@oliof I have been looking at the code, and it looks like the PA only uses a linear relationship of pressure to movement speed. Over Christmas I will be looking at ways to add different algorithms to the PA to change it from a linear relationship to maybe accepting a quadratic based algorithm instead
Best posts made by Notepad
-
RE: Issues with pressure advance since RRF 3.4
-
RE: Issues with pressure advance since RRF 3.4
@tommyb
Definitely some luck. I have identified a specific line of code that can be modified to allow for a user settable exponent based PA value. I still fully intend to get this tested yet I have unfortunately had external distractions eating up all my spare time to actually implement the change and to validate if it works.In theory, the change will allow for all current PA values to be respected (by having the exponent default to 1) so behaviour will be unchanged. But when the exponent is raised past 1, it can start becoming more aggressive to match the magnitude of the pressure differential.
It is by no means a perfect process, as I have yet to fully work though all the maths that can be contributing factors, But hopefully it will at least minimise the problem.The biggest improvement I have made actually came by accident as I was analysing a new nozzle design I have been working on. The long and short of it is just to reduce the maximum flow rate to minimise the amount of cold core printing. This didn't fix the issue, but it did let the issue subside slightly back into the lowest acceptable standard.
-
RE: Issues with pressure advance since RRF 3.4
Quick update from my side.
I have concluded that this issue is definitely a filament behaviour / Pressure Advance component.
The largest improvement I have witnessed by tweaking every setting is to simply increasing the jerk value. I previously used a pretty conservative jerk of 4mm/s but with high accelerations 6000->12000mm/s^2. I have now changed the profiles to be of a much higher jerk 10->14mm/s with middling accelerations of 3000-5000mm/s^2.
This change overall has maximised the speed the nozzle can take the corners at, and thus has reduced the bulge amount considerably. The other change I have done is to manually change the STL files I use to round most corners by a minimum of 0.8mm in a bid to again, increase the overall speed the nozzle can take a corner at.
What I have found interesting during my testing is there seems to be a minimum jerk to acceleration ratio where if the nozzle cannot take a corner with sufficient speed. the outer layer quality on curved surfaces drops considerably (even with extremely high quality STL files and arc welder enabled). There is another thread that seems to be having the same issue: https://forum.duet3d.com/topic/33903/jerk-settings-at-different-speeds
Though, in order to keep this thread focused on the PA aspect (which I am now 100% certain there is) I will probably be creating a new thread for the Jerk/Accel behaviour oddities separately in the next few weeks while I gather more data.
-
RE: Issues with pressure advance since RRF 3.4
Im not entirely sure what point you are trying to make here. But its definitely not on topic, nor is it helpful to identifying the cause and effect which causes these surface irregularities.
My slightly modified Creality 5 pro prints FLAWLESS prints in ASA within +/- .03mm, and perfect corners every time, and has printed several hundred of these prints for two years now in numerous sizes and detail; with no PA or IS applied.
Im happy that your machine works flawlessly for your use case and speed requirements, however this is the equivalent of saying "it works on my machine so it should work for yours"
A CR-5-pro is an entirely different class of machine compared to most people who are having issues. And comparing a Bowden system to a Direct Drive is already a stretch.If you’re relying on current software to fully tune out harmonic and flow issues of your printer, you simply need to slow down your outer perimeter print speed and accept the limitations of your printer.
The issue most of us are having is not flow issues, its flow control issues. I can happily print my outer walls at 100mm/s using only 30% of my nozzles max flow, but the issue still persists. Thats because its not directly a speed or flow rate problem. The issue still exists at 60mm/s or 40mm/s.
Just making a post about how "slow your machine down and you'll be fine" just derails the entire conversation and the research and identification that the other users in this thread are documenting.You will be more successful and efficient resolving resonance/pressure issues by improving the physical design of the printer than you will be expecting software to resolve it.
If you have any advice to give I would gladly welcome it. I am currently using a Hemera XS with a copper heat block using a CHT Volcano nozzle. I would like to think that my choice in nozzle, extruder and tertiary materials are all aiding in making a better printed output. I would expect most people in this chat are probably using different extruders, with different nozzles from various brands and qualities.
If you really want super speed, buy a Bamboo and deal with their proprietary bullshit.
I hope its not just me, but I find this defeatist attitude extremely offensive towards all the hard work Duet have been putting in to making their ecosystem, to any advanced user (Voron, VZbot, Etc) who spent lots of time and money to make a fantastic machine, or even to the other companies like Bondtech or E3D with their high flow hot ends or parts designed for higher grade machines than your creality 5 pro.
Your response is just "buy a Bamboo and deal with their proprietary bullshit".
Im sorry if this comes across a bit ranty, and you are correct in your statement that without full feedback, no software will be 100% able to compensate for every situation, but that doesn't mean we shouldn't try. Trying to improve stuff with the community is pretty much the definition of Open Source and everyone here is documenting their experiences in a bid to help reduce or mitigate the issue.
-
RE: How To Update Splashscreen On PD3.4.0?
Issue Found!
Turns out the copy instruction is native to windows and not the program itself. The documentation doesn't make it very clear that this instruction line shouldn't be prepend with 'bmp2c.exe'. Or it could just be me a bit slow on the uptake.anyway, all sorted!
-
RE: Sensorless homing with 1HCL / Quad motor XY setup
@dc42
Thank you for this. Now you're getting me excited for 3.6 before 3.5 has even launched -
RE: Issues with pressure advance since RRF 3.4
Just chiming in for visibility as I can also confirm to be noticing what appears to be PA oddities in regards to corner bulging, but also material deposition before and after a corner.
The hardware I use is a HemeraXS with a CHT 0.6mm Volcano nozzle with a InputShaper of EI2@20hz or EI2@48hz. Pictured is PA0.035The issue is more noticeable with a 0.6 nozzle compared to the 0.4.
I need to use a PA value of 0.045 to get non bulged corners, however by doing so I get inconsistent extrusion before and after the corner.There appears to be a correlation with flow rate and the severity of the behaviour. When pushing flow rates (>35mm^3/s) close to the limits of the nozzle, the behaviour is at its worst. which makes sense as the pressure differential would be the greatest. What is puzzling is that the issue also appears in smaller parts where the flow rates are below 60% of the nozzle capacity.
I am currently using PA of 0.025 which is a middle ground of bulged corners and under/over extrusion.
I have not currently found any changes in this behaviour by changing the Input Shaper frequency nor shaper type
For extruder settings I was running a jerk value of 4mm/s (250mm/min) and an acceleration of 66mm/s2 (4000mm/min) - I am currently testing higher jerk and acceleration settings however data has been inconclusive so far -
RE: PETG as a support interface layer for PLA
I have got PLA / PETG printing working, the trick is to make sure the interface layer is 100% dense, and to slow down the print speed of that interface layer so it gets a better chance of adhering, and lastly to increase the support density overall, as more surface for the PETG to stick to massively improves reliability.
If you are using a Bambu machine (or similar one hot end multimaterial) then you MUST purge more filament than you expect, and once you have a value you are happy with, DOUBLE IT. Any PETG contamination will compromise the structural strength of the print, as PLA on PLA layers will refuse to stick to each other.
-
RE: Duet 3 Scanning Z probe
@fcwilt
Nope, thats on me, My wording skills are pretty poor at times.I probe before every print because I want to reduce any risk of an imperfect first level. A failed print because I tried to save 4 minutes in skipping a probe sequence just isn't worth it IMO.
And when you take a step back and look at the average time to complete a print (say 4 hours) the probing time is literally only 1% of the total time. So I'm happy to do the probing before every print.
But on the flip side of the coin, from a user experience aspect, the initial wait for 4 minutes just feels sluggishly slow, especially when doing really small prints (say less than 1 hour) the probing time just eats at the mind as I wait for it to complete.
my feelings are juxtaposed between happy with what I currently have, but also disappointed by the completely acceptable 4 minutes it takes.
What I am definitely looking forward to is the ability to do increase resolution height maps without any time impact. And after using a beacon probe, it does become a tipping point where once you have experienced those speeds, you never really can look at the slower methods the same.
Hope that makes more sense.
-
RE: Duet 3 Scanning Z probe
Brain dump regarding using the probe as the only probe.
In theory it would be possible to use the probe as the only sensing equipment as it includes the only variable you need. which is an amplitude reading from the probe (which is then converted to distance)
The issue is there is no datum point, as depending on the scanned surface material, the activation distance could vary.
HOWEVER, assuming there is no minimum distance to the sensing coils, you could put the probe directly against the print surface and take a measurement to create a "zero" reference measurement and store that for later use.
In reality you may want to take various zero point measurements at different coil temperatures and create a look up table of sensitivity vs temperature.Then once the coils are mounted you would need to set a single Z-offset and that should be good enough.
There is another way off the top of my head, but I cannot remember the algorithms name where you can repeatedly move a reference plate (in this case the bed) up and down a known amount (say 1mm and 2mm) and then you can work out the differences in measurements to gain a pretty accurate depth to probe value. The math was complex the last time I checked, but once I find the name of it, ill post it here as it may be helpful.
-
RE: Issues with pressure advance since RRF 3.4
This problem can be replicated on Hbot and other similar Cartesian designs. Its only more obvious because CoreXY tends to have higher speeds thus requiring higher flow rates. Overall, its almost entirely a flow control issue instead of a kinematic issue.
Klipper has a feature called smoothing time which works with their pressure advance - My current belief is that this smoothing time is the key to fixing this issue. I am currently building a test machine that will use klipper for an apples to apples comparison between RRF and Klippers implementation. But ill be having to do it after SMRRF as time constraints are tight right now.