Input Shaper with Klipper Testprint
With the Mk4, the frequencies are universal in the firmware and there is currently no way to use an ADXL sensor. I would be interested in how the individual printers differ or whether they are relatively close to each other. One could use the test print from Klipper to determine the settings. This might also be interesting for Prusa. Maybe completely superfluous because apparently input shaping works well with the MK4.
RE: Input Shaper with Klipper Testprint
I have 2 GCode for the MK4 high with PLA and PETG.
It is used to measure the resonance.
In this video everything is explained in detail.
Here you can simply enter your measurement and get the frequencies calculated.
Simply post the results here. Would be a contribution to show the effects of input shaper
RE: Input Shaper with Klipper Testprint
For me the test showed the following.
X44.6
Y37.8
Belt tension a little too tight. Measured with the app.
Printer stands on a stone slab.
No filament rollers on the printer
RE: Input Shaper with Klipper Testprint
With the PLA gcode you posted, the speed varies quite a bit - about 90 to 170mm/s.
Using mostly default IS settings with the latest PrusaSlicer, the speed varies from about 89 to 109mm/s.
In Chris's video, he shows a constant 100mm/s speed at timestamp 6m44s.
How much do you think the speed differences matter? One thing I noticed is that with Jessie PLA @170mm/s, the surface is fairly matte, making it difficult to see the ringing artifacts. At 100mm/s, the print is more shiny, making both the ringing and the "Vibration artefacts on Y-Axis" mentioned in another thread easier to see.
RE: Input Shaper with Klipper Testprint
I found the trick to having a constant speed for the entire layer is to disable "Enable dynamic overhang speeds" in Print Settings / Speed.
RE: Input Shaper with Klipper Testprint
Did you use my GCode?
When I look at the speed of my Gcode in Prusa Slicer, the PLA variant 170mm shows continuouslyThe PETG almost 160mm.
The absolute speed does not matter except that the artifacts see better at greater speed.The speed should be even.
Just for understanding, you print my Gcode. Then you are looking for the best areas for X and Y.
You give the measured value on the website that calculates the frequencies.
You can then change this in your input Shaper printer profile. If you are now slicing the test yourself, the artifacts should be better.
RE: Input Shaper with Klipper Testprint
Yeah, I printed with your gcode and I'm also trying with my own. You can see the slowdowns in your screenshot. Where the shape is not flat the color turns yellow or orange.
I imagine the speed changes are not a big deal, but in my initial test prints, it's hard to determine exactly which layer is best, so I'm attempting to adjust the slicing to get the most pronounced ringing artifacts while also reducing (what I presume to be) non-ringing artifacts I'm getting that look like what is being discussed in the other thread.
RE: Input Shaper with Klipper Testprint
Actually, all speed brakes are turned off. But really it's all about provoking the artifacts.
The area where the input shaper works is quite large. That's the only reason why it can work as prusa plans.
The fewer vertical artifacts, the better you can see the ringing. And they don't go away with input shaping. It has another reason.
RE: Input Shaper with Klipper Testprint
lol this is all to much complicated for me. haha ^^
RE: Input Shaper with Klipper Testprint
That's why prusa does that with the profiles.
It doesn't get any easier with an accelerometer. And not necessarily better.
The problem with the bed getting heavier has nothing to do with the sensor at all. prusa tries to compensate for this by lowering the frequency with increasing weight. The acceleration is also taken back.
Just print out the test print. Best with shiny material. And not black or white. Then you can see how input shaper works.
RE: Input Shaper with Klipper Testprint
@Rainer - I printed the test object using your PETG gcode and Prusament Anthracite Gray, firmware "5.0.0-RC+11963".
I am a little confused by the X and Y markings on the test part. They seem reversed. But all the instructions say use X-marked leg for X (even though it is printed on Y axis) and use Y-marked leg for Y (even though it is printed on X axis).
The leg marked X... Best quality was around 32mm, which calculator says is 38.94Hz. (Firmware default is 50Hz)
The leg marked Y... Best quality was around 39mm, which calculator says is 44.24Hz. (Firmware default is 40Hz)
My Mk4 is on a 38cm x 38cm x 50cm concrete block, with stock rubber feet between Mk4 and the block. The block is in a Prusa Enclosure, with (6) of these between the block and the enclosure floor: https://www.amazon.com/gp/product/B075KK8WXM
Not sure if I should try those calculated values in firmware or not, as they seem very different from the stock values.
RE: Input Shaper with Klipper Testprint
The marking with and Y fit like this. It is also mentioned in the linked video.
Your values are also plausible. I also have the printer on a stone slab and dampers underneath. I have similar values too.
But you can also see the large area in which the input shaper works.
Try it. Just print the test print with a normal IS profile. You don't need to print the whole height. Change only the frequencies in the printer profile for comparison.
RE: Input Shaper with Klipper Testprint
Okay, I gave the custom settings a try. There was a slight improvement on X, while Y looked unchanged. Details below.
As mentioned in my earlier post, the test gcode (with graduated frequency changes) suggested X=38.9Hz and Y=44.2Hz, while the stock values in the PrusaSlicer profiles are X=50.7 and Y=40.6).
I sliced and printed the test part with Prusa's stock settings, Prusament PETG, "0.20 SPEED @MK4IS" (just changed to Vase mode). The print looked really good, virtually no ringing. Even with the shiny gray PETG, I had to look really hard, use magnification, and get just the right light to see some very, very light echoes on the X-axis (the leg marked Y). I really could not detect any ringing on Y (the leg marked X).
Next, I changed the M593 frequencies in the Start gcode to X=38.9 and Y=44.2, re-sliced, and printed again. The Y axis (the leg marked X) looked identical to the stock settings print. The X axis (the leg marked Y) was very slightly improved -- the very light echoes seen in the stock print were still there, but even lighter and harder to see.
So, the improvement was small -- but it was there. I am curious what the test would have shown before placing the printer on the concrete block and isolation dampers. Unfortunately, the block and printer are in an enclosure now -- and my curiosity is not strong enough to warrant the effort of taking it out. But if I ever have it out of the enclosure for maintenance, I will repeat the test to see if the suspension difference changes the frequencies.
RE: Input Shaper with Klipper Testprint
Great that you did the test 👍
This also makes it clear why the 😭 is largely not justified due to the missing acerolometer.As you can see, the changes are not very big.
Especially by the Bambu fans, it is always shown as a big disadvantage.
I have already tried the other input shaper that is possible. But with not so good result.
RE:
To be fair to those questioning the lack of accelerometers, that test print has very small mass. I know Prusa is interpolating frequency values at layer change to compensate for mass, but do we know how well that compensation works with more mass on the bed (i.e., a large print)?
I find it likely that a static value for X, based on manufacturing samples, is going to work pretty well (at least for unmodified extruder and carriage). The X-carriage mass doesn't change during the print.
I have slightly less faith in the frequency interpolation method Prusa is using on Y for weight change. Maybe it works great -- but I would like to see some data from larger prints. In any case, the changing bed mass makes me think an accelerometer (and firmware to use it) for the Y-axis would be of more potential benefit than one for X-axis.
RE: Input Shaper with Klipper Testprint
However, the acerlorometer does not change this aspect.
This is the construction-related disadvantage of a bedslinger compared to coreyx. That's why the MK isn't any bigger. In order to keep the mass of the bed as small as possible.
It is always only the state at the beginning that is measured. Never during printing.
The sensor would have no influence on the increase in weight due to the print.
Prusa compensates for it by reducing resonance and acceleration. Whether it works, I have no idea.
I couldn't see any disadvantages in my prints. On the contrary, I am always enthusiastic. But I have to admit that I don't use the whole volume.
RE: Input Shaper with Klipper Testprint
I've just performed this test twice, once with squash balls and once without. Its really hard to see any ringing in both cases, however there does appear to be a slight difference between the two (which you'd expect.)
With squash ball suspension: X = 37.42; Y = 41.21
Without squash ball suspension: X = 29.85; Y = 18.48
This is running the latest IS RC firmware and standard Prusa 0.4mm brass nozzle.
There's one thing I'm not clear on: if I added the GCODE with these specific frequencies through PrusaSlicer, would they apply or does the machine just use the Prusa hardcoded values in the firmware regardless?
RE:
There's one thing I'm not clear on: if I added the GCODE with these specific frequencies through PrusaSlicer, would they apply or does the machine just use the Prusa hardcoded values in the firmware regardless?
The printer boots up with the values as stored in firmware. These are the values that would be used if a Gcode file was selected that did not have any M593 commands in it. Here's an M593 response immediately after booting, before any print has been run:
Send: M593 Recv: echo:axis X type=MZV freq=50.700001 damp=0.100000 vr=20.000000 Recv: echo:axis Y type=MZV freq=40.599998 damp=0.100000 vr=20.000000 Recv: echo:axis Z disabled Recv: echo:weight_adjust y freq_delta=-20.000000 mass_limit=800.000000 Recv: ok
The PrusaSlicer printer profiles have M593 commands in the "Start Gcode" which override the printer defaults (though, at least for now, using the same values). This is where you would place your custom values, assuming you are using PrusaSlicer.
M593 X T2 F50.7 M593 Y T2 F40.6
The PrusaSlicer printer profiles also have "End Gcode" that contains M593 commands to entirely disable IS after the print is complete:
M593 X T2 F0 ; disable IS M593 Y T2 F0 ; disable IS
Here's an M593 response after those commands are sent:
Send: M593 Recv: echo:axis X disabled Recv: echo:axis Y disabled Recv: echo:axis Z disabled Recv: echo:weight_adjust y freq_delta=-20.000000 mass_limit=800.000000 Recv: ok
RE: Input Shaper with Klipper Testprint
Perfect - thanks very much. Will have a play and see what difference it makes...
RE: Input Shaper with Klipper Testprint
This topic has been super helpful.
I have an MK3S+ that was just upgraded to an MK4, but it has its own unique wrinkles because I've chosen to leave the squashballs on the printer, plus I've opted for the misumi bearings and rod upgrades. There is no way a set of stock values are going to best represent this piece of hardware, so crafting a test print via the above linked video allowed me to find the right g-code to add to the 'After layer change G-code' section:
M593 F{(layer_num < 2 ? 0 : 15 + 45.0 * (layer_num - 2) / 297)} ; Hz Input Shaping Test
G92 E0
The video describes these settings well. I recommend starting with the existing MK4 0.4 IS printer definition in PrusaSlicer and make the suggested print changes. With all of the correct settings, I believe the test print is cranked out in 30 minutes.
M593 X T2 F35.9 M593 Y T2 F35.1
It would be nice if Prusa published a defacto MK4 Gcode test print so that we don't have to go through the whole process of figuring out how to slice these on our own. We should be able to download something pre-sliced that we can crack open to find the appropriate input shaper values.