XO Wave: EQ Distortion

No EQ is perfect. This page attempts to explain why. If you would like to know more, digital signal processing texts cover these issues in greater depth.

People generally talk about EQ, (or equalization, which is sometimes called "filtering") in terms of adding or reducing the amount of sound at certain frequencies. This is the "frequency domain" perspective of EQ. There is another important perspective to consider regarding EQ, though: the "time domain". Designers of filters (especially digital filters) often spend a great deal of time thinking about the time domain, but it's sometimes important to non-engineers as well. For one thing, a filter can have great frequency response but still produce undesirable results because of effects in the time domain. These problems can be hard to describe, but people tend to say they hear problems with "stereo image", "clarity", or "naturalness" of the resulting audio.

Technically, time domain issues are generally caused by problems with "non-linear phase shift." Without going into mathematical detail, phase shift is an artifact of virtually all EQ (both analog and digital), and can be a serious problem in some contexts. But because phase shift is so common and so difficult to avoid, audio engineers tend to regard it as a necessary evil. Digital filters can be designed to avoid phase shift, but such filters tend to be hundreds or even thousands of times more computationally intensive. Unless carefully designed, these "linear phase" filters can have problems of their own, including pre-echoes, "jumpy" frequency response, extreme latency, and parameters that are impossible to automate.

Time domain aside, all filters have some problems in the frequency domain as well. Some filters have uneven or "rippled" frequency response, others allow too much signal through the stop band, and still others eliminate desired signals in the pass band. In fact, it is mathematically impossible to design a filter with perfect frequency selectivity, which usually makes filter design and selection a case of picking the best approximation to the filter you really want.

Obviously, picking a filter can be tricky business, and there are a wide variety of filters which offer various trade-off in both frequency and time domain response. Probably the most popular filters used in audio are Butterworth filters. They offer a smooth transition between pass band and stop band, have no ripple, and are computationally efficient. On the other hand, Butterworth filters have a fairly strong phase shift, which may not be desired.