The idea
Measure the force required for extrusion during transient.
The approach is not new, we tackle the task similar to

• Duet forum thread
• Renkforce rf2000 uses a similar system

How we do it - mechanics

• Separate extruder from cold/hot end with load cells TAL220G
• A first fixture is used to screw the extruder on top of the load cells. The extruder has a M7 thread wich is used with an aluminium tube to fix it to the fixture and with a short PTFE tube inside to guide the filament to the coldend.
• A second fixture is an angled bracket used to screw the load cells on the MGN12H carriage.
• A third fixture represents the cold end and aligns with the filament path from the extruder. The 4mm hole at the top is used to provide for another short PTFE tube that guides the filament to the heatbreak. A M7 thread at the bottom allows to attach a E3D V6 hotend.

All three mentioned fixtures were manufactured by 3CTS and it became possible for me to realize this idea at all. Thanks guys!

How we do it - electronics + software

• One of the load cells is attached to a HX711, which in turn is connected to a raspberry pi zero w. There is four load cells, as I was not sure for how much force I needed to prepare for and the 1.5kg sensors were cheap and readily available (the 3kg option was not in stock at the time, so I went for it this way). I guess it could make sense to remove two of the four sensors and improve the sensitivity - but I find the result satisfying already.
• Software readout is done according to Github link
• The HX711 is supplied with 5V from the Pi and takes 80 samples per second, which seems to be sufficient for useful readings, but does not give insight if you want to see what is happening during a retract (which takes only ~10ms in my setup).
• Display the readings at a webpage and embed it within the Duet Web Control.

Gallery - first and second fixture

Gallery - third fixture/coldend

Gallery - full assembly

Exemplary reading
Here you can take a look at the readings from an exemplary print with three retractions taking place. unit=grams

Well, I am still playing with it trying to find out what information can be gained from this system. For now, it helped me dial in retraction and pressure advance, but in the end it does not waive calibration prints. Any suggestions welcome!

Okay. So I split up both of my expansions (one of which already had a groove and was easy, the other one was more work) and installed both "P6" to the Maestro.

Everything is running fine now!

Well, this concludes it for me. I have two leftover P5 - if someone is interested.

@zapta said in Filament force measurement system:

@wurstkarton said in Filament force measurement system:

The HX711 is supplied with 5V from the Pi and takes 80 samples per second

It's not that difficult to increase the sampling rate up to 140 samples/sec by adding a crystal and changing a jumper. The data sheet has the necessary information. For example, I increased mine to 100Hz with a 13824000 HZ crystal.

Thanks for the info, but I think also 140 samples/sec are not enough. I'd guess I need 1k samples a second to make a real difference for taking a closer look at e.g. retraction.
Nothing that could not be solved with a reasonable DAC and instrumentation amplifier, but nothing that I'll do soon...

@o_lampe said in Filament force measurement system:

@wurstkarton
What would it take to integrate the loadcell readings into RRF?

For now, I just use the webcam functionality of DWC to display the readings.

Would it make sense to establish a constant pressure print mode? (like constant volume slicing)

Not really, I do not think that it would help. At every corner or turn the nozzle takes, the pressure will not (and must not) be constant (you need to decelerate there and hence extrusion rate will reduce and hence pressure too).

I also see opportunities to detect skipped extruder-steps and help tuning filament temp.

Yes, this should work.

@zapta said in Filament force measurement system:

@wurstkarton said in Filament force measurement system:

For now, it helped me dial in retraction and pressure advance,

Very interesting. Can you explain how you use these force measurements to calibrate pressure advance and retraction?

Well, I am still trying to understand it myself.
The idea is to perform the typical K-factor calibration test where you vary the speed on a straight line and check for the change in extrusion force. You want the force to change quickly to its next static value without much overshoot.
For retraction it's similar (you want the force to disappear (it will be a bit negative) after retraction and be back at its static value after unretract), you can also check the Airtripper webpage. Something similar was done there, but for a bowden system.

Also, would be nice to correlate the force vs time graph with the motor position, in microstep ('electrical' position) and with an actual rotational decoder ('mechanical' position).

I guess you could do that, but I'd need some sort of information from the Duet to process in the Pi to align time with the motor position.

EDIT: One more question, does the force correlate proportionally to the pressure at the nozzle?

Yes, pretty much. This depends on how you apply the principle of intersection. The load cell is placed between coldend and extruder. You will see any force that can move the load cell:

• The force required to "press" filament through the nozzle+heatbreak.
• The force required to pull filament from the roll.

What you will not see in the measurement:

• The force from touching the heatbed with the nozzle. (in comparison the renkforce rf2000 fixture would see this)

EDIT2: Is the force them same if you actually print or print in the air, with nozzle not touching the bed? (To understand the potential back pressure from the print itself)

I think you can see a difference, but I did not pay attention to it (yet). If the nozzle is too close to the bed, then you need much more force to press the filament through. Hence, we should see that in the measurement (but you cannot measure the distance to the bed just by touching it with the nozzle).

The idea
Measure the force required for extrusion during transient.
The approach is not new, we tackle the task similar to

• Duet forum thread
• Renkforce rf2000 uses a similar system

How we do it - mechanics

• Separate extruder from cold/hot end with load cells TAL220G
• A first fixture is used to screw the extruder on top of the load cells. The extruder has a M7 thread wich is used with an aluminium tube to fix it to the fixture and with a short PTFE tube inside to guide the filament to the coldend.
• A second fixture is an angled bracket used to screw the load cells on the MGN12H carriage.
• A third fixture represents the cold end and aligns with the filament path from the extruder. The 4mm hole at the top is used to provide for another short PTFE tube that guides the filament to the heatbreak. A M7 thread at the bottom allows to attach a E3D V6 hotend.

All three mentioned fixtures were manufactured by 3CTS and it became possible for me to realize this idea at all. Thanks guys!

How we do it - electronics + software

• One of the load cells is attached to a HX711, which in turn is connected to a raspberry pi zero w. There is four load cells, as I was not sure for how much force I needed to prepare for and the 1.5kg sensors were cheap and readily available (the 3kg option was not in stock at the time, so I went for it this way). I guess it could make sense to remove two of the four sensors and improve the sensitivity - but I find the result satisfying already.
• Software readout is done according to Github link
• The HX711 is supplied with 5V from the Pi and takes 80 samples per second, which seems to be sufficient for useful readings, but does not give insight if you want to see what is happening during a retract (which takes only ~10ms in my setup).
• Display the readings at a webpage and embed it within the Duet Web Control.

Gallery - first and second fixture

Gallery - third fixture/coldend

Gallery - full assembly

Exemplary reading
Here you can take a look at the readings from an exemplary print with three retractions taking place. unit=grams

Well, I am still playing with it trying to find out what information can be gained from this system. For now, it helped me dial in retraction and pressure advance, but in the end it does not waive calibration prints. Any suggestions welcome!

@deckingman said in Firmware retraction question:

@Wurstkarton Yes. All parameters that are used in M207 apply to that tool and drives are assigned to tools. So it will apply to all drives.

Thanks for this. Now I know that I must refrain from using the additional unretract length.

The rationale for using extra length on the un-retract is that the hobbed bolt might grind the filament away slightly during the retraction, so the filament diameter would be reduced when the un-retract happens. Given that all filaments are retracted together, then it makes sense that they should all be un-retracted the same way.

What do you do if you have extruders that behave differently? I am inclined to say that this is not the full picture why you would apply an additional unretract length. It could also be that you do not have oozing under control (and may never have depending on the material and hotend that you use). In that case, it may make sense to "refill" your nozzle with the lost material according to your mixing ratio.

This could be something for a change request: Apply additional unretract length according to mixing ratio (when you specify an additional parameter in the M207)

Anyways. Now that I know how it behaves, I can also do that in my gcode with a script.

@deckingman I understand that and I have 6 tools in total defined (I just copied a single tool for illustration, maybe I should have only put a single M207 command there for better understanding). This is not the question.

The question is not about the P parameter, but the R parameter. Does the R parameter affect all drives involved for the respective tool?
If yes, you will unintentionally mix a different color into your current tool.

Lately, I have been experiencing some blobs after printing with my mixing hotend (e3d cyclops). It is known that for mixing hotends, it is beneficial to use firmware retraction, as you can retract from all inputs to take the pressure away.

So my setup looks like this (RRF 3.1.1, Duet 2 Maestro):

M563 P0 D0:1 H1 F0 S"MIDDLE_LEFT"                               ; define tool 0
M567 P0 E1:0                                                    ; set mixing ratios for tool 0

Lately, I have introduced an additional restart length:

M207 P0 S2.1 F3000 T3000 R0.1 Z0                                ; firmware retraction settings
M207 P1 S2.5 F3000 T3000 R0.3 Z0                                ; firmware retraction settings

Now to the question:
When firmware retraction is done, does the additional unretract length apply to all drives involved in the tool?
If yes, this would be bad, as I am - at each G10+G11 - start to mix in another color. This could explain the behavior I am seeing.
If not, I need to look somewhere else...

Thanks!

@BotLawson said in Duet Maesto Stepper Driver Expansion Schematic:

Looking at the layout render, it looks like the culprit is the Vin trace to C6. It cuts off the direct path to ground and forces current from R3 to loop around the outside of the pcb. Personally, I'd put R3 and R2 between the TMC chip and stepper connector and give them 2-4 vias to the ground plane and the thermal die pad of the TMC. The motor phase connections can then sneak under or around the 805 resistor.

Exactly. There is this "unconnected" peninsula to the right of R3/C2. Great job to both of you guys. I guess this fixes it.

@dc42 Thank you for this analysis. At the moment, I do not have the need to try it out (as I have two P6). But I am sure that others may find it useful!

@dc42 from what I could tell, yes. But that really does not mean much.

The layout of output 1 (which is the output that generates the problem) looks a bit different, but I am not sure if this really would be a problem. Would be easier to see with the actual layout in a design tool.