Bass and subwoofers

[Sokel]

My point is that there seems an implication that there can be major, difficult if not impossible to correct, imbalances in frequency response in that part of the spectrum so the image presented is inevitably skewed by that. If the q is assessing the degree to which stereo bass is perceptible, surely you have to correct for this, just as you correct for volume differences in other such tests, otherwise you are not just testing stereo Vs mono.

Put another way, how do you even test if what you hear is from stereo Vs mono if you can't normalise the level/response?

The implication goes the other way around as well.

The correction for summed lows is based on the assumption (not any more after all this analysis though) that recordings are indeed mono in amplitude and phase, which are not.

Other signals than the traditional sweeps may reveal another another set of imbalances, as phase differences (like in music) can work destructive or not when played together.

Professional acousticians use very different signals when fixing a room (and laugh at our amateur endeavors) .
As with any other test the results depend on the signal quality and suitability.

And another.
The fact that some findings that do not correlate with the (established but poorly documented) "principles" get so much push back should be alarming in a science based forum. People should jump on testing, not theorizing or defending their time and investment when turns out not ideal.

Just like the FSAF measurement for example, which reveals concerning results at smaller gear so people avoid it like plague.

We should be glad we have the chance to search for ways for the "holy" "as the artist indented" even to a smaller degree and without all the pains and cost of a room with 200 liters flush-mount monitors.

 

[3ll3d00d]

The correction for summed lows is based on the assumption (not any more after all this analysis though) that recordings are indeed mono in amplitude and phase, which are not.

Not really, what you call correction for summed lows in a just correction for a single channel which is the same as for every single channel in the entire system. Whether such an approach is complete when content as a whole is multichannel is another question across the entire spectrum.

You didn't answer the question though, which has nothing to do with pushing back on the concept, it's purely a practical concern. To reliably test whether what you hear is stereo bass, you must need to normalise the frequency response from each channel otherwise you might just be listening to those imbalances. Was this done for any test reporting a difference? I didn't see it mentioned so it's unclear if/how this was done in a real room.

 

[NTK]

I don’t think it’s that simple. For mono bass to be effective against a given room mode you would need to locate the L and R subs in high pressure zones of different polarities for that mode. What happens then if you switch to the «stereo preset»? If the signal was truly stereo, we can imagine the sound hard panned to, say, L. This would excite the mode. It would be even worse if L and R played the same offending frequency exactly out of phase.

Most rooms in real life aren't rectangular cuboids, and you have next to no idea on what the mode frequencies are and the locations of their correspopnding modes and nulls. Multi-sub optimization schemes such as MSO or Harman's SFM can get good results with subs that are not located at the (usually unknown) "optimal" locations.

As for interference, there is really not much one can do. If you equalize your subs "cooperatively", the EQ will likely be suboptimal for uncorrelated signals. So all you should probably do is EQ each channel independently, which would then be likely suboptimal for the correlated mono bass signals. The stereo bass proponents say maximally flat amplitude response is not the be-all and end-all.

Source: https://www.audioholics.com/room-acoustics/history-of-multi-sub-sfm

As for

 

[Sokel]

Not really, what you call correction for summed lows in a just correction for a single channel which is the same as for every single channel in the entire system. Whether such an approach is complete when content as a whole is multichannel is another question across the entire spectrum.

You didn't answer the question though, which has nothing to do with pushing back on the concept, it's purely a practical concern. To reliably test whether what you hear is stereo bass, you must need to normalise the frequency response from each channel otherwise you might just be listening to those imbalances. Was this done for any test reporting a difference? I didn't see it mentioned so it's unclear if/how this was done in a real room.

Correct me if I'm wrong, but the prevailing way of correcting, is using L+R signal together down low.
Other methods do use individual test signals for each.

@NTK post above sums it.

As for testing, two full range speakers and test signals like @bmc0 just above (and lots of others scattered between the three threads about it) can do the trick for on the fly testing.

 

[Flaesh]

pains and cost of a room with 200 liters flush-mount monitors.

As the owner of a pair of compact 40-liter (internal volume, plus a waveguide on top) monitors, usually placed in corners (but I can almost easy arrange everything differently and listen to Chesky's samba (which I'm generally indifferent to)), and four (for now) 200-liter (currently wall-mounted) subwoofers, I can say I haven't experienced any pain, suffering, or significant losses from all of this. I'll probably add another four 200-liter subwoofers this summer. IB subwoofers are even cheaper and less painful (if possible).

I'll respond to Keith's message later.

BTW, I would be sincerely grateful to anyone who explores the envlopment and stereo bass, for example, in F. Dupré's "Ruins"

 

[Sokel]

As the owner of a pair of compact 40-liter (internal volume, plus a waveguide on top), usually placed in corners (but I can almost easy arrange everything differently and listen to Chesky's samba (which I'm generally indifferent to)), and four (for now) 200-liter (currently wall-mounted) subwoofers, I can say I haven't experienced any pain, suffering, or significant losses from all of this. I'll probably add another four 200-liter subwoofers this summer. IB subwoofers are even cheaper and less painful (if possible).

I'll respond to Keith's message later.

The "pain" goes for the room and its proper construction mostly, as to be both functional and nice.
But yes, in general, anything goes.

 

[markstein]

You can EQ the peaks for each channel separately, but you’re stuck when it comes to nulls during playback of recordings that have mono bass baked in.
I don’t let that worry me, though. For me, once the peaks are tamed with EQ, the system becomes quite listenable and enjoyable. Seems I’m not as picky as some.

I have recently changed from a 2.2 system with 2x mono subwoofers to a 2.0 system (no subs), partly cardioid (Ascilab C8C). After some experimentation and listening I am no longer trying to equalize the (stereo) bass perfectly to a target, just pushing down the 2-3 major modal peaks below 150Hz. The result sounds much better to my ears than the graphs might suggest.

 

[Flaesh]

much better to my ears than the graphs might suggest.

What graphs exactly?) And how exactly is it better? And I don't think that all\every listener should have preferences identical to the 54% statistical preferences..

 

[janbth]

Most rooms in real life aren't rectangular cuboids, and you have next to no idea on what the mode frequencies are and the locations of their correspopnding modes and nulls. Multi-sub optimization schemes such as MSO or Harman's SFM can get good results with subs that are not located at the (usually unknown) "optimal" locations.

OK, I was thinking about one specific mode manipulation technique, like the one described in Sec. 14.5.4, «Reduce the Energy Delivered to Bothersome Modes by Using Combined Coupling of Two Sources to Mode», in Floyd Toole’s Sound Reproduction, 4th Edition. I could have been more clear about that. There are of course other techniques, like MSO or SFM, where sub localization can be more flexible.

As for interference, there is really not much one can do. If you equalize your subs "cooperatively", the EQ will likely be suboptimal for uncorrelated signals. So all you should probably do is EQ each channel independently, which would then be likely suboptimal for the correlated mono bass signals. The stereo bass proponents say maximally flat amplitude response is not the be-all and end-all.

I agree with this, but I would go a step further: Those of us who suffer from nasty room modes will most likely not be able to achieve AEV, neither by applying EQ or otherwise. I could be wrong, of course, but I haven’t seen any convincing reports from people with both AEV and problematic room modes.

 

[bmc0]

All four combinations are attached:

1. Cross-correlation of 1 at all frequencies (mono_hf_mono_lf)
2. Cross-correlation of 1 above 80Hz and ~0 below 80Hz (mono_hf_uncor_lf)
3. Cross-correlation of ~0 above 80Hz and 1 below 80Hz (uncor_hf_mono_lf)
4. Cross-correlation of ~0 at all frequencies (uncor_hf_uncor_lf)

I decided to experiment with creating the above signals in a different way: directly in the frequency domain. This ensures identical magnitude spectra. I've also increased the slope to 15th order (90dB/oct) and additionally generated signals with a 60Hz crossover frequency.

 

[markstein]

What graphs exactly?) And how exactly is it better? And I don't think that all\every listener should have preferences identical to the 54% statistical preferences..

I am referring to the SPL graph specifically. With my mono subwoofers and MSO optimization I tried an as smooth as possible SPL response. With my current stereo setup, I allow for a little bit irregularities and just reduce the 2-3 main peaks due to room modes below 150Hz. I concentrate on left and right channel eq separately and do not care too much about the L+R SPL curve which has some deep nulls. The result sounds full, without being too "thick" which could easily happen with my subwoofers. REW auto EQ to a target curve sounds less pleasing to my ears - either too "thin" or too "thick", less spacious.
I was always wondering how to optimize the transition area (80-150...200Hz) between subwoofers and mains. Using 4x subwoofers you can very much optimize the sub 80Hz area, the 80-150...200Hz area remains a mess in many setups. With my stereo setup I do not care at the moment, as everything below 200Hz looks like a mess (but sounds ok) ;-)

 

[bmc0]

One more collection of test signals. This one is a vaguely pipe organ-esque sound produced by filtering Gaussian white noise through a parallel bank of narrow 4th-order bandpass filters, each having a bandwidth of ~4Hz. Given uncorrelated noise, the interchannel level and phase vary randomly at a rate corresponding to the filter bandwidth. The pitches are the notes of an A0 minor triad (A0=27.5Hz, C1=32.703Hz, and E1=41.203Hz), plus the 2nd, 3rd, 4th, and 5th 2nd, 4th, 8th, and 16th harmonics of each fundamental (decreasing in level as frequency increases). Three files are included in the zip:

• a0_minor_triad.flac — uncorrelated at all frequencies
• a0_minor_triad_mono_lf_65Hz.flac — pitches 65.406Hz (2nd harmonic of C1) and below are mono
• a0_minor_triad_mono_lf_82Hz.flac — pitches 82.406Hz (2nd harmonic of E1) and below are mono

Edit: Oops, I wrote the script for this a while ago and evidently forgot exactly what I did. The harmonics are actually octaves above the fundamental—the 2nd, 4th, 8th, and 16th harmonics, in other words.

 

[levimax]

One more collection of test signals. This one is a vaguely pipe organ-esque sound produced by filtering Gaussian white noise through a parallel bank of narrow 4th-order bandpass filters, each having a bandwidth of ~4Hz. Given uncorrelated noise, the interchannel level and phase vary randomly at a rate corresponding to the filter bandwidth. The pitches are the notes of an A0 minor triad (A0=27.5Hz, C1=32.703Hz, and E1=41.203Hz), plus the 2nd, 3rd, 4th, and 5th 2nd, 4th, 8th, and 16th harmonics of each fundamental (decreasing in level as frequency increases). Three files are included in the zip:

• a0_minor_triad.flac — uncorrelated at all frequencies
• a0_minor_triad_mono_lf_65Hz.flac — pitches 65.406Hz (2nd harmonic of C1) and below are mono
• a0_minor_triad_mono_lf_82Hz.flac — pitches 82.406Hz (2nd harmonic of E1) and below are mono

Edit: Oops, I wrote the script for this a while ago and evidently forgot exactly what I did. The harmonics are actually octaves above the fundamental—the 2nd, 4th, 8th, and 16th harmonics, in other words.

It is pretty easy to tell the mono tracks from the original as the LF sound is noticeably louder on the mono tracks. When you "sum to mono" did you allow for the 3 dB boost? If yes then these tracks show "sum to mono" certainly sounds different, can't say for sure if I was enveloped or not though.

 

[quattro98]

Feel free to delete this post, but wanted to share this.

R.I.P Prof. Jens Blauert

One of the pioneers of psychoacoustics has passed away, Prof. Dr-ing Jens Blauert. He started scientific research on binaural listening, HRTF, perception of ambience and understanding localization as early as of 1969. His theory on preferred elevation angles for boosted narrow frequency bands...

www.audiosciencereview.com

Jens Blauert - Wikipedia

en.wikipedia.org

Obituary for Professor Jens Blauert

With great sadness, we bid farewell to Professor Jens Blauert, who passed away on March 14, 2026, at the age of 87.

www.oeaw.ac.at

 

[bmc0]

It is pretty easy to tell the mono tracks from the original as the LF sound is noticeably louder on the mono tracks. When you "sum to mono" did you allow for the 3 dB boost?

No summing to mono was done; the bandpass filters corresponding to the lowest pitches were instead fed with correlated noise having the same level as the uncorrelated noise used for the higher frequencies. The unweighted RMS levels are within ±0.1dB between the three files. Spectra of the uncorrelated and mono-below-82Hz signals:

I don't measure a gross level discrepancy nor do I hear one using my speaker setup.

 

[levimax]

No summing to mono was done; the bandpass filters corresponding to the lowest pitches were instead fed with correlated noise having the same level as the uncorrelated noise used for the higher frequencies. The unweighted RMS levels are within ±0.1dB between the three files. Spectra of the uncorrelated and mono-below-82Hz signals:
View attachment 524884
I don't measure a gross level discrepancy nor do I hear one using my speaker setup.

OK, not sure if I completely understand how the files were made but I got 10 of 10 right in foobar ABX and I make no golden ear claims. I heard something in the very low bass that sounded different and once I keyed in on that it was relatively easy. I could not reliably hear any difference in one of the noise files you posted earlier.

 

[Flaesh]

Maybe he could share his measurements if we ask nicely

No need for asking ; I'll publish it someday. I hope to do so in the foreseeable future, but for some reason the subwoofers aren't connected right now. And so far there hasn’t been time (in particular yesterday) to do it as it was. I had some old measurements taken at 5-6 points near the MP, but I don't remember where the file is. I'm also curious to see how much the frequency response varies in different locations.
Let me remind that this is just 2+2 subwoofers in a living room with 4 windows, 3 doors, and some furniture.

ps: Real life is interesting, but anyone can simulate it in REW in a couple of minutes

I was too lazy-busy to do more than 1+1 sub.

 

[bmc0]

the LF sound is noticeably louder on the mono tracks

The unweighted RMS levels are within ±0.1dB between the three files.

Well, it turns out I made a silly mistake... The levels of L and R are virtually identical when analyzed individually, but the LF in the mono ones is indeed louder when reproduced over loudspeakers (assuming fairly closely matched channels which sum mostly coherently for mono signals). The levels are correct in the case of headphones.
As to why I didn't hear this, I guess I can only blame seasonal allergies currently compromising my ability to hear LF properly...

 

[levimax]

Well, it turns out I made a silly mistake... The levels of L and R are virtually identical when analyzed individually, but the LF in the mono ones is indeed louder when reproduced over loudspeakers (assuming fairly closely matched channels which sum mostly coherently for mono signals). The levels are correct in the case of headphones.
As to why I didn't hear this, I guess I can only blame seasonal allergies currently compromising my ability to hear LF properly...

Interesting, seems to me like the "circle of confusion" for "sum to mono" has a lot of confusion. Electronic summing is different than speaker summing is different than headphone summing.

 

[bmc0]

Electronic summing is different than speaker summing is different than headphone summing.

Indeed, the "correct" level adjustment when summing to mono depends on both the input signal and the reproducing system. For loudspeakers, it's often frequency dependent to a considerable degree even at very low frequencies because of the interaction with the room.