Until Synergy, the filtering needs for a traditional FFT analyzer and scopes were very distinct. Almost all competitive instruments are designed to favor one domain or the other, and therefore will exhibit large and uncorrectable errors in half a users applications.

The steep roll off required to maximize FFT performance causes a large overshoot in Time Domain step response. Conversely, the wide bandwidth required for good waveform fidelity in DSO applications is poorly suited for FFT use due to aliasing in the frequency domain.

The waveform shown below is a 1 kHz square wave as reproduced through a steep FIR filter or sigma-delta converter. The distortion, called Gibb’s phenomenon, is so severe it makes the square wave almost unrecognisable. It may seem counter-intuitive that a filter could “add” oscillations, but recall that in theory a square wave is constructed of an infinite series of harmonic sine waves. The overshoot effect is actually caused by phase non-linearity and the lack of the higher frequencies which would “fill in” the ripples.

The wide bandwidth required for good waveform fidelity on a DSO is poorly suited for FFT use due to aliasing in the frequency domain. The figure below shows the FFT of a square wave beginning at about 10 kHz and sweeping upward in frequency. Since a square wave’s harmonics extend to infinity, the components above 100 kHz are “folded back” into the spectrum and falsely appear as lower frequencies. For example, the 11th harmonic at 110 kHz appears in the display to be 90 kHz. All the spectrum ridges that lean to the left are alias frequencies that did not exist in the original signal.

Rather than permanently compromise one mode like competitive instruments, Synergy offers software selection of filter characteristics to best suit varying tests. The following examples show when to use each type. The figure below shows a linear frequency sweep increasing to 100 kHz. As seen in the upper plot an oscilloscope-like time-optimized filter gradually attenuates the signal beginning at a low frequency, increasing to -3 dB (a 29% amplitude error) at its cutoff frequency. Even frequencies at 1/10th the 3 dB point are reduced in amplitude by several percent. In contrast, the steep-cutoff FFT filter provides very accurate amplitude readings up to 80 kHz but strongly attenuates all signals above 100 kHz to avoid aliasing. When you need the flattest frequency response such as for modal analysis or acoustic studies, always use the FFT filter type.

The figure below shows a square wave input, the classical test of an oscilloscope’s quality. An accurate square wave requires a wide, smooth bandwidth with gentle roll-off. In this case notice the time-domain filter is much more accurate when you zoom in on the edge. The more aggressive FFT filter causes overshoot and aberrations that did not exist in the signal. When accurate peak measurements are needed on a fast signal such as a ballistics signal, an actuator response or an engine combustion waveform, a time-domain filter should always be used and the highest possible sample rate. Systems which use sigma-delta A-D converters are unsuitable for this class of measurements and can lead to large uncorrectable errors.

Left in the “Auto” mode, Synergy’s filters provide automatic anti-aliasing protection of the optimal bandwidth for all sample rates. Manual selection of filter frequencies is also provided to further clean up noisy signals.