Terrible tolerances/dimensional accuracy on stock profiles H2D

Bambu Lab H2 Series
1

Hi everyone,

We are currently using the H2D printer calibrated with the Vision Encoder in conjunction with Bambu filaments. However, we are experiencing significant issues with tolerances on the stock profiles. For example, when printing a part that should measure 150mm, we’re seeing inaccuracies of over a millimeter, resulting in parts that measure only 149mm. For instance, when using Bambu ABS, we’ve found that a shrinkage value of 99 is necessary. This means that, with the stock profiles, a part intended to be 300x300mm actually prints at 297x297mm, resulting in a tolerance discrepancy of 3mm. Additionally, the holes we print are consistently undersized.

Despite using properly dried Bambu filaments and Bambu Studio latest version and filament profiles, we find ourselves having to manually calibrate each filament. After calibration, we can achieve tolerances of ±0.05mm, which is excellent. However, the time and cost associated with calibrating each filament and nozzle size adds up quickly. Our primary focus during calibration is on filament shrinkage and the auto circle contour hole offset (which needs to be set to -0.15 i.e. for ABS).

Has anyone else encountered similar issues? I must admit I’m quite disappointed. I had hoped that using Bambu filaments (including PLA, ABS, PA6-GF, and PETG) would minimize the need for extensive calibration. From now one we are switching to cheaper alternatives since we need to do the work anyway.

1 Like
Shrinkage not compatible across printers? (weird results in this thread...)
Hole with triagular structure inside - dental model printing with p2s
2

The profiles from Bambu are primarily intended to adjust the print settings for the filament.

The fact that filament shrinks after cooling is nothing new and depends on the material and its composition – for ABS, it’s said to be up to 1.5%.

Therefore, I don’t see the problem with the printer but with the filament, and if you want top-notch results, you’ll probably have to calibrate each filament and conduct appropriate tests.

EDIT: I just looked through the ABS profiles of several filaments I use, and they all say shrinkage = 100%.

1 Like
3

Thank you for your response.

Shouldn’t the shrinkage of the filament be calibrated by Bambu? What’s the point of using Bambu filament if it isn’t calibrated in Bambu Studio?

Moreover, a dimensional inaccuracy of 1% is significant. While I don’t expect perfect, ultra-tight tolerances straight out of the box, I certainly anticipate better than 1%. I mean, it’s marketed as a prosumer device.

1 Like
4

Shrinking due to cooling down really isn’t a question of quality, but an inherent physical property of all materials, not only in 3D printing and not only plastics but almost all materials.

I too can’t understand, why BambuLab don’t measure shrinking and integrate the correction in their profiles. Even if not perfect it would enhance dimensional accuracy tremendously.

On the other hand, the calibration should not be too much effort. I created a dedicated model, that typically gets you very close in a single iteration and prints quickly:

Maybe give it a try?

On the circular correction: I’m very sceptical about that feature. In my experience, a single value for adjusting hole size is not sufficient. Small holes shrink far more than bigger holes. So if you calibrate with 3mm holes, smaller holes probably still are too small and suddenly larger holes will be too large / over compensated.

Even worse, the compensation only applies to closed holes. E.g. when you cut a slit it suddenly is not compensated anymore even though it shrinks just the same as a closed hole. And if you have a square hole, it also gets compensated although it almost doesn’t shrink on top if regular shrinking. So I would always fear that when I compensate on one feature, I will throw other features out of accuracy.

I performed a small test series with different hole sizes and different materials and found a very linear correlation between curvature and wall offset. In my opinion, setting back any wall depending on the curvature would be the right way to correct this problem. This would also work on any shape of holes, open and closed holes and even outside concave walls.

Unfortunately I’m not aware of a slicer that has incorporated such a compensation yet. So for critical holes, I make test prints and compensate holes in CAD.

Admittedly, I haven’t tried the latest smart hole compensation in BambuStudio that is available for a few dedicated materials. But I’m not holding my breath.

1 Like
5

Because shrinkage is not as predictable as people might think. And it highly depends on the actual geometry of the model.

Commercial softwares charge much more than our printers to calculate and simulate the shrinkage, just so you understand why it’s hard.

Cheapest and easiest way to resist shrinkage is just to use fibre reinforced filament.

4 Likes
6

While I acknowledge that shrinkage is a complex issue, I believe that Bambu filament should be calibrated to the extent that a printed cube measuring 100x100x100mm should not end up as 99x99x99mm.

When a company markets its own filament specifically for its printers, touting advanced features like a vision encoder for the H2D, one would reasonably expect accurate prints using that filament without extensive calibration. In a professional environment, the costs associated with calibration can quickly exceed the price of the material itself. This is precisely why we opted for Bambu filament; we anticipated needing less calibration.

While I don’t expect out-of-the-box tolerances of 0.05mm, I also don’t think it’s unreasonable to expect that prints should not deviate by several tenths of a millimeter with stock profiles.

Overall, we are quite satisfied with the printer itself. However, in a professional setting, investing in the entire Bambu ecosystem and purchasing their filament may not be worthwhile. It may be more cost-effective to choose higher-quality filament from other manufacturers rather than pay a premium for Bambu spools.

After thorough calibration, we have achieved an impressive accuracy of ±0.05, which is remarkable for a 3D printer in this price range. While we have no complaints regarding this performance, it is crucial for users to have a clear understanding of this capability.
Despite the marketing claims, users should not expect to achieve such high levels of accuracy right out of the box. I am somewhat disappointed that Bambu has not made more efforts to optimize their stock profiles; there is certainly significant room for improvement.

2 Likes
7

The printer is not a professional printer, and also not marketed as such.

If you buy a professional printer of €10K plus, you can expect a bit more.

2 Likes
8

You wrote that in the other thread too and I still don’t agree :squinting_face_with_tongue: In my view shrinkage itself ( in FDM!) is very consistent. It can have secondary effects like warping, holes shrinking extra due to rubber band effect and more, but parts becoming smaller in X / Y due to shrinking has been very consistent for me.

When you look at injection molding, that is an entirely different story. There I fully agree. the effects of material shrinking has really mean effects on the entire geometry and adjusting the mold to compensate to me seems seems like a mixture of art and science, needing complicated simulations, adjustments to process parameters and several iterations to get it right.

In FDM printing, you can easily improve overall accuracy by an order of magnitude with the simple compensation values available in BambuStudio and other slicers. Why BBL don’t take advantage of that is beyond my understanding. Even if they implement values in the rough ballpark, it would greatly improve the results for 99% of users. So even if results will not be perfect, why not vastly improve them for everyone?

9

However, in injection molding, the shrinkage rate is factored in and thus taken into account from the outset.

10

image
image1920×669 148 KB

This certainly sounds like it is marketed as a professional printer.

If Bambu promotes their products with the promise of injection-molded accuracy, they should also ensure that their filament profiles are calibrated accordingly. The H2D is indeed capable of achieving exceptional precision, but it’s frustrating that their factory profiles are not optimized for this level of performance. It wouldn’t require much effort on their part to calibrate for shrinkage and enhance overall accuracy.

1 Like
11

How is that? In my understanding, you start with the model that you expect as end result. Of course you have to respect design rules like slanted walls, avoid material accumulation etc. but the dimensions match the expected final outcome.

Then the mold maker creates an initial mold design and simulates the process. From that, the geometry and the process parameters are adjusted until the outcome matches the original design. Then a first test batch is being produced. Usually you still need some further adjustments to process parameters and maybe even rework the mold until you receive parts with close tolerances.

I would say, the shrinkage compensation in the slicer is much more from the outset than in injection molding.

12

Just quoting this so I can give it a try later. Thanks for this upload!

13

Have you seen some Chinese molds?

But you are not comparing apples with apples here.

The vision encoder is to calibrate the XY motion system, not the filament profiles.

14

Could you clarify what you mean by “Chinese molds”? The Chinese are widely recognized as leaders in mold making and injection molding today, so I’m not entirely sure I understand your point.

Regarding the H2D, along with the vision encoder, it is indeed marketed as a semi-professional machine. I can personally vouch for its quality; we have one at our company, and I also own one privately and overall we happy with it. However, it’s disappointing that Bambu doesn’t calibrate their filaments. If they did, we would continue purchasing Bambu filaments in the future, as it would save us a lot of time and money. As it stands, the only advantage of Bambu filaments is the convenience of the RFID chip, which, honestly, doesn’t justify the higher price.

15

Maybe I should have accentuated “some” a bit more.

I know Chinese molds are fine quality in general. Therefore I used the word “some”, because I have seen poor quality products also.

Nevertheless, the vision encoder improves the positioning accuracy of the motion system, but doesn’t make it a professional 3D printer.

And yes, you are totally free in choosing any other brand filament besides Bambu, but you might be spending more time to tune in all those other profile settings, besides the shrinkage percentage to get a good result. And if you make a lot with the same filament, it might be worthwhile to invest the time in it.

But be aware that nozzle size, layer thickness, nozzle temperature, layer time, part shape and many more factors also influence shrinkage. So one uniform shrinkage factor that solves everything is an utopia.

1 Like
16

I would agree if I was printing only simple geometric shapes with 100 % infill but there are far too many variables (shape, infill type and amount, etc.) that influence shrinking that I consider it close to impossible to provide a perfect “one fits all” shrinking compensation factor.

Before FEM calculation is part of the slicer (which I don’t expect in the next 10 years or so) I doubt that there can be a fixed shrinking compensation that works in all situations.

1 Like
17

In the uk if you buy 10 reels of pla/petg/abs it works out to be £10.25 a reel which is pretty cheap I think. I personally don’t bother and get all mine from aliexpress for even cheaper.

18

If the shrinkage number for a specific plastic is 1%, and you had a printer with a 3000 mm bed, would you expect to see a 3000mm line come up at 2970mm? Of course not- the extruder moved 3000 mm. There’s going to be plastic the full length of that line, and it doesn’t all cool at the same time, it starts cooling the second it leaves the extruder nozzle. The shrinkage numbers quoted for plastics are bulk numbers that apply to injection molding, not to the FDM print process.

The steps/mm of the mechanism is determined exactly by the steps/rev of the motors and the tooth count of the drive pulleys, and belt pitch. Size differences in prints mostly come down to variations in filament diameter and geometry. Back in the good ol’ days when I was building my own printers, I had to measure the diameter at several places along a length of filament, calculate an average, and enter the diameter into the slicer. Variations were compensated for by using offsets in the design, typically +0.2 to 0.3mm diameter for holes. That sort of offset would allow tight press-fit of bearings, etc. Outside dimensions of objects were typically accurate within 0.05-0.1mm regardless of the material printed, so no compensation was used.

These days, the diameter of even the cheapest, no-name brands are 0.02 or 0.03mm tolerance, making it unnecessary to measure the filament. I don’t think there’s going to be a lot of difference between the shrinkage of one brand of PETG or another, but maybe some differences between PETG and PETG-CF, or whatever other fiber might be in the stuff. It’s fairly quick to print a test block, measure it, and adjust the compensation in BS. I don’t think that’s asking too much, but then I am a hobbyist (no, not a cosplay/starwars toy/tugboat printing hobbyist, I print machine parts, etc.), not running a print farm and trying to do mass production. I would print and measure one block for each different type of filament, not each brand, or each spool. If you need higher accuracy than that yields, maybe you should be using a different fabrication process.

For small holes that need to be exact sizes I usually print them undersized and run a drill through them after the print finishes.

Until I saw this thread I hadn’t thought to check the XY calibration of my H2C, so I’m printing a 100mm square PETG block with round holes in it now. Then I’ll print the same block with ABS and compare the two.

2 Likes
19

Of course you will never get perfect dimensions, but when using XY shrinkage compensation ( don’t use datasheet values but determine shrinking with an actually printed test part), dimensional accuracy for me improved a lot. Bigger parts were off in the ball park of millimeters but with compensation, the same parts are now within 0,2 .. 0,3 mm to the design. And that is very consistent. Even when using the same material from different manufacturers, I don’t bother measuring each individually anymore , because shrinkage figures are so close, that final dimensions are plenty good enough.

I don’t see any disadvantage of applying a general compensation factor for each material. So why not make everyone but the most critical users happy?

Do you have any evidence that shrinking in FDM is so highly depending on geometry, infill etc.? I didn’t observe that at all. Everything just got better with shrinkage compensation.

It doesn’t solve all issues, especially size of small holes, but why don’t you seize the benefits it provides - in my opinion without drawbacks?

And on the mechanisms at work: I still claim, that geometry has much more influence on dimensional accuracy in injection molding than in FDM printing. My theory is, that FDM cools down very locally, so that global geometry has far less influence, while injection molded parts cool globally at once, so that shrinking in one area much more affects the stresses in other areas. That in turn causes much more deformation depending on geometry. But that is just my theory without evidence.

1 Like
20

FDM printing has inherent limitations on accuracy. As discussed at length, plastic changes size with temperature, and in directions influenced by the specific shape of the model.

I have made a lot of functional parts over the years (started printing in 2010). I basically do what metal printing does. For parts that need to have tight tolerances, I print the surfaces I need to be accurate a little bit over-size on purpose, and then I take the surface down to the tolerance I want with my Taig CNC benchtop mill.

ABS is one of the worst plastics for this. It has a very high CoTE compared to other plastics like PETG. It’s one of the reasons ABS is so hard to print without warping. Even a small reduction in temperature can result in high internal stresses that deform the print.

3 Likes