Extruder Calibration

[mastory 44]

So I just used Buildmemarlin to install a new copy of marlin. I changed the extruder constant to 865.888 as I have the newer extruder bolt. I made a couple prints (with skeinpypy and pronterface) and thought the threads looked narrow. I decided to check the extruder calibration. with limited code skills, I hacked together the following to extrude 100mm of filament.

G21 ;metric values

G90 ;absolute positioning

G21

G92 X0 Y0 Z0 E0 ;reset software position

G21

G92 E0 ;zero the extruded length

G21

M1 ;Pause for user input

G1 F75 E100 ;extrude 100mm of feed stock

M1 ;Clean the nozzle then press YES to continue...

G1 F75 E0 ;reverse feed back to zero

G92 E0 ;zero the extruded length again

G21

M84

There is probably some extra stuff in there, but it did work as I intended - except the resulting feed was only about 71mm. to measure, I took the filament out of the nozzle, pulled the bowden tube and measured the displacement right above the extruder.

I built another marlin with 926.5 as the extruder constant and redid the test. This time I extruded only about 79mm of filament.

I did find that when I extrude virgin filament, the displacement is slightly longer than when I extrude filament that's already marker by the extruder. Perhaps it feeds 5% more with virgin filament.

If I calculate a constant by ratio, I get an extruder constant that would be about 1200 to move 100mm filament. By looking at my print, it shouldn't need a correction equating to 130% of the current setting.

Is there something flawed in my understanding of calibrating the extruder? Something wrong with my gcode? Or is Pronterface altering the output?

Your thoughts please...

[mm_build 1]

I've been meaning to get around to this for a while, so thanks for giving me an opportunity.

I repeated this test several times each at speeds of 40, 75, and 200. All the results were too close for me to notice any difference. I measured ~93mm of extrusion in virgin filament, and ~94.5mm in previously extruded filament. (with an E value of 865.888)

My method was to disassemble the hot end, mark the filament with a razor knife flush to the bottom of the bowden tube, extrude 100mm, repeat. I took the filament afterwards and put it flat against a straight millimetre ruler. I would have tried your method too but I couldn't really follow from your description what you were doing. If your numbers are that far off I suspect it's something to do with how you are measuring, but I could be wrong. There is a range of extrusion rates that your print will still look OK with, even if it is not ideal.

According to my calculations I should be using 931 as my E value, which is closer to the value for the old hobbed (which I don't know off hand).

If you are confident in your method of testing, then I would suggest trying out the number it gives you and see how your prints look. I will be doing some more testing myself, in fact I will be putting my new values in and extruding 1000mm to see where that gets me.

The code I used to extrude is below

 

G21			(metric)G90			(absolute positioning)G92 E0			(zero the extruded length)G1 F75 E100		(extrude 100mm of filament)	G92 E0			(zero the extruded length again)

 

[mastory 44]

Thanks for your insight MM. The instructions in the wiki are dodgy in my opinion, and don't address the different software options.

To be more clear about my measurement, basically I pulled the bowden tube from the fitting at the extruder mechanism so the filament could come right out the top. I unlocked the drive mechanism and pulled the filament thru until I could close it on 'virgin' material. I closed the drive clamp and checked for 'proper' bolt tension. Wrapped a piece of tape around the filament flush with the push lock fitting. Ran the gcode to extrude 100mm. Then measured the displacement of the tape with dial calipers.

I will give 1200 a try as the constant. It just seems way out of line from the norm. By the way, I did check that the bolt is clean before testing.

Matt

[ddurant 0]

This stuff should really be deterministic and it's been bugging me for like 9 months that the math doesn't work. Feel free to look through

http://reprap.org/wiki/Volumetric_Dimension_settings

and see if you guys can make it work..

 

The instructions in the wiki are dodgy in my opinion, and don't address the different software options.

Er.. How so?

[mm_build 1]

To be more clear about my measurement, basically I pulled the bowden tube from the fitting at the extruder mechanism so the filament could come right out the top.

This could be a problem.. I've been trying to get my bowden tube out of the fitting for months. Just about destroyed it trying just now. Any tips would be appreciated.

BTW I did my extrusion test at 1000mm (speed of 40) and the length came out to ~927.5mm, which is pretty much right on target for what I was measuring with 100mm lengths. I will be setting my steps per mm in the firmware to 933.572

[mastory 44]

The push lock fitting should be quick to attach and detach. The tube may need to be fully seated before you can pull it out. Gently see if the tube is fully bottomed in the fitting. There is a small ring at the tube entrance. To release the tube , you have to push and hold the ring down and the tube should be released. Should pull right out.

I used a ratio to find my extruder constant and changed it 1179 and am now within .5mm of 100mm every time I advance 100. It is strange that it is so much greater than the norm. The fill print volume is a a lot higher now, but the layer adhesion seems to be better. I have been running at 220-230C. My gut tells me from watching it I can bring the temp down a bit now. So I was saying the fill looks a bit heavy - its like it gets 'fuzzy' between the rows of threads. to help that I think I will dial down the pack density.

Matt

[mm_build 1]

This stuff should really be deterministic and it's been bugging me for like 9 months that the math doesn't work. Feel free to look through

http://reprap.org/wiki/Volumetric_Dimension_settings

and see if you guys can make it work..

 

micro steps per rev * gear ratio / ( pinch wheel diameter * pi)

Alright, so microsteps per rev are 3200 (1/16 stepping, 200*16), Ultimaker's gear ratio is 49/8, or 6.125. Pinch wheel diameter I measured to between 7.5 and 7.9, let's just say 7.6 because I can measure the peaks but probably not all the valleys.

3200 * 6.125 / (7.6 * ~3.14)

=19600 / 23.8761

=820.90444331609173186055835844771

Ok, I see that is supposed to be the number of steps it takes to extrude 1mm of filament? But clearly that isn't the case.

[mastory 44]

A 5% error in the pitch diameter estimate puts it right in the ball park of the 865.888, But that doesn't explain why I have to set at 1179 to get an accurate feed.

I was getting prints with the standard setting and obviously better with the 962.5. I think that alot of people haven't calibrated, or maybe compensating some other way like pack density.

[mm_build 1]

Ok, so I think I'm onto something here. I backed off my tensioner all the way, then tightened it just a hair, literally just enough so it was touching ...barely at all. Then I did my test again extruding 100mm, I haven't changed anything in my firmware yet. Lo and behold... Wow, it's bang on! I tried the test 2 more times... yep it's extruding exactly 100mm. That's with 865.888 in the firmware.

Then I tightened my extruder back up to where I thought it felt about right. Did the test again.... 94mm this time. Ok so tighten it again... down to 90mm. Alright, let's tighten the hell out of this thing... suddenly I'm down to extruding 82mm instead of 100. So it appears that the amount of tension is causing some slippage of the bolt against the filament. This could actually be used as a metric to determine how tight people have their feeders.

mastory, I was able to finally get my bowden tube out thanks to your help. It does not look happy though, looks pretty beat up now I have to say.

I would suggest backing off your tensioner a bit and see where that gets you. There might still be another problem with your setup though, because I don't think I could have gotten it any tighter and I still wasn't reporting as short as you are. Maybe that combined with packing density. Have you manually measured your filament diameter yet? That could be affecting flow as well.

[Daid 306]

So it appears that the amount of tension is causing some slippage of the bolt against the filament. This could actually be used as a metric to determine how tight people have their feeders.

Or, the tension is causing the filament to compress into the bolt, which causes a smaller diameter to be used at bolt radius.

[mastory 44]

Well, I can only say from printing for a short time (a few weeks) that in order to get a positive drive without stripping the filament it needs to be tight. At least on my machine. Its not scientific, but on the high side of finger tight.

If the math above is definitive, working it backwards from the constant hat is yielding a consistent result (1179), the effective pitch diameter of my drive bolt is 5.59mm. The root diameter of the knurling is about 7mm+/-.25. Its difficult o measure. I don't it is possible for the pitch diameter to be smaller than the root dia.

Regardless, now that I have extruder drive that is calibrated, the print is looking really nice including the infill with an adjusted packing density of .95. I will try .90 next and see how that looks. Maybe I can bring the temp down a bit too.

That would be great if my bowden tube hadn't just popped out of the hot end. I guess the extra filament displacement and the positive drive worked together and found the weakest point.

Matt

[mastory 44]

This stuff should really be deterministic and it's been bugging me for like 9 months that the math doesn't work. Feel free to look through

http://reprap.org/wiki/Volumetric_Dimension_settings

and see if you guys can make it work..

 

The instructions in the wiki are dodgy in my opinion, and don't address the different software options.

Er.. How so?

ddurant,

I had not seen the page you are linking until now. It makes sense to me. I need to refer to rerap wiki more often. I was referring to the ultimaker wiki page that covers extruder calibration http://wiki.ultimaker.com/Calibration_and_Print_quality_improvements

What does 10,000 units refer to? There is no mention about M92 E14 when running Netfabb. That's a whole other topic, but with the mechanics the same, it is hard to balance that this calibration factor can be so concrete over many machines (with variable feeder tensions and users that seem to make big differences) just by changing software.

I really do appreciate the help yourself and the other experienced users offer here. My first comment above may have come across inflammatory. Better words could have been chosen. This learning curve is little frustrating, but I am on the road. As in many things, lesson from failure are most valuable. I think there are a lot more lessons ahead.

Thanks.

[Daid 306]

ddurant,

I had not seen the page you are linking until now. It makes sense to me. I need to refer to rerap wiki more often. I was referring to the ultimaker wiki page that covers extruder calibration http://wiki.ultimaker.com/Calibration_and_Print_quality_improvements

Never seen that page before. But I would ignore it if I was you. It doens't cover the new "volumetric" printing where 1 unit of E is 1 mm extrusion.

I just measured my extruder for the first time (had to take a few pics for my calibration wizard for SkeinPyPy) but my default of 865.888 seems to be quite good.

[ddurant 0]

Ok, I see that is supposed to be the number of steps it takes to extrude 1mm of filament? But clearly that isn't the case.

No, though I expect you have the right answer but mis-spoke.. It's not the steps needed to extrude 1mm, it's the steps needed to pull 1mm of filament into the extruder.

The stuff about the correct esteps value varying based on the little tensioner screw is interesting.. I would very much like to understand why that's happening.

[ddurant 0]

...My first comment above may have come across inflammatory. ..

Was it? I didn't get that from it but will go back and read it again to see if I can feel offended..

My response was just asking for more details, since choices are really only "volumetric 5D" and "not volumetric 5D". Software either has it or it doesn't so there's really not much per-software stuff to this.

And, as Daid said, now that you're up to volumetric 5D, avoid instructions that talk about not-volumetric-5D.

[mastory 44]

And, as Daid said, now that you're up to volumetric 5D, avoid instructions that talk about not-volumetric-5D.

Good point and thanks.

Next I will back off my tensioner to something way less than it is and see what kind of output I get and if the extruder can transmit enough force without stripping the filament. Of course, first I have to fix the hot end of my bowden tube that popped out as a result of the increased filament displacement. I am going to use some of the improvements by others or engineer my own improvement for this.

Thinking about the extruder... If an extruder drive could be designed that does not bite into the filament, it could be engineered to slip when overdriving the hot end, but provide a positive feed otherwise. It seems there is a race between delivering enough filament to the hot end and the strength of the bowden tube connection.

Am I pushing too hard or is the bowden tube not strong enough?

I thought I was just tweaking my machine to have the right amount of feed. So far I have not had melted plastic leak between my bowden and the brass screw.

What is prevents this from happening? Is it just not hot enough in that area?

Matt

[mm_build 1]

So it appears that the amount of tension is causing some slippage of the bolt against the filament. This could actually be used as a metric to determine how tight people have their feeders.

Or, the tension is causing the filament to compress into the bolt, which causes a smaller diameter to be used at bolt radius.

That was my first thought as well, but I pushed 1000 ACTUAL mm of filament with my old tension (92.75% effective rate), and it weighed 7.8 grams. Then I backed off the tension until I was pushing filament at 100% effective rate and weighed 1000mm of filament again. On my scale the two lengths weigh the same, or within 0.1 gram of each other. At 92.75% effective rate, my metre of compressed filament should have weighed ~7.23 grams if the issue was due to the filament being of smaller diameter.

However as I compressed my filament more, the smaller diameter started to become apparent. Weighing a 1000mm of filament pushed at 83.25% effective rate weighed 6.8 grams on my scale. So the smaller diameter does come into effect at some point, but it is not proportional to the amount of difference being seen in the extrusion rates. It would appear that when the filament is deformed going through the hobbed bolt, most of the volume is being pushed sideways, creating a filament that is wider and flatter, but overall about the same volume. Only after compressing it past a certain point it would appear the compression starts to extend length-ways down the filament, resulting in less volume of extruded material. At 83.25% ratio, the volume was still approximately 87% of total.

To get any more precise I would need a more accurate scale, or a lot more filament.

[mm_build 1]

It's not the steps needed to extrude 1mm, it's the steps needed to pull 1mm of filament into the extruder.

You are of course, technically correct. (That's the best kind of correct)

[joergen 2]

ddurant,

I had not seen the page you are linking until now. It makes sense to me. I need to refer to rerap wiki more often. I was referring to the ultimaker wiki page that covers extruder calibration http://wiki.ultimaker.com/Calibration_and_Print_quality_improvements

Never seen that page before. But I would ignore it if I was you. It doens't cover the new "volumetric" printing where 1 unit of E is 1 mm extrusion.

I just measured my extruder for the first time (had to take a few pics for my calibration wizard for SkeinPyPy) but my default of 865.888 seems to be quite good.

As the original 'author' of the 865.888 value, it was derived from running the filament through the cold UM without nozzle with the new bolt and proper bolt-filament friction, or in other words, without any hot extrusion pressure load.

when adding the nozzle and this whole molten plastic stuff into the equation, it might very well be the wrong value, or at least YMMV... but I think I always said that one should make their own measurements, and I was very surprised to see this value used so often (not like the XY steps/mm setting, which should be the same with everybody).

My theory for having a value that is significantly different from 865.888 is that the setup might have too much friction (see the videos I posted in the google group), or the temp is too low to extrude 100mm stock with the given steps, or the stepper driver pot is set too low. it might be that you are skipping steps, not like 10-100 at once (which would be noticeable skipping), but 10-30% overall.

[ddurant 0]

I wonder if it'd be a worthwhile experiment to switch the extruder motor to 1/8 microstepping and see if that also varies with tension..

[ddurant 0]

It's not the steps needed to extrude 1mm, it's the steps needed to pull 1mm of filament into the extruder.

You are of course, technically correct. (That's the best kind of correct)

Usually (sometimes.. occasionally) I'm not that picky but in this case, it's a distinction I'm careful to make because non-volumetric-5D is about length out of the nozzle and volumetric-5D is about length into the extruder drive..

[joergen 2]

I wonder if it'd be a worthwhile experiment to switch the extruder motor to 1/8 microstepping and see if that also varies with tension..

I don't really see the reason why the extruder is running with 1/16 instead of 1/8 steps, considering this step of the print process needs far more torque than precision, while the z-stepper could easily run with 1/16 instead of 1/8 steps, and benefit from the extra resolution.

[ddurant 0]

I wonder if it'd be a worthwhile experiment to switch the extruder motor to 1/8 microstepping and see if that also varies with tension..

I don't really see the reason why the extruder is running with 1/16 instead of 1/8 steps, considering this step of the print process needs far more torque than precision, while the z-stepper could easily run with 1/16 instead of 1/8 steps, and benefit from the extra resolution.

Er.. What?

Are you saying testing esteps vs tension at 1/8th is good idea or a bad idea??

[joergen 2]

I wonder if it'd be a worthwhile experiment to switch the extruder motor to 1/8 microstepping and see if that also varies with tension..

I don't really see the reason why the extruder is running with 1/16 instead of 1/8 steps, considering this step of the print process needs far more torque than precision, while the z-stepper could easily run with 1/16 instead of 1/8 steps, and benefit from the extra resolution.

Er.. What?

Are you saying testing esteps vs tension at 1/8th is good idea or a bad idea??

I am saying 1/8 steps for resulting in something like E433 would be nice... I would almost venture that 1/4 steps (resulting in E216) would be sufficient (and lead to greater torque and maybe even greater speed for E).

I am also saying that it would be good to change the Z to 1/16, resulting in Z1066.

[mm_build 1]

I wonder if it'd be a worthwhile experiment to switch the extruder motor to 1/8 microstepping and see if that also varies with tension..

You are thinking the difference might be due to lost micro-steps? I have nothing better to do right now so I'll check it out.