The first 75 ms of a contraction is a small window. At 40 Hz it contains three samples. At 80 Hz it contains six. At 250 Hz it contains nineteen. Same contraction, same patient, same window: the only thing that changes is the resolution of the slope you can compute through it. The review literature on explosive force production makes the same point with a sharper edge: rapid contractions should be treated as a time series, not as a small set of sampled points, because the time windows themselves are arbitrary thresholds rather than physiological events (Del Vecchio 2022).
Peak force, plateau steadiness, and the LSI between sides are mostly insensitive to this. They average across hundreds of samples over multiple seconds and don't depend on resolving fast transients. RTD does. A single outlier in a three-sample regression can shift the slope by 20 to 30 percent. The same outlier in a nineteen-sample regression barely moves it.
ForceIQ measures the effective rate on every connection (or every uploaded file) and surfaces it as a reliability badge on the RTD tiles. This article walks through what each tier looks like, which devices land in which tier, and what to do when the badge is yellow or red.
Low: below about 54 Hz
The first plot below is a clean MVIC captured at roughly 40 Hz. The trace looks fine because the contraction itself is multi-second; you cannot tell from the plateau that the rise was undersampled. The RTD Early window (the orange band) is where the limitation shows up. There are only three samples inside it, which is enough to fit a regression line but not enough to be confident the line reflects the patient's actual rate of force development rather than the noise on any one of those three samples.
What you'll see in the app. Every RTD tile on a session at this rate shows a red Low Confidence badge and an inline warning explaining the sample density.
What lands here. Two kinds of recording, mostly. Devices whose transmitted rate is lower than their advertised rate, which is common enough to be worth assuming until measured: a rate is often quoted for the sensor and not for what reaches your software, and a firmware option that adds a feature sometimes costs transmission rate to pay for it. And files that were downsampled somewhere between the sensor and the export.
Grip trainers sit further down the same scale. A device with an event-driven readout rather than a continuous stream can put well under one sample inside the RTD Early window. That is a design decision rather than a defect, and those devices serve the screening purpose they were built for. They are simply not instruments for rate of force development, and the platform flags them accordingly.
The science underneath is unchanged regardless of which device produced the signal. A 75 ms window holds three samples at 40 Hz, period. Peak force, plateau steadiness (nRMSE, CV, Yank), and bilateral symmetry on any of those measures are unaffected by the rate and remain fully valid.
Moderate: about 54 to 80 Hz
Between the low and high tiers there's a soft middle ground. Four or five samples in the RTD Early window is enough for a regression that is usable but not durable. ForceIQ marks these readings as Mod. Confidence in yellow. The number is shown the same as any other; the badge is your reminder that it deserves less weight in the read than a high-confidence number would.
The band is set by the sample count rather than by the rate itself: four samples is the moderate floor, which a 75 ms window reaches at about 53.3 Hz, and six is the high threshold, which it reaches at exactly 80 Hz. That is a narrow band, and in practice few devices sit inside it. Most land either below 54 Hz or comfortably above 80. The moderate badge is a quantitative statement about the regression: four or five samples is enough to fit a line, but the line is more sensitive to any one of those samples than the same fit at higher rates.
A high rate does not by itself earn a good RTD number. Some devices stream fast and report in coarse steps, and the reliability badge on this page only describes sample density. A device reporting in 0.1 kg or 1 lb increments produces a staircase, and the derivative RTD depends on turns into a series of spikes separated by flat stretches no matter how many samples per second arrive. That limitation is tracked separately, as a steadiness reliability flag rather than a rate flag; see What the steadiness reliability badge means. Read the two together when you are judging whether a fast but coarse device suits explosive testing.
High: 80 Hz and above
The plot below is the same MVIC captured at 80 Hz. Six samples inside the RTD Early window, eight inside the RTD Late window (100 to 200 ms). The regression has enough resolution that a single noisy sample cannot dominate it.
What you'll see in the app. No badge on the RTD tiles. The reliability flag exists on the session document either way; high-reliability sessions simply don't surface it visually.
What lands here. Most hand-held tension gauges built for clinical testing land in this band or just above it, and custom rigs around lab-grade load cells typically stream here or higher. The 80 Hz threshold is also where most published return-to-sport literature on rate-of-force-development sits. One property is worth checking separately from the rate: a device that stamps every sample in hardware gives you a true time base, and a device sitting right at 80 Hz with per-sample hardware timing is a better foundation for any time-derivative metric than the modest rate suggests. That combination is rarer than it should be, and it is almost never on the spec sheet.
Very high: 250 Hz and above
At very high rates the RTD Early window contains many samples (seventeen at 225 Hz, nineteen at 250 Hz, twenty-five at 333 Hz, fifty at 500+ Hz). The slope estimate is stable; the limiting factor on RTD precision shifts from sample count to the patient's own motor-unit firing variability between trials, which is a real biological signal rather than a measurement artifact.
What you'll see in the app. No badge. The chart looks visibly denser inside the RTD windows, but the metrics output is unchanged in format from the 80 Hz case.
What lands here. Isokinetic dynamometers commonly export between 100 and 500 Hz, and some research-grade strain-gauge rigs sample at 1000 Hz or higher. A number of Bluetooth force gauges reach this range too, by sending several samples in each packet rather than one, so this is not only an upload tier.
Files uploaded from any of these systems land in this tier when the CSV preserves the native rate. If an export was downsampled before being saved, the platform will measure the post-downsample rate and flag accordingly; the file's history matters, not the device's catalog spec.
This tier is also where the caveat from the moderate section bites hardest. A device can earn a high rate flag on sample density alone while reporting in 0.1 kg or 1 lb increments, so the signal it delivers is a staircase rather than a smooth rise. A high rate flag on the RTD tiles means the slope had enough points to fit; it does not certify that the underlying signal was smooth enough for the slope to mean what you want it to. The steadiness reliability badge is the one that speaks to that, and at this tier it is the one to read first.
Why the platform measures the rate instead of trusting the spec sheet
A device's listed sample rate and its effective streaming rate often diverge. Firmware updates change the transmission rate. Bluetooth congestion drops packets. Some devices buffer and burst rather than streaming evenly. ForceIQ measures the rate from the actual inter-sample intervals over the first two seconds of every recording and uses that measured value to decide the badge tier. The spec sheet is never consulted.
The same applies to uploaded files. If a CSV exported at 200 Hz was post-processed and saved at 100 Hz, the platform sees 100 Hz and treats it that way. There's no metadata header that overrides the actual timestamps. The methodological lineage on this point is well established in the rapid-force literature: the same zero-phase 20 Hz low-pass filtering on natively sampled signal used in Del Vecchio et al. (2019) is what the platform applies under the hood today.
The other reason the rate matters is onset placement. Threshold-based onset detection at low sample rates can place the onset off by 24 to 30 ms (Tillin et al. 2013), and that error budget falls almost entirely inside the RTD Early window. Higher sample rates do not solve onset detection on their own, but they make it possible to do better.
When to act on a low-confidence badge
The badge does not mean the RTD value is wrong. It means the value depends on so few samples that one noisy reading can swing it by a clinically meaningful amount. Whether that matters depends on what you're doing with the number:
- Return-to-sport decisioning that hinges on RTD asymmetry. A low-confidence RTD value is a noisy estimate of the patient's actual rate of force development. The clinical question to ask is whether the noise floor of a three-sample regression is compatible with the threshold you're using to clear the patient; that's a clinical judgment, not a platform one.
- Within-patient longitudinal tracking on the same recording method. Low-confidence trends are still informative as long as the rate hasn't changed between visits. The noise floor is consistent across recordings, so the direction of change is preserved even when the absolute value isn't durable.
- Plateau and peak metrics. Peak, nRMSE, CV, and Yank are computed over hundreds of samples across multiple seconds. Their precision is not meaningfully affected by the rate tier.
What to do next
- Check the badge before reading the RTD numbers, not after. The reliability tier is the context everything else sits in.
- After any device firmware update, restart the device once and reconnect. Firmware changes can move the transmission rate in either direction. The first session afterward measures and stores the new effective rate, and the badge tier updates from there.
- If you're doing high-stakes RTD work and only have a 40 Hz device, document it in the session notes. The number still ships and is useful as a relative measure, but the next clinician reviewing the patient should know which tier the data came from.
Sample rate is one of three device properties that decide which metrics a recording can support. The other two, the size of the step in the reported number and where the timestamps come from, are worked through in which dynamometer specs actually decide your metrics.
References
- Del Vecchio A. Neuromechanics of the rate of force development. Exerc Sport Sci Rev. 2022. doi:10.1249/JES.0000000000000306
- Del Vecchio A, Negro F, Holobar A, et al. You are as fast as your motor neurons: speed of recruitment and maximal discharge of motor neurons determine the maximal rate of force development in humans. J Physiol. 2019;597(9):2445-2456. doi:10.1113/JP277396
- Tillin NA, Pain MTG, Folland JP. Identification of contraction onset during explosive contractions. Response to Thompson et al. J Electromyogr Kinesiol. 2013;23(4):991-994. doi:10.1016/j.jelekin.2013.04.015