@ekaggrat
Yes, but the problem is that when you do it like that (with the effector centered) errors in one carriage can compensate errors from other carriages, becoming into x and/or y errors and not being very visible in a z measurement. That is why it is a better method to to that with the effector close to each tower, so you isolate each carriage's movement. About the formula, of course you can calculate it, but I think there will always be some error due to different diameters, line tension, etc. Maybe it is not significant, I haven't gone into the detail of calculating the influence, but I can give you the values I have after calibrating each tower so you get an idea of the differences that are possible:
#define DEFAULT_AXIS_STEPS_PER_UNIT {53.015, 53.270, 52.512, 511}
@hercek
Thanks again for your insight with my problem, I have finally adjusted the printer and it is working fine. Initially I tried with your Maxima notebook and while I did get some results initially, after building a z-probe and trying again, I got to a point where I couldn't get the notebook to solve properly and was giving me some errors. I tried RichCattel's Marlin autotuning, but it didn't converge either. But it was a good working base to modify and do some further tests.
My next step, and one I'm pretty happy with was that I implemented a genetic algorithm python script to tune the printer. It worked very well and adjusted the geometry in the way you predicted. Actually it was a nice solution for me because not having a lot of time I implemented the script and left it running continuously while I continued with my busy life. But it did take a long time to find a good solution, genetic algorithms are solid, but slow.
However, there still was a 0.2mm error on the opposite of each tower that I couldn't get rid of... and it was all due to the fact that in that position, the geometry of the rods is less robust than inside of the triangle formed by the towers. So, some flexing was happening because I had an endstop based z-probe. So, I think a good tip with delta printers, or at least with the 3DR is to use a z-probe with hall effect sensor. I know someone will ask which endstop did I use, so it is an Omron D2F-F.
By the way, in the end my z-tower is 2.02mm away from the center.
Yes, but the problem is that when you do it like that (with the effector centered) errors in one carriage can compensate errors from other carriages, becoming into x and/or y errors and not being very visible in a z measurement. That is why it is a better method to to that with the effector close to each tower, so you isolate each carriage's movement. About the formula, of course you can calculate it, but I think there will always be some error due to different diameters, line tension, etc. Maybe it is not significant, I haven't gone into the detail of calculating the influence, but I can give you the values I have after calibrating each tower so you get an idea of the differences that are possible:
#define DEFAULT_AXIS_STEPS_PER_UNIT {53.015, 53.270, 52.512, 511}
@hercek
Thanks again for your insight with my problem, I have finally adjusted the printer and it is working fine. Initially I tried with your Maxima notebook and while I did get some results initially, after building a z-probe and trying again, I got to a point where I couldn't get the notebook to solve properly and was giving me some errors. I tried RichCattel's Marlin autotuning, but it didn't converge either. But it was a good working base to modify and do some further tests.
My next step, and one I'm pretty happy with was that I implemented a genetic algorithm python script to tune the printer. It worked very well and adjusted the geometry in the way you predicted. Actually it was a nice solution for me because not having a lot of time I implemented the script and left it running continuously while I continued with my busy life. But it did take a long time to find a good solution, genetic algorithms are solid, but slow.
However, there still was a 0.2mm error on the opposite of each tower that I couldn't get rid of... and it was all due to the fact that in that position, the geometry of the rods is less robust than inside of the triangle formed by the towers. So, some flexing was happening because I had an endstop based z-probe. So, I think a good tip with delta printers, or at least with the 3DR is to use a z-probe with hall effect sensor. I know someone will ask which endstop did I use, so it is an Omron D2F-F.
By the way, in the end my z-tower is 2.02mm away from the center.