Awesome! One thing I've done doing something like this is randomizing my measurements points. Ive been able to really see if there is a drift (thermal expansion, or some other error) in the system. As you will have an early measurement next to a later measurement. Super awesome work, love what you are doing!
Controlling the room temp will be the bane of this. Winter is coming up and with the big temp change from outside to inside the garage I’ll be running a space heater that will be intermittently turning on to heat the air. Maybe I should get a good temperature sensor, weld a small metal box around it (so as not to measure the air) and attach it to some part on the CNC close to where the laser is attached. That way I know the temp of the CNC. I could then calibrate and account for thermal expansion of the steel and concrete of the CNC machine making it less important to control the room temp.
What are your thoughts on smaller x & longer y axis for stiffer X gantry. Also if someone were to implement your X gantry design for PrintNC mini, how would they proceed? Love the videos.
Smaller X than Y would be more rigid but I do prefer having a wider X and having the CNC machine face me when opening the door and easier access to everything. The gantry with the double tubes filled with grout in the middle could be done on the mini with the motors inside. You could do the same with the Y tubes.
Using the same mechanism to do the machining and the measurement is going to create issues where the same mechanical errors will show up in both with a fixed amount of offset. You may be able to make some kind of compensation map if you can properly quantify it.
Yeah and I honestly don’t know how useful it would all be. I think if I wanted to machine a surface flat, one way to do it would be to map out in machine space in XY using the webcam sensor while it’s not touching anything. That would give me the grid of points that represent the offset in Z to the laser beam I could use to correct for. I could then use that to machine a flat surface. The sensor touching the surface could be used to verify if the surface is flat to the laser after. I think that would work, and you get to run a more conventional machining op instead of a whole bunch of circles.
Neat! Would it help to gate the data points and reject any that are impossible? I mean while you're taking measurements, so it'll re-check any screwy points.
It’s doable but would only help on the lowest ones. In the probing of this coin, I had the area in blue that is outside the coin not touch any surface so it’s just air sampling. What would be good is recording an average surface deviation for the last X number of points for that row. If a point deviates by some percentage beyond the average, it’s likely an erroneous sample that needs to be redone right than. Trying the find a solution that is on the fly to find artifacts is beneficial because if I try to sample them after the main job, the thermal expansion will make it near impossible to sample a single point and have it fit correctly with the surrounding local points. When all else fails points can always be moved to the average of the 6 local points surrounding it, but it'd be great to have every sample be a real one.
Hello, is there some way to reach you?? I dont have a problem making a ''donation'' for your time spent. RUclips doesnt allow me to post my email here, so if there is a place to reach you, let me know. thanks
Yeah gonna do that as a last ditch effort when all else fails. I should be able to detect bad samples by tracking the last X number of samples and comparing the height deltas. If the change is exceeded by some percentage then I can resample it right there. Those bad samples seem to be very infrequent like around 0.025% out of 100,000 so I could probably make it overly cautious when detecting them.
Awesome! One thing I've done doing something like this is randomizing my measurements points. Ive been able to really see if there is a drift (thermal expansion, or some other error) in the system. As you will have an early measurement next to a later measurement. Super awesome work, love what you are doing!
Controlling the room temp will be the bane of this. Winter is coming up and with the big temp change from outside to inside the garage I’ll be running a space heater that will be intermittently turning on to heat the air. Maybe I should get a good temperature sensor, weld a small metal box around it (so as not to measure the air) and attach it to some part on the CNC close to where the laser is attached. That way I know the temp of the CNC.
I could then calibrate and account for thermal expansion of the steel and concrete of the CNC machine making it less important to control the room temp.
What are your thoughts on smaller x & longer y axis for stiffer X gantry. Also if someone were to implement your X gantry design for PrintNC mini, how would they proceed? Love the videos.
Smaller X than Y would be more rigid but I do prefer having a wider X and having the CNC machine face me when opening the door and easier access to everything. The gantry with the double tubes filled with grout in the middle could be done on the mini with the motors inside. You could do the same with the Y tubes.
Using the same mechanism to do the machining and the measurement is going to create issues where the same mechanical errors will show up in both with a fixed amount of offset. You may be able to make some kind of compensation map if you can properly quantify it.
Yeah and I honestly don’t know how useful it would all be. I think if I wanted to machine a surface flat, one way to do it would be to map out in machine space in XY using the webcam sensor while it’s not touching anything. That would give me the grid of points that represent the offset in Z to the laser beam I could use to correct for. I could then use that to machine a flat surface. The sensor touching the surface could be used to verify if the surface is flat to the laser after. I think that would work, and you get to run a more conventional machining op instead of a whole bunch of circles.
Neat! Would it help to gate the data points and reject any that are impossible? I mean while you're taking measurements, so it'll re-check any screwy points.
It’s doable but would only help on the lowest ones. In the probing of this coin, I had the area in blue that is outside the coin not touch any surface so it’s just air sampling.
What would be good is recording an average surface deviation for the last X number of points for that row. If a point deviates by some percentage beyond the average, it’s likely an erroneous sample that needs to be redone right than.
Trying the find a solution that is on the fly to find artifacts is beneficial because if I try to sample them after the main job, the thermal expansion will make it near impossible to sample a single point and have it fit correctly with the surrounding local points.
When all else fails points can always be moved to the average of the 6 local points surrounding it, but it'd be great to have every sample be a real one.
What lib do you use for 3d plotting?
plotly
Hello, is there some way to reach you?? I dont have a problem making a ''donation'' for your time spent. RUclips doesnt allow me to post my email here, so if there is a place to reach you, let me know. thanks
Yeah, you can get my email on the RUclips about tab on my channel
i would just delete the outliers and lerp from adjacents.
Yeah gonna do that as a last ditch effort when all else fails. I should be able to detect bad samples by tracking the last X number of samples and comparing the height deltas. If the change is exceeded by some percentage then I can resample it right there. Those bad samples seem to be very infrequent like around 0.025% out of 100,000 so I could probably make it overly cautious when detecting them.