How audible is LFE phase shift?

[krabapple]

(Normally I'd google it, but a very quick try turned up nothing. )

What I'm talking about is something being obseved in the quadraphonicquad forum:
quite a few surround sound releases (but certainly not all) have what looks to be an error in their LFE contents. These are errors printed into the audio, visible in waveform view, not playback errors

The LFE content is delayed between a few and as much as 10-20 milliseconds compared to the other channels. As if during production there was a variable latency issue confined to the LFE, that was not caught by quality control.

QQ members there are manually (or using software) doing 'fixes' on the audio to bring the LFE into phase.

What I'm wondering is how bad this has to be, to be audible.

Things I know:
I know that phase discrepancy audibility is frequency and magnitude dependent
I know that LFE channels I have analyzed have contained anything from just low bass thuds, to full 20Hz-20kHz content (which is insane, but it happens).
I know the observed LFE latencies are typically between low single digit and low double digit (milliseconds).
I know a typical home system (i.e. with an AVR as a head unit) will always apply a low pass filter to the LFE channel before output , usually with a 125 Hz set point (as per Dolby spec).

So, knowing all that....?


Here's some examples:

 

[DVDdoug]

Interesting observation...

The main "danger" is if the subwoofer & main speakers are out-of-phase in the crossover range where they are both operating. In that case the soundwaves will cancel, or partially cancel depending on the phase difference.

Many subs have a phase reversal (polarity) switch and some subwoofers and room correction software have a continuously variable phase/delay adjustment. But that's mainly to compensate for speaker location and it's not going to help if there's an issue in the recording....

 

[krabapple]

Yes, the issue is in the recording, I hope I was clear on that.

(I imagine a phase reversal switch on the sub would only work, if at all, if the subwoofer input was purely LFE content. But typically it's LFE + lowpassed bass from other channels. And if a variable phase control worked, it would have to be readjusted on an album-by-album basis...sometimes even track-by-track)

The remedy being applied over on QQ is to import the audio into software, nudge the LFE leftward until it's in phase, then save the corrected file. An intreprid member there has even automated the process.

I'm wondering HOW AND WHEN IS IT AUDIBLY NECESSARY? That's the topic of this thread.


FWIW, total OOP occurs too, and not just in LFE channels.
Here's Beck's Sea Change in 5.1. The front L/R are completely OOP with the C and surround L/R. And the LFE is a little delayed.

 

[Cogito]

Is this a result of mastering errors or component processing? If using with a wireless subwoofer setup, this issue can be compounded with more introduced latency.

 

[krabapple]

Again, IT IS PART OF THE RECORDING. It's not a playback or 'component processing' issue.

These are waveforms of audio ripped directly from the discs.

 

[krabapple]

This is a question for audio science , not about how this happened -- the question is about how audible it is.



(Though interestingly, or not, the front channel total OOP issue shown on the Sea Change example, only occured on the SACD or BluRay of the mix. The DVD-A did not show it.
All of them have the delayed LFE, though.
SACD was the first format this mix was released on, in 2002. DVD-A was in 2003. Blu-Ray in 2013)

 

[Cogito]

Again, IT IS PART OF THE RECORDING. It's not a playback or 'component processing' issue.

These are waveforms of audio ripped directly from the discs.

Sorry, your clearification posted while I was responding.

 

[dannut]

Here's some examples:

Those examples could be misleading. LFE is lowpassed and others are full range. You can't just put a marker at the signal peak and declare that those should be aligned.

Mostly agree - music production practices are.. all over the place. Mixing same signal to LFE and front channels and not expecting problems with bass management is just one of them.

 

[Keith_W]

Let's see what Toole has to say about LF phase distortion.

He states that phase shift in low frequencies may potentially be audible, although he is generally sceptical about the idea. In section 4.8.2 of his book, he notes that minimum phase high pass filters, which have the effect of rotating the phase of low frequencies, are applied at multiple points in the music reproduction process - starting at the microphone, through various electronic devices, and the mixing process. The loudspeaker itself also rotates phase, and so do reflections in a listening room. The signal is already corrupted by phase distortion even before we attempt to play it back. He says that the cumulative phase shift must be “enormous”. So the question really is - does phase correction of the last part of the chain - the loudspeaker and listening room - have much of an effect?

Toole reported that he attended a demonstration that demonstrated audible low frequency phase distortion. However, this was heard with specially recorded music and a contrived signal, and even then he reports that the difference was “quite subtle”. He then cites several studies, but notes that none of them were controlled, double-blind listening tests.

The only psychoacoustic study I was able to find (that was not mentioned by Toole) was published by Liski et al: link. Note that the test was done with test tones and headphones. Loudspeakers and group delay were simulated with DSP. Regrettably, low bass or thresholds > 10ms was not studied. Examination of the published curves among the three speakers simulated (illustration above) shows that any trends above 10ms at 40-50Hz were not studied and the curves had an exponential appearance towards the low frequencies. Note the limitation of this study - test tones, and headphones only. Not real music (which already has phase distortion built-in), and not in a listening room (which will add even more phase distortion).

As far as the evidence I have found goes, I would say that the jury is out on this one. For sure some degree of phase distortion would be audible, e.g. if the group delay of your subwoofer was an hour behind your main speakers. You would surely hear that!

I would like to see what @j_j thinks about this.

 

[krabapple]

Those examples could be misleading. LFE is lowpassed and others are full range. You can't just put a marker at the signal peak and declare that those should be aligned.

If it's visually consistent and obvious over the entire length of the track, why not?

(And , there are plenty of releases that don't show any such apparent delay)

(And, a poster had demonstrated how incorrectly using a filtering plug-in can product a very similar effect

None of which gets to my actual question: when is it audible?

 

[krabapple]

Let's see what Toole has to say about LF phase distortion.

Thank you!

Prediction of what one might hear at home is complicated when systems use bass management, which sums lowpassed bass + LFE as the subwoofer input. The LFE delay, if it exists, is baked into the source audio. The lowpassed bass presumably isn't delayed, so if the LFE is delayed, there will be phase interference already at subwoofer input.

 

[dannut]

None of which gets to my actual question: when is it audible?

• If the LFE signal is correlated to the main channels and delayed, the bass management will make the situation more interesting.
• If the LFE is an uncorrelated source - no problem even with large delays.
• If the LFE is a portion of a signal from the 'main' channels, (lowpassed goes to LFE, highpassed to main and the associated group delay) - there could be audible timing issues. Just like adding a +1 way to a speaker system. Difference with point nr1 is that there are no cancellation issues in the overlap region.

It all depends how the production is done. All over the place.

 

[Keith_W]

BTW, I have heard JJ say in other threads that phase distortion of more than 15 degrees per ERB is audible.

The ERB can be calculated with 24.7 + (0.108 * f). So if f = 100Hz, the ERB is 35.5Hz.

15 degrees of phase shift at 100Hz:
- 1 period (360/360) = 1/100 seconds
- Therefore 15 deg of phase shift is 15/(360*100)

The result is 0.42ms which seems suspiciously low to me unless my maths has gone horribly wrong.

 

[krabapple]

BTW, I have heard JJ say in other threads that phase distortion of more than 15 degrees per ERB is audible.

View attachment 406179

The ERB can be calculated with 24.7 + (0.108 * f). So if f = 100Hz, the ERB is 35.5Hz.

15 degrees of phase shift at 100Hz:
- 1 period (360/360) = 1/100 seconds
- Therefore 15 deg of phase shift is 15/(360*100)

The result is 0.42ms which seems suspiciously low to me unless my maths has gone horribly wrong.

(for those reading along, ERB = Equivalent Rectanglar Bandwidth = an approximation to the bandwidths of the filters in human hearing, -https://ccrma.stanford.edu/~jos/bbt/Equivalent_Rectangular_Bandwidth.html)


NB 100 Hz is at the upper end of most subwoofer input/output, which is typically rolled off starting at 120 Hz
(100Hz is also at the upper end of much -- but certainly not all -- LFE contents themselves).

This thread also addresses phase shift audibility