ReaEQ shows incorrect phase plot? (FIXED)

[bezusheist]

what exactly is the phase plot showing you on ReaEQ ?
is it supposed to show the actual phase response of the filter ?
and where are the "degree" markings ?
i have checked impulse responses from ReaEQ in other analysis apps and have found the phase plots on ReaEQ to be incorrect, or at least they do not correspond.
has anyone else noticed this ?
is this a bug or intended behavior ?

[Bri1]

@bezusheist >

Cockos-please reaspond-thinking it's quite important to investigate all these things further?
Even a little description would be usefull

[bezusheist]

well, i never look at a phase plot while actually using an EQ...nor have i ever felt the need to. so it's not "important" to me if it works correctly or not.
my only concern is that if something is technically "wrong", then it should be fixed, regardless of it's relevance to me or anyone else.
(...i feel the same way about Reaper's "normalization" feature being wrong.)

[bezusheist]

Nobody else cares that this is incorrect ?

[EvilDragon]

Apparently?

[Fergler]

It probably looks different from other EQs because not all EQs follow the same representation of frequencies. The X axis of the graph is not linear.

Also be sure you are comparing filters perfectly. Can you be sure that the 0-4 'Q' (bandwidth) setting on one EQ is equal to the 0-1 Q setting on another? What if the other one uses 0.025-40, like Pro Q 2?

[bezusheist]

the way i check the phase response is with an impulse response and measurement software.
so i am not comparing to any other EQ, i am comparing ReaEQ's phase plot to the phase plot generated by a 3rd party software (which is correct).

[Fergler]

You will have to go into more details (pics?) of this measureing process.

[bezusheist]

ReaEQ's phase plot...


Fuzzmeasure phase plot...


i realize that one is up-side down, but the important thing to note is where 90 and 180 degrees are on both plots...
ReaEQ does not show degrees, but the vertical line is 180 degrees...and in the wrong place...the corner should be 90 degrees for this filter.

[Robert Randolph]

So, how do we know that Fuzzmeasure's plot isn't incorrect?

What's the predicted phase response for this filter?

[bezusheist]

Originally Posted by Robert Randolph

So, how do we know that Fuzzmeasure's plot isn't incorrect?

What's the predicted phase response for this filter?

Fuzzmeasure gives me consistent measurements with various (brand) filters and seems to be in sync with "ideal" plots. (seems to work fine for IIR filters.) theres a bunch of stuff you can check/make plots for PC, but i am mac so i use what i can get. R.E.W. gives me the same plots.

the filter is a 12 dB Butterworth filter, so you can probably find a bode plot somewhere for the ideal phase response. or even the math.
but like i said, the important thing to note here is for this type of filter's phase response, the corner is at 90 degrees...ReaEQ (with butters Q) shows it at 180.

[bezusheist]

another way to confirm is with a sine wave...(sine Hz = HPF Fc)
100Hz sine wave > 100Hz HPF = 90 degrees phase shift...


edit: i will note i have not checked all of the filter types on ReaEQ, only the HPF so far...

[Robert Randolph]

So I think there is definitely a bug with the phase shift display in ReaEQ.



Take this example with a notch filter at 1khz. A notch filter should present a 90° phase shift at 1khz since the correlation being measured is no signal to full amplitude signal, which is 90° rotation.

Taking a measurement confirms this



This is taken with REW and shows the expected 90° shift at 1khz.

Testing against DMG Equilibrium, which has a free phase design mode, confirms the REW measurements are correct.

[Justin]

Thanks, fixing this -- two issues: the range was not useful, and we had a sin(w) rather than sin(-w), oops.