EQ Phase Responses

[Mlewis]

Everybody talks about the importance of an EQs phase response but what does it actually look like in general?

Do State variables have significantly different phase response graphs to say Constant Amplitude Phase shift circuits or is it simply a case of certain EQ circuits have more or less of a standard shape phase shift for similar looking frequency responses?

Has anybody ever created phase response graphs of various famous EQs? Got a link for further reading?

 

[mikep]

As far as I know, in the analog world, phase response is inextricably linked to frequency response. if the frequency response of two differently constructed filters is exactly the same, so will be the phase. there are digital filters that do not conform to this. also, in some marketing departments this fundemental law can be temporarily suspended.

there are some differences between a CAPS EQ and a state variable and a biquad, all producing the exact same filter response. but not related to phase. things like control law, internal headroom, noise, parameter interaction, adjacent band interaction, possible stability issues, etc.

mike p

 

[SSLtech]

Yep. What he said.

The phase response will basically look like a line which slowly rises as the ANGLE of the amplitude response rises, then dive swiftly down through zero degrees as the amount of boost levels off, then as the amount of boost falls, the (now negative) phase shift will slowly return to the zero line.

It's related to the integrated ANGLE of the amplitude response curve at any given point.

Now here's how things CAN sound different: change the 'Q' of the filter in the sidechain and vary the AMOUNT of the filter added to the pass signal. With comparable amplitude responses, the impulse response can be VERY different. This is what I think contributes more to the "sound" of an EQ than anything else.

Keith

 

[Mlewis]

[quote author="SSLtech"]
It's related to the integrated ANGLE of the amplitude response curve at any given point.
Keith[/quote]

That was quick thanks...

That seems to make sense in an intuitive kinda way, but if phase is related to an integral of frequency response, how do all-pass filters alter phase without 'differentiating' back from the integral to alter the frequency response? eg:

http://www.forsselltech.com/LF%20Phase%20Test.pdf

 

[Mlewis]

...hang-on. I thing i might be getting my calculus tied up in knots. Did that question make sense?

Edit: Yeah... i think it does make sense. I don't understand how something related to the intergral of the angle of amplitude response can alter without altering the the differential of the integral ie: the angle of the amplitude response.

 

[Boswell]

To be pedantic for a moment: the steady-state response of a system evaluated against frequency is a complex quantity. The complex quantity can be expressed in different ways, e.g. as real and imaginary components (X-Y) or polar form (radius and angle). It is this latter form that when represented as two separate functions against frequency is called a Bode plot. The radial component is the amplitude of the response and the angular component is the phase. The amplitude response is commonly called the "frequency response", but that term should include both the amplitude and phase components. The two components are orthogonal, and can change independently, although it is usual for a physical system to show changes in both. The "all-pass" network is an example of a system where the amplitude response does not change (over a specified frequency range) but phase does.

So to go back to the original question, a true second-order resonance (or anti-resonance) is defined by a second-order equation, and that dictates both the amplitude and phase response. EQ circuits that implement exactly the second-order equation should all sound the same, but they don't. The differences probably lie in the way the circuit is implemented, the quality and tolerances of the components used and the circuit's non-linear behaviour, invoked during transients.

 

[SSLtech]

Well, with a phase-shift all-pass filter, the amplitude/frequency response will change when the two paths are combined, or if any regeneration/feedback is applied.

If you look at the topology of most EQ circuits -like the DIY Calrec EQ, you should see what I mean.

Keith

 

[CJ]

Don't forget the transformers.
There are three of them in the EQP 1A.

So on top of whatever filter phase eggnog you have, add that to the sometimes non linear phase response of those three transformers, then give me the bode plot.

 

[Mlewis]

That was exactly the info that i was hoping to find... thank you all!!! :sam: :sam: :sam:

 

[Boswell]

For my previous post, I was hunting around for a link to a Michael Gerzon article "Why do equalisers sound different?", but failed to find it. Got it now:
http://www.audiosignal.co.uk/Resources/Why_do_equalisers_sound_different_A4.pdf

 

[SSLtech]

Fanastic. I was looking for that article myself... in fact I used it as a reference when I had an article in Studio Sound in 1991... -I forget the publishing writer, but it was about stuff based on investigations and discoveries made when I built the AAD Equaliser which retrofitted into the SSL E and G-series consoles.

Keith

 

[Marik]

Since we are on the topic, do digital EQs (both, stand alone and software) affect phase response?

 

[SSLtech]

The question is broad.

Basically yes, though there are finite impuse response (FIR) and Infinite Impulse Response (IIR) types.

It is POSSIBLE to build a phase-linear-at-all-times EQ in digital. This was initially held up as a shining jewel in the "let's all go digital" crown in the 1980's. unfortunately, as I understand it these types of EQ are prone to increased processing latency (*note: corrections welcome here*) but they DO sound different... and very good, I understand.

However, the much publicised ability of digital to do phase-linear EQ has led to me finding a lot of people who think that digital EQ is therefore phase linear.

(sigh)

It seems it will take a lot of re-educatin to correct some of these miscomprehensions.

Keith

 

[Samuel Groner]

As I understand it these types of EQ are prone to increased processing latency.

Filters with linear phase have inherent delay as linear phase implies a symmetrical impulse response (example at random: impulse_response.jpg) and half of the impulse response length is the delay.

In fact in digital it is possible to choose any phase response--so it would be possible to build an EQ which could seamlessly blend between minimum and linear phase response.

Samuel

 

[SSLtech]

Cool, thank you.

I was hoping that I wasn't way off-base. -I'd hate to add to the pool of misinformation!

Keith

 

[Larrchild]

Here's John Oram's take on some of this:
http://www.soundonsound.com/sos/may98/articles/JohnOram.html

 

[Michael Tibes]

Trident asked if I could also design a compressor, which I did, using a single 2N3819 FET as the shunt element in the attenuator, and it worked fine. That kind of circuit traditionally suffers from bad distortion, but I found that if I mixed some of the input signal with the control voltage, I could actually cancel it out.

That's what always gives me mixed feelings about John Oram, to me this sounds like he 'reinvented' the 1176...

I might understand things wrong though.

Michael

 

[gyraf]

..sad to say, I think you read him correctly.. :roll:

 

[SSLtech]

Yep. It's only because he felt sorry for Mr. Ohm that we don't call it 'Oram's law'...

Well, if you believe everything he says, that is...

Keef

 

[Larrchild]

Nothing is more contentious in gear fora-culture, than mentioning the "O" word, I know. It's like tossing a live grenade into a thread.

But the statement about the phase response of the hi's and lows and the broad eq "Q" statements intrigued me.

Any comments on those?