Originally Posted by The Byre
Look at it this way - you take two channels on you oscilloscope and feed a sine wave directly into one and feed the other into the EQ (or whatever you are looking at) and then fiddle with the controls of said EQ. If the second signal starts to drift with any changes, then it is out of phase with the original. Even the better older designs tend to drift, so if the user is mixing a signal that is EQ'ed with one that is not EQ'ed (e.g. bass mic on a piano, just about anything on drums) one part of the mix will be out of phase with the other.
The A&H does not drift at all. Not even very, very slightly.
That is to say - any two filters, given different frequency responses, still maintain identical phase responses.
Byre, thank you for indulging me so far. I do enjoy discourse like this, and it's nice to be enjoying it with friendly folks.
I do understand what you are saying, but a filter that behaved like that could only be a linear-phase equalizer, something achievable only in the digital domain. It is possible the whole thing is done on DSPs with analog controls, but I find that chance to be remote given the somewhat stiff processing requirements of the algorithms responsible. But, I cannot rule that out, given what you have stated so far.
An analog filter, though, must cause phase lead or lag. It is simply not possible to avoid this, nor can I envision a design which could correct for all possible control settings to "null" the phase change of the filter. The number of all-pass filters necessary would be literally infinite. Indeed, if you combine an exact copy of an original signal with a filtered one, you will definitely get some comb filtering and ripple. In fact, this is the basic principal of most guitar phaser pedal circuits, although they use in general 4 filter stages that vary in center frequency slowly using an LFO. The filters' changing phase response causes many slowly changing (and interesting sounding) comb filter artifacts when combined back with the original signal.
(As a side note, there are some crossover designs that couple a first order filter with an all-pass filter to compensate for phase change, but these designs are not adjustable and suffer from severe passband ripple.)
Analog Devices has a good link that is unfortunately a bit math-y, but Figure 3 shows the phase change of a static sine wave for a first order high-pass and low-pass. They also show how phase alters with differing Q values and gains. Fader8's blog also has a nice sampling of different frequency and phase plots of various equalizers, and although those are plugins, the principals in analog are absolutely identical.
Interestingly enough, you can see how some of those plugins derive their sound, too. Some of them are really quite inaccurate, although I can see where they're going with the vintage stuff, trying to model transformer and inductor low-end losses and ripple. Either way, though, the gain changes are completely synchronous with phase changes, just like in any other filter. Phase Response in Active Filters: Analog Dialogue: Analog Devices Example EQ Phase Plots