I cooked churros today and it helped me a lot to understand pressure advance. Basically the process involves pushing dough through a stars shaped hole. The pressure is applied by turning a screw that moves a piston. The first I noticed was that when you stop turning, dough keeps coming out for about 10 seconds. Then I noticed that the opposite thing happens when you first start turning the screw. Kind of random but it was very funny to me (but difficult to explain why to the other people in the kitchen).
You are making physics interesting! I love how I can apply what I learned at uni last semester to my favorite hobby and get better print results out of it.
Thanks for this. Breaking down the barriers between my brain and higher understanding, bit by little bit. Thanks for keeping the content coming so regularly! Love it!
Excellent video! I’d love to see a video like this on input shaping. How the accelerometer reads the resonances and how the input shapers impact the motion path.
Have yet to work with this feature. Appreciate your detailed analysis. Essentially an open PID loop with feed forward parameter? It would be handy if there were some form of real-time feedback in scalable stepper motor/driver current or strain gage with the correct span and resolution to close the loop. Relying on constants should keep the influencing parameters realistic between under and over extrusion with methods to verify cause and effect for tuning associated variables.
Compliance due to extruder force is not just the filament but all components subjected to the force eg housings, Bowdens in tension or compression etc. It all stacks up in series.
Excellent video! I love the graphs. You talk about the smooth PA somewhat rounding the edges of the green graph (for a lack of better word). At this point would it make sense to replace the trapezoidal motion with something similar to what S-curve was attempting to do in Marlin? (Technically I can't use S-curve in Marlin because it doesn't have the required PA implementation) While S-curve aims at reducing instant changes in acceleration (3rd derivative?), we could have the motion follow the blue graph instead for a similar effect, but with the added advantage of reaching an ideal extrusion. This would slightly increase print time as a side effect. On a separate note, in practice moves can be fairly quick, perhaps under 0.05s in some cases. It would be awesome to see an actual graph of what would be going on if several such short moves at sharp angles occurred one after another. I bet the graph would look very different, but I don't have a quick way of simulating it myself.
I just thought of the same. I am not into 3d printer software but love the engineering behind it and asked myself, if the movement of the print head can be matched to the flow rate of the extruder. Maybe the motion system of a printer has a faster response and can compensate for the lagging flow.
Today i hadmortar in my mastic gun it seem far more elastic or compressible than silicone and there was a lag when i pressed the trigger and it continued to ooze long after i released the pressure. Seems pressure advance in essence is overcomming lag and ooze by applying and releasing pressure. In advance of requirement. Like ignition advance in a petrol engine
Cool thanks again Eddie for the education. Could we not have 2 reels of the exact same filament feed into separate heat zones but one nozzle, via 2 identical extruders, then then required acceleration /deceleration would be split between two motors, this main mean going back to bowden, and step backwards until a better system is created. Or idex with different nozzles, same material, (the reason I bought a bcn3d sigma), so infill and skins printed quickly by large nozzle and small details by small nozzle, (seems software isn't quite there yet). Just spit balling ideas for anybody to shoot me down so I don't waste any time chasing thats already done please
That would be cool, especially with the two different nozzle sizes for IDEX! I'll have to try and see if I can do that once my IDEX printer is a bit more refined
I’m curious why use a smoothing function on the extruded acceleration, which produces a near constant jerk(mm/s3) rather than an equivalent effective constant acceleration. Is there something I’m missing? The idea would be to take the raw PA acceleration, and a acceleration limited version of if, and integrate to determine the missing impulse, then distribute that before and after the acceleration limited region up to the acceleration limits. By doing smoothing you are limited in the extruder by the torque available at a very short peak in the smoothed PA.
I just found this series and I absolutely love it! I can't wait for the next installment! Since it's been a minute since you released them, I'm hoping the tool has been uploaded by now, but I haven't looked yet. Anyway, question time. Your assumed spring force of 70N/mm, is that a spring force for a single type of filament, or an average across many filament types? Either way, what/which filament(s) did you base this assumption on?
That’s a great question! The spring force will vary based on the filament for sure. In the model I have, the spring force ends up kind of being a dependent variable (or maybe opposite, the PA is dependent on the spring force). If you don’t pair them together correctly the extruded amount vs time doesn’t match up with reality at all!
@@eddietheengineer that makes sense. PA is technically designed solely to compensate for this spring force, so it makes sense that there would be dependcies between the two. Now that I think about, I suppose it really doesn't matter what the assumed spring force is based off of. It just needs to be something to make the tool visually represent what PA and smoothing does in an accurate and easy to understand way. It's not a replacement for an actual PA tune after all.
@@whistlinturbo , so we use different PA for different filament types, right? E.g. with PETG we probably interested in a greater smoothing compared to let's say PC as later one is more rigid (less springy)? And I'm assuming that extruder motor has enough torque to push either of those filaments.
So if I understand correctly Eddie, if I use a different filaments or different temp, because it will not be the same fluidity, I would have to readjust PA
I tested my titan style extruder which has a single sided filament gear plus a bearing on the other side to see what the actual impact of the gear itself is. So i printed a rectangle which was near exactly one rotation of the extruder gear per layer. I used clear petg so I could really see the difference in extrusion with and the teeth itself are really visible. I think this is commonly discussed as issue 6. Now I’m thinking about a way to compensate for it by measuring the actual feed of the filament itself, or calibrating the feed to the gear orientation and then measuring it with a rotational sensor. A closed loop extruder gear in other words. Would be great to see if it’s working.. at least with a modified gcode or so. However, in general it’s interesting what the actual impact of the extruder gears is and how it’s better with smaller teeth and bigger diameter.
You have me thinking--it would be interesting to have an encoder right above the hotend on a bowden style printer, and map what the filament position at the hotend is vs. the driven position at the extruder is. That should be able to help experimentally validate the model!
@@eddietheengineer Yes, this is what I mean.. I´ve looked for some hardware and sensors already. You´re on voron discord too, or not? If you like to, leave me a message and we can exchange some ideas. I´m making pcb´s and microcontroller stuff.
Hello! The spring force/flow I believe is in Extruder 102: ruclips.net/video/Hc-Mhtaxse8/видео.html It's been a little while since I've reviewed these videos so I'm a bit rusty on exactly what calculations I did!
I find it odd that people go into this level of detail to squeeze the most out of their printer, yet they are using a bondtech extruder with a massive filament path.
The filament path is the reason pressure advance exists in the first place. And what's wrong with Bondtech extruders? Also, I'd hardly call the Voron Afterburner filament path "massive"
I cooked churros today and it helped me a lot to understand pressure advance.
Basically the process involves pushing dough through a stars shaped hole. The pressure is applied by turning a screw that moves a piston. The first I noticed was that when you stop turning, dough keeps coming out for about 10 seconds. Then I noticed that the opposite thing happens when you first start turning the screw.
Kind of random but it was very funny to me (but difficult to explain why to the other people in the kitchen).
This made my day!
@@eddietheengineer thank you! It made mine to!
Oh yeah. You're right. Also applying caulk or mastic with one of those tube guns.
You are making physics interesting! I love how I can apply what I learned at uni last semester to my favorite hobby and get better print results out of it.
Thanks for this. Breaking down the barriers between my brain and higher understanding, bit by little bit. Thanks for keeping the content coming so regularly! Love it!
After watching many videos about PA this is the first one (I saw) explaining it properly, thanks!
There’s not a lot of videos with 3D printing content that is this good. I get that this is not for everyone, but I love it.
Keep learning about extruder! Great knowledges.
Excellent video! I’d love to see a video like this on input shaping. How the accelerometer reads the resonances and how the input shapers impact the motion path.
As always, very informative video. Good job!
Have yet to work with this feature. Appreciate your detailed analysis. Essentially an open PID loop with feed forward parameter? It would be handy if there were some form of real-time feedback in scalable stepper motor/driver current or strain gage with the correct span and resolution to close the loop.
Relying on constants should keep the influencing parameters realistic between under and over extrusion with methods to verify cause and effect for tuning associated variables.
Compliance due to extruder force is not just the filament but all components subjected to the force eg housings, Bowdens in tension or compression etc. It all stacks up in series.
Excellent video! I love the graphs. You talk about the smooth PA somewhat rounding the edges of the green graph (for a lack of better word). At this point would it make sense to replace the trapezoidal motion with something similar to what S-curve was attempting to do in Marlin? (Technically I can't use S-curve in Marlin because it doesn't have the required PA implementation) While S-curve aims at reducing instant changes in acceleration (3rd derivative?), we could have the motion follow the blue graph instead for a similar effect, but with the added advantage of reaching an ideal extrusion. This would slightly increase print time as a side effect.
On a separate note, in practice moves can be fairly quick, perhaps under 0.05s in some cases. It would be awesome to see an actual graph of what would be going on if several such short moves at sharp angles occurred one after another. I bet the graph would look very different, but I don't have a quick way of simulating it myself.
I just thought of the same. I am not into 3d printer software but love the engineering behind it and asked myself, if the movement of the print head can be matched to the flow rate of the extruder. Maybe the motion system of a printer has a faster response and can compensate for the lagging flow.
Great stuff!
Today i hadmortar in my mastic gun it seem far more elastic or compressible than silicone and there was a lag when i pressed the trigger and it continued to ooze long after i released the pressure. Seems pressure advance in essence is overcomming lag and ooze by applying and releasing pressure. In advance of requirement. Like ignition advance in a petrol engine
Cool thanks again Eddie for the education.
Could we not have 2 reels of the exact same filament feed into separate heat zones but one nozzle, via 2 identical extruders, then then required acceleration /deceleration would be split between two motors, this main mean going back to bowden, and step backwards until a better system is created.
Or idex with different nozzles, same material, (the reason I bought a bcn3d sigma), so infill and skins printed quickly by large nozzle and small details by small nozzle, (seems software isn't quite there yet).
Just spit balling ideas for anybody to shoot me down so I don't waste any time chasing thats already done please
That would be cool, especially with the two different nozzle sizes for IDEX! I'll have to try and see if I can do that once my IDEX printer is a bit more refined
I’m curious why use a smoothing function on the extruded acceleration, which produces a near constant jerk(mm/s3) rather than an equivalent effective constant acceleration. Is there something I’m missing?
The idea would be to take the raw PA acceleration, and a acceleration limited version of if, and integrate to determine the missing impulse, then distribute that before and after the acceleration limited region up to the acceleration limits. By doing smoothing you are limited in the extruder by the torque available at a very short peak in the smoothed PA.
I just found this series and I absolutely love it! I can't wait for the next installment! Since it's been a minute since you released them, I'm hoping the tool has been uploaded by now, but I haven't looked yet.
Anyway, question time. Your assumed spring force of 70N/mm, is that a spring force for a single type of filament, or an average across many filament types? Either way, what/which filament(s) did you base this assumption on?
That’s a great question! The spring force will vary based on the filament for sure. In the model I have, the spring force ends up kind of being a dependent variable (or maybe opposite, the PA is dependent on the spring force). If you don’t pair them together correctly the extruded amount vs time doesn’t match up with reality at all!
@@eddietheengineer that makes sense. PA is technically designed solely to compensate for this spring force, so it makes sense that there would be dependcies between the two.
Now that I think about, I suppose it really doesn't matter what the assumed spring force is based off of. It just needs to be something to make the tool visually represent what PA and smoothing does in an accurate and easy to understand way. It's not a replacement for an actual PA tune after all.
@@whistlinturbo , so we use different PA for different filament types, right? E.g. with PETG we probably interested in a greater smoothing compared to let's say PC as later one is more rigid (less springy)? And I'm assuming that extruder motor has enough torque to push either of those filaments.
So if I understand correctly Eddie, if I use a different filaments or different temp, because it will not be the same fluidity, I would have to readjust PA
Correct!
I tested my titan style extruder which has a single sided filament gear plus a bearing on the other side to see what the actual impact of the gear itself is. So i printed a rectangle which was near exactly one rotation of the extruder gear per layer. I used clear petg so I could really see the difference in extrusion with and the teeth itself are really visible. I think this is commonly discussed as issue 6. Now I’m thinking about a way to compensate for it by measuring the actual feed of the filament itself, or calibrating the feed to the gear orientation and then measuring it with a rotational sensor. A closed loop extruder gear in other words. Would be great to see if it’s working.. at least with a modified gcode or so. However, in general it’s interesting what the actual impact of the extruder gears is and how it’s better with smaller teeth and bigger diameter.
You have me thinking--it would be interesting to have an encoder right above the hotend on a bowden style printer, and map what the filament position at the hotend is vs. the driven position at the extruder is. That should be able to help experimentally validate the model!
@@eddietheengineer Yes, this is what I mean.. I´ve looked for some hardware and sensors already. You´re on voron discord too, or not? If you like to, leave me a message and we can exchange some ideas. I´m making pcb´s and microcontroller stuff.
@@randomname3894 yes! Feel free to message me, I’m eddie on Discord 👍🏼
Hi, may i ask , how did u find the formula for the Flow ? und Could u please link the Video in whcih u calculated the Spring force
Hello! The spring force/flow I believe is in Extruder 102: ruclips.net/video/Hc-Mhtaxse8/видео.html
It's been a little while since I've reviewed these videos so I'm a bit rusty on exactly what calculations I did!
I find it odd that people go into this level of detail to squeeze the most out of their printer, yet they are using a bondtech extruder with a massive filament path.
The filament path is the reason pressure advance exists in the first place. And what's wrong with Bondtech extruders? Also, I'd hardly call the Voron Afterburner filament path "massive"