Sampling Frequency: How much is enough? 

Accurate biomechanical measurements depend on not only precise measurements but also the speed (specifically, the sampling frequency) of these measurements. Sampling frequency is measured by the number of measurements taken in one second, known as “hertz” (Hz). While modern technology enables very high sampling rates – up to thousands of hertz – the real question is: how much is enough?

While modern technology enables very high sampling rates…the real question is: how much is enough?

This blog unpacks the science behind sampling frequency and explains its impact on data quality based on research-based insights. This will allow you to determine the minimum sampling rates you need for different types of movements and, in turn, which technologies – such as force plates, fixed-frame dynamometers and handheld dynamometers – may be best for assessing them.

Why Sampling Frequency Matters 

Sampling frequency refers to how often we record data from movements such as jumping or lifting. The higher the frequency, the more snapshots we have of the action, leading to a clearer understanding of the movement.

However, there is a limit beyond which we do not have any clearer understanding of what is happening.

In the table below, a sinusoidal signal sampling at 500Hz has been deliberately downsampled by taking every 5^th, 10^th, 25^th, 50^th and 260^th point to demonstrate the influence of sampling frequency on the signal quality.

Example	Effect on Data Quality	Effects Illustrated
Every 5th and 10th point	Lots of samples around the peak values, which suggests oversampling at 500Hz.
Every 25th point	Less frequent sampling but the peaks of interest are still identified accurately.
Every 50th point	Too few samples are taken, and the peak values are lost. However, the overall shape of the original signal is still recognizable.
Every 260th point	Too few samples are taken making the original signal unrecognizable.

 

 

The Frequency of Human Movement

Most everyday human motion, from walking to waving, occurs at frequencies below 20Hz.

Why does this 20Hz threshold matter? In essence, it represents the typical upper limit of the frequency spectrum for everyday human movements.

While most human movements fall below the 20Hz threshold, the precision required for elite performance analysis or the detection of subtle rehabilitation progress may necessitate higher sampling rates. This includes rapid movements, such as the impact phase of a drop jump or landing from a countermovement jump (CMJ).

While most human movements fall below the 20Hz threshold, the precision required for elite performance analysis or the detection of subtle rehabilitation progress may necessitate higher sampling rates.

This is particularly true for capturing the rapid onset of force or the nuances of explosive movements (e.g., metrics such as rate of force development (RFD)) that may have critical components occurring at higher frequencies. If data is not captured at sufficiently high frequencies, crucial details can be missed.

How Much is Enough? 

Enter the Nyquist theorem, a fundamental principle in signal processing. It states that to capture a movement accurately, you need to sample at least twice as fast as the highest frequency present in that movement. This “Nyquist limit” helps ensure we capture the movement without introducing errors or “aliases” – false frequencies that can mislead our interpretation.

In practical applications involving human movement analysis, sampling at exactly twice the highest frequency is often insufficient. Many researchers and professionals in the field recommend sampling at rates significantly higher than the Nyquist rate to ensure accurate data capture. A common guideline is to sample at 5 to 10 times the highest frequency of interest.

…sampling at exactly twice the highest frequency is often insufficient. Many researchers [recommend sampling] at 5 to 10 times the highest frequency of interest.

Fast, non-cyclical movements have frequency components at higher sampling rates (e.g., isometric mid-thigh pull (IMTP)), meaning we need to measure at higher sampling rates to capture information relevant to assessing rapid force production.

Modern devices, including force plates and fixed-frame dynamometers, can capture the amount of force produced over time, but how much is enough to ensure nothing is missed?

The table below summarizes suggested minimum sampling rates across different tests and metrics. While each example may result in a similar amount of force over time, the minimum sampling rate depends on the specific test and metric being analyzed.

Next, we will unpack the reasoning behind the ideal sampling rates for each of these movements based on the evidence we have available.

The Debate in Literature

No one-size-fits-all sampling frequency exists in biomechanics. Research varies, with studies examining everything from jumping to isometric strength tests. The following studies have examined isometric and jumping tests to determine the appropriate sampling frequency:

1. Hori et al. (2009): This study found that key metrics like peak power and force from jumps remain reliable across a range of sampling frequencies, but accuracy dips significantly below 200Hz. It suggests higher frequencies ensure more precise data, particularly for dynamic movements like jumping.
2. Dos'Santos et al. (2019): This research in rugby players showed that 500Hz frequency provided reliable and accurate measurements of peak force, time-specific force values and RFD for an IMTP when compared to data sampled at 2,000Hz.
3. Maffiuletti et al. (2016): This paper underscores the importance of high sampling rates for accurately measuring RFD, which is crucial for understanding explosive strength and performance and recommends at least 1,000Hz when measuring force at the motor unit level in a laboratory environment.
4. Vanrenterghem et al. (2001): This study emphasized the impact of sampling and processing techniques on measuring jump height, concluding that the magnitude and variance in the absolute difference in jump height between sampling rates increased substantially when sampling at less than 100Hz compared to 1,000Hz. However, it was fine above 100Hz.
5. Street et al. (2001): This study explored errors in measuring jump height using the impulse method from a jump. It showed that sampling at under 300Hz causes significant jump height percent differences compared to jump heights calculated from measurements at 2,700Hz.
6. Renner et al. (2022): Compared to variables obtained while recording at 1,920Hz, the study found that for drop landings, 95% accuracy in peak impact force can be achieved with a sampling frequency of 96Hz, while 99.5% accuracy in impulse metrics and 95% accuracy in average loading rate can be obtained at 148Hz and 384Hz, respectively. These findings suggest that lower sampling frequencies than traditionally used can still provide highly accurate measurements for specific landing metrics.

VALD's Internal Testing

While academic research provides invaluable insights, the real-world application of these findings can sometimes be a different story. To address these uncertainties, VALD conducted case studies on isometric hip abduction and drop landing tests, with the results summarized in the charts below.

Collectively, these results demonstrate that the determination of peak force is not significantly impacted by sampling rate, especially beyond a certain threshold.

For the isometric hip abduction analysis, data obtained at a high sampling rate of 400Hz were systematically downsampled in increments of 50Hz to evaluate the effect on peak force.

Both the RMSE and R-squared metrics indicated that the accuracy of force measurements significantly improves as the sampling frequency increases up to 50Hz. Beyond this point, increases in sampling frequency result in diminishing returns regarding measurement accuracy.

…accuracy of force measurements significantly improves as the sampling frequency increases up to 50Hz. Beyond this point, increases in sampling frequency result in diminishing returns…

This suggests that a sampling frequency of around 50Hz is sufficient for capturing highly accurate isometric force data. However, this is highly dependent on the force profile of the movement being assessed.

Force produced with a single peak, such as during an isometric test, can rely on lower sampling frequencies compared to force traces with rapid peaks and troughs, such as in a drop landing.

In the drop landing test analysis, peak force calculations remained consistent across various sampling rates, from 1,000Hz down to 142Hz, despite reductions in the number of samples captured and variations in the window of capture.

This indicates that even at lower sampling rates, the determination of peak force in dynamic movements such as drop landings is not meaningfully impacted, but the minimum sampling frequency required is higher than in isometric tests.

Overall, these findings suggest that for both static isometric tests and dynamic movements like drop landings, a sampling rate of around 50Hz is adequate for accurately determining peak force.

…for both static isometric tests and dynamic movements like drop landings, a sampling rate of around 50Hz is adequate for accurately determining peak force.

However, higher sampling rates may deliver more accurate data when assessing metrics involving both force and time (e.g., impulse and RFD). This challenges the assumption that higher sampling rates are always necessary for precise biomechanical measurements, offering practical implications for biomechanical research and the design of measurement devices.

Final Thoughts

So, where does this leave us? There are several implications for practitioners to keep in mind:

• It is commonly assumed that sampling at very high frequencies (e.g., >1,000Hz) is needed for measuring force over time in all applications. However, recent evidence shows that 1,000Hz is significantly more than is required for many applications, with as little as 50-100Hz often being sufficient.
• At VALD, we provide products capable of high sampling rates in order to meet industry standards. However, we believe that users should not overemphasize the importance of high sampling rates to the accuracy of their measurements. 
• Practitioners should ensure, first and foremost, that they are using good protocols and validated technologies to control the quality of their assessments. These factors likely play a much larger role in maximizing the accuracy of results than a faster sampling rate.

If you would like to learn more about how sampling frequency impacts accurate and reliable objective musculoskeletal analysis, please reach out here.

References

1. Hori, N., Newton, R. U., Kawamori, N., McGuigan, M. R., Kraemer, W. J., & Nosaka, K. (2009). Reliability of performance measurements derived from ground reaction force data during countermovement jump and the influence of sampling frequency. The Journal of Strength & Conditioning Research, 23(3), 874–882. https://doi.org/10.1519/jsc.0b013e3181a00ca2
2. Dos’Santos, T., Jones, P. A., Kelly, J., McMahon, J. J., Comfort, P., & Thomas, C. (2019). Effect of sampling frequency on isometric midthigh-pull kinetics. International Journal of Sports Physiology and Performance, 14(4), 525–530. https://doi.org/10.1123/ijspp.2015-0222
3. Maffiuletti, N. A., Aagaard, P., Blazevich, A. J., Folland, J., Tillin, N., & Duchateau, J. (2016). Rate of force development: physiological and methodological considerations. European Journal of Applied Physiology, 116(6), 1091–1116. https://doi.org/10.1007/s00421-016-3346-6
4. Vanrenterghem, J., De Clercq, D., & Cleven, P. V. (2001). Necessary precautions in measuring correct vertical jumping height by means of force plate measurements. Ergonomics, 44(8), 814–818. https://doi.org/10.1080/00140130118100
5. Street, G., McMillan, S., Board, W., Rasmussen, M., & Heneghan, J. M. (2001). Sources of error in determining countermovement jump height with the impulse method. Journal of Applied Biomechanics, 17(1), 43–54. https://doi.org/10.1123/jab.17.1.43
6. Renner, K. E., Peebles, A. T., Socha, J. J., & Queen, R. M. (2022). The impact of sampling frequency on ground reaction force variables. Journal of Biomechanics, 135, 111034. https://doi.org/10.1016/j.jbiomech.2022.111034