Correcting Room Interference Dips with Minimum Phase PEQ

[NTK]

There are discussions in the miniDSP Tide16 thread (started about here) regarding whether or not to correct a room response dip, and I can't help myself and want to add my 2¢ Since the topic is not really related to the miniDSP product, I think a new thread would help me stay on the good side of the mods.

My thinking is that, the main cause of an interference dip is due a reflection (or reflections) that destructively interfere with the direct sound or each other. The result of the interference is probably not (should not be?) a minimum phase response, and cannot be fully corrected with a minimum phase PEQ.

To illustrate this, I will use a simple single front wall reflection (SBIR) which causes comb filtering cancellation dips. I borrowed the pictures below from Genelec to show how these interference dips are created.

Monitor Placement - Genelec.com

To create the ideal listening experience, several aspects need to be taken into account when placing the monitors within the listening space. First, you need to identify the listening area within the space and consider where the listener will be placed in relation to the space as well as the...

www.genelec.com

In my simulation I had a perfectly reflective wall, and the sound source (omni point source) is placed 0.8575 m from the wall so that f1 = 100 Hz. Below are my simulation results. I used a "measurement" position of 4 m from the front wall but it is not important. The blue curve shows the response before EQ. The cancellation dips aren't infinitely deep because the reflection image has to travel a longer distance to the measurement point and are therefore lower in amplitude than the direct sound.

Two biquad PEQs are used to correct this dip based on the "measurement". In the plot, the PEQ is shown inverted to match the dip. I'd have to say the correction looked surprisingly effective, and minimum phase PEQ may indeed be useful remedy for these interference dips. They WILL REQUIRE A LOT OF POWER, though. The PDF of my Mathematica notebook is also attached.

 

[ppataki]

Wouldn't that correction also increase the speaker's distortion?

 

[NTK]

Wouldn't that correction also increase the speaker's distortion?

Sure. In my example, the PEQ has a 14 dB gain (25 times power). You aren't getting any free lunch here.

 

[ppataki]

Sure. In my example, the PEQ has a 14 dB gain. You aren't getting any free lunch here.

I have tried this with all my systems and it works in real-life too with certain dips

The FR got flat indeed but in my case there were two issues:

- increased distortion (which can actually be a non-issue in the bottom octave since our ears are not that sensitive about distortion in that region)
- increased decay time - this was unfortunately a showstopper for me, ringing in the time domain got unbearably high

But it is definitely worth a try in any case!

 

[Keith_W]

I decided to replicate your experiment with my own simulation.

I created two 100Hz sine waves. I delayed the second sine wave so that it is 180deg out of phase with the first sine wave.

Sum the two sine waves, and we get the classic comb filtering pattern.

Extract the excess phase and view the group delay (lower graph). All those spikes in the EP GD plot are non-minimum phase, and all the flat regions are minimum-phase.

In theory, this dip can not be filled since we are attempting to invert a zero outside the unit circle. But I decided to try anyway. In pink is a PEQ. The Q is 50 (!!!!), and the gain is 50dB (!!!).

When the PEQ is convolved with the comb filtered wave, the result is surprisingly good. Ignore those fuzzy lines, it has something to do with how Acourate displays graphs. But the gain on that PEQ is 50dB!!

 

[markstein]

View attachment 505815

When the PEQ is convolved with the comb filtered wave, the result is surprisingly good. Ignore those fuzzy lines, it has something to do with how Acourate displays graphs. But the gain on that PEQ is 50dB!

Yes, but your example is theoretical and in reality, depending on how the room is constructed (concrete, wood, dry walls/ floor/ ceiling) the dip(s) may be shallower and Q may be wider.

Nevertheless I would add a gain of only some dB to equalize a dip, something between +3...+6dB max.

 

[kemmler3D]

I'm able to correct low frequency dips in my office system in practice, but I also have more than 15dB headroom to do so... So this tracks.

 

[TurtlePaul]

When the PEQ is convolved with the comb filtered wave, the result is surprisingly good. Ignore those fuzzy lines, it has something to do with how Acourate displays graphs. But the gain on that PEQ is 50dB!!

In most real world rooms, the back wall bounce is 3-6 dB attunated relative to the direct sound. This is primarily because if you are listening at 2 meters and you speakers are a meter from the wall, then the back wall reflection needs to go a meter back to the wall, then a meter back to the speaker, then two meters to your listening position for four meters total. This is double the distance of the direct sound, so it at most -6 dB.

If you reduce the delayed 100 hz tone to half amplitude, you will still have the cancellation but it won't asymptote to negative infinity and therefore won't require a 50 dB filter.

 

[thewas]

I have tried this with all my systems and it works in real-life too with certain dips

The FR got flat indeed but in my case there were two issues:

- increased distortion (which can actually be a non-issue in the bottom octave since our ears are not that sensitive about distortion in that region)
- increased decay time - this was unfortunately a showstopper for me, ringing in the time domain got unbearably high

But it is definitely worth a try in any case!

I had tried similar several times in that past and can confirm it usually doesn't sound good even if distortion stays low, due to the high ringing of such a sharp positive filter. The response at the LP where the correction is performed may flatten but the filter usually causes bass peaks in other locations in the rooms which also can still be perceived as boomy bass at the LP. This seems to be also the reason that experienced companies like Genelec, Neumann warn from such placement and their room correction tools do not create such filters.

 

[CT Man]

How about installing bass traps in corners behind your speakers and in back wall corners? Bass is far more even in my room after doing that. Powerful with clear articulation at the listening position. It also tames mid and high frequency reflections without too much deadening.

I mostly listen to get what sounds "right" to me. I wonder to what extent psycho-acoustical effects on perceived sound "fixes" certain problems that appear in measurements.