the file with pressure advance now, then I'll run without.

https://gist.github.com/jschall/98f9459590cfad6bbe413f3b49c266eb

No underruns.

Hi, bought an E3D ToolChanger about a month ago, totally new to RRF and duet.

https://github.com/Duet3D/RepRapFirmware/issues/421

I have my esteps calibrated correctly. I can have my printer extrude a given length of filament and the extruded filament will be just under the expected length and it will weigh what I expect it to weigh.

I reduced my extruder jerk, acceleration and max speed to ensure no missed steps.

Nonlinear extrusion is disabled.

However, when I print this part with pressure advance on (set to 1.0, calibrated using this script), it weighs 20% more than I would expect it to weigh based on the total extrusion in the file. If I have pressure advance set to zero, it weighs just under the expected weight.

I've tried with and without retractions, with the slicer set to generate relative moves, and with the slicer set to generate absolute moves.

At the end of the print, the reported extruder position is almost exactly what it is supposed to be.

So, questions:

• Which variable exactly in the code does the reported extruder position come from?
• Can someone give me an introduction to the codebase? Is there developer documentation somewhere?
• Is "COMPENSATE_SPEED_CHANGES" typically enabled by default? I have whatever firmware came on the board.
• Is "stepsPerMm" logged somewhere?
• Is there a way to count total steps on the extruder drive?

jschall created this issue in Duet3D/RepRapFirmware

closed Part 20% over weight predicted by slicer #421

@phaedrux so, accelerometer required? No way to tune without? Where do I buy one compatible with my duet 2 wifi?

Things improved greatly with input shaping disabled. How do I properly tune input shaping? I don't have an accelerometer.

Input shaping was actually enabled (Input shaping 'zvd' at 50.0Hz damping factor 0.10, min. acceleration 10.0, impulses 0.334 0.822 with durations (ms) 10.05 10.05) - I am re-running the test.

Physical setup: E3D ToolChanger with direct Hemera extruder, PolyMax PC @ 285C.

RRF 3.4.0, input shaping and nonlinear extrusion off (should be, not configured - will check)

I am tuning pressure advance using the python script - that part is working - the print is changing as the Z axis gets higher - for example, when pressure advance is too low, there are the expected holes where the nozzle is accelerating. The pictured sample is from 0.0 to 0.2. I've gone 0.0 to 1.0, and it starts to totally fail just beyond 0.2.

However, on every acceleration and deceleration, there is a thin spot regardless of the pressure advance setting. What's weird is that it is on both sides - I'd expect a thin spot on accel and a thick spot on decel with PA too low, and the opposite with PA too high. I can't think of anything that would cause this. I tried tightening belts as well - they're tightened up to 8lbs.

What is interesting is that this doesn't happen nearly as much with my bowden extruders - it is like there's some bug or issue or missing tuning feature in RRF that is being covered up by the slow response of the bowden extruders.

Any suggestions?

@deckingman said in RRF 2.03 pressure advance causes 20% overextrusion:

One last contentious comment is that as mass increases, the resonant frequency decreases. Is it purely coincidence that I never get "ringing" or "ghosting" when using my heavy hot ends but people with lightweight hot ends offen have those problems?

What is your jerk set to?

@deckingman So why the hell is bowden so popular? It isn't simpler, it isn't cheaper, it isn't better.

@deckingman said in RRF 2.03 pressure advance causes 20% overextrusion:

If a Nema 17 is perfectly capable of attaining printing speeds of >300mm/sec with a 2Kg mass, then it is implicit that it must have sufficient torque to accelerate that mass up to that speed in less than half the axis length.

So you are actually getting achieved speeds of 300 mm/s with a NEMA 17 pushing around 2kg?

Still, acceleration matters a lot. It dominates when printing smaller parts. What kind of accelerations are you achieving?

@jschall Can those servos meet the speed/acceleration achieved by a stepper? Not saying they can't I just have no idea of what they can do.

Yes. Regardless, not suggesting the use of those specific servos.

@Veti said in RRF 2.03 pressure advance causes 20% overextrusion:

i think the biggest problem is that they cant do a full revolution

That's just because of mechanical stops in the gearbox to prevent the potentiometer from having problems. In an extruder application, you'd have a somewhat different configuration. Remember, this is just more of an example of how cheap, small and light an integrated servo motor could be, not suggesting that these are options for our application without significant modification.

@JRDM said in RRF 2.03 pressure advance causes 20% overextrusion:

Hobby servos are very limited on what they can do for that kind of use because the type of input doesn’t allow for the level of control needed.

@JRDM said in RRF 2.03 pressure advance causes 20% overextrusion:

Hobby servos also use deep multistage gearboxes, which induce excessive backlash.

I'm not talking about using hobby servos. I'm using hobby servos as an example in a first-principles argument that steppers are dumb in extruders. Backlash not important for extruder. Just means you need very slightly more retraction to take up the backlash.

@deckingman said in RRF 2.03 pressure advance causes 20% overextrusion:

So if carriage mass isn't a limiting factor, then why reduce it? In fact I have demonstrated this by printing at up to 300mm/ sec with a moving carriage mass of around 2Kgs driven by modest NEMA 17s. I'll make a other contentious statement that adding mass reduces the resonant frequency - I don't get ringing -ever.

It's more about acceleration than speed. A NEMA 17 could accelerate a locomotive to 300 mm/s, assuming low enough friction. It'd just take longer.

@jschall said in RRF 2.03 pressure advance causes 20% overextrusion:

@Veti

Yeah totally agree.

Bowden is fking stupid. The whole premise of bowden is to REMOVE a motor that needs to output LESS THAN 1W OF MECHANICAL POWER in order to SAVE MASS on something that weighs in at like 500 grams. Hey, here's an idea: how about instead of that, we stop using half-pound 1970s-era stepper motors?

Here's a servo that outputs more torque than the geared titan extruder. It costs a whopping $10 - which is cheaper than the stepper. It weighs 58g vs 280g for the stepper. It includes the gears, which add not-insignificant mass as well. I'm not trying to say you can buy that servo and stick it on an extruder and it will work unmodified - it won't. I'm pointing out from a first principles point of view that you can cut out 80% of the mass of the extruder, without bowden, and it doesn't have to be expensive.

https://hobbyking.com/en_us/hobbykingtm-high-torque-servo-mg-bb-w-proof-12-8kg-0-22sec-58g.html?queryID=76697295b4ca978c54e06a560255a667&objectID=47699&indexName=hbk_live_magento_en_us_products_hbk_price_stock_2_group_0_desc

@Phaedrux said in RRF 2.03 pressure advance causes 20% overextrusion:

Doesn't it seem silly to replace a one of the walls with a travel move though? It's still taking the time to move around the entire perimeter it's just not extruding anything.

Not really, because the only case where it isn't extruding anything is the case where the wall thickness is exactly equal to the line thickness. But yeah, could be optimized.

What slicer does the absolute best with thin walls? I bought Simplify3D and was horribly disappointed. Cura so far seems to do the best job overall.