Topre & EC Dome Force Curves: Weight, Tactility, and Force-Wall Onset

Updated for the complete public release: Force Curve Bench fc-3.4, EC Parts Builder lib-6.0, and Beyond Snap Ratio Preprint 2.0, published 29 August 2026. The preprint is not independently peer reviewed.

A force–travel curve is not a complete model of how a keyboard feels. It is a controlled mechanical measurement: the reaction force recorded while a key mechanism moves through a stated displacement path. The curve can be reduced into useful quantities, but no single quantity is a universal measure of perceived weight, tactility, sound, or typing preference.

That distinction matters for Topre and Topre-compatible electrocapacitive (EC) domes. Topre's classic 30 gf and 45 gf designations refer to peak press-load targets, while electrical actuation is a separate event.^[1][2] Aftermarket names do not necessarily follow the same convention.

The current Force Curve Bench fc-3.4 therefore reports several complementary measurements. It also provides two plain-language percentile indices derived from an exploratory 25-dome subjective pilot:

• Weight Index is where a dome's collapse force ranks among the 68 domes in the released reference fleet, from 0 (lightest fleet endpoint) to 100 (heaviest fleet endpoint).
• Tactility Index is where its force drop ranks in the same tested fleet, from 0 (least-sharp fleet endpoint) to 100 (sharpest fleet endpoint).

The indices make comparisons easier; they do not turn perception into a universal unit. They are percentile positions, so 80 is not twice 40.

What is current in fc-3.4

The frozen release contains 76 semantic records: 68 dome-baseline records used by the percentile indices and eight part-assembly records kept outside that calibration population. It closes over 184 retained run bindings representing 180 unique acquisitions. These counts describe the released evidence, not the number of independently sampled production populations.

The bench and house protocol

The Dome Lab rig is based on bluepylons' open-source Open-Switch-Curve-Meter Gen 2 and was commissioned as a custom build.^[3] The source design uses a motorized linear stage and load cell. The Dome Lab adds its own Topre/EC fixture, acquisition procedure, intake policy, and analysis pipeline.

The house protocol records:

• a nominal 0.005 mm displacement increment;
• a quasi-static press near 0.15 mm/s, subject to the released speed-cohort checks;
• load-cell force calibrated and checked with known weights under the project procedure;
• both press and controlled return strokes; and
• at least two retained matching runs for every released cohort.

Each physical run is analyzed independently. The run-level event positions and nonlinear metrics are then averaged at full precision. The visible mean curve is constructed separately for comparison; it is not re-analyzed to create the canonical Collapse, Valley, steepest-drop, or force-wall values. That is why a mean marker does not always sit exactly on the feature that appears sharpest on the averaged line.

The nominal 0.005 mm increment is a commanded step size, not a claim of 0.005 mm measurement accuracy. Results depend on calibration, alignment, drift, fixture geometry, specimen condition, and the complete tested assembly. A released curve characterizes that specimen and configuration under this house protocol; it is not a manufacturing tolerance or an accredited calibration certificate.

First identify the mechanical landmarks

Let F(x) be press force in gram-force at recorded press displacement x. The main landmarks are:

Mechanical collapse, the post-Collapse Valley, electrical actuation, detected force-wall onset, and the recorded end of a test are different events. The bench does not infer electrical actuation from the force curve. That separation is especially important because current REALFORCE APC models can assign several electrical actuation depths to the same key mechanism.^[2]

Detected force-wall onset is not “Travel”

The former public label Travel has been retired. Starting after Valley, the fc-3.4 detector looks for the earliest point that is at least 10 gf and begins three consecutive 0.005 mm steps whose force increase is greater than 1% of the force at the start of each step.

This measured onset is affected by the complete assembly. It is not nominal dome height, physical slider travel, electrical actuation travel, or a universal contact coordinate. A shorter or longer value is not automatically better. If no sample meets the declared rule, the viewer displays Not detected. It never substitutes the final sample or recorded turnaround.

The force-curve view draws the detected onset as a vertical marker. All released comparisons use the same absolute 0–4.5 mm displacement scale, which covers every released force-wall marker without allowing longer recorded test limits to stretch the chart.

The weight-family measurements

The single-dome readout groups three measurements that described perceived weight most strongly in the exploratory pilot:

Collapse force (gf)

Collapse force is the force at the selected mechanical tactile peak. It was the strongest observed rank match to perceived weight in the 25-dome pilot. It is also the only input to Weight Index.

Ramp (gf/mm)

Ramp describes the central pre-Collapse force buildup. The method takes the median analysis force in the 0.05–0.15 mm baseline band, finds the last upward crossings at 10% and 90% of the baseline-to-Collapse rise, and divides the middle 80% force rise by the interpolated travel between those crossings.

Ramp excludes the initial seating transient. It is a supporting descriptor, not an additional term in Weight Index.

Pre-collapse work (gf·mm)

Pre-collapse work is the one-way mechanical work integral from the recorded press start to the interpolated Collapse position:

Wpre = integral from 0 to xc of F(x) dx

It describes work supplied before the tactile peak under this low-rate test. It is not a universal definition of “actual weight,” and it is not added to the index. For unit context, 1 gf·mm equals 9.80665 microjoules.^[6]

The tactility-sharpness measurements

Let Fc and Fv be Collapse and Valley force, and let xc and xv be their positions.

Drop (gf)

Drop = Fc - Fv

Drop is the absolute amount of supporting force lost between Collapse and Valley. It was the strongest observed rank match to perceived tactility sharpness in the pilot and is the only input to Tactility Index.

Steepest 0.10 mm drop (gf/mm)

For each eligible 0.10 mm interval between Collapse and Valley, the method calculates the average force-loss rate:

[F(q) - F(q + 0.10 mm)] / 0.10 mm

The largest eligible value is reported. This fixed-width descriptor captures the sharpest local portion of the descent while being less noise-sensitive than an adjacent-sample derivative. It is a supporting descriptor and not an index input.

Drop rate (gf/mm)

Drop rate = (Fc - Fv) / (xv - xc)

Drop rate is the average peak-to-Valley force loss per millimetre. It describes the full descent rather than its sharpest 0.10 mm section.

Snap (%)

Snap = 100 × (Fc - Fv) / Fc

Snap reports the force loss as a percentage of Collapse force. This makes it a useful normalized descriptor, but it contains no travel term and is not a complete tactility score. Similar peak-to-contact ratios appear as “click ratio” in silicone-keypad terminology.^[4]

These four descriptors share landmarks and are correlated. Their pilot associations do not isolate independent causal contributions. Combining them also produced a weaker match than simply ranking force Drop, so the Tactility Index does not blend several metrics together.

Where the two indices came from

The released subjective-pilot-v1 used one rater, 25 selected domes, and three fixed-order sessions. Each dome used the same Topre housing, slider, and conical spring. Weight was scored from 1 (feather weight) to 10 (heaviest in the panel), and tactility sharpness from 1 (linear/off) to 10 (sharpest in the panel). The analysis used each dome's arithmetic mean across its three sessions, so the effective experimental count is 25—not 75.

Collapse force had the strongest observed Spearman rank association with mean perceived-weight rank (rho = 0.952). Force Drop had the strongest association with mean tactility-sharpness rank (rho = 0.930). Those findings selected the mechanical input for each index. The correlation values are not coefficients in the index equations.

The final indices rank the full-precision Collapse-force or Drop value within the frozen 68-dome reference arrays. Ties use the average occupied rank; between-reference values are interpolated; values outside the fleet clamp to its 0 or 100 endpoint. The calculation is fully specified in the released method documentation.

The pilot is exploratory:

• it has one rater;
• the 25 domes were a selected panel, not a representative random sample;
• all sessions used the same fixed order;
• it did not use an independent holdout panel;
• it did not vary housings, sliders, springs, or precompression; and
• no tactility rating was 1, so it cannot validate linear/off classification.

The correlations therefore describe this panel and rater. They do not prove causation or predict every person's experience. Part-assembly experiments are outside the calibrated dome population and correctly display Not calibrated.

The complete pilot observations and analysis are available in the subjective-pilot release.

How to compare domes responsibly

Start with the question you are trying to answer:

• For perceived-weight context, read Weight Index, then compare Collapse force, Ramp, and pre-collapse work.
• For perceived collapse sharpness, read Tactility Index, then compare Drop, steepest 0.10 mm drop, Drop rate, and Snap %.
• For the measured assembly's hard-rise region, inspect the force curve and detected force-wall onset separately.
• For acquisition extent, use recorded turnaround only as a test limit.
• For sound, return feel, impact, and natural typing, remember that speed, hysteresis, vibration, keycap, housing, spring, silencing parts, and user technique are not reducible to these quasi-static press descriptors.

The viewer offers four comparison modes:

1. Force curves overlays the measured press and return traces on one absolute displacement scale.
2. Weight profile places Collapse force, Ramp, and pre-collapse work at their separate percentile positions in the reference fleet.
3. Tactility profile does the same for Drop, steepest 0.10 mm drop, Drop rate, and Snap %.
4. Weight vs tactility plots Weight Index on the horizontal axis and Tactility Index on the vertical axis.

The profile percentiles make different units visually comparable; they do not change the index equations. A high or low coordinate is descriptive, not a quality judgment.

Keep the assembly boundary visible. A dome baseline, a complete-key test, and a slider/ring experiment are not interchangeable. Aged rubber specimens and aftermarket parts can vary. Repeated runs from one specimen improve confidence in that measurement but do not measure production-line variation.

Force slope and natural typing speed

Drop rate and steepest 0.10 mm drop are spatial slopes: change in force per change in displacement. They are not force change per second. During motion, the relationship is:

dF/dt = (dF/dx)(dx/dt)

The same spatial force slope can therefore create a different time-domain force rate at a different press speed. Prior work on experimental click switches supports reporting force-drop magnitude and force-drop rate separately, but it does not validate a universal Topre perceptual score.^[8] Rubber keyboard mechanisms are also rate dependent; quasi-static curves should not be treated as exact predictions of force, vibration, sound, or impact during natural typing.^[9]

Data processing and reproducibility

Raw press and return CSVs, the generated viewer, method documentation, release tests, and canonical evidence are maintained in the Dome Lab repository.^[10]

The current rule is compute per retained run, then aggregate. Scalar values use full precision internally and are rounded only for display. Missing values are not converted to zero:

• Not detected means no force wall met the declared detection rule.
• Not available means a metric is absent or outside its valid domain.
• Not calibrated means a perception index does not apply to that record.

intake-qc-v1.4, rather than the retired v4.1 median filter, is the retention authority. It checks package integrity, acquisition shape, target-speed cohort, and replicate agreement under the versioned policy. Every result should be cited with its displayed build and evidence identity because later evidence or method versions can change the measured population or a derived value without changing an old release.

What changed from the previous article

The earlier article documented an intermediate metrics-v4.1 state. The fc-3.4 revision makes several material corrections:

• metrics-v4.2 and intake-qc-v1.4 replace the older method/filter wording;
• Travel is renamed Detected force-wall onset and kept distinct from recorded turnaround;
• missing force walls remain null and display Not detected;
• full-stroke press work and normalized drop rate remain auditable supporting evidence but are no longer headline consumer comparisons;
• the completed 25-dome subjective pilot replaces the former statement that controlled human work was only prospective;
• Weight and Tactility Indices communicate the pilot-selected Collapse-force and force-Drop ranks within the frozen 68-dome fleet; and
• the released evidence now contains 68 dome baselines, eight part-assembly records, and 184 retained run bindings.

Practical conclusion

For the force needed to reach a dome's tactile peak, begin with Collapse force and Weight Index. For the amount of supporting force released during collapse, begin with Drop and Tactility Index. Use Ramp, pre-collapse work, steepest 0.10 mm drop, Drop rate, and Snap % to understand the surrounding curve shape. Keep detected force-wall onset separate from both indices and from recorded test turnaround.

The Dome Lab does not replace subjective language with a universal score. It connects clearly named mechanical measurements to transparent, limited perceptual evidence—and keeps those claims versioned so they can be tested and improved.

Related public releases

• Force Curve Bench fc-3.4 — the released interactive viewer.
• Dome Lab — the public hub for the Bench, EC Parts Builder, and research.
• EC Parts Builder lib-6.0 — an evidence-aware assembly and compatibility tool. Compatibility remains evidence beta.
• Beyond Snap Ratio overview — plain-language context, limitations, citation, and permanent DOI links.
• Reproducibility data and analysis code — the versioned compendium for Preprint 2.0.

Sources and method notes

1. T. Nagashima / Topre Corporation, Keyboard, Japanese Patent Application Publication JP2001-216070A (2001), JST J-GLOBAL record.
2. Topre Corporation, REALFORCE Features.
3. bluepylons, Open-Switch-Curve-Meter.
4. Shin-Etsu Polymer Europe, Force–travel characteristic.
5. ASTM International, ASTM F2592-16, withdrawn in 2023 with no replacement. Historical context only; the Dome Lab does not claim ASTM compliance.
6. A. Thompson and B. N. Taylor, NIST Special Publication 811.
7. Topre Corporation, JP2012-138254A.
8. J. Hyeong and J. Lee, “Effects of Design Parameters of Haptic Profiles on the Feedback Clarity at a Switch,” Journal of the Ergonomics Society of Korea 40(3), 161–171 (2021), doi:10.5143/JESK.2021.40.3.161.
9. M. L. Nagurka and R. W. Marklin Jr., “Measurement of Stiffness and Damping Characteristics of Computer Keyboard Keys,” Journal of Dynamic Systems, Measurement, and Control 127(2), 283–288 (2005), doi:10.1115/1.1902823.
10. BuddyOG / Unreal Keyboards, Topre Force Curves — Dome Lab repository, release fc-3.4.
11. Brian Gebo, Beyond Snap Ratio: Reproducible Force–Travel Measurement and Exploratory Perceptual Correlates of Weight and Collapse Sharpness in Electrocapacitive Domes, Preprint 2.0 (2026), doi:10.5281/zenodo.22167065.

Method version described here: metrics-v4.2 + intake-qc-v1.4 + perception-rank-v1, bound to the identities listed above.

© 2026 Brian “BuddyOG” Gebo — Unreal Keyboards. All rights reserved.