My blue plates should be here any day. Ill let you know how they go. I know the x1c ones are great.
1 Like
My first thoughts are that this is being somewhat over-complicated.
Trying to characterize how the entire printer expands at different temperatures and gradients, with different materials throughout, is essentially impossible. But, I don’t think we need that data… as long as the printer doesn’t distort when heating up (skew, etc).
The biggest thing we should be concerned about is how much the belt pitch changes with temperature. My guess is that this is very little. I went ahead and sent Gates an email asking, as I can’t find any public data about the coefficient of thermal expansion of timing belts. We will see if they can help.
But, even if the expansion is above negligeable, I think it could easily just be handled within the material shrinkage settings. Let the plate calibrate for any steps/mm cumulative error, skew, etc.; and let material shrinkage handle the sum total of belt expansion and material shrinkage.
Depends on if they used high temp belts or not.
Nothing I make has to be 100% perfect, but its a fun conversation lol
Maybe, maybe not.
The high temp belts are rated to retain strength and stiffness at higher temps. But, do you have any data saying that the thermal expansion is different?
I haven’t looked, but there is a chance that the max chamber temps don’t even fall outside the temp rating for a standard belt.
I dont have any data. Just stuff ive read over the years. I also dont know the limits of the standard belt.
From what ive read the loss of stiffness is what causes excess stretch though. Besides my bambus, i calibrate with the cube at temp, so never even had to think about it. Prusa has pulled it of until now, because no enclosure
Now that i think of it, this is easy to test. Do a full calibration for pla, and then print a large model (maximum size your calipers can still measure). Then do the same for PPS-CF or whatever other high temp filament you have on hand and using the same print model. Then compare caliper measurements for each.
1 Like
I bet its not enough to matter unless youre making some crazy parts for a submarine or something lol. But fun experiment
But, you are conflating material shrinkage and machine related deviations.
I think we could start to get some data by running a room temp encoder cal, and then running it again to see what values is reports on the 1st and 2nd run. Basically, does it report the same/similar avg and max deviations on both runs.
Then, heat soak the bed and the encoder plate to something a bit above the max chamber temp you are interested in, but leave the door open to avoid heating up the belts/frame/etc. Run it again, ideally with the bed still on and the door still open. Whatever difference is observed can be attributed to the expansion of the plate.
Then, heat soak the whole chamber and plate and run a final time. Any difference from the last one could be attributed to the belts/frame/etc response to the temp.
1 Like
In my opinion, a different temperature for printing compared to vision plate calibration shouldn’t matter too much:
1.) I looked for a spec of the temperature coefficient of gates GT2 belts but didn’t find anything. Actually, I wouldn’t be surprised if there isn’t much influence. I don’t know which type of fibers is inside those belts, but probably they have less elongation than steel. And while the bed could be anywhere between 20 and 120 °C, the belts won’t see more than 20°C - 60°C range.
2.) I see the main benefit of the vision encoder plate in correcting non-linearities of the motion system and skew correction. When you do shrinkage compensation afterwards, you will eliminate scaling errors anyway.
Only thing I could find was above 70c
Which ones did you get? I’m running the glacier right now with some PPA-CF. Did they release the frostbite yet?
Agreed.
Plus, it should actually compensate for some runout on the stepper pulleys.
But if you don’t test it, it’s still just a theory. 
Maybe a good theory. But you won’t know for sure if it’s true or not without the empirical test.
History is littered with good theories that turned out to be false.
Test what?
. . . . . . . . . . .
Good discussion here and a lot of things can be taken into account to get out the perfect model.
Main question here is: do we need to get that much in detail to get a good print?
Lets say you live in the perfect world and you are able to compensate all the positional deviations that occur in a heated 3D printer, then there is one more factor to take into account and that is the shrinkage of the part.
And that is not a linear shrinkage, that you can adjust with a single number, but is heavily depending on the geometry of the part and the direction in which it is printed.
The shrinkage of the layer that lays on top of layers that have already been partially cooled will generate stress in the part and causes a non-uniform shrinkage.
Though this simulation (image below) is done for a powder bed 3D printer for metals with higher melting temperatures than our FDM process, it clearly shows the te non-uniform shrinkage of parts.
And here the question again is, do we need that amount of accuracy for our parts?
But coming back to the point where the discussion is if it is useful to have a vision calibration plate that is accurate on 65C, I still think the answer is no.
To the question if it is useful to do a motion calibration with the standard vision calibration plate on 65C bed in a chamber of 65C, I would say yes, if at this temperature significant skew would introduced that would require compensation for the accuracy you seek.
I personally think that will not be the case, but the problem is that with the current firmware, we do not know what the skew correction is that the vision plate is correcting.
If you would decide to calibration at a higher temperature, you face the problem that you would have to redo this at lower temperatures again if you would be printing at those lower temperatures.
It would be useful if the firmware of the printer would be able to store the correction for different temperatures, and that the correct data could be retrieved when required.
I have currently done something similar for my bed mesh on my Voron.
For temperature intervals of 10C I have stored the bed mesh in the controller and select the bed mesh that belongs to a certain temperature.
Conclusion:
If the change in skew between 20C (ambient) and 65C (max heated chamber) is within acceptable limits, I would not worry about calibrating at different temperatures.
If you would do so, you would eventually end up with a slightly different shrinkage factor because you calibrated te motion system at a different temperature.
1 Like
I can think of a number of things where the more accuracy, the better. One would be large gears that you want to mesh together extremely well. Maybe you want to print the gears directly, or maybe you want to print a model that can be used for casting such a gear.
I think the more accuracy you have, the more clever people will come up with ways to put it to use. If you wanted to 3d print a zero backlash gearbox, I’m pretty sure it’s going to take above average accuracy:
For me, it made an incredible difference.
I was getting acceptable, but not great, finishes on many of my prints, and spent about 30 hours trying to get a decent ironing profile sorted out. Never could succeed with any ironing settings that worked with HF nozzles, only standard (thread on Ironing reached the consensus that cooling was overpowerful and ruining the ironing lines), so I gave up.
Few days later, got my vision plate, ran the calibration test, and printed a box for a geocache. The surface finish of the inside feels like its ironed:
Its not the 100% glossy perfect my tuned X1C could get, but it’s consistent. The X1C would give you beautiful ironing, or it would choose to just spit out garbage, and there was no way to tell ahead of time.
1 Like
The low textured glacier plates. Cant wait for the higher textured frostbite, but still not available
If only there was a calibration that took all of these things into account 
and also didnt require firmware access