Filament Drives

[Neotko]

I think it's related to the test of the ruby nozzle somehow. Anyway. When other tester 'test it' better. Maintly the concert for me was that you can see erros on the print than e3d blurs, not constant flow it's one of them. And again, it prints so nice

[johan]

I can imaging different reasons why the result of the Olsson 'bling' nozzles are better than the E3D nozzles. I would guess the better thermal conductivity of the material used is the main reason for this difference. To keep the plastic flowing smoothly through the orifice keeping the plastic at an even temperature there is crucial. Using steel or other materials with low thermal conductivity get more temperature fluctuations and an uneven extrusion. This difference in temperature and therefore in flow characteristics would impact the desired flow rate settings.

[ivan.akapulko]

With that it could be posible to run calibration according to each temp/speed and generate profiles for the esteps on each case. I would love an encoder actually controlling the real amount of filament move.

{offtop on}
It can count a filament on a roll, and make some warnings if he close to end?
{offtop off}

[Neotko]

There's a sollution for that ivan http://tunell.us/

And yes you are right johan we are going way offtopic XD

[Meduza]

Maybe some mod (that means LePaul right now) can split this thread into a new one about filament sensors and servo loop filament drive

[LePaul]

I don't know if I can split it up...I'd have to see what my admin account can do.

OR...someone could just start filament drive thread

(I'll check about option 1)

[LePaul]

I'll see if I can move the discussion that went off topic (from Hardened Nozzles) to here...but no promises

Please....resume the Filament Drive chat here =)

[Meduza]

So, how to do a reliable, not too expensive (to get adoption) accurate filament sensor?

I have been thinking about using a rotary encoder like this: http://www.ebay.com/itm/221765812822, cut out a small notch in the bowden tube, add some thin silicone heat shrink tubing to the encoder axle and spring load it against the filament.

[johan]

I want give a rough intuition about the problem of getting the extrusion amount just right. First of all we don't want to bother with the case of failure where the filament slips or we have missed steps, those are real problems of 3d-printing but not of importance when discussing the nature of the problem with setting the e-steps right. The question of whether we have a bowden setup or not is also not necessary to address because that introduces related but not the same type of problems. The only thing we actually need to consider is the length of filament you actually feed through the feeder (due to conservation of mass ).

The reason for the problem of too little filament fed is really a fundamental one, whenever you apply a force to some object the object will deform. How much depends on the force applied.

When we push filament through a 3D-printer nozzle we usually apply a large force to the filament to overcome not only the force to obtain the pressure necessary to push the molten plastic through the nozzle but also frictional and similar forces to feed the filament through the entire system. To realize the problem we only need to ascertain two simple facts, first of all that depending on how fast we try to push filament through the nozzle and at what temperature the force required to feed the filament will vary. Secondly, when we feed the filament the force we employ will deform the filament at the contact point and the deformation will depend on the force applied and this deformation will mean the filament is fed less forward. It is a bit of a leap that this is the source of the problem but with measured differences of material fed by 25 % depending on the rate of extrusion without observed missed steps or noticeable slippage gets me to the conclusion that this is a real effect.

The main point I want to make is that there is a fundamental relation between extrusion rates and the amount the filament that is actually fed when your feeder turns a certain distance. You will therefore need to change the amount of movement of your feeder motor depending on the aggregated back pressure of your extrusion system. The back pressure will depend on your extrusion rate, the design of your system and will also change over time when parts of your system wears out, the temperature of your hot end fluctuates, heat flows back up your filament path etc. I would say it is practically impossible to create a model which simulates all these time varying processes to attain a feed forward solution to the flow rate/e-steps parameter. Therefore I would say you will need a feedback filament feeder to address this problem in a satisfactory fashion.

Sorry for the wall of text .

ps. I would guess @Meduza agrees with what I have said above.

[Meduza]

I totally agree with Johan, i believe that a lot of the problems with slight underextrusion and such we see today is because of deformation and micro-slippage on the surface of the filament, factors that most people seem to just ignore or even deny that they exist.

A high resolution filament sensor right in the print head's cold part and running the feeder in some kind of servo loop is one possibility to actually know what is going into the print head instead of guessing... just throwing out ideas here

[Anders Olsson]

I can imaging different reasons why the result of the Olsson 'bling' nozzles are better than the E3D nozzles. I would guess the better thermal conductivity of the material used is the main reason for this difference. To keep the plastic flowing smoothly through the orifice keeping the plastic at an even temperature there is crucial. Using steel or other materials with low thermal conductivity get more temperature fluctuations and an uneven extrusion. This difference in temperature and therefore in flow characteristics would impact the desired flow rate settings.

The thermal conductivity of the ruby is actually closer to stainless steel than to brass. but since most of the ruby nozzle is made from brass the thermal conductivity of the orifice is not that important.
I believe that the improvement in overhangs comes from a combination of reduced friction (plastic vs. ruby), smaller shoulder and much higher manufacturing precision than for the brass nozzles.
(Sorry for the off topic but that post was moved here )

When it comes to feeder precision my view on things after using the direct feeder with 1.75 mm filament is that those two things are absolutely crucial to get decent precision. It is almost unbelievable what difference there is in the response from the feeder compared to the original bowden 3 mm setup.
I will record a video when I am finished with the XT-CF20 testing.

Apart from that a filament movement sensor would be nice of course.
I somehow would like a pressure sensor that measures the feeding pressure too and decreases the feedrate if it is obviously too high, like when starting a print with the nozzle too close to the platform. But all those things would require quite some clever programming and lots of testing.

[Neotko]

This looks like a cheap and micro rotary counter. http://sparklab-shop.de/parts/sparepart ... ing-system

[Meduza]

That could be a option if it good enough resolution

It would be interesting to just attach one of those just on top of the feeder, and log requested and measured extrusion distance with timestamps to see what the difference really are.

[Amedee]

From the doc it is a HAL sensor so I don't expect high resolution -- but probably good enough for looking at differences on a print.

I didn't knew Repetier firmware had that capability built in, interesting...

[Meduza]

I would like to have at least the same resolution as a 1/4step of the feeder motor, to be able to really see the deformation of the surface.