New to Input Shaping
-
I've been working with the LC1 on my custom printhead (CoreXY) and the input shaper. I have had good results so far using zvddd. I do seem to have a rather high amplitude abrupt excursion/spike in the x axis at one small point centered around 47-59 Hz, depending on the speed (not sure the cause yet).
I am seeing an improvement in the test prints as far as reduced ringing.
I also would like to try testing with some slightly different belt tension adjustments to see if it makes any difference.
One thing I am curious about:
Would it be of more benefit the to repeat the input shaper test at various typical print speeds to get a better idea of what range of frequencies I want to try and tune out?... Or does it get all of the data it needs with the higher acceleration and max speed? I noticed on the default test that it just uses the maximum set speed and acceleration written in my config.g. The peak frequencies obviously vary in Hz based on the speed that the steppers are moving the printhead.
-
@br7408 the resonances do not depend on the print speed, however the speed and acceleration can affect how much those resonances are excited. This means that if you have more than one significant resonance, changing the speed or acceleration may affect the relative amplitudes of those resonances, but not their frequencies.
Changing the belt tension will affect the resonance frequencies, however as the frequency varies approximately with the square root of the tension, you would have to change the tension a lot to shift the frequency significantly.
-
@dc42 said in New to Input Shaping:
@br7408 the resonances do not depend on the print speed, however the speed and acceleration can affect how much those resonances are excited. This means that if you have more than one significant resonance, changing the speed or acceleration may affect the relative amplitudes of those resonances, but not their frequencies.
Changing the belt tension will affect the resonance frequencies, however as the frequency varies approximately with the square root of the tension, you would have to change the tension a lot to shift the frequency significantly.
After further testing and reading, I believe that I have determined that the issue I identified is not an input shaping issue. Its more inherant to shorter distances and speeds that are not able to be measured or cancelled by input shaping, and they are more likely unavoidable and based on physical characteristics of the printer. I am working through some more tests with longer accelerations/decellerations and radius curves to capture a transient resonance that I can truly test using input shaping. I will report back.
-
@dc42
I cannot thank you enough. Your selflessness, knowledge, and advice you've given over the last few years has helped me immensely and taught me so much. There is always a lazy answer, but you always take the time to explain. I hope Duet is paying you extremely well. I'll run into a snag or two, and I archive the forum, and its usually your advice that always leads me to a solution. Very rarely have I seen such efficient/ problem solving subject matter experts, and very few like you left anymore. Thanks. -
@br7408 thanks for the compliment!
-
@dc42 said in New to Input Shaping:
@br7408 the resonances do not depend on the print speed, however the speed and acceleration can affect how much those resonances are excited. This means that if you have more than one significant resonance, changing the speed or acceleration may affect the relative amplitudes of those resonances, but not their frequencies.
Changing the belt tension will affect the resonance frequencies, however as the frequency varies approximately with the square root of the tension, you would have to change the tension a lot to shift the frequency significantly.
So re-visiting this... I am seeing an a very clear and obvious change in resonant frequency that is based on printhead feed speed in the Y axis, where the resonant frequency varies based on the feed speed. The X axis resonant frequency seems to remain the same, regardless of speed/accel/jerk, although the amplitude does change. This does not seem right, according to your last statement.
-
@br7408 How are you measuring/determining this change in resonance?
-
Correction, I had it backwards... I am using M203 to gradually increase speed. As I increase the speed, the resonant frequency increases on the X axis on a single X axis move. The Y resonant frequency remains the same at all speeds on y axis moves. amplitude varies with speed for both.
I also do not understand the entire logic behind the option of capturing data during or after the move. I have been mostly capturing during the move. If I capture data after the move, I only get significant peaks in the Y axis, with a minimal amount on X. It would be great if there were a more in-depth explanation of this feature.
I am also confused as to why the motion profile only allows input for negative to positive axis movement for each move, but not the reverse of that.
I have both a 1LC and a Scanning Probe board on my printhead, and both accelerometer readings are similar and consistent on all tests.
One of my main print resonance issues seems to stem mostly from stepper vibration (big AWD corexy). I am primarily concerned with finding an outer wall print speed that will minimize this resonance. My theory behind this was to find an axis feed rate that evokes the lowest amplitude of resonance on the accelerometer at a practical printing axis feed rate, and then also apply IS to the frequency as well. My printer evokes very little resonance at higher axis feed rates, but these speeds are not practical speeds for printing under all circumstances. This is where I found the resonant frequency on the X axis varying based on axis feed rate. Trying to utilize a practical axis feed rate to read accelerometer amplitude and compare with what I see visually seems like it could be of some benefit.
Perhaps I'm off in that thought process, but until the developers provide a more comprehensive explanation of how Duet IS and motion profiles work and how it can be practically applied... I have scoured the internet for weeks for any answers I can find and I am good at figuring things out. There are still many great unanswered IS questions in this forum and I find the explanation given on the Duet website not sufficient enough to fully understand it.
What would also be phenomenal is some sort of implementation for using accelerometer data to compare and check belt tension on corexy printers, similar to what Klipper uses.