Bambu X1C vs H2D dimensional accuracy: measured engineering test

1. What was tested - and what was not

Dimensional accuracy in FDM is often discussed as though it were a property of the machine. In practice it is the output of a chain: CAD geometry, slicer path generation, nominal nozzle size, selected line width, extrusion calibration, pressure control, kinematics, cooling, thermal contraction, local feature geometry, part orientation and finally the measurement method. A printer can therefore be very accurate on one class of geometry and noticeably biased on another.

This study was built around that distinction. Instead of printing a conventional calibration cube and reducing the result to X, Y and Z error, three copies of the D&L PQ-01 v1.2 dimensional qualification coupon were produced and measured. The coupon deliberately places internal holes, rectangular apertures, external cylindrical pins, thin walls, reference prisms, Z-height pads and two larger round-profile features on the same part. The intention is to expose process behaviour by feature class.

The principal like-for-like comparison is the Bambu Lab X1C with a 0.4 mm nozzle against the Bambu Lab H2D with a 0.4 mm nozzle, both at a recorded 0.20 mm layer height. A third coupon was printed on the H2D with a 0.2 mm nozzle, again at 0.20 mm layer height, to show how the H2D process changed when the nozzle configuration changed while the recorded layer height remained constant.

This is not a formal machine-capability study. There is one coupon per configuration, two caliper reseatings for the recorded dimensions, and no claim of traceable dimensional uncertainty or statistical process capability. The value of the test is comparative and diagnostic: it shows what happened on these three controlled coupons, and it shows why a single cube dimension can be misleading.

Printer platform context

Bambu Lab specifies the X1-Carbon with a 256 × 256 × 256 mm build volume and a 0.4 mm hardened-steel nozzle as standard, with 0.2, 0.6 and 0.8 mm options. The H2D is a newer, larger dual-nozzle platform; Bambu lists a 0.4 mm included nozzle and support for 0.2, 0.4, 0.6 and 0.8 mm diameters. Those manufacturer specifications provide context only. None of the dimensional results below are manufacturer figures - they are D&L measurements from the PQ-01 coupons.

2. Test method and measurement model

PQ-01 v1.2 geometry

How error is reported

For each row in the measurement workbook, the mean of the repeated readings is used as the measured dimension. Signed error is:

A negative value therefore indicates an undersize feature; a positive value indicates oversize. Feature-group mean absolute error (MAE) is:

MAE is useful for comparing the magnitude of error without positive and negative values cancelling. It should not be confused with measurement uncertainty, capability indices such as Cpk, or the mean dimensional error of an entire printer. Each group contains a different number and type of measurements.

Small holes deserve an additional caution. PQ-01 retains Ø2-Ø3.5 mm holes because they are useful practical features, but ordinary digital-caliper internal jaws are not ideal bore metrology at those diameters. For that reason the article reports the complete hole bank and also a secondary statistic for holes ≥4 mm, where the caliper method is less compromised by jaw-tip geometry.

3. Summary results

The table is enough to reject the idea of a single “most accurate” machine. The H2D 0.4 mm configuration is the best of the three on the reference block/bar and Z pads, yet it is the worst on holes, external pins, the R20 boss and the C12 profile-cylinder group. The X1C 0.4 mm is stronger on the external pin bank and substantially stronger on holes than the H2D 0.4 mm. The H2D 0.2 mm configuration is exceptional on thin walls and brings hole error down to the same practical region as the X1C for holes ≥4 mm.

4. Overall X/Y scale and reference geometry

The overall coupon envelope was slightly undersize in X and Y on all three prints. The X1C measured 139.755 mm across a nominal 140 mm X dimension and 89.875 mm across nominal 90 mm Y. The H2D 0.4 mm measured 139.710 × 89.695 mm; the H2D 0.2 mm measured 139.730 × 89.790 mm. Taken alone, those values might suggest that a small global XY scale correction would improve the parts.

The reference geometry shows why that conclusion would be too crude.

On the H2D 0.4 mm coupon, the 20 mm reference block was measured at 20.000 mm in X, 20.000 mm in Y and 10.000 mm in Z. The 40 mm reference bar was 39.965 mm. Its reference-group MAE was only 0.009 mm. In other words, the same print that produced near-nominal orthogonal reference geometry produced circular holes that were hundreds of microns undersize.

The X1C reference group was less exact than the H2D 0.4 mm group - 0.078 mm MAE - but still close enough to show that its better hole and pin behaviour did not come from an obviously superior global scale calibration. The H2D 0.2 mm reference group landed between the two at 0.044 mm MAE.

5. Circular-hole behaviour

This was the clearest separation between the 0.4 mm configurations. Every diameter in the H2D 0.4 mm hole bank was undersize, and the bias remained large even after the sub-4 mm holes were excluded from the statistic.

Across the full ten-size bank, hole MAE was 0.160 mm on the X1C 0.4 mm, 0.430 mm on the H2D 0.4 mm and 0.142 mm on the H2D 0.2 mm. Because every group-level signed hole error was negative, MAE and the magnitude of mean signed error are the same for this dataset.

The larger-hole subset is especially useful. For holes from 4 to 10 mm, where standard caliper internal jaws are a more credible method, MAE was 0.114 mm for the X1C 0.4 mm, 0.394 mm for the H2D 0.4 mm and 0.115 mm for the H2D 0.2 mm.

That means the H2D nozzle-configuration change reduced full-bank hole MAE by 66.9%. On the ≥4 mm subset the reduction was 70.8%, bringing the H2D 0.2 mm result to within about 0.001 mm of the X1C 0.4 mm group MAE. This should not be read as sub-micron equivalence - the measurement method does not justify that level of precision - but it does show that the broad hole bias seen with the H2D 0.4 mm configuration was largely absent with the 0.2 mm configuration.

The Ø5 mm feature makes the difference easy to see. Its mean X/Y diameter was 4.885 mm on the X1C, 4.573 mm on the H2D 0.4 mm and 4.863 mm on the H2D 0.2 mm. The H2D 0.4 mm print therefore put the same nominal bore roughly 0.43 mm undersize while its 20 mm reference block was effectively nominal.

X/Y anisotropy in the holes

The mean absolute difference between X- and Y-oriented hole measurements was 0.103 mm for the X1C 0.4 mm, 0.066 mm for the H2D 0.4 mm and 0.031 mm for the H2D 0.2 mm. The 0.2 mm H2D configuration therefore produced not only smaller hole bias but also the lowest X/Y difference across the bank in this test.

It would be tempting to attribute the H2D 0.4 mm hole result to one mechanism - flow calibration, arc discretisation, extrusion pressure, filament shrinkage or a slicer hole-compensation policy. The coupon does not identify root cause. What it does establish is the pattern: the error is geometry-specific, systematically negative for circular internal features, and strongly altered by the nozzle configuration.

6. Rectangular apertures

Rectangular internal geometry behaved very differently from the circular holes. Aperture MAE was 0.038 mm for the X1C, 0.025 mm for the H2D 0.4 mm and 0.033 mm for the H2D 0.2 mm. All three are small compared with the H2D 0.4 mm circular-hole bias.

This contrast is one of the most useful outcomes from PQ-01. The H2D 0.4 mm configuration can produce a 10 mm square aperture and an 8 × 14 mm rectangular aperture close to nominal while producing a Ø5 mm circular hole more than four tenths undersize. The process therefore cannot be characterised adequately by saying “internal dimensions are undersize”. Geometry and toolpath type matter.

7. External pins and the R20 round boss

The pin bank produced another non-intuitive result. The X1C 0.4 mm configuration had the lowest pin MAE at 0.078 mm. The H2D 0.4 mm and H2D 0.2 mm results were 0.166 mm and 0.154 mm respectively. Moving to the 0.2 mm nozzle configuration improved the H2D pin MAE only modestly, even though it transformed the hole and thin-wall results.

The H2D 0.4 mm pins show a broadly increasing undersize tendency at the larger diameters: P6 averaged 5.783 mm, P8 7.780 mm and P10 9.718 mm. The X1C equivalents were 5.975, 7.880 and 9.878 mm. The H2D 0.2 mm configuration did not collapse this difference in the way it did for the hole bank.

20 mm round boss

R20 reinforces the external-round-feature pattern. Mean absolute error across X and Y was 0.197 mm on the X1C, 0.277 mm on the H2D 0.4 mm and 0.150 mm on the H2D 0.2 mm. X/Y ovality - the absolute difference between the measured X and Y diameters at the controlled mid-height plane - was 0.075, 0.135 and 0.060 mm respectively.

That matters because it argues against treating circular internal and circular external dimensions as one correction problem. The H2D 0.2 mm configuration largely corrected the hole-bank behaviour, but its pin bank remained materially undersize. Any later compensation study should therefore keep internal and external round-feature correction separate.

8. Height-dependent cylinder profile: the C12 result

PQ-01 v1.2 added the C12 profile cylinder because the earlier v1.1 work exposed a practical measurement problem: the apparent diameter of an external boss changed depending on where the caliper was placed vertically. Rather than trying to hide that behaviour by prescribing one arbitrary measurement height, v1.2 measures it explicitly.

The X1C C12 mean diameter was 11.873 mm at the lower plane, 11.873 mm at mid-height and 11.920 mm at the upper plane, giving a lower-to-maximum range of 0.047 mm. The H2D 0.4 mm cylinder changed much more: 11.723 mm lower, 11.833 mm mid and 11.868 mm upper, a 0.145 mm range. The H2D 0.2 mm result was 11.858, 11.858 and 11.920 mm, a 0.063 mm range.

The H2D 0.4 mm result is particularly useful because the diameter increases as the measurement plane moves away from the base. The coupon does not prove why. Near-bed thermal history, first-layer/early-layer process conditions, local cooling, contour strategy and extrusion dynamics are all plausible contributors. What can be said from the measurement alone is that the cylinder was not dimensionally uniform through its height, and the effect was appreciably larger on the H2D 0.4 mm coupon than on the other two.

9. Thin-wall behaviour and the 0.2 mm nozzle configuration

Thin walls produced the strongest benefit from the H2D 0.2 mm nozzle configuration. Wall-group MAE fell from 0.088 mm with the H2D 0.4 mm configuration to 0.004 mm with the 0.2 mm configuration - a reduction of approximately 95.5% in this dataset.

The individual results are more informative than the average. The nominal 0.4 mm wall measured 0.690 mm on the X1C 0.4 mm and 0.665 mm on the H2D 0.4 mm, but 0.395 mm on the H2D 0.2 mm. The nominal 0.6 mm wall measured 0.800, 0.760 and 0.600 mm respectively. From 0.8 to 1.2 mm, the H2D 0.2 mm walls were essentially nominal within the resolution of the recorded measurements.

This is not simply a statement about “printer accuracy”. A wall close to nozzle/line-width scale is a toolpath-generation problem as much as a machine-positioning problem. With a 0.4 mm nozzle, a nominal 0.4 mm wall may be represented by a single extrusion line whose actual selected line width is larger than the nozzle or larger than the CAD wall. The 0.2 mm configuration gives the slicer a different set of representable line widths and path choices. The result seen here is therefore best described as feature-resolution behaviour of the nozzle/profile combination.

That distinction also explains why the 0.2 mm configuration can be almost exact on walls while remaining undersize on the external pin bank. Finer representable wall geometry does not automatically remove every contour bias.

10. Z-height behaviour

The v1.2 edge-pad measurement method is one of the quiet successes of the test. Earlier coupon work produced implausible intermediate Z readings because the measurement geometry was ambiguous. In v1.2, total thickness is measured from the coupon underside to each edge-pad top and the measured base thickness is subtracted.

Using that method, Z-edge MAE was 0.031 mm for the X1C 0.4 mm, 0.006 mm for the H2D 0.4 mm and 0.011 mm for the H2D 0.2 mm. The H2D 0.4 mm Z pads were particularly close to nominal: 0.990, 1.985, 4.000, 6.000 and 10.005 mm after base-thickness subtraction.

These values are consistent with a measurement method that is now doing what it was intended to do. They also provide another reason not to use the H2D 0.4 mm hole result as evidence of a generally inaccurate machine. The same coupon shows very small Z error and near-perfect orthogonal reference geometry.

11. Measurement repeatability and what the duplicate readings tell us

Each recorded dimension was measured twice with the caliper reseated between readings. The mean absolute difference between Reading 1 and Reading 2 was 0.012 mm for the X1C coupon, 0.009 mm for the H2D 0.4 mm coupon and 0.012 mm for the H2D 0.2 mm coupon. Maximum observed pair differences were 0.090, 0.080 and 0.070 mm respectively.

Those figures are not a gauge R&R study and should not be converted into a formal measurement uncertainty. They are nevertheless useful as a sanity check. A systematic H2D 0.4 mm hole bias of roughly 0.35-0.57 mm across many diameters is far larger than the typical reseating difference seen in the workbook. The main hole conclusion therefore cannot reasonably be explained by one poor caliper placement.

The larger repeat differences occurred on features that are inherently less convenient to access with standard calipers - for example the long rectangular aperture or an overall/base-thickness measurement. That is exactly why PQ-01 v1.2 increased spacing around the external bosses and defined measurement planes for round features.

12. Engineering interpretation: what the pattern suggests

12.1 “Accuracy” is not one axis scale factor

The strongest conclusion is structural rather than brand-specific. A single dimensional scale correction cannot explain this dataset. On the H2D 0.4 mm coupon, a 20 mm square reference block is effectively nominal, rectangular apertures are close to nominal, Z pads are close to nominal, yet circular holes are systematically undersize and external cylinders are also undersize. The error changes with feature type.

For engineering parts this matters more than a cube score. A housing can have correct outside dimensions while screw clearance holes are tight. A locating pin can be undersize even though a rectangular connector opening is correct. A thin rib can be overbuilt with one nozzle profile and nearly nominal with another. Qualification therefore needs geometry that resembles the functional features in the actual product.

12.2 The 0.2 mm H2D result is a nozzle-configuration result, not a pure nozzle-diameter constant

The recorded layer height, material and temperatures are the same between the two H2D coupons, and the nozzle diameter changes from 0.4 to 0.2 mm. In a real slicer, however, changing nozzle configuration also changes the set of viable line widths and can change toolpath construction or profile defaults. The improvement should therefore be attributed to the 0.2 mm nozzle configuration, not to a simplistic claim that halving nozzle diameter mechanically quarters dimensional error.

Within that constraint the result is still striking. Hole MAE fell by 66.9%, and thin-wall MAE fell by 95.5%. The same configuration change had only a small effect on external-pin MAE. The feature class again determines the benefit.

12.3 The hole result is suitable for compensation work - but not for a universal correction value

Because the H2D 0.4 mm holes are consistently undersize, the dataset is a sensible starting point for a second-stage compensation experiment. It would be poor practice, however, to take the mean -0.430 mm signed error and blindly add 0.430 mm to every CAD hole. The error is diameter-dependent, the smallest holes are measurement-limited, and slicer compensation operates on path geometry rather than on a simple measured-diameter offset in every case.

A better follow-up would keep the CAD coupon neutral and apply controlled slicer hole compensation to a second print, then compare the residual error curve across diameter. That preserves PQ-01 as a qualification artefact rather than baking one printer’s correction into the model.

12.4 External round features need their own treatment

The external pin and R20 data show why internal-hole compensation cannot be assumed to fix external cylinders. The H2D 0.2 mm configuration largely closes the gap on holes but leaves the external pins with roughly 0.154 mm MAE. The C12 cylinder also remains slightly more biased than the X1C result. Internal and external contours should therefore be characterised separately whenever fit depends on both.

12.5 Measurement height can be part of the process signature

The C12 data demonstrates that “the diameter” of a printed cylinder may not be one number. For the H2D 0.4 mm print, mean diameter changes by 0.145 mm between the lowest and highest prescribed planes. That variation is larger than the mean duplicate-reading difference by more than an order of magnitude. It is therefore not just a caliper-reseating effect in this dataset.

13. Limitations and boundaries of the study

• One coupon per configuration. The experiment does not quantify print-to-print variance, machine-to-machine variance, long-term drift or process capability.
• Two measurement reseatings. Duplicate readings improve confidence in gross effects but do not constitute a formal gauge repeatability and reproducibility study.
• Caliper bore measurement. Ø2-Ø3.5 mm holes are retained as practical indicators, but gauge pins or a bore measurement system would be preferable for metrological claims.
• Profile identification is incomplete. The workbook records “Bambu” rather than the exact Bambu Studio profile name and detailed line-width/flow settings.
• Filament traceability is limited. The recorded material is PLA, but the workbook does not preserve manufacturer, colour, batch, moisture condition or spool history.
• No environmental log. Ambient temperature, chamber temperature and humidity were not recorded as experimental variables.
• No root-cause instrumentation. G-code/path analysis, extrusion-flow logging and thermal imaging were outside the scope of this first study. Mechanistic explanations are therefore hypotheses, not measured causes.
• No formal uncertainty budget. Values are reported to the precision recorded in the workbook, but the article does not claim traceable accuracy at 0.01 mm.

These limitations are intentional rather than hidden. PQ-01 is aimed first at practical engineering process characterisation. A later capability study would require replicated coupons, controlled filament conditioning, a defined measurement system and statistical treatment of within-run and between-run variation.

14. Practical implications for engineering users

If a printed component contains only generous external geometry, a calibration cube can be reassuring. If the part contains close screw clearances, bearing seats, dowel features, connector openings, thin ribs or mating cylinders, it is not enough. The measured results here suggest a more useful qualification sequence:

1. Establish broad X/Y/Z scale using orthogonal reference features.
2. Measure internal circles separately from rectangular apertures.
3. Measure external round features separately from internal holes.
4. Include feature sizes around the expected line-width/nozzle-resolution threshold.
5. Control the height at which external cylinders are measured, or deliberately characterise diameter versus height.
6. Apply compensation only after the uncorrected process signature is understood.
7. Re-run the coupon after changing nozzle, filament family, slicer profile or compensation settings.

That workflow is more work than printing a cube, but it answers the questions that matter when CAD geometry has to assemble with real components.

15. Conclusion

The most important result from this PQ-01 study is that dimensional accuracy is feature-specific. The H2D 0.4 mm configuration was exceptionally good on the 20 mm reference block and the Z-height pads, yet it produced the largest circular-hole and external-round-feature errors in the test. The X1C 0.4 mm configuration produced better holes and external pins despite less exact reference geometry. The H2D 0.2 mm configuration substantially reduced hole error and almost eliminated thin-wall error, while leaving a measurable external-pin bias.

That combination of results is exactly why the coupon was designed around multiple geometry classes. If the test had used only a 20 mm cube, the H2D 0.4 mm print would have looked nearly perfect. If it had used only a 5 mm hole, the same machine/profile would have looked badly calibrated. Both impressions would have been incomplete.

For the three coupons measured here, the engineering conclusion is not that one printer is universally more accurate than another. It is that the relationship between CAD and printed geometry depends strongly on feature class and nozzle/profile configuration. For functional prototyping, that is the behaviour worth qualifying.

Sources and disclosure

The dimensional measurements, calculations and interpretations on this page are D&L Design Engineering original test data from the PQ-01 v1.2 workbook. Manufacturer specifications are used only for platform context:

• Bambu Lab X1-Carbon technical specifications
• Bambu Lab H2D technical specifications

This research article contains no affiliate links in the results, analysis or conclusion. Commercial links elsewhere on Engineering Tools Finder do not determine the test outcome or ranking presented here.