I think the settings in Bambu Studio are on the very safe side when it comes to how much filament is flushed during a change. I wanted a way to test what would be optimum.
I thought about how best to approach this and came up with a plan. I wanted to test the worst case scenario which is black filament to white.
In order to do this I made a two part print which comprises of two halves, one a single layer (black) and the other two layers (white). This will be printed white/black/white. That means it needs to flush sufficient of both for each half to be the correct colour.
I considered the settings in Bambu Studio and if I should just look at static numbers for the flush amount. If you flush to find the optimum flush amount for white from black in mm3 then it should be good for any colour, right? The issue with this approach is you will over-flush for many colours when it's not needed. The second approach isn't perfect but I decided to focus on the flush multiplier as the means for this test. I selected the colours for my filaments(in Bambu Studio) and used the auto-calc at flush rate of 1. I printed without any prime/purge block. This was the baseline and I'm using the settings the Bambu Lab team come up with for the flush value for each colour. I've included photos of the tests.
I then change the flush multiplier to try find the optimum value, such that there's no discernible bleed between the black and white. I went to the middle first, 0.5 which looked good. I then halved that again, 0.25. Bleed was clearly visible. I then moved up to 0.33, still bleed showing. Up to 0.42, slight bleed visible. I then tried 0.46 and to my eyes this looked good. I will still use a prime tower for quality so that's a margin for error in the findings.
I printed a test part which looked good. I'm happy and will use this now for future prints. It's a good saving and worth the small effort. My thinking is that if the calc function is working then the multiplier will allow the best control for any colour and reduce filament waste.
Hope you find this useful.
I've included my test part and 3mf.
UPDATE: thanks to comments about the possibility of the algorithm changing between version of the slicer I should add the tests were done with version 1.1. I believe the values are the same in 1.3. This is the baseline so you can compare future versions.
EDIT: I have also added a 16 colour test process.
https://www.printables.com/model/318506-bambu-lab-ams-test-prints-16-colours
The author marked this model as their own original creation.