Optimal filtering of constant velocity torque data.

The purpose of this investigation was to implement an optimal filtering strategy for processing in vivo dynamometric data. The validity of employing commonly accepted analog smoothing methods was also appraised. An inert gravitational model was used to assess the filtering requirements of two Cybex II constant velocity dynamometers at 10 pre-set speeds with three selected loads. Speed settings were recorded as percentages of the servomechanism's maximum tachometer feedback voltage (10 to 100% Vfb max). Spectral analyses of unsmoothed torque and associated angular displacement curves, followed by optimized low-pass digital filtering, revealed the presence of two superimposed contaminating influences: a damped oscillation, representing successive sudden braking and releasing of the servomechanism control system; a relatively stationary oscillatory series, which was attributed to the Cybex motor. The optimal cutoff frequency for any data set was principally a positive function of % Vfb max. This association was represented for each machine by a different, but reliable, third order least-squares polynomial, which could be used to accurately predict the correct smoothing required for any speed setting. Unacceptable errors may be induced, especially when measuring peak torques, if data are inappropriately filtered. Over-smoothing disguises inertial artefacts. The use of Cybex recorder damping settings should be discouraged. Optimal filtering is a minimal requirement of valid data processing.