Does Phase Distortion/Shift Matter in Audio? (no*)

[gberchin]

You might be interested in participating in the constant beamwidth transducer thread. I watched Don Keele's video. He said he got his inspiration from submarine sonar systems.

Already on it; thanks.
https://www.audiosciencereview.com/...ducer-cbt-speakers.12060/page-13#post-1956297
https://audiosciencereview.com/foru...th-transducer-cbt-speakers.12060/post-1956297
I guess it's up to the world at large to continue Don's work.

 

[j_j]

Having worked with SONAR beamformers in a previous life, I know that there is an entire mature technology that is only just starting to be exploited for audio.

Just barely. Also being exploited, but just barely, is, if you will, anti-beamformers for diffuse rendering. It has some interesting behaviors in PA systems where the diffuse nature is a good thing.

 

[Holmz]

Just barely. Also being exploited, but just barely, is, if you will, anti-beamformers for diffuse rendering. It has some interesting behaviors in PA systems where the diffuse nature is a good thing.

It would probably would be good in a church or cathedral to have one of the speakers wired out of phase.

 

[j_j]

It would probably would be good in a church or cathedral to have one of the speakers wired out of phase.

No, not really that at all. The issue is substantially more complex than "out of phase".

 

[mdsimon2]

The processing of both FIR and IIR was exactly the same. Both were time aligned, but the IIR had a massive time discrepancy compared to the FIR - 35ms to align the subwoofer as opposed to 9ms, no doubt due to phase rotation at 50Hz.

Why would you have different delays for minimum phase and linear phase LR4 crossovers? They should be exactly the same and without identical delays you are not comparing apples-to-apples, but rather just showing how poor phase alignment impacts magnitude response.

Michael

 

[j_j]

Why would you have different delays for minimum phase and linear phase LR4 crossovers? T

That's how the filters work.

 

[gberchin]

Why would you have different delays for minimum phase and linear phase LR4 crossovers?

Linkwitz-Riley crossovers sum-to-allpass, which is not minimum phase.

 

[mdsimon2]

Linkwitz-Riley crossovers sum-to-allpass, which is not minimum phase.

IIUC, @Keith_W was comparing a linear phase LR4 (FIR) to a minimum phase LR4 (IIR). Making a linear phase LR filter is easy in programs such as rephase but I am not sure what was used to generate the filter.

I made some FIR and IIR filters and compared them with your test signal. Both were LR4's with a subwoofer XO point of 50Hz.

Michael

 

[mdsimon2]

That's how the filters work.

It is possible I misunderstood @Keith_W's statement. I agree that delay would need to be added minimum phase filter in order to get it match the delay of the linear phase filter.

I took it as delay was added to one side of the minimum phase filter which in retrospect doesn't make much sense. Maybe @Keith_W can clarify his approach. Does 26 ms correspond to the delay of the linear phase filter?

Michael

 

[gberchin]

IIUC, @Keith_W was comparing a linear phase LR4 (FIR) to a minimum phase LR4 (IIR).

Though an IIR LR4 is composed of two minimum-phase filters, the sum of the outputs is not minimum-phase. I'm not certain from context whether he was comparing the responses of only the two lowpass filters, or comparing the summed outputs from the lowpass and highpass sections.

 

[mdsimon2]

Though an IIR LR4 is composed of two minimum-phase filters, the sum of the outputs is not minimum-phase.

100% agree.

I'm not certain from context whether he was comparing the responses of only the two lowpass filters, or comparing the summed outputs from the lowpass and highpass sections.

I agree, would be helpful to understand exactly what was being compared.

Michael

 

[j_j]

Though an IIR LR4 is composed of two minimum-phase filters, the sum of the outputs is not minimum-phase. I'm not certain from context whether he was comparing the responses of only the two lowpass filters, or comparing the summed outputs from the lowpass and highpass sections.

This is correct, but because it's not power conserving, it can have a different (assymetric) delay. Same thing for biorthogonal wavelets. Tried that a few years ago for coding, the lack of power complimentary ate it alive for coding.

 

[Keith_W]

Linkwitz-Riley crossovers sum-to-allpass, which is not minimum phase.

Said it in far fewer words than I could have. Thank you.

All pass = max phase. It introduces additional delay. So the time alignment procedure needs to be repeated.

 

[gnarly]

All pass = max phase. It introduces additional delay. So the time alignment procedure needs to be repeated.

I don't use all-pass much, finding complementary linear-phase xovers and min-phase Eq's (for individuals drivers), are all i need to make time and phase aligned DIY speakers.
So i may be wrong here.....
But I think max phase = inverse all pass.

 

[Keith_W]

I don't use all-pass much, finding complementary linear-phase xovers and min-phase Eq's (for individuals drivers), are all i need to make time and phase aligned DIY speakers.
So i may be wrong here.....
But I think max phase = inverse all pass.

I was wrong. I received an email correcting me. Max phase is the opposite of min phase, so it is like a reversed min phase. I simulated it. A minphase has the impulse peak near the beginning of the filter, maxphase has the peak at the end.

 

[gnarly]

was wrong. I received an email correcting me. Max phase is the opposite of min phase, so it is like a reversed min phase. I simulated it. A minphase has the impulse peak near the beginning of the filter, maxphase has the peak at the end.

Thanks for that follow up. I can use all the confirmation/correction I can get when I wade away from my well know technique.

 

[3ll3d00d]

Fwiw https://dsp.stackexchange.com/questions/85787/are-allpass-filters-maximum-phase-systems may be of interest

 

[mdsimon2]

It introduces additional delay. So the time alignment procedure needs to be repeated.

Can you expand on what you specifically mean here? Time alignment between the drivers needs to be repeated? Can you describe exactly what filters you are applying?

As of now I assume you are doing the following.

Case 1 - Linear Phase LR4 50 Hz Crossover
-Apply linear phase LR4 LPF @ 50 Hz to sub
-Apply linear phase LR4 HPF @ 50 Hz to woofer
-Result is linear phase behavior

Case 2 - Minimum Phase LR4 50 Hz Crossover
-Apply minimum phase LR4 LPF @ 50 Hz to sub
-Apply minimum phase LR4 HPF @ 50 Hz to woofer
-Result is all-pass phase behavior, phase is 180 deg at 50 Hz

I don't see why any time alignment would be needed unless you are referring to adding global delay to case 2 to match the delay caused by the linear phase filter in case 1. There certainly shouldn't be any need to adjust the delay between the sub and woofer.

For reference, here are some electrical measurements of the above cases, all done at 48 kHz sample rate. Measurements are combined output. For case 1 I used 16384 tap FIR filters, for case 2 I used IIR. Source is TOSLINK output of a MacBook Pro. All processing done in CamillaDSP with a MOTU Ultralite Mk5, using digital input / output. REW capture device is a UR23 TOSLINK to USB card. Right channel is used as a loopback for timing reference. For reference here is the CamillaDSP pipeline of case 1.

First looking at magnitude response, as expected both are completely flat.

Next looking at phase response. Here I removed the 170.67 ms delay from case 1 which is caused by the FIR filter to eliminate unnecessary phase wraps. Clearly case 1 is linear phase, case 2 shows all pass behavior.

Next looking at impulse response for purposes of delay. Case 2 is delayed 170.67 ms due to the FIR filter latency compared to case 1.

If I add 170.67 ms delay to case 2 to match the delay of case 1 and re-meeasure, I get the following impulse response. Impulse peaks are exactly aligned.

Michael

 

[gnarly]

Fwiw https://dsp.stackexchange.com/questions/85787/are-allpass-filters-maximum-phase-systems may be of interest

Thanks, yes that was interesting. Math was over my head, but I could follow the 'definitional' discussions pretty well.

 

[gnarly]

As of now I assume you are doing the following.

Case 1 - Linear Phase LR4 50 Hz Crossover
-Apply linear phase LR4 LPF @ 50 Hz to sub
-Apply linear phase LR4 HPF @ 50 Hz to woofer
-Result is linear phase behavior

Case 2 - Minimum Phase LR4 50 Hz Crossover
-Apply minimum phase LR4 LPF @ 50 Hz to sub
-Apply minimum phase LR4 HPF @ 50 Hz to woofer
-Result is all-pass phase behavior, phase is 180 deg at 50 Hz

I don't see why any time alignment would be needed unless you are referring to adding global delay to case 2 to match the delay caused by the linear phase filter in case 1. There certainly shouldn't be any need to adjust the delay between the sub and woofer.

For those two cases, I don't see why any time alignment would be needed either......other than like you say, to match the constant delay of the lin-phase filter, which really wouldn't have anything to do with the sub-to-woofer alone xover.