Force Plate Baseline Drift: Zeroing and Tare Fixes for Clean Session Data

You zero the plate at 8am and everything reads clean. By the third block, an hour and a half later, with the gym now ten degrees warmer, the quiet-standing trace before every jump sits eight or nine newtons above where it started. Nobody touched a setting or bumped a cable. The athlete's body weight obviously hasn't changed between rep three and rep thirty, but the plate now insists it has. If your software has flagged movement onset a few frames early, or a jump-height number has drifted lower across a session even though the athlete's vertical looks identical on video, you've met baseline drift. It isn't a malfunction. It's a strain-gauge plate doing exactly what strain-gauge hardware does when temperature or residual load shifts underneath it.

The fix isn't a full recalibration between every set. It's a short, repeatable re-zeroing habit built into the session, plus knowing which drift is ordinary noise and which means stop and check the hardware. That's what this covers: why the baseline moves, how a zero problem differs from a tare problem, and a checklist you can run mid-session in under thirty seconds per block.

Why the Baseline Moves Mid-Session

A force plate's zero reading is just what the amplifier reports with nothing on the plate. That number isn't a fixed physical constant. It's the output of strain gauges bonded to a metal structure, and both respond to their environment in ways that have nothing to do with an athlete standing on top of them.

Temperature is the biggest everyday driver. Strain gauge load cells carry a manufacturer-specified thermal zero shift, typically 0.01-0.02% of rated capacity per degree Celsius. That sounds negligible until you run the numbers: a plate rated to 10,000N with a 0.015%FS/°C coefficient drifts roughly 1.5N per degree. A field house that swings 9°C from an early-morning session to a midday one accumulates about 13.5N of pure thermal offset, enough on its own to push a quiet-standing baseline past the few-newton noise floor most jump-onset detection is tuned around, and enough to nudge a calculated body weight, and everything derived from it, in a direction that has nothing to do with the athlete.

Residual load is the second driver, and it's more physical than electrical. Load cells exhibit creep, a slow return toward true zero after a heavy load is removed, so zeroing right after a max-effort squat or a loaded jump catches the cell mid-recovery rather than at rest. Debris does the same thing more crudely: a strip of athletic tape, a film of sweat, a water bottle set on the corner between reps, or a heel resting fractionally off the plate's edge all add a small, constant mass the system reads as part of the baseline.

Zero and Tare Are Not the Same Fix

Coaches use these two words interchangeably, and that's part of why sessions run into trouble. A true zero sets the reading to 0 with the plate genuinely unloaded: no fixture, no mat, no athlete. A tare subtracts whatever is currently sitting on the plate as a new offset, which is what you want when a jump-mat bracket or fixed attachment stays on the plate permanently and the system should ignore its weight going forward.

The failure mode is running a tare when a true zero was needed, or the reverse. Taring over debris bakes that debris into the new baseline instead of removing it, quietly stealing a few newtons of resolution from every measurement that follows. Zeroing with a fixture still attached forces the software to report a negative reading the instant it's removed. Before either operation, physically confirm what's actually sitting on the plate. Don't assume yesterday's setup matches today's.

Warm-Up and Pre-Session Setup

Most of a session's drift budget gets spent in the first twenty minutes, before anyone has jumped. Amplifiers and load cells self-heat once powered on, and that warm-up curve is steepest right after power-up. Testing immediately captures a zero that keeps moving for the next fifteen to twenty minutes, so the fix is unglamorous but effective: power on the system twenty minutes before the first athlete arrives, and treat that window as inspection time rather than dead time.

Use it to check the plate surface for tape residue, chalk, or moisture; confirm cables are fully seated at both ends, since a loose connector produces its own erratic offset that mimics drift without any temperature involved; and if the plate was just mounted or pulled from storage, give the hardware ten to fifteen minutes to settle before capturing the session's first true zero. A plate zeroed the moment it's bolted down often reads differently ten minutes later purely from mechanical settling.

The Mid-Session Re-Zero Checklist

Run this on a fixed schedule rather than waiting for the numbers to look wrong. By the time drift is visible on screen, several trials have already been recorded against a shifted baseline.

1. Capture a reference baseline at the true start of the session: unloaded plate, 5 seconds of quiet data, mean and standard deviation recorded.
2. Re-check every 15-20 minutes or every 8-10 trials, whichever comes first. Clear the plate, wait 30-60 seconds if the prior trial was a heavy or maximal effort, then capture another 5-second unloaded sample.
3. Compare against the reference. Within roughly 0.1% of the athlete's typical body weight, about 0.8N for an 80kg athlete, treat it as noise. Between 0.1% and 0.5%, re-zero at the next natural break. Beyond 0.5%, re-zero immediately, before another trial counts as data.
4. Log the offset and ambient temperature if the system allows notes. An offset that tracks room temperature confirms thermal drift; one that appears suddenly with no temperature change points at debris or a connector instead.
5. Re-zero, don't just tare, once the plate is fully unloaded for the check. Save taring for genuine fixture changes.

For a two-hour combine-style session with four or five blocks, this adds well under three minutes of downtime and catches drift before it changes which athlete looks like they improved between blocks.

What the Research Says About Baseline Error

Street, McMillan, Board, Rasmussen, and Heneghan (2001), writing in the Journal of Applied Biomechanics, modeled how errors in the impulse-momentum method for countermovement jump height propagate from the initial body-weight/baseline measurement. Because the method integrates force twice, once to velocity and once to displacement, even a small, sub-percent offset in that starting baseline compounds through both integrations, producing jump-height errors well out of proportion to the size of the original offset. Their practical recommendation, since widely adopted in CMJ protocols, was a genuinely quiet standing phase of roughly one second immediately before the jump, used to establish a clean baseline rather than assume the system's last zero still holds. The limitation worth carrying forward: the error propagation was demonstrated through simulated force-time data rather than a real plate drifting under variable temperature, so it establishes the mechanism, amplification through double integration, more than a numeric tolerance for every device.

Owen, Watkins, Kilduff, Bevan, and Bennett (2014), in the Journal of Strength and Conditioning Research, compared methods of establishing body weight for CMJ power calculations, including deriving it from a pre-jump quiet-standing force average versus other estimation approaches. They reported that peak power output changed by several percent depending on which method was used, and recommended flagging movement onset using a threshold roughly five standard deviations above the noise in that quiet baseline, a threshold that only works if the baseline hasn't already drifted out from under it. Their comparison was limited to a single jump type and software implementation, so the exact percentages don't necessarily transfer to every plate.

How Much Drift Is Actually a Problem

Not every wobble in the baseline needs a stoppage. These bands describe a practical way to triage what shows up mid-session, scaled to the athlete's own body weight rather than a fixed newton value, since an 8N shift means very different things for a 55kg gymnast and a 140kg lineman.

Mistakes That Bake Drift Into Your Data

When Re-Zeroing Doesn't Fix It

If a fresh, unloaded zero won't hold, drifting again within a couple of trials instead of staying put for the usual 15-20 minutes, the problem has moved past ordinary thermal behavior. Check the cable run for a pinch point or a connector that's slightly backed out; both produce exactly this pattern of a zero that refuses to stabilize. If the cable checks out clean, isolate whether the drift tracks one specific corner of the plate rather than the whole reading, which usually points at a single failing load cell rather than an environmental cause.

At that point the fix isn't a protocol change, it's a service call. Continuing to test against a baseline that won't hold still produces data that looks plausible on screen while being unusable beyond a same-day comparison. Flag the unit, note the failure pattern for the technician, and fall back to a known-good plate for anything tracked across weeks rather than within a single workout.