New beta firmware 1.20.1RC2 and DuetWiFiServer 1.20+1

burtoogle

I have done a few prints using the non-linear extrusion feature and can confirm that it's working OK. I no longer have the workaround in place that I did previously (extruder was set to 104% and first layer was printed at 95% extrusion). I now have the extruder set to 100% for all layers and can see no evidence of over or underextrusion. Looking good.

deckingman

Glad it's working for you. Personally, I've never had a need for extrusion multiplier greater than about 98% but then, I'm seeing quite different results on the static tests. Do you by any chance have any pics of with and without?

JohnOCFII

@dc42:

Please report good, bad or neutral feedback on this release here. Unless any serious bugs are reported, these binaries will be re-versioned as 1.20.1 by the end of this week.

I installed both DuetWifiFirmware and DuetWiFiServer yesterday. Basic WiFi connectivity was stable for 10+ hours of operation. Basic features (movement, heaters, fans, homing, calibration) all seemed fine on my Kossel mini

Dougal1957

installed yesterday about 15:00 printer has been on and stable connection since then currently 3+ hours into a print with no issues

burtoogle

Do you by any chance have any pics of with and without?

Sorry, no.

To be honest, I don't think there's going to be much visible differences because we're only talking about a few percent here and there. Having said that, I do think that it's worth using because now I know that for any extrusion rate up to the point at which it starts skipping I can be reasonably confident that it will print the amount of plastic demanded by the gcode. Whereas, before, I was getting -10% at the high rates.

joestefano

2 WiFi's and 1 Ethernet board all printing fine. No WiFi disconnects. Seems very stable

deckingman

@burtoogle:

Do you by any chance have any pics of with and without?

Sorry, no.

To be honest, I don't think there's going to be much visible differences because we're only talking about a few percent here and there. Having said that, I do think that it's worth using because now I know that for any extrusion rate up to the point at which it starts skipping I can be reasonably confident that it will print the amount of plastic demanded by the gcode. Whereas, before, I was getting -10% at the high rates.

That's a shame (no pics I mean).

I'm a bit confused though. You said in a post above that you were running 95% extrusion on the first layer but 104% on subsequent layers and that now you use 100% throughout. The 104% I can understand because it indicates that you were seeing some under extrusion issue, presumably caused by slippage and for which you have now compensated a different way. It's the first layer that I'm trying to get my head around. It would indicate that you were seeing 5% over extrusion before, because you were using 95% extrusion multiplier. Yet you've added a positive compensation plus taken out the 5% negative compensation. Assuming you print the first layer at half speed of other layers, then the slippage compensation would be around 2% so you must be laying down 7% more filament on the first layer than you were before, which ought to be hugely noticeable. It would on my machine anyway. Even with no slippage compensation, changing the extrusion multiplier from 95% to 100% is a pretty big step. Something very odd going on….

burtoogle

Look, it's simple. Before I started using this new feature, I noticed that at my typical infill/skin speed I was getting underextrusion. So I upped the steps/mm for the extruder to 104% but then because I print my first layer walls really slow (so little underextrusion was occurring), I was then getting fat lines. So in the slicer I reduced the first layer extrusion to 95% to compensate. This way I was getting closer to 100% for all layers. But it was a gross simplification. The non-linear extrusion provides a much more accurate means reducing the effect of feedrate changes on extruded volume.

burtoogle

Hi David, I am now trying out M592 on my Kossel XL which is fitted with a e3d v6 and a flex3drive extruder. Filament is PETG at 240C. I'm getting very odd results:

mm/s   A0 B0    A0.015 B0.0012
1        52.5        52
2        51          50.5
3        50          48
4        47          44
5        44          38

The larger I make A or B, the worse it gets!

I am not perceiving any extruder skipping or other odd noises.

The flex3drive extruder uses a lot of gearing so it's running at around 2250 steps/mm with the stepper set to 8 microsteps.

Any thoughts?

deckingman

@burtoogle:

Look, it's simple. Before I started using this new feature, I noticed that at my typical infill/skin speed I was getting underextrusion. So I upped the steps/mm for the extruder to 104% but then because I print my first layer walls really slow (so little underextrusion was occurring), I was then getting fat lines. So in the slicer I reduced the first layer extrusion to 95% to compensate. This way I was getting closer to 100% for all layers. But it was a gross simplification. The non-linear extrusion provides a much more accurate means reducing the effect of feedrate changes on extruded volume.

Ahh, sorry. But you didn't make any mention in your earlier post that you'd changed the extruder steps per mm at the same time, hence the reason I was confused. So your 95% for the first layer was actually 95% of 104% which makes a lot more sense.

dc42

@burtoogle:

Hi David, I am now trying out M592 on my Kossel XL which is fitted with a e3d v6 and a flex3drive extruder. Filament is PETG at 240C. I'm getting very odd results:
mm/s   A0 B0    A0.015 B0.0012
1        52.5        52
2        51          50.5
3        50          48
4        47          44
5        44          38
The larger I make A or B, the worse it gets!

I am not perceiving any extruder skipping or other odd noises.

The flex3drive extruder uses a lot of gearing so it's running at around 2250 steps/mm with the stepper set to 8 microsteps.

Any thoughts?

Yes, that looks odd.

1. If you revert to A=0 B=0, does it revert to the speeds you measured before at A0 B0?

2. If you send M592 D0 does it respond with the values you configured, and a sensible value for L?

3. Please confirm that when doing these tests, you are doing head movement as well as extrusion.

burtoogle

Yes to all 3 of your questions.

I can't fathom what's going on here. I'm getting some variation in the results from run to run.

A question: do the A and B get scaled by the steps/mm value for the extruder? i.e. if the steps/mm value is large will that mean that the values for A and B will be larger compared to the values used for an extruder with lower steps/mm?

dc42

The calculation it does is:

	const float dv = dda.directionVector[drive];
	float stepsPerMm = reprap.GetPlatform().DriveStepsPerUnit(drive) * fabsf(dv);
	const size_t extruder = drive - reprap.GetGCodes().GetTotalAxes();

#if NONLINEAR_EXTRUSION
	if (dda.isPrintingMove)
	{
		float a, b, limit;
		if (reprap.GetPlatform().GetExtrusionCoefficients(extruder, a, b, limit))
		{
			const float averageExtrusionSpeed = (dda.totalDistance * dv * DDA::stepClockRate)/dda.clocksNeeded;
			stepsPerMm *= 1.0 + min<float>((averageExtrusionSpeed * a) + (averageExtrusionSpeed * averageExtrusionSpeed * b), limit);
		}
	}
#endif</float>

Let me know if you spot a problem with it. dda.totalDistance is always positive, and dda.directionVector[drive] will be positive if a positive amount of extrusion is commanded. I guess the calulation would go wrong if you did a simultaneous movement and retraction.

burtoogle

I did look at that code but I'm not familiar with the RRF source so I can only guess as to how it is working. I do agree that the calculation code looks like it is doing the right thing.

I will continue to play with my XL - let's assume that my odd numbers above are due to just clumsiness on my part. I'll get back when I have done more testing and have come to any kind of conclusion.

Captain_sq

I'm not very familiar with C++ but should't a, b and limit be pass by reference rather than pass by value?

dc42

Passing them by value is best here because their values are read but not changed.

Captain_sq

@dc42:

Passing them by value is best here because their values are read but not changed.

Yes, I see, but are the variables initialized right here? My guess is that their values is supposed to be set in these two statements.
a, b and limit are passed by value here and not by reference. My guess is that the GetExtrusionCoefficients() method expects a pointer
to be able to write the values into the variables, like this:

    float a, b, limit;
    if (reprap.GetPlatform().GetExtrusionCoefficients(extruder, &a, &b, &limit))

dc42

I'm sorry, I forgot what that bit of code looked like. The a, b and limit parameters of function GetExtrusionCoefficients are already passed by reference, because they are declared as reference parameters in the declaration of GetExtrusionCoefficients. This is a C++ feature not available in C.

burtoogle

Hi David, as I said, I'm not familiar with the RRF code so these may be dumb questions but in this code snippet, why are you multiplying totalDistance by dv and not fabsf(dv) and why do that multiply at all? Isn't dv a distance also?

const float averageExtrusionSpeed = (dda.totalDistance * dv * DDA::stepClockRate)/dda.clocksNeeded;

dc42

Variable dv is the proportion of totalDistance that applies to the current drive, that's why totalDistance is multiplied by it. It shouldn't matter whether we multiply by dv or fabsf(dv) because nonlinear extrusion is only applied to extruding moves, for which dv is positive.