I would possibly say that the order of 2 and 3 could be reversed.
In other words, this test model allows for inside and outside measurements, that eliminate the under/over extrusion.
So scaling could already be compensated before flow calibration.
But possibly the order of the last 2 don’t matter that much.
1 Like
Right, in theory, but given the choice, would there be any reason to put step 3 ahead of step 2?
Likewise, if, say, an uncalibrated H2D were wildly overextruding some new filament, to the point where the shrinkage compensation model started to look like a droopy Dr. Seuss pudding cake… I’m not confident @Alex_vG 's model would correct for that extreme amount of overextrusion. So, that’s why I would put #2 before #3.
However, if you have a counter-argument to that, my mind is still open and I’m willing to be persuaded.
1 Like
I can fully agree with your argument.
And then of course, step #2 and #3 basically need to be repeated for each material again, depending upon how much you require that accuracy.
1 Like
Just like now, with the vision encoder. Only its 3 seperate steps for each filament. Instead of 1.
This is how we made perfectly sized parts for years. Until x1c showed up and people decided it was ok to be a little off.
Just pointing out that there was a way to get perfect parts before the encoder plate came along. The encoder plate only calibrates motion. Missing all other factors. A one trick pony
I have a feeling that if bambu went the estep route, people would be saying it was brilliant
I disagree.
Motion accuracy calibration is not required every time. Only after a longer period or for instance a belt tensioning procedure.
Dialing in your filament, and scaling are basically one process, that normally has to be done for each filament, one way or the other.
That can be, but that was before there was a vision encoder.
The vision encoder removes one variable from the calibration process, making it easier to dial in the remaining ones.
To my opinion the vision encoder for a 3D printer is the same as laser calibration of CNC machine tools. It improves the positional accuracy.
Still other factors like spindle backlash, tool deflection etc. determine the end result.
It is evolution. The argument “we always did it like that in the past” is not an excuse.
I was saying that when using the plate, youll still have to calibrate each filament.
The plate was most likely added for the 10w laser
Its basically just automated scaling that isnt filament specific(the reason why its most effective for the laser). So more scaling may be needed still. It measures two points on the plate, moves and measures again. Finding the scale factor. The cube does the same thing but jnstead of trusting a camera, we trust calipers. It says, for every 20mm of movement.add or subtract (number here). Also works for the extruder
One method gets you straight to the end result you want for any specific filament and doesnt cost as much as a cheap printer.
As mentioned earlier, dialing in a 20mm cube says nothing about a 200mm part you will be printing after that.
1 Like
^^^This. This is the key point.
“for every 20mm” Does the plate measure an infinite number of distances? or a couple x and a couple y?
Im guessing youre not understanding how calibrating esteps work. Its ok.
The H2D will aparently be the first accurate (movements atleast) printer.
Not infinite, but it appears to measure all x and all y–optically–by doing sweeps of the vision encoder plate. Otherwise, I don’t think it could promise accuracy to <0.03mm for x and y coordinates.
How about z? Wish there was more info on it. Or any
I believe we are all guessing
No guessing.
I know what E-step calibration is. Essentially it is dialing in the # steps/mm.
And yes: you can do linear calibration on the axis with that.
And no: you cannot with only a cube of 20mm. You just need more sophisticated models than just a cube.
The vision encoder calibrates only X/Y, but it measures over much longer distances (more accurate) and corrects skew also.
The vision encoder does nothing on Z
There E-steps is currently the only way to dial that in.
And yes: that can be measured on 2 printed models (one high and one low) to eliminate 1st layer deviation because in Z the only shrinkage involved is the one in the last layer.
Compensation of skew in ZX or YZ depends heavily upon the build quality of the printer and how hard it is dropped during transport.
Most of this is already has been said somewhere earlier.
Measures longer distances but not all distances. Same as the cube. It supplies a scaling factor. It could get the same result with a 30mm cube. or a 325mm cube. I had a feeling you didnt understand it. All good.
"per 20mm)
Does bambu have a calibration for z? Or requires a calilantern print and an hour of spare time? For each filament
Here.
XYZ and E-step calibration in 3D printers involves adjusting the machine’s firmware to ensure accurate movement and filament extrusion. This is done by measuring the actual movement of the axes and the amount of filament extruded and then recalculating the steps per millimeter value.
Here’s a breakdown of how it works:
- XYZ Calibration (Axes Movement):
- Purpose:
XYZ calibration ensures the printer’s axes (X, Y, and Z) move accurately and are perpendicular to each other. This is crucial for precise prints and avoiding skewed or distorted objects.
- Process:
The printer measures the actual movement of the print head along each axis and compares it to the expected movement (e.g., 100 mm). This is often done using a ruler or digital calipers.
- Adjustment:
If there’s a discrepancy, the firmware’s steps per millimeter value for that axis is adjusted to correct the error. For example, if the printer moves 101 mm when commanded to move 100 mm, the steps per millimeter value is decreased to compensate.
- E-step Calibration (Filament Extrusion):
- Purpose:
E-step calibration ensures that the printer extrudes the correct amount of filament, which directly impacts the dimensions of the printed parts.
- Process:
A known length of filament is manually measured and then extruded by the printer. The amount of filament actually extruded is then compared to the amount the printer was supposed to extrude.
- Adjustment:
The E-steps value (steps per millimeter for the extruder) is adjusted in the printer’s firmware to match the actual extrusion with the desired extrusion.
- Calculating New Steps per Millimeter:
- Formula:
The new steps per millimeter value is calculated using the formula: New Steps/mm = (Current Steps/mm * Desired Measurement) / Measured Measurement.
- Example:
If the printer is supposed to move 20 mm but only moves 19.5 mm, the new steps per millimeter value is calculated to account for the difference.
- Saving Settings:
Once the new steps per millimeter values are calculated, they are saved in the printer’s firmware or EEPROM (a special memory chip) to be used for subsequent prints.
In essence, XYZ and E-step calibration involves calibrating the printer’s movement and filament extrusion to match the printer’s commands, ensuring accurate prints.
Have you actually read what ChatGPT or Copilot is telling you here?
Who measures? The printer?
And with what? With a caliper or ruler?
Possibly with a Vision Encoder then 
Its ok man. You dont have to admit it. all good.
The plate is a great tool for new users. And in Bambus case, necessary because the firmware is locked. Its not perfect though. Just run it twice in a row and see that the results will be slightly different each time. To me, this says both results cant both be accurate.
You were so excited that you thought you had me cornered, that you didnt even finish reading the post. Im just here to give my 2 pennies. Take it or leave it.
You obviously didnt know how the estep cal worked, and nobody truly knows how the plate works. Besides Bambu. It sounded great though
I read your post and didn’t want to corner you.
And I never said that your method was wrong.
I have done E-step calibration on my self built Voron 2.4 also because there was nothing else for it. And for most cases it is still enough.
Same for skew correction. Had to print calibration models for that too.
And which method you use and how much time/money you (are willing to) spend on calibration depends on the accuracy that you require.
I rest my case.
1 Like