By Alan Macy, BIOPAC Systems, Inc.
This blog continues the discussion of Electrocardiogram waveshape.
In some situations, it can be helpful to craft a compromise between maximizing ECG signal bandwidth, preserving the essential ECG waveshape, reducing the impact of skeletal myo-electrical noise and removing mains interference. If this collective compromise is not well-considered, certain problems can arise. For example, if notch filters are used to remove mains interference in the ECG signal processing channel, significant distortion can be introduced into the ECG waveshape. Accordingly, notch filters are generally avoided or only used in combinations with lowpass filters to minimize undesired effects.
When constraining the ECG signal bandwidth, it becomes increasingly important to be aware of the phase linearity of the filters used to bandlimit the channel. For example, if a 35 Hz, second order, lowpass, Butterworth filter is used instead of one at 150 Hz, then an observable “ringing” will be evident after the QRS interval. This ringing occurs because of the phase non-linearity of the 35Hz Butterworth filter. However, if the 35 Hz filter is adjusted to have more “Bessel-like” phase qualities, the problematic ringing can be eliminated. When limiting the ECG signal frequency band to less than 0.05-150 Hz, reasonably phase linear highpass and lowpass filters will preserve the essential waveshape of the ECG. Phase linearity implies constant group delay. Constant group delay means that all frequencies in the filter passband are delayed by the same amount of time. Phase linear analog filters, and their digital equivalents, can be designed to process an ECG signal with minimal, filter-induced, distortion.
In addition to the above-listed simple strategies, a host of digital signal processing methods are available to optimize the previously mentioned compromise. Methods include linear and non-linear adaptive filters, synchronous adaptive noise cancellation, dynamic neural networks, coherent averaging, transform domain singular value decomposition and Legendre moments. 
BIOPAC Systems, Inc. provides life science researchers and educators with data acquisition and analysis systems that inspire people and enable greater discovery about life. Visit us at www.biopac.com.