3D Print Tolerance Calculator: Slicer Settings & CAD Fits

Fit each X or Y dataset separately. All calibration sizes are in millimetres internally; division by a 1 mm reference length makes regression coefficients numerically well scaled.

For outside widths and pins, equal-weight least squares fits $M_o=sN+b$. Four measurement pairs estimate the common scale $s$ and a diametral (two-wall) bias $b$. One-size mode instead fixes $b=0$ and takes $s=M_1/N_1$; that assumption cannot distinguish shrinkage from edge error.

After fixing $s$, fit hole residuals by least squares: $M_h(d)=sd+h+k/d$. The two fitted coefficients come from a line in $1/d$. This inverse-diameter term is an empirical interpolation chosen here, not a universal polymer shrinkage law or nophead's polyhole formula. In one-hole mode $k=0$ and only the measured hole bias is used. Hole and outside RMS residuals describe model mismatch, not manufacturing tolerance.

Invert the increasing branch to get the required baseline hole path: $z_h(T)=[T-h+\sqrt{(T-h)^2-4sk}]/(2s)$. The baseline pin path is $z_p(T)=(T-b)/s$. Reject nonpositive, nonreal or decreasing solutions. The diameter-dependent correction is kept in CAD because these slicer fields supply a constant offset.

The recommended scale is $q=1/s$. Orca/Bambu retained size is $100s$%; Prusa/Cura object scale is $100/s$%. Correct the outside with $c=-b/(2s)$ per wall. At the chosen hole anchor $A$, Orca/Bambu hole compensation is $a=[z_h(A)-A/s]/2$. Positive hole offset opens the bore. Cura first expands the solid by $c$, closing holes by that radius, so Hole Horizontal Expansion is $a+c$. Set its Max Diameter field to 0; nonzero values taper the correction and are outside this model. PrusaSlicer XY Size Compensation is $c$ and its equivalent hole offset is $-c$; it has no separate hole field in this workflow. Unavailable hole-only fields are omitted from the selected-slicer setting table.

A horizontal through-bore intersects each layer as open contours rather than a closed hole polygon. In that mode all four slicers use effective bore offset $a=-c$, and no hole-only compensation is credited. Positive solid expansion narrows the transverse XY opening; negative solid expansion widens it. This mapping does not describe Z height or roof sag.

For the baseline CAD workflow use $q=1,c=a=0$. For the corrected workflow use the selected slicer's $q,c,a$. Then $d_{CAD,h}=(z_h(T)-2a)/q$ and $d_{CAD,p}=(z_p(T)-2c)/q$. The predicted round-trip clearance verifies that scale and offsets are applied once. For a target finished pin $D$ and diametral clearance $C$, set $T=D+C$. The table's pin column targets $D$ for every fit class.

If each finished diameter can vary by at most $u$, the worst-case clearance interval is $[C-2u,C+2u]$. This is an arithmetic bound, not a confidence interval. UMD guide 306, page 7, gives the same two-part tolerance stacking example. It is used here without adopting layer height as a reliable tolerance estimate. Settings conventions were checked against the linked documentation and source code on 2026-10-27.

Numerical residuals are rounded to 0.0000000001 mm to suppress floating-point subtraction noise in the unit audit. This has no practical metrology significance.

Checks against measured calibration data

The CaliCross check passes all 15 published nominal/measured pairs through the production outside-fit function. It gives 99.8581% retained size, 100.1421% object scale and -0.1052 mm intercept, agreeing with the worksheet's 99.86%, 100.14% and -0.11 mm displays. This is a comparison at published rounding precision, not a claim that every worksheet averaging step is identical.

LuckyPants' original v2 workbook is checked separately as a transfer: its published approximate scale is 100.6730863% and radial correction -0.02541074735 mm. That legacy calculation averages mixed inner/outer and X/Y data. Its scale expression uses a first-order approximation instead of an exact reciprocal; it is reproduced only for verification. The production method is unchanged. A second check fits four outer X readings and predicts the independently measured 2 mm web: about 2.0134 mm versus 2.040 mm. The 0.05 mm comparison band is a screening criterion, not measurement uncertainty or validation over the extrapolated interval.

Sudin et al. (2016), Table 3, provides four measured ABS cylinder sizes and four hole sizes in the same X direction. Fit the cylinders and only the 5, 10 and 20 mm holes. The withheld 15 mm bore is predicted as 14.7692 mm versus the published 14.7630 mm: about 0.0062 mm difference. That tests interpolation on independent measurements; it does not qualify another printer, orientation or filament.

The outside extrapolation warning compares every required baseline path $(D-b)/s$ with the nominal outside calibration range, before applying any chosen CAD workflow. The diagram caps its visual gap enlargement to retain a visible housing wall; the CAD calculation is never capped.

Use an uncorrected baseline: object scale 100%, Orca/Bambu retained-size entry 100%, and XY/hole offsets zero. If your test used existing compensation, first print an uncompensated test; these are replacement baseline settings, not increments to an unknown preset.

One dataset represents one printer, filament condition, nozzle, process profile, measurement axis and orientation. Repeat for X and Y. Shared XY shrinkage and offsets cannot independently remove anisotropy; compare the two runs before choosing a shared value, or use independent object-axis scales and keep filament shrinkage at 100%. Do not combine reciprocal object scaling with retained-size scaling a second time.

For horizontal through-bores the measured X or Y direction must be transverse to the bore axis. Their open contours respond to the solid contour offset, not the hole-only offset. Outside pins and flat dimensions are assumed to share one scale and edge bias. Use pin-only outside measurements if that assumption fails. The empirical response is assumed stable when the slicer changes the nominal boundary. Shrinkage scaling precedes the stated contour offsets. Avoid fuzzy-skin painting and multi-material painted objects that can disable compensation.

Fit names are practical labels for editable trial clearances, not ISO classes. Default fit targets and measurements are illustrative engineering starting points, not copied material tables or a qualified press-fit design. Running is used here for a freer moving fit than slip; terminology differs among guides. Formlabs' 2017 fit measurements concern specific SLA resins and are not transferred to FDM defaults.

Use the curved fit near the measured hole-diameter range, on its increasing branch. Extrapolation flags are screening guidance: a good residual or a quiet warning box does not validate a different shape, wall thickness or temperature. Four outside points leave two residual degrees of freedom; three holes leave one; one-size modes cannot assess model mismatch.

Print and measure at room temperature after cooling. Correct flow, mechanics, skew and elephant foot separately. Positive clearance is a gap; negative clearance is interference. Press-fit force, stress, cracking, material compliance, layer adhesion, creep, lubrication and service temperature are not calculated. A small negative clearance may still be unsuitable for brittle or thin parts.

The entered per-part bound must include repeat-print scatter and measurement uncertainty after mean correction. Never interpret zero RMS residual as perfect accuracy. Test the proposed CAD sizes in a mating coupon, then update the record.

For horizontal through-bores this calculator describes only the measured X or Y width transverse to the bore axis. Its corrected workflow uses the solid contour offset; a hole-only setting is ignored. Hole height, unsupported roof sag, bridging, first-layer fusion and separate Z gaps require physical tests. Print-in-place values alone do not guarantee release. Do not enable polyhole conversion or change mesh faceting after calibration without repeating the test.