Phase correction in an audio blend circuit

[frozenfrogz]

Dear all,

currently I am working on a small signal splitter/blending circuit based on a TL074.
Its main purpose is to be used as an effects loop for guitar effect pedals.

The signal is split after the input buffer U1A. U1C acts as a variable-impedance phase flipper. You can put a guitar effect pedal in the send-return loop and blend between the original signal and effected signal via RV1. The phase flipper is needed because some pedals work as inverting amplifiers and some as non-inverting.

Here is my question though: What possible circuits are there to explore if I want to achieve granular phase correction. So instead of flipping 180° say shift 90° or some other value between 0° and 180°.
I was thinking 2nd order Bessel-allpass, but phase shifting would be frequency selective in that case of course.

I should add that I am not an EE, but just started to discover the magical world of op amps.

Any pointers are highly welcome!

Edit: Added question mark as message icon.

[duak]

Here is a link to an article on a phase shifter that works over a 1:10 range of frequencies: https://www.edn.com/electronics-blogs/living-analog/4375814/All-pass-filter-phase-shifter
I expect that if more stages with different frequencies are added, the range of frequencies can be increased.  Please note that this circuit works by delaying the input signal by varying amounts in two chains and then taking the differences between the chains.  Since the the phase modified signal is delayed, summing it with the original undelayed signal may not give you the effect you wish, although it may be an interesting effect.

I've attached another phase shift network, but I don't know anything about it.

The phase shifter is also known as a mathematical function called a Hilbert Transformer which these days is done digitally with a DSP.

[T3sl4co1l]

Phase at what frequencies?

See, the hard thing to do is constant phase (see above!) -- but if you're not correcting for something that's constant phase, you certainly don't need it for the corrector!

Electrically, lumped filters are the easiest.  If you're correcting for the phase shift of an electronic filter, you need to use its phase dual (which, by the way, is nontrivial to synthesize -- at least, I expect so, I haven't seen how it's done).

If you're correcting for cable delays, forget about it -- you need miles of cable for that to matter.  (Ma Bell figured that all out, back in the 1930s!  The journal articles they wrote on the subject are free, by the way -- if rather technical and dry.)  Basically irrelevant in a normal audio system.

If you're correcting for delay, again you don't want constant phase shift -- indeed, you need linear phase shift.  Just think of how many sine waves fit in a given length.  It's proportionally more at higher frequencies, and phase is the angle in a single cycle, therefore -- linear phase shift.   Now, this is rather hard to create through passive electrical means: it would take enough cable to, well, build a telephone company!  You can make a lumped transmission line equivalent, but that sucks (the number of components goes as delay * bandwidth^2!).  So, it's done either through acoustics (where the speed of sound is naturally lower), or digitally (where the signal can be stored in a RAM chip, delayed arbitrarily).  Both are common in effects today.

(There's also the analog equivalent of the digital delay: a bucket brigade device (BBD), now a relic of times passed; you'd make the discrete equivalent with a chain of op-amps (to regenerate the signal) and switched capacitors.  Needless to say, no one sane does this. )

Tim

[capt bullshot]

There's a design tool for allpass based phase shifters:

http://www.tonnesoftware.com/quad.html

[frozenfrogz]

I guess linear phase shift would be what I am looking for. A micro delay line for alignement correction to get the wet and dry signal perfectly in sync. But maybe I am overthinking the whole operation and working for a solution to a rather non existing problem. XD
Differential delay seems to be the correct term (not a native speaker here - please excuse).
In the studio I have to deal with this a lot when working with audio signals from for example two microphones on the same guitar speaker.
Here I have the option to look at the signal and physically move the microphone closer or further away to compensate, but with delay that is introduced by the electronics in an effects pedal I would need to have other means.

I should go and measure a bunch of pedals to have some hard numbers on the amount of delay introduced.

Thanks for bearing with me, there is a lot of stuff to wrap my head around.

[Benta]

Your microphone placement is a good example of what you might be looking for.

There's a 1:1 relationship between (linear) phase and (constant) delay.

If you delay all frequencies equally, you'll have completely linear phase shift. This is only possible digitally (bucket brigades are gone, as Tesla says)

You could approximate it with an analog allpass filter, but be aware that a second order type has double phase shift compared to a lowpass filter.

I suggest experimenting with a Bessel allpass filter with the possibility of moving the transition frequency by switching the caps.