Did the high precision calibrations make a difference?

[NeverDie]

I just now ran through all of them, including the encoder plate and the light/dark PLA calibration (I didn’t have black and white, so I used yellow and green, which I hope is good enough–it only asked for a light-dark pairing), and the one where you remove the build plate and it scans the markers on the heated plate. After the vision plate calibration, it said the average improvement compared to no calibration was 156microns, with a maximum of 301 microns.

I did some XY measurements beforehand, and I’ll be re-running those tests afterward to see whether I notice any improvement. While I’m waiting for that to complete, I thought I’d just inquire as to whether anyone else has noted any improvements of any kind after running these calibrations. i.e. compare notes, if anyone is able or willing. At least so far, other than one skeptical comment a youtuber made, I’ve seen no commentary on this aspect of H2D.

I’m also re-runing the built-in speed benchy to see whether it has netted any improvement in the visual fidelity of that print:

[NeverDie]

Well, first result is that I noticed zero difference in the visual fidelity of the built-in speed benchy.

Currently checking whether the measured difference in X and Y shrinkage will reduce after this calibration. In theory they should be identical (or so I’m told), so any apparent measured difference should be a reflection of differences in the X and Y distances, which in a perfect world would be zero difference. Measured before this fancy pants calibration, the difference was greater than on my X1C. After running all the available built-in accuracy improvements, I’m re-printing and will see whether my Mitutoyo calipers measure any improvement in the shrinkage calibration print. For better or worse, the Bambu slicer allows only one shrinkage compensation number for both X and Y, whereas on some other slicer there can be separate shrinkage compensation numbers for X and Y individually. This makes it even more important to stamp out any measured X and Y differences, by whatever means necessary, including skew compensation, which the visual build plate encoder is alleged to achieve, according to Dr. Tao in his recent interview with CNCKitchen.

If it doesn’t squash the problem, then I guess my next move would be to run the calilantern print and use measurements from that to quantify all 3 skews.

Anyhow, my shrinkage calibration print just finished, and I do the measurements after it slowly cools down to ambient:

However, if nobody else has anything meaningful to contribute to this thread, then I’ll assume there’s no actual genuine interest. In that event, I’ll just truncate further information and not waste any more of my time posting on this topic.

[Leonilian]

I would like to test this as I’m designing a print with 0.1mm tolerances specifically for the H2D with the vision encoder plate in mind. My printer doesn’t ship out until the 25th, but I’ll do a before VE / after VE / A1 test to see how fit is for the parts when it comes in

[NoahKatz]

I think there are a lot of us who are interested in whether the encoder can provide a meaningful improvement in tolerances.

Something I’ve wondered is why didn’t bamboo just engrave the encoder pattern onto the top of the heatbed.

Or offer that as an option, which I expect would cost a fraction of a separate encoder plate.

[just4memike]

I’m thinking they have to create them to a higher standard than the basic plate. If you add it to the plate and manufacture the plate to +/- 0.1mm, the encoder is totally useless.

They have to pay more money to manufacture it to a far greater accuracy level. So, that’s why I believe the cost of the plate is so high and why it wouldn’t work well added to a $30 plate.

In most things, manufacturing to .1mm is relatively cheap, but manufacturing to .01mm is very expensive.

[NoahKatz]

@just4memike

If you add it to the plate and manufacture the plate to +/- 0.1mm, the encoder is totally useless.

The only thing that matters for the heat bed is the flatness of the top surface, which is already required to be good, and in any case flatness is Z and its X/Y accuracy that matters here.

I’m assuming that the encoder plate bar codes are laser engraved and that that could just be done on the top surface of the heat bed as well as a separate piece of sheet metal.

[NeverDie]

@NoahKatz

I think there are a lot of us who are interested in whether the encoder can provide a meaningful improvement in tolerances.

I tested the accuracy of my shrinkage compensation both before and after doing the encoder calibration. Considering X and Y separately, the shrinkage compensation got slightly worse for X and considerably better for Y. Averaging the two was nearly dead nuts.

The measurements and calculations were as prescribed by:

ShrinkingTest V4.1 by Alex_vG by Alex_vG

If I had it to do over, I would have done the before and after measurements on a calilantern, if only because it measures more things and casts a wider net. However, I had to run this test anyway because I was dialing in a new filament, so I just went with printing the model I was already using for my other purposes.

[NoahKatz]

@NeverDie

Averaging the two was nearly dead nuts.

So to be clear, the encoder plate didn’t do much for you - right?

[just4memike]

Depends on the user… to some people, those results are worth the cost and efforts, to others, it would be a waste of time and money. Perspective means everything. His numbers look scarily repeatable, even if not perfectly accurate, and that would be very sought after by some.

My worry for this is for people who use Makerworld. There will be a lot of trial and error when it comes to calibrated and uncalibrated machines printing the same thing. People complaining something doesn’t fit or is too loose. However, looking at his findings that may not be as big of a deal as I first thought.

[Deafcat22]

Guys, part shrinkage compensation is a material-specific offset. This is different than machine XY compensation with the vision encoder (though, machine XY inaccuracies will confound the test!) Also, there are more variables to contend with in the overall process of delivering repeatable, consistent, accurate results, there is no one fix that solves the entire chain of variables… It is over-simplification to believe otherwise.

In conventional manufacturing, the chain goes like so: Machine, Tooling, Material, Programming, Setup, Operator, Operations, Inspection.

The same chain of events can be generalized to 3D printing, and the Vision encoder is designed to solve some big variables at the first stage only (it’s implied you should solve them one at a time after another, rather than trying to solve all at once!). You still have to get all the following steps “right” if you want to minimize the variables and get highly accurate, repeatable results consistently. You, the operator (and programmer!) are the biggest variable by far, along with the material, but you’re in charge of compensating for that.

The shrinkage test shared earlier is intended to address that material variable, and is pretty good.

Note: I ran the Vision Encoder calibration and results were 119 average improvement, 247 max improvement, versus no calibration. Glad I got that outta there! Now I can focus on the material compensations, and really put this machine to work on the expensive jobs.

[NeverDie]

Let’s simplify this. Let’s say you set your shrinkage compensation to 100%, meaning that there effectively is no shrinkage comopensation. This is the default value by the way. Say you print a test object, whether it be a cube or whatever, where in your CAD modeling software the X (length) and Y (width) should be the same. Then you print it and measure each. Now, in theory, X should equal Y. I thought that the visual encoder was supposed to adjust for any inaccuracies so that X really would equal Y to within, say, 30 micron (the claimed accuracy), which is equal to 0.03mm. Right? Is it, or is it not doing that? I guess technically it’s only guaranteeing that the nozzle is within 0.03mm of where it is supposed to be, but that should translate into the print not differing by more than that either (other than shrinkage effects, which are alleged to be the same in X and Y).

I probably could say this more accurately, but I hope the meaning conveys. If someone wants to re-state it in a better way, go for it. I guess I’m saying that, yes there will be shirnkage effects, but if it’s true that X and Y shrink by the same amount, then the measured values of X and Y should still measure equally, to within 0.03mm or better.

[Deafcat22]

Yes, the XY compensation is separate from shrinkage compensation, though the sum of them is measurable inaccuracies on the part, and individually errors from either are measurable inaccuracies. Both have to be solved separately. Neither can be solved completely, just looking to minimize each variable when pursuing best results within reason.

[APEX86]

This really sums it up well.

@NeverDie if a part, including the path planning, is symmetric about X and Y, the shrinkage should also be equal. It’s still possible that due to asymmetries in cooling, it may not be. Once the part in question has big differences along X and Y the shrinkage for each will likely start to deviate. Even the same part, printed with different infill pattern/density, wall number, etc; will probably shrink differently.

But, correcting the underlying machine motion first is always going to be the best approach.

[NeverDie]

@APEX86 OK, thanks, that gives me an idea then. I need to look into it, but there may be an asymmetry in the infill pattern I’m using on the webbing. If I switch that infill from rectilinear to concentric, then I expect that should make it the 100% symmetric.

I retained the beveling of the original makerworld model, which was a brilliant idea by the original author, but I added half circles directly into the print as further prevention against possible warping in the final print, on the theory that they approximate mouse ears. This may or may not be overkill, or even the most desirable geometry, but I don’t mind spending a little extra plastic if it might help. After cooling, if I transfer the printed model onto a piece of float glass (such as the top of the lid of the H2D), it doesn’t appear to teeter totter at all, or bend, no matter where I press on it. At least none that I can feel by hand. By that I mean, I don’t sense any detectable warping in the final print. It seems to be as perfectly flat, as near as I can tell using that method.

Also, I’ll try rotating it 90 degrees and print another, and compare. In theory it would then have no difference, but if it does, then maybe I made a mistake somewhere in my fork of the model.

Well, that’s on me to fix and try again. I’ll do it as time allows. These are good things to print whenever a printer might otherwise sit unused for a stretch of time.

But it won’t be that useful to others, so meanwhile I’ll try printing a calilatern. I presently have v2, but there’s a v3 out, which I’m entitled to as a free upgrade, so I’ll request that and print v3. I did this in the past for v1 on my X1C, and one other person posted their numbers for their X1C, and our numbers were quite similar. If anyone else here would care to print v3 on their H2D, we could compare our H2D numbers. Unfortunately, unless I were to do a full factory reset (which I’d rather not), I can no longer do a before and after vision encoder print of the calilantern. Only an after. However, if someone new, who hasn’t yet run their vision encoder, wanted to print a calilantern v3 (or whatever similar thing you normally do), it would be interesting to know how before and after compare.

By the way, when I import my calibration object from a .stp file, I’m using the maximum triangles the Bambu slicer allows:

That results in the following comment by the slicer:

However, I decline its invitation to simplify the model, as I’m not sure what that does. I’m assuming that “more triangles = more better”, but maybe there’s a point at which more triangles actually becomes worse? Perhaps the motion isn’t as smooth? No idea, but just giving full disclosure in case anyone has an opinion on that.

I do allow “arc fitting” in the slicer, which in general does cut down on print time–sometimes by a lot–but I’ve never investigated whether it introduces inaccuracy or not:

Given the current model and slicer settings, including 8 walls, and concentric top, bottom, and solid infil settings, it takes 1 hour 10 minutes to print, which is fine by me. Using 8 walls means there is no sparse infill, so no assymetry introduced by that. If I were to turn off arc fitting, then on this print it would take 1 hour 11 minutes, which is a minor difference, so I’ll try my next print with it off as a “just in case” pre-caution.

So, given all of that, if you look at how it prints, long uninterrupted lines go away, and so that source of warping (AFAIK, the main source of any warping) is eliminated:

This is the view from the bottom. Thus the advantage of this particular geometry, at least from my POV. I’d be curious whether anyone else agrees or disagrees with that.

[NeverDie]

Well, not to beat a dead horse, but just re-interpreting my original measurements, the result is exactly what one might hope for in the utility of the vision encoder plate:

Explanation: the original author’s method is similar to the most accurate method of measuring the distance between two pins. For those not familiar with that method, I’ll let chatgpt explain, and then I’ll continue afterward:

chatGPT:

Me again:
So, how does that relate to the original author’s model and method of measuring? You can think of it like measuring the distance between 2 pair of 2 pairs of pins, where the centers of the basic pin centers are spaced exactly 145mm apart (i.e. what it would be if you measured it in the original CAD model). The original author provides two such pairs for measurement in X, and the same for Y, presumably on the theory that averaging will help compensate for inevitable measurement error.

So, when looked at that way, only the original Y measurement (prior to the vision encoder plate calibration) was out by more than 0.03mm. After the vision encoder plate calibration, though, both X and Y are within that 0.03mm tolerance that one would expect.

Looked at this way, all is good! The vision encoder plate, at least in this instance, achieves what it claims.

Granted, this is only one datapoint, but if the rest of you run the same or similar kind of measurement, and post your results, as I have, then by pooling our datapoints we can have more or less confidence, depending on how your measurements turn out on your respective machines. So, I would encourage anyone who is not a hopeless sponge to do so. The effort required is quite minimal. Just download the original authors model (I linked to it earlier in this thread), and post your measurements. The measuring process is easily less than 2 minutes tops.

Or, if you prefer to print a more thorough calibration model that casts a wider net, even better. More power to you.

P.S. I should perhaps add that the two models I printed were both the same third iteration print, where I had incrementally improved the shrinkage compensation factor for Sunlu HS-PLA that I used in the print to 99.7932% i.e. I used the same shrinkage compensation factor for both the “before” and “after” prints. As you would expect, not much shrinkage with PLA, which is partly why it is so popular. The net effect is that whatever I now print using that shrinkage compensation factor for that filament on the H2D (assuming the filament is equally dry) will be dimensionally accurate in X and Y. That’s also why I keep the humidity in my AMS2 very low: 0% as reported by the AMS2

though more likely 7%RH, as reported by the Yolink and Switchbot RH sensors that I sealed in the AMS2 as a check on the accuracy of the AMS2’s reported humidity level. Most likely Bambu used a cheaper, less accurate humidity sensor for the AMS2, maybe one that doesn’t measure at all below 10%, but that’s a different topic for another day on some other thread.

[Alex_vG]

I think the vision encoder plate would also be perfectly suited to correct skew.

I will stick to the X1C but I’m curious. Has anyone tested skew before and after calibration?

[APEX86]

Yes, no doubt. Fine tuning skew with the encoder plate should be effective.

But, as a mechanical engineer, I would always recommend correcting skew mechanically to a close to true as possible, first.

[Tev]

I also bought the precision plate. But in all the videos covering the topic, people don’t seem to notice any improvement in print quality. I think they didn’t push the calibration as far as you did.

So my plate is still packed in its box, and I’m hesitating—maybe I should return it and go for an AMS PRO instead ^^

Anyway, hats off to you for the level of detail in your tests!

[APEX86]

“Print quality” is really about a lack of observable defects. The calibration plate will do nothing for that.

[NeverDie]

It’s for when you want tight fighting parts to fit together precisely, possibly even more so if the parts were printed on different machines. If you’re just printing small one-piece whimsical figurines, you most likely don’t need it. If you’re printing assemblies, or something that requires precision, like gears, then maybe you do need it.

I think it’s cool that we’re seeing more and more mechanical tools, contraptions, and jigs that are downloadable. If that’s your thing, maybe you’ll be glad you have one. And hope that the author has one too, or something similar, or else it might only fit if it’s printed on his machine because his CAD dimensions don’t align with his printed dimensions.

A bit of trivia: did you know that without gauge blocks from Sweden, the US couldn’t have made all the stuff it did in different factories during WW2 and have them actually fit together? Kinda like that.