Diamond hot end potential print speeds and pressure advance testing
-
David,
Ref pressure advance - I had read that but it was a long time ago and I don't recall seeing the mathematical explanation - must have been short of caffeine at the time. Anyway, thanks for the link which explains it perfectly.
Ref Jerk - Thanks for the clarification. That's what I expected.
Ian
-
Simon,
That's interesting. I take it that was with your 0.4mm nozzle. Do you plan on doing a repeat with a larger nozzle?
Ian
-
Yes 0.4mm. If I get chance I will try it with a big nozzle, probably next time I need to print a large strong part (not that often to be fair).
-
Just been experimenting with pressure advance. Man it's like some magic switch! Cured all the roughness and the blobs are history.
Two things.
My Bowden tubes are only about 165mm so given that the Wiki states typical values are 0.1 to 0.2 for Bowden set ups, previously I had tried settings from 0.01 up 0.2. None of which made any noticeable difference. So I assumed it was maybe something to do with using a mixing hot end - some sort of interaction between filament inputs perhaps. It turns out that a Diamond hot end with Titan extruders and 165mm Bowden tubes needs a pressure advance setting of 0.5 for each of the 3 extruders!
I needed to increase extruder Jerk a lot. It did slow down the short 30mm Y moves when I had it set to 600. Changed it to 1200 and there was a noticeable increase in speed. Almost up to the same speed as not having any pressure advance but not quite. Changing it to 2400 made no further difference. Switching to tool3 (all 3 extruders) and there was a further increase and I'd say there is no noticeable difference in speed between no pressure advance and a lot of pressure advance (using all 3 extruders). Long moves aren't much affected as it's only the acceleration phase of the extruders. I am of course using very high pressure advance and the short Y moves are speed limited by carriage acceleration in any case, so it's not an issue (especially as the quality of the print is so much improved).
I tested up to 150mm/sec and down to 40mm/sec with a single extruder and from 40 up to 250 with all 3 extruders. The same pressure advance setting works well for all configurations.
Well done David! I don't know how you came up with the algorithm but once the right value has been found for a particular machine, it seem that it works for all speeds and for single filaments and mixing of 3 filaments.
I'll do a bit or a write up on my blog, with another little video to accompany it.
Ian
PS. I'm seriously thinking that 200mm/sec plus print speeds with no reduction in quality might actually be achievable. Not an issue for most people, but when you have a 355mm x 340mm x 760mm print volume, speed starts to become important.
-
I'm glad it's working well for you. Did you also reduce the amount of retraction you are using? Be careful if you are using E3DV6 hot ends (not the Lite version), because pressure advance has the effect of retracting before the ends of some printing moves, so you can easily end up with too much retraction in total for the V6 if you don't reduce your retraction settings.
I am surprised that with Bowden tubes only 165mm long you needed pressure advance as high as 0.5. I guess I need to do some more experimentation.
-
I've seen an interesting effect with a circular thin-walled object printed with extremely elastic flex41 filament. There was a large chunk missing from one side. I can't recall whether it was at the start of extrusion or end so I guess I had too much or too little advance.
Ian I'd suggest playing with advance printing a single perimeter thick object as this seems the most sensitive to the level set, especially with flexibles where it's really on a knife edge with this type of object. Circular seems logical as there are no corners to accelerate or decelerate around and jerk comes into play more. 1200 doesn't sound too crazy for a titan, I'm having to play it differently on my delta running flex3drive as acceleration and jerk have to be set very differently to optimise retraction.
-
Hi David,
I'll likely need less retraction but haven't got around to tuning it yet - the test object only does one retraction on layer change. Not sure about your comments - had you forgotten that it's a Diamond hot end? If you were just referring to the heat sinks, then they are E3D V6 Lites.
Yes, the high number surprised me too. Maybe its a feature of the Diamond hot end because it has 3 inputs? Or maybe it's the Titans?
I'll be posting a video soon - the extruder action looks absolutely crazy but it prints beautifully.
Ian
-
I've seen an interesting effect with a circular thin-walled object printed with extremely elastic flex41 filament. There was a large chunk missing from one side. I can't recall whether it was at the start of extrusion or end so I guess I had too much or too little advance.
Ian I'd suggest playing with advance printing a single perimeter thick object as this seems the most sensitive to the level set, especially with flexibles where it's really on a knife edge with this type of object. Circular seems logical as there are no corners to accelerate or decelerate around and jerk comes into play more. 1200 doesn't sound too crazy for a titan, I'm having to play it differently on my delta running flex3drive as acceleration and jerk have to be set very differently to optimise retraction.
Hi Simon,
Yes I'll have a play around with some "real life" objects at some time. Obviously different filaments may need different pressure advance settings, just as they need different temperatures, speeds, etc. Easy enough to do using slicer start code or macros. For now, I'm just experimenting with maximum melt rates.
-
Hi David,
I'll likely need less retraction but haven't got around to tuning it yet - the test object only does one retraction on layer change. Not sure about your comments - had you forgotten that it's a Diamond hot end? If you were just referring to the heat sinks, then they are E3D V6 Lites.
Yes, the high number surprised me too. Maybe its a feature of the Diamond hot end because it has 3 inputs? Or maybe it's the Titans?
I'll be posting a video soon - the extruder action looks absolutely crazy but it prints beautifully.
Ian
If you were only using retraction on layer change, then without pressure advance you are likely to get a blob at the start of each travel move tat follows a printing move anyway. Is that where the blobs are?
I hadn't forgotten that you were using a Diamond, but I understand it can use the E3DV6 or E3D Lite6 heatsink and other parts. My concern was that if you were using the V6 heatsinks then you need to be careful not to over-retract. The Lite 6 with its PTFE liner is much more tolerant of retraction and probably a better choice for printing PLA.
-
Ian, another great blog post and excellent discussion here. There is a lot to learn from the issues/characteristics you highlight at different points as the filament is being extruded. Also the discussion on pressure advance suggests it could fix a pattern I see at certain points on some of my prints.
I agree with your write up on the blog that real world printing is what matters, but would still like to know what the baseline is for just the hotend and extruder - mark a point on your filament tight against the frame before the extruder, extrude 50mm into air, and find the maximum speed where that mark still moves 50mm.
Have you tried your results with 'Autospeed' in Slic3r? (Print settings -> Speed -> Autospeed (advaned))
-
If you were only using retraction on layer change, then without pressure advance you are likely to get a blob at the start of each travel move tat follows a printing move anyway. Is that where the blobs are?
I hadn't forgotten that you were using a Diamond, but I understand it can use the E3DV6 or E3D Lite6 heatsink and other parts. My concern was that if you were using the V6 heatsinks then you need to be careful not to over-retract. The Lite 6 with its PTFE liner is much more tolerant of retraction and probably a better choice for printing PLA.
Hi David,
Without pressure advance but with a lot of (firmware) retraction (about 4mm with 165mm long Bowden tubes), the blobs started to appear at above 70mm/sec or so print speed. It was as if the filament was oozing and got deposited in blobs along the length of the long (300mm plus) non-print move. In fact, that's exactly how I visually observed it happening. As the print speed was increased, the blobs got worse (more pressure build up before the long non-print move). Enabling pressure advance has completely eliminated them. i.e there is no oozing during the long non-print move. Trying to control that by only using retraction would mean having to have more and more retraction as the print speed is increased.
Thanks for the heads up ref the heatsinks. RepRap.me recommend using the lite version rather than the V6 so that's what I've always used. It could be I end with zero or near zero retraction - don't care as long it prints well.
The video is being rendered as I write this. I'll post a link as soon as I get it uploaded to YouTube. I think you'll find it interesting if not amazing to watch. Yes, the extruders run backwards at high speeds but just give a bit of jerk at "normal speeds". At high speeds it looks (and sounds) crazy, but prints beautifully.
Ian
-
…..............................would still like to know what the baseline is for just the hotend and extruder - mark a point on your filament tight against the frame before the extruder, extrude 50mm into air, and find the maximum speed where that mark still moves 50mm.
Have you tried your results with 'Autospeed' in Slic3r? (Print settings -> Speed -> Autospeed (advaned))
Hi Rob,
Extruding in air is on my list (everyone want me to try something).
Haven't tried the slic3r autospeed feature. It's experimental which has always put me off. I prefer to set my own speeds for perimeters (large and small), infill, solid layers etc etc. but I might give it a go one day.
Ian
-
Blog is done - link here https://somei3deas.wordpress.com/2017/06/25/duet-pressure-advance-experiments/
If you don't want to read the words, the (crazy) video is here https://www.youtube.com/watch?v=lnYYNfVoxmQ&t=304s
This one is much shorter - less than 8 minutes.
Ian
-
Ian, I finally found time to watch your videos. I'm very impressed by the printing speeds you have been able to achieve! I'm glad that pressure advance helped you. Your video is an excellent demonstration of what pressure advance does and how the extruder movement appears.
Is the carriage that carries the Titans driven, or towed by the main carriage?
-
Hi David,
Thanks for the feedback.
The extruder carriage is just dragged around by the main carriage. If you look closely at the video you'll see one of the 7/02 guide wires had snapped - need to replace it with 16/02 or some such. Not surprising given that I've been chucking 1.6kg around at up to 300mm/sec. (Y is worse at over 3kg). The wires are to keep the Bowden tubes in compression during large or fast moves. There is a bit of an explanation of it all here https://somei3deas.wordpress.com/my-corexy-printer-build/ and a video here https://www.youtube.com/watch?v=5RptHCX7z0o&t=184s
I initially thought about putting motors, belts and pulleys on the extruder carriage and mirroring the main XY motors but dismissed it due to complexity. It's also complicated by the fact that the extruder Y carriage and Y axis are wider than the main carriage.
Effectively there is some decoupling between the axes so when the main carriage is doing short moves, the extruder carriage remains more or less stationary (see video). The downside to that is that when doing big fast moves, the extruder carriage lags behind the main carriage slightly so gets jerked at the start, then at the end of the move, the inertia tries to carriy it on while the main carriage is decelerating, resulting in another jerk at the end. However, the extruder carriage is light compared to the main carriage and even at those silly speeds, I didn't notice anything particularly horrible happening. The wires prevent the Bowden tubes from pulling the hot end mount up on it's sliding dowels.
Time will tell. If I end up having to drive the extruder carriage as well as the main carriage, I'd be inclined to use a shaft with pulleys top and bottom connected to each set of belts and a central pulley connected to the motor, probably with 1:2 gearing (1 on the motor, 2 on the shaft).
It continues to evolve….............
-
I have given this some thought, and come to the conclusion that the Diamond needs 3 times as much pressure advance as a similar single-filament hot end would. Here's why.
Let's suppose that you are extruding with only 1 filament. When you add extra steps to pressurise the single extruder, the extra pressure in the mixing chamber will make filament flow back through the two inactive heat breaks, compressing the filaments in those Bowden tubes until the pressure is equialised. Ideally we should pressurise both of them too. But since we don't, we will need 3 times the pressure advance movement on the main extruder drive to allow for the back flow and get the same hot end pressure.
If you extrude with all 3 filaments equally, this is a better situation. However, because each extruder is only extruding 1/3 of the total filament, the amount of pressure advance applied to each extruder is only 1/3 of what is required. So again you need 3x the pressure advance.
Ideally, we should calculate the pressure advance based on the total speed of extrusion of all 3 filaments, and apply that amount of pressure advance to all extruder drives regardless of the mix. I may implement this in a future firmware release.
-
Hi David
Good thought but I have to disagree. What you probably missed is that when using a single filament, I actually set the mixing ratio to 0.98:0.01:0:01. That is to say, I always keep the other two filaments moving, albeit slowly. So I find it hard to imagine that the pressure on one input is imparting a significant back pressure on the other two inputs, given that they are being (slightly) driven as well.
My personal gut feel is that it's simply a function of the size of the melt zone (which is big on a Diamond). If you have a large volume containing a fluid at a given pressure, then poke a 0.4mm hole in it, it will take longer for the pressure to normalise than a if you had a container of much smaller dimensions. Similarly, if you push fluid into a large volume container at a given pressure, it will take more time to pressurise that container than it would a smaller container. I'm not explaining it well but it's just how hydraulics works.
Or another analogy but in reverse. When a fluid is heated it expands. If the system containing the fluid isn't vented, then an expansion vessel is needed (a kind of diaphragm and spring arrangement). Typically on a domestic heating system, the expansion vessel needs to be about 10% of the total volume. So the bigger the system, the bigger the expansion vessel needs to be otherwise pressure will build up to an unacceptable level.
I personally find it hard to believe that it's possible to compress filament when it's in a solid state. It might buckle slightly inside the tube but that's about it.
What we need is someone to test pressure advance with an E3D volcano which has similar internal dimensions to one input of a Diamond. I'm willing to bet that would need similar pressure advance values to what I ended up with.
Please don't mess with the algorithm. The same pressure advance value works fine for both single filament and all three filaments, at high print speeds and at low print speeds.
Ian
-
If you have an extrusion multiplier of .01, there could be many times when those extruders never produce a single step. Look at a typical gcode file you print, see the distance the E axis is driven, and calculate how many steps that is at your e steps/mm. If it is less than 100, your other filaments are likely not being driven often.
-
Ian, it is certainly possible to compress filament in a Bowden tube. With filament loaded but the hot end cold, turn the extruder motor current off and turn the gear wheel by hand. You will find that from the point where it starts resisting, you can compress it a lot.
There is a lot of spare space around 1.75mm filament in a 2mm ID tube, The filament has a cross sectional area of 2.41mm^2 while the tube has an internal area of 3.14mm^2. That's 30% more space than the filament occupies. The filament can take up some of this space by assuming a helical shape. The tighter the pitch of the helix, the more filament you can fit in a given length of tube, and the more the filament will resist this. If you use 1.8mm ID tubing instead (which might work for your short Bowden tubes), you may find that you need less pressure advance.
When you are using just one extruder, whether the other 2 extruders are stationary or they move a little is irrelevant. The pressure in the melt chamber will be varying a lot, and the filament in the unused tubes will compress accordingly. That pressure needs to be either resisted by applying full pressure advance to the other 2 extruders, or compensated for by using 3 times the pressure advance on the main extruder.
I don't think the volume of the melt zone is relevant. Fluids are generally regarded as incompressible (the exception would be filament with steam bubbles in it), and pressure waves are transmitted in fluids very quickly.
Your point about the expansion vessel might be relevant if the temperature in the melt zone was varying a lot; but it should be more or less constant at the hot end temperature you set. There will be a small variation due to the varying melt rate, but I think the amount of expansion will be small. Whereas it's clear from your video that pressure advance is applying several mm of additional movement to maintain pressure when you are extruding quickly. That's several cubic mm of extra filament going into the Bowden tube and coming back out again. I can't believe that it's all disappearing into the melt chamber and not coming out somewhere else.
Trying a Volcano would be interesting. My expectation is that at a given extrusion temperature and filament diameter, and using the same Bowden tube length and ID, a larger nozzle will need less pressure advance, because the back pressure per unit extruder drive speed is less. So a Volcano with a 0.5mm nozzle should I think need 1/3 the pressure advance of your Diamond, and if you fit a larger nozzle then it should need less. All this assumes that the same nozzle temperature (not indicated temperature) for both hot ends.
-
@bot:
If you have an extrusion multiplier of .01, there could be many times when those extruders never produce a single step. Look at a typical gcode file you print, see the distance the E axis is driven, and calculate how many steps that is at your e steps/mm. If it is less than 100, your other filaments are likely not being driven often.
I was talking mixing ratios, not extrusion multiplier bit I think I see where you are coming from. I guess for very small moves, the 1 % won't actually be acted on as it would be less that 1 micro step, which would potentially lead to maybe a maximum of 2% under extrusion on very small moves. In practical terms, I can't say that this is visible in the final printed object.
