Flow sensor for the UM2+ (and even perhaps for the UM3) - Call for pilot users

[foehnsturm 970]

This is a project by a group of community members which was also involved in the Mark2 dual extrusion upgrade.

More precisely, it's me coming up with an outside the box approach / weird idea for a certain unresolved problem. Smart people like @gr5, @Anders Olsson, @Dim3nsioneer, @rooiejoris throwing in ideas and @tinkergnome who implants the stuff into firmware.

 

My impression of the current state of development when I started this was as follows. There have been filament monitor projects since the beginning of reprap. Only very few made it to some kind of product state, like the one by Aaron Tunell. Manufacturers like Prusa and others recently introduced some kind of filament monitors, with mixed success / reliability issues. The Duet3D guys set their hardware research (laser-based and rotating) on hold because they were experiencing inaccuracies of +/-20%. Well and then there was Ultimaker ... until yesterday with the S5   All these efforts have been or still are struggling to fulfill the most important objective: NO FALSE ALERTS. Otherwise any filament sensor would quickly render itself useless.

 

What we want to achieve

 

Objectives, the obvious part:

• zero false alerts
• detect filament runout ("nothing there")
• detect filament grinding ("nothing/very little moves")

 

Objectives, the challenging part:

• detect first layer issues (see video below)
• detect when real flow leaves a certain safe process window and starts to compromise part quality
  (first, inter layer adhesion will suffer, then classical under extrusion will be visible)
• and try to counteract, that's where the real fun starts 
• ...

 

 

Current state of development

 

We chose an encoder and there's a reliably working prototype for an easy to attach external flow sensor, mounted to the entry side of the feeder. Resolution is in the range of 0.015 mm. It's integrated in Tinkerware with a dedicated menu and we (well, he) implemented a gcode command:

M591 T0 S1 E0.5000 L0.01695 R35:130 A0.3 P100.00

I leave the parameter interpretation as a little quiz here.

 

Right now I'm working on a modified design which, besides the encoder, doesn't need some parts which cannot be printed and are in the +30€ range to have them manufactured. But most likely some parts will still not be FFF printable.

 

How can I get this?

 

First give us some more time to test and evaluate. If everything works like intended we might proceed like with the Mark2 project.

If we should offer this as a product I'd expect a price tag between 70-100 €. 

 

And the UM3?

 

That's the BIG question. Like @Daid recently stated their main market is already different. And indeed, has anyone seen any kind of (hardware) upgrade for the UM3 so far?

Quote

As far as I know, the flow sensor won't become available as upgrade kit for the UM3. We had that as option on the table, but I though the latest design no longer fits directly on an UM3 mechanically. And the market that we are in right now isn't super fond on upgrade kits, even if you guys are ?

 

Feeders are the same, mechanically our sensor fits. Electronics, not sure. Ultimaker originally wanted to use a serial interface on the UM3. For the UM2+ we simply connect the sensor's quadrature output signal to free I/O pins, there are enough left (4) for two sensors for a Mark2 dual extrusion UM2. Ultimaker won't do anything to support a sensor on the UM3. Anyway, if a large number of UM3 users would show interest, they might at least not impede a development ...

Edited June 26, 2018 by foehnsturm

[neotko 1,417]

Main issue from UM3 is that the firmware must be signed by UM (like apple firmware). Ofc this can be circumvent as posted here. The problem anyhow is that is way too complex for a printer to do that, and there’s a Huge (and by pms from 3 um3 users, and my experience when I had one, is way too easy to brick the um3 by just writing a bad comma or dot).

 

Check Oliver instructions to upload an unsigned firmware

 

http://www.3dprintingforum.us/viewtopic.php?f=34&t=12&start=930#p9062

 

Edited April 24, 2018 by neotko

[robinmdh 100]

2 hours ago, foehnsturm said:

until yesterday with the S5

Well we've been working on it since before the UM3 launched, the UM3 was supposed to have filament flow sensors but you guessed it it was rife with inaccuracy.

I'd also say there have been significant SW upgrades to the UM3 but there certainly is place for a flow sensor there, some (maybe all) UM3's even have extra wires going to the feeders for that purpose.
 

2 hours ago, neotko said:

Main issue from UM3 is that the firmware must be signed by UM (like apple firmware). Ofc this can be circumvent as posted here. The problem anyhow is that is way too complex for a printer to do that, and there’s a Huge (and by pms from 3 um3 users, and my experience when I had one, is way too easy to brick the um3 by just writing a bad comma or dot).

With the 5.x.x version we are working to make unbricking/installing SW by USB stick possible, so that could become less of a hassle... (but also less secure as no signature required, this is a side-effect of some other changes else)

[rigs 735]

First, thank @foehnsturm ( and all ) for this development work, now we will never be blind again

I print 1/2 kg parts every day and have always wondered "why is the weight lighter than the weight in the slicer ? "

 

Here is a simple test of several printed samples ( the same ones ) that I weighed.

In Red: without retractions
In blue: with retractions

Carbon/PETG Composite
Nozzle 0.4 mm
Bondtech QR

 

 

This under-extrusion is a combination of several factors :
- viscosity
- shear rate
- Fourier heat exchange
- motor torque
and more.....

 

I think it would be interesting to view the sensor information via USB --> curve graph.

 

I can't wait to use this sensor

 

Edited April 24, 2018 by rigs

[uncle_bob 21]

Hi

 

Why not just stuff a little MCU into the mix and make your sensor look like their S5 sensor? As long as they have unified firmware across the platforms ( do they ???? ) it should just be a line in a config section to turn it on or off. 

 

Yes, this very much requires you have a bit of info about what their sensor "looks like" in terms of it's I/O data stream. Anybody taken a hammer to an S5 yet?  

 

Bob

[foehnsturm 970]

Feeding 40 mm filament with the filament tip just leaving the feeder, no resistance due to bowden or hotend.

Blue: extruder e-position  = number of steps the extruder performed  / steps per e

Green: measured e-position = number of flow sensor steps / flow sensor resolution 

 

When looking at longer e-distances there are minor, periodically repeating deviations which indicate a slight eccentricity of the flow sensor drive gear which should be fixed with the next iteration.

 

Edited April 30, 2018 by foehnsturm

[conny_g 251]

Interested. Following this. ?

[neotko 1,417]

4 hours ago, foehnsturm said:

Feeding 40 mm filament with the filament tip just leaving the feeder, no resistance due to bowden or hotend.

Blue: extruder e-position  = number of steps the extruder performed  / steps per e

Green: measured e-position = number of flow sensor steps / flow sensor resolution 

 

When looking at longer e-distances there are minor, periodically repeating deviations which indicate a slight eccentricity of the flow sensor drive gear which should be fixed with the next iteration.

 

 

The other value making micro errors could be how much the tooths cave in the filament (depends from filament to filament) changing the actual dia/rotation

[uncle_bob 21]

Hi

 

There's also a tolerance on the wheel / gear diameters that can get into this. If the filament is not quite round ( it never is perfect) a twist between the wheels could result in errors in either direction.  As a practical matter - do I really *care* below some percentage? My. guess is that the don't care threshold is somewhere in the 1% region. I must admit this is the fist time I've seen a way to *test* that theory.

 

Bob

 

[foehnsturm 970]

14 hours ago, neotko said:

The other value making micro errors could be how much the tooths cave in the filament (depends from filament to filament) changing the actual dia/rotation

Good point. As far as I can tell that makes the e.g. difference between GreenTec and regular PLA. The GreenTec seems to deform a little more on the drive gear due to the back pressure. I actually see a few percent less extruded than with other PLA at modest speed. Which matches the 105% flow setting you recommend for it

[robinmdh 100]

7 minutes ago, foehnsturm said:

deform a little more on the drive gear

why not just measure the filament before the deformation has occurred?

 

and how much the feeder moves the fillament was alwais a calibrated value and varies between materials a bit due to how much the feeder wheel digs into the material. you could compensate for that but you might need more information about why under extrusion is happening...

perhaps material profiles including calibration data for actual e steps?

Edited May 1, 2018 by robinmdh

[foehnsturm 970]

20 minutes ago, robinmdh said:

why not just measure the filament before the deformation has occurred?

 

That's what we are doing. I wanted to say, when the feeder dents dig into the GreenTec and there is back pressure the material deforms more than other PLA (dent marks in the filament soon getting closer than they ideally should be) .

 

We already discussed the idea of a self calibration function: Do another retract before the purge / prime at the beginning of a print. Then feed the retracted length again and calibrate the flow sensor to 100%. But then this 100% would never match the requested extrusion length in reality because we're already ignoring a few percent which might be lost in the extrusion train. Or we could do it the other way round: Do the calibration and increase flow until 5mm requested feed really transforms to 5mm moved filament. Then print with the increased relative flow.

[conny_g 251]

1 hour ago, foehnsturm said:

Good point. As far as I can tell that makes the e.g. difference between GreenTec and regular PLA. The GreenTec seems to deform a little more on the drive gear due to the back pressure. I actually see a few percent less extruded than with other PLA at modest speed. Which matches the 105% flow setting you recommend for it

 

I have quite some issues with GreenTec. I seem to have strong underextrusion with it. I print it with 115% flow, below that I have gaps in the print between perimeters and infill and sometimes even between perimeters.

[foehnsturm 970]

same here, maybe below 110%. The above mentioned 105 were the minimum recommendation I think.

[conny_g 251]

3 hours ago, foehnsturm said:

same here, maybe below 110%. The above mentioned 105 were the minimum recommendation I think.

 

Weird to say "glad to hear that". But the last few days printing a larger mechanical object (CO2 laser tube holder ?) the GreenTec was driving me crazy searching for the right print parameters. Good to see I am not alone ?
I had checked the nozzle if it was clogged, the feeder if it grips and whatever else.
And finally it's just soft for the feeder and of lower viscosity for the nozzle, pretty challenging combination.
 

I researched this in the web and did not read a lot of these issues about it. Actually very little reviews / tests / experience with GreenTec. Makes the impression it's still low in reach in the community.
 

[starbuck 25]

I'd just like to chime in here as an interested UM3 operator and say that I'd need 15 filament flow sensors...  

[ColoRich 1]

Hopefully this thread isn't dead, but would this work?  https://dyzedesign.com/shop/filament-detector/filament-sentinel-2/

[starbuck 25]

On 5/29/2018 at 12:13 PM, ColoRich said:

Hopefully this thread isn't dead, but would this work?  https://dyzedesign.com/shop/filament-detector/filament-sentinel-2/

Unfortunately, I read in another thread that it does not integrate directly to the UM3 (yet). Something like the developers couldn't figure out the existing interface to the control board.

[foehnsturm 970]

Not dead at all ?. We had holidays here in Bavaria.

Currently a few guys are testing the flow sensor on UM2+ (standard feeder, medusa feeder, Bondtech QR).

 

The feedback suggests that these features already work very well 

• Spool end detection.
• First layer underextrusion.
• Clogged nozzle
• Temperature limit too low ( viscosity too high )

 

However volunteers for a UM3 hardware/firmware integration still have to be found.

[foehnsturm 970]

Call for pilot users - let's make this thing see the light of day!

 

I would like make a first small batch of flow sensors available by the end of July.

Here are the details:

• fully assembled and tested, ready to use
• comes with pre-assembled cables
• designed for the UM2+
• twist-lock mount on the entry side of the feeder
• works as well with Bondtech feeders 
• based on a capacitive encoder with 2048 steps /rev
• full firmware integration by Tinkergnome

Introductory price: € 65 (plus tax and shipping)

 

The sensor has been tested for some 100 hrs so far. There are no known hardware issues, however this is still work under development. Especially the firmware part will see improvements over the next months.

 

Interested? Then just post here!

 

 

 

 

[Dim3nsioneer 558]

18 hours ago, foehnsturm said:

Call for pilot users - let's make this thing see the light of day!

Interested? Then just post here!

 

 

 

 

 

Count me in.

[kmanstudios 1,120]

19 hours ago, foehnsturm said:

Interested? Then just post here!

Count me in for two...I have two UM3E's and would like to make them both the same. :) So cool!!

 

[Dim3nsioneer 558]

@kmanstudios The firmware is only available for the UM2+ at this time. In order to make it work with the UM3, the source of the UM3 firmware would be needed.

[cjs 34]

Count me in, too ?

[kmanstudios 1,120]

13 minutes ago, Dim3nsioneer said:

@kmanstudios The firmware is only available for the UM2+ at this time. In order to make it work with the UM3, the source of the UM3 firmware would be needed.

But....But....he included the UM3....oh phooey :(