Is there any fundamental difference between filtering and EQ? The filtering I'm particularly interested in is linear-phase filtering. Could you explain the theoretical difference between the two?
SOS Forum Post
Technical Editor Hugh Robjohns replies: The difference is largely tied up in terminology. In conventional terms, 'EQ' refers to tonal equalisation and involves modest cuts or boosts of amplitude across the frequency spectrum to technically or subjectively 'improve' the signal's frequency response. Equalisation generally consists of filter stages that exhibit shelf or bell-shaped slopes, normally of 6dB per octave and rarely exceeding 12dB per octave. Simple EQ is provided with just a cut/boost control for each frequency band. More complex designs go on to offer adjustable turnover frequencies and possibly some adjustment of Q or bandwidth as well. The most flexible form of EQ is parametric EQ, which provides frequency, bandwidth and cut/boost controls.
Passive equalisers place the equalisation circuits either before or after a fixed-gain amplifier — in which case the amp makes up for the inherent loss in the EQ circuit, effectively boosting the frequency range(s) that haven't been cut. Active equalisers incorporate the EQ circuitry in the feedback loop around the amp, which makes its gain frequency-dependent. This is the more common approach in modern EQ designs, but is often claimed to sound inferior.
Filters are generally described as cut-only devices that are intended to remove some portion of the frequency spectrum. They come in three flavours: high-pass, low-pass and band-pass. Most filters have slopes of 12dB per octave or higher, with the steepest found in audio systems generally being 24dB per octave (although there are steeper versions for specialist applications). The filter slope is sometimes designed to incorporate a peak in the amplitude response immediately prior to the steep cutoff slope, and this tends to emphasise the frequency region close to the cutoff area with a kind of 'resonance'. If allowed to resonate excessively the filter will self-oscillate — in other words, it will generate a pure tone at a pitch related to the cutoff frequency — and this is a common facility found in many synthesizer VCF sections.
Every filter and EQ has a phase response which varies in some way relative to the frequency, just as the amplitude response may vary relative to frequency. The process of filtering inherently imposes a small delay, and this is what creates the phase shift. But the important thing is how that delay (phase) varies relative to the frequency. This is the phase response.
Most analogue audio equalisers and filters are 'minimum phase' designs — the well known Butterworth filter designs, for example — where some frequencies experience a different amount of 'processing' delay to others. The steeper the filter, the worse the phase-response variations become, which inherently distorts the waveform shape.
In contrast, a filter with a 'linear phase' response provides a constant time delay for all frequencies. Linear-phase filters are difficult to create in the analogue domain (the Bessel filter is the closest, but cannot have a steep slope between the pass and stop bands), but are fairly straightforward to achieve in the digital domain. The anti-aliasing and reconstruction filters used in most digital converters tend to employ linear-phase filters to minimise waveform distortion.