Pressure advance calibration tool

[legend069]

new pressure advance calibration tool ! - yay :)
i've read how orca does this, and i wanted it to do it differently. not a fan of the "custom gcode lines"
per_objects_gcode i'm hoping to be a powerfull tool that can be used in later features / calibration tools
example;

• user configures it to add the 'camera snapshot macro' it would take a snapshot per that object
• user configures it to make the printer 'beep' when it prints that object.
• configures it to call a klipper macro to change toolhead.

theres a fair few areas where it can be used.

still alot of work to-do!

generates and slices model just fine.
few bugs to workout, mainly with CalibrationPressureAdvDialog.cpp model sizing, handling default values, converting float/percent values, hard coding in where retraction/seams would be, having the values be use input choices(not combo boxes)
per_objects_gcode later can be used to insert the smooth_time bits aswell.

can i please get help with inserting the per_objects_gcode where it's supposed to be ? 🙏
i'm not entirely sure how the Gcode generation works, and how to get it to process each object/volume and layers
would also need to add the "first layer pa"first_pa for the first layer only too(i've noticed during all my PA prints the first layers PA value needs to get set for the first layer speed, otherwise it looks like it has bad layer squish. but it's actually PA doing that.)

-currently generates the model and only writes the PA values at the start of the file(helps verifying it's correct)
note: cmake will throw a fit if you don't rebuild the deps 🙃

[supermerill]

Thanks, i'll review it tomorrow.
Do you want it to be merged on the 2.4, 2.5 or directly into the 2.7 branch?
Currently the master is on the 2.4.
and master_dev is on the 2.5.
Maybe I should just rename them master_24 and master_25 like prusa does.

[legend069]

Do you want it to be merged on the 2.4, 2.5 or directly into the 2.7 branch?

that's a good question!
currently alot of users are just jumping from release to release. so i think 2.7 would be best, just to help with debugging and the devs jumping between versions.

[legend069]

another bug i've found is the models will need to be scaled correctly so the selected extrusion roles widths will work correctly(although this might be already fixed in the 2.7 merge) but then scaling the models will "skew" the model slightly so it will no longer be a "perfect" 90° bend

pretty pictures :)

to add with the main calibration configure page i would like it to have a " dropdown" row so the user can select the extra calibration tools to make based on selection for "numbers of tests to create" i'm not at all familiar with wxWidgets 😢

[supermerill]

what do you mean by "dropdown row" ?
something like the flow help/details in the width& flow tab in print settings?

[legend069]

what do you mean by "dropdown row" ? something like the flow help/details in the width& flow tab in print settings?

theres been requests to be able to have several PA calibration tests on a single plate.
easiest way i've found is to have the user select them on the main GUI (screenshot below).
number of tests to create = 3 would create two separate rows for the user to adjust parameters for.
clicking generate would then create 3 PA tests each model would have the parameters for the selected ER roles.

[supermerill]

what you can do:
have a box were you select the number of items.
set a callback on this box

    my_combo_box->Bind(wxEVT_COMBOBOX,  [this, my_combo_box](wxCommandEvent&) {
        this->redraw_bars(my_combo_box->GetSelection())
    }, my_combo_box->GetId());

and when the value change:
. if more than before, add a row
. if less than before, remove a row

I think the best is to create a wxFlexGridSizer to store your table.
just set the number of cols.
when you add a line, add all item of the row (same number of Add as the col number)
when you remove a line, call Remove(GetItemCount()-1) the number of col you set.
I think it could work.

[legend069]

I think the best is to create a wxFlexGridSizer to store your table. just set the number of cols. when you add a line, add all item of the row (same number of Add as the col number) when you remove a line, call Remove(GetItemCount()-1) the number of col you set. I think it could work.

sorry for the delay, i haven't had much spare time recently.
i'm not familiar with WxWidgets so this is a real challenge for me. I'm going to need to put aside some more time and mess around with wxFormBuilder to see what would work for the GUI

[supermerill]

You can left it to me, but then you'll have to describe exactly the behaviour you need (input, output, drawing, etc..).

[FoldZero]

can you make it work for both .4 and .6 noxxles?
Orca currently has an issue with the .6 - looks to be a purely mechanical issue with proper line widths not being set in the generator properly.

[legend069]

can you make it work for both .4 and .6 noxxles? Orca currently has an issue with the .6 - looks to be a purely mechanical issue with proper line widths not being set in the generator properly.

thanks for mentioning that. i've changed nozzle sizes and everything breaks 🙃
with my new changes,
selecting a 0.4mm nozzle the wall count/ect is good
using a 0.8mm nozzle the wall count gets calculated to 6(not my target of 4)the model itself doubles in size, not ideal i know..

but i'll keep messing around with the math to try keep the models the same size+scale for widths for regardless of nozzle_size

[legend069]

You can left it to me, but then you'll have to describe exactly the behaviour you need (input, output, drawing, etc..).

thank you, I should be able to figure out the math/scaling issue to get everything aligned up.
due to the scaling/positioning issues, i think it would be 'easier' to create several models for each nozzle size. this way the models don't change size in xy, they will just have different 'thickness'

as for the GUI, it will just need to have that "dropdown box" so users can select other variables for the other PA test to run(so they can run several tests for 1 plate), how it 'looks' doesn't really matter that much.
planning on including "smooth_time" with it too, but that's an easy enough change with a bool checkbox and some text changes.

[supermerill]

working on it.

• i'll rework object_gcode so it work, and add region_gcode.
• i'll fix the acceleration (currently they won't split fills) and put the ironing one into the object.
  when it's finished, i'll retry your calibration.

[supermerill]

I've merged your latest changes into the /dev branch
03906fa
I applied this path of mine:

• move object & region-based config into the object, so the current main config won't change that much.
• apply ensure_on_bed() so there isn't any sinking model
• "update plater" only once.
  btw, on my win10, i didn't need to resize my window for the field to appear.

[legend069]

thank you, i think this tool is almost finished, just need more feedback from the users now.
i can't think of any other improvements for it.( until i get the PA injection per ER role completed that is. )

move object & region-based config into the object, so the current main config won't change that much.

i was planning on switching this over aswell, but wasn't sure what other configs might need to be changed. (still need more feedback from users)

apply ensure_on_bed() so there isn't any sinking model

might need to adjust the z scale/position for the numbers

btw, on my win10, i didn't need to resize my window for the field to appear.

that was fixed with the void CalibrationPressureAdvDialog::on_row_change() function, i don't think the flashing can be fixed though

[TheAndr0id]

i was planning on switching this over aswell, but wasn't sure what other configs might need to be changed. (still need more feedback from users)

I don't know if this is the correct place to provide the above feedback, but I guess if anything my feedback will be connected to the PR for the change.

I've been using Ellis' PA tool for a while now and while it works well enough, having something in the slicer that would pull all the settings in sounds great. It's close, but I have some issues/concerns/suggestions.

• The trailing 0s on the labels are a waste of filament, and the labels in general are too big. A major benefit of Ellis' tests is they are fast and don't use much filament unlike the tower print tests for PA/LA in OrcaSlicer. With the label size this benefit is lost.
• The test generation doesn't take bed size into consideration. Generates tests that can't be printed because they don't fit on the bed.
• No control over the pattern spacing like Ellis' tests. I can place more tests in the row with closer spacing and thus spend less time doing multiple prints to cover the range I want to test. (for example I use a 2mm spacing with the Ellis' test)
• Additional description on what exactly the Extrusion Roles are. Is it just using the speeds/acceleration for the specified role? Flow rate overrides? Sorry, I'm just not understanding...
• Last used values need to be saved. Having to reenter all the settings every time gets real old quickly.
• First layers LA should default to off (or "0").
• I understand pictures are still a todo item, but how to interpret the tests is very important. I'm glad you included the link to Ellis' page as there is some description/direction there.

[legend069]

* The trailing 0s on the labels are a waste of filament, and the labels in general are too big.  A major benefit of Ellis' tests is they are fast and don't use much filament unlike the tower print tests for PA/LA in OrcaSlicer.  With the label size this benefit is lost.

the trailing '0's have been limited to '2'. the numbers can't get any smaller if they do then it would get considered as a "thin wall" and parts of it won't get sliced/printed.

* The test generation doesn't take bed size into consideration.  Generates tests that can't be printed because they don't fit on the bed.

why should it take the bed size into consideration? if the user generates a model that's too big then that's their fault not the software?

* No control over the pattern spacing like Ellis' tests.  I can place more tests in the row with closer spacing and thus spend less time doing multiple prints to cover the range I want to test.  (for example I use a 2mm spacing with the Ellis' test)

this has been fixed and the spacing is calculated based on the thickness of the 90_bend model / other factors.

* Additional description on what exactly the Extrusion Roles are.  Is it just using the speeds/acceleration for the specified role?  Flow rate overrides?  Sorry, I'm just not understanding...

extrusion role = external/internal perimeter ect. you can set different speed, width, fan cooling, layer heights ect per extrusion role. most users calibrate PA for the external and leave it as that, but this negatively impacts the other roles, especially supports/ internal infill if they're done at different heights.

* Last used values need to be saved.  Having to reenter all the settings every time gets real old quickly.

when the PA per extrusion role code is merged i can make the tool pull the values the user has currently saved in their config.

* First layers LA should default to off (or "0").

please explain why.

* I understand pictures are still a todo item, but how to interpret the tests is very important.  I'm glad you included the link to Ellis' page as there is some description/direction there.

i don't have the time/hardware to take high res /close up photos :(
this tool is similar to ellis' but it's still different due to how the spacing/width is done. so the method to interpret the printed results is technically different.

[TheAndr0id]

the trailing '0's have been limited to '2'. the numbers can't get any smaller if they do then it would get considered as a "thin wall" and parts of it won't get sliced/printed.

I just opened an issue ticket #4738 that shows the issue with the extra zeros. With the current release it is printing 4 extra zeros on the labels. I didn't see a way to change that.

* The test generation doesn't take bed size into consideration.  Generate tests that can't be printed because they don't fit on the bed.

why should it take the bed size into consideration? if the user generates a model that's too big then that's their fault not the software?

Initially I would say user experience. It's frustrating to go through all the settings to find out the generated test is not going to work and have to start over. If the user doesn't know how big the generated test is going to be is that the fault of the software or the user? 😄

extrusion role = external/internal perimeter ect. you can set different speed, width, fan cooling, layer heights ect per extrusion role. most users calibrate PA for the external and leave it as that, but this negatively impacts the other roles, especially supports/ internal infill if they're done at different heights.

Thank you, that helps. I've also been playing around a bunch with the test today so I've gotten a better feel for what these roles are.

* First layers LA should default to off (or "0").

please explain why.

If you are running the test to determine LA, how would you know what value to set the first layer LA to? Also since (usually) the first layer is printed very slowly for bed adhesion LA doesn't really do anything at slow print speeds.

* I understand pictures are still a todo item, but how to interpret the tests is very important.  I'm glad you included the link to Ellis' page as there is some description/direction there.

i don't have the time/hardware to take high res /close up photos :( this tool is similar to ellis' but it's still different due to how the spacing/width is done. so the method to interpret the printed results is technically different.

Maybe a diagram with a description? Sometimes photos are hard to explain what one should look for.

If this test has different results than the Ellis test what is the process for interpreting it?

[legend069]

why should it take the bed size into consideration? if the user generates a model that's too big then that's their fault not the software?

Initially I would say user experience. It's frustrating to go through all the settings to find out the generated test is not going to work and have to start over. If the user doesn't know how big the generated test is going to be is that the fault of the software or the user? 😄

lets say that the code gets added to calculate the size of the test, it will then prompt an error and the user will still have to adjust the values again.
the window can be dragged out of the way while the main UI is still accessible, i think the help window says this too.

• First layers LA should default to off (or "0").

please explain why.

If you are running the test to determine LA, how would you know what value to set the first layer LA to? Also since (usually) the first layer is printed very slowly for bed adhesion LA doesn't really do anything at slow print speeds.

there are some users that print the first layer at 200mm/s
since the first layer is printed at a different speed regardless of the amount it will need it's own LA value.(i would like to get data on users with larger nozzle sizes and how this works here too.)

I understand pictures are still a todo item, but how to interpret the tests is very important. I'm glad you included the link to Ellis' page as there is some description/direction there.

i don't have the time/hardware to take high res /close up photos :( this tool is similar to ellis' but it's still different due to how the spacing/width is done. so the method to interpret the printed results is technically different.

Maybe a diagram with a description? Sometimes photos are hard to explain what one should look for.
If this test has different results than the Ellis test what is the process for interpreting it?

between extrusions 2 and 3 there would normally be "internal infill" there is no overlap in this area they're "perfectly" spaced so the extrusions will be barely touching, if in a perfect world. (i don't think this is in the current release, i had to do some complex math to figure out how to get it like this.)
with more time and experimentation this could be used as a way to "calibrate it". as i've found the gap "fades" through the printed model.
the user should be selecting the value that best resembles "sharp corners", none of the rounded stuff.
the biggest problem

[TheAndr0id]

lets say that the code gets added to calculate the size of the test, it will then prompt an error and the user will still have to adjust the values again. the window can be dragged out of the way while the main UI is still accessible, i think the help window says this too.

I may very well have missed that help message, and that's on me - but if I missed it, others will as well. I was thinking of a single line status message stating something like "test exceeds available print surface" on the bottom of the window in red/yellow to make it clear that the test isn't going to fit.

It would also help with determining the resolution of the test being created as well. For example, I want to test 0.02 -> 0.03 values, but can I go in increments of 0.005 or even 0.001? I select the value and determining if it will fit would be a simple formula to calculate how long and wide that test would be. Compare that to rect_size(0) and rect_size(1) and report an error/warning message.
(please PLEASE don't make it blink... 😄 )

It makes it much easier to quickly generate test parameters without having to generate the test, find out it doesn't fit, clear the print surface, go back into the test window change values and repeat. Even moving the window out of the way (I work on a smaller screen, there's no way to have the print bed and the test config windows up at the same time...) is a bit of a hassle for the sake of a single line warning message.

• First layers LA should default to off (or "0").

please explain why.

If you are running the test to determine LA, how would you know what value to set the first layer LA to? Also since (usually) the first layer is printed very slowly for bed adhesion LA doesn't really do anything at slow print speeds.

there are some users that print the first layer at 200mm/s since the first layer is printed at a different speed regardless of the amount it will need it's own LA value.(i would like to get data on users with larger nozzle sizes and how this works here too.)

And that's fair, but wouldn't a value of zero be the safest default? I'm going to change it to zero every time I run the test since I don't want weirdness happening because I used a value I shouldn't.

I may very well be interpreting this test incorrectly, but I've never even considered the first layer in this type of test. I'm always looking at the results of the top layer to see what/where gaps are sharpness of the corners, etc.

In reality, I am currently using K values in the 0.22 - 0.24 range. I have NO idea what a K value of 0.04 will do... (I'm guessing it would not be good)

I understand pictures are still a todo item, but how to interpret the tests is very important. I'm glad you included the link to Ellis' page as there is some description/direction there.

between extrusions 2 and 3 there would normally be "internal infill" there is no overlap in this area they're "perfectly" spaced so the extrusions will be barely touching, if in a perfect world. (i don't think this is in the current release, i had to do some complex math to figure out how to get it like this.) with more time and experimentation this could be used as a way to "calibrate it". as i've found the gap "fades" through the printed model. the user should be selecting the value that best resembles "sharp corners", none of the rounded stuff. the biggest problem

This sounds very similar to the method described in the Ellis test. I wouldn't worry about getting "high def" photos of the tests. If you use modeled extrusion paths (like the image used above) showing a bad result and a good result that would give a starting point of what to look for.