• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

DRC-FIR & excess phase

StefanR

Member
Joined
Jan 17, 2020
Messages
10
Likes
4
Location
Sweden
HI all,

I have an (almost) full range 2-channel system (2 large passive floorstanders) that I have measured with an amplitude- and phase-calibrated microphone and corrected in two steps: first with REW's EQ, filter exported as wav and applied in convolver on original measured IR and then I fed the resulting EQd measurement to DRC-FIR by Denis Sbragion. The graph shows the resulting final FR (EQ+DRC) in blue, for one channel, pretty good. In dotted red is the unwrapped phase, in black excess phase and in grey minimum phase.

I am wondering why DRC has not reduced the excess phase? It is seemingly way off, a lot of phase rotation (thousands of degrees). In the graph there is zero IR delay, i.e. the impulse peak is at t=0.

In my second system (second graph with FR in green and phase in dotted blue, for one channel), with two active digital/dsp monitors and one active sub, I get the same phase pattern, sharp fall from sub bass to treble. Sub is time aligned. Again nice FR but is phase OK like this?

(never mind the absolut FR curve, it is without mic calibration)

(I did watch Amir's video where it is stated that deficiencies in absolute phase in loudspeaker audio reproduction is basically inaudible and what matters is channel symmetry and that phase changes smoothly with frequency, so maybe my question is not important? Or maybe my measurement result does not reflect reality when it comes to delay vs. frequency?)

1727712369737.png
1727713176485.png
 
Last edited:
why DRC has not reduced the excess phase?

That would just cause audible artifacts e.g. pre-ringing.

Excess phase correction should mostly be performed with the room windowed out of the equation — ideally anechoic. You could maybe do a bit more in the low bass when trying to improve integration between multiple speakers/drivers with different phase profiles in-room — but even that has to be realistically limited.

Even the creator of DRC says his “normal” correction preset could already be a bit too much in some systems — recollection from discussions in the archived mailing list.
 
First of all, do you understand that excess phase is caused by the room? When a speaker is measured in the room, what is measured is the minimum phase response of the speaker, together with the excess phase response of the room. One of the first steps is to extract the minimum phase response through the Hilbert-Bode transform. Correction is done on the minimum phase response only, since it is the only response that can be inverted and remain stable. The excess phase response can not be inverted because the zeros reside outside of the unit circle, and attempting to invert this will create a pole which will create an unbounded output. This will result in asking your amp to supply infinite current, which isn't good for the amp to put it mildly.

Excess phase is the result of a cascade of minimum phase filters, e.g. minimum phase crossover + minimum phase driver = excess phase. Another example is that a single reflection is minimum phase, but multiple reflections add up and create excess phase. Although I have lumped them together, they are not the same. The former is a convolution (multiplication) function, whereas the latter is a sum. The difference is that zero times one is zero, but zero plus one is one. Thus a minimum phase cascade that is produced by multiplication is more likely to retain its minimum phase characteristics (including the important characteristic of inversion) than a cascade that is produced by addition.

Up to a point, it is possible to correct excess phase with a reverse all pass filter and there are tools available to do so, e.g. RePhase. As ernestcarl points out, there is a risk of creating pre-ringing.

But it is virtually impossible to correct excess phase created by the room, it would be akin to trying to move walls with DSP. Because it is a physical problem, it requires a physical solution: reposition your speakers to create more symmetry, think about moving furniture, acoustic treatment, etc.

Re: whether excess phase is audible. Well, we have to be careful about what we are talking about here. If you have one speaker jammed up against a wall, with the other into a room opening, the excess phase will be asymmetrical creating inter-channel phase asymmetry. You will easily hear that. But what if the speakers are symmetrical and the EP response looks the same between the speakers? Whether this is audible or not is somewhat controversial. Ohm's Acoustic Law states that it is not audible. Amir's video states it is not audible. JJ seems to think that if it is less than 15 degrees per ERB, it is not audible. My personal belief is that it is audible.

My take-away from all this reading is: above all, make sure the EP response between left and right speaker is symmetrical. This is best achieved by ensuring symmetry when the speakers are positioned, because attempting to DSP your way out of badly positioned speakers is a fool's errand.
 
Back
Top Bottom