[gregorio]

Part of the discussion in the "Testing Audiophile Myths and Claims" centred on how binaural recordings are superior to multi-mic'ed recordings because they achieve (or can achieve) a coherent sound stage whereas multi-mic'ed recordings by definition cannot. I don't entirely agree with that assertion! While that assertion is technically correct, in practice we can create an illusion with multi-mic'ing which although technically is not accurate, is good enough to fool even quite discerning ears.

I assert that one of the main reasons why many believe that binaural recording is so superior, and that multi-mic'ing is effectively viewed as a recording technique developed by an audiophile anti-christ, is simply due to an apples vs oranges comparison. The comparison of "mix" specifically/exclusively for headphones vs a mix primarily designed for speakers. A multi-mic'ed recording, mixed with a more equal emphasis/consideration of HP playback, would be somewhat more of an apples to apples comparison but such mixes are exceptionally rare, so I present a short except below for your feedback.

If possible, I would like you to provide feedback on the following considerations:
1. Does it sound realistic, somewhat like "actually being there"?
2. Does the acoustic space sound coherent or is it a bit of a nonsensical mess? Does it sound like a real space or one artificially generated by a reverb unit?
3. What sort of space (real or artificial) does it sound like? IE. A studio, a cathedral, a concert hall or somewhere else?
4. Do you perceive the sound all between your ears (IE., a flat left/right stereo image) or does it feel more in front, above or behind you?
5. If you have the opportunity, what does it sound like when you play it on speakers and what differences did you notice with HP playback?
6. Any other observations, impressions or questions you may have.

This recording was an atypical situation for me but I don't want to discuss that too much at this stage, as I don't want to prejudice your listening impressions more than I already have. I'll explain after/if there is some feedback.

Lastly, I am providing this short except with kind permission of the copyright holder, please delete it when you have finished listening to it and do not distribute it, post it elsewhere, play or give it to anyone else.

Multi-Mic_Test.zip

Thanks, G

 

[71 dB]

Part of the discussion in the "Testing Audiophile Myths and Claims" centred on how binaural recordings are superior to multi-mic'ed recordings because they achieve (or can achieve) a coherent sound stage whereas multi-mic'ed recordings by definition cannot. I don't entirely agree with that assertion! While that assertion is technically correct, in practice we can create an illusion with multi-mic'ing which although technically is not accurate, is good enough to fool even quite discerning ears.

I assert that one of the main reasons why many believe that binaural recording is so superior, and that multi-mic'ing is effectively viewed as a recording technique developed by an audiophile anti-christ, is simply due to an apples vs oranges comparison. The comparison of "mix" specifically/exclusively for headphones vs a mix primarily designed for speakers. A multi-mic'ed recording, mixed with a more equal emphasis/consideration of HP playback, would be somewhat more of an apples to apples comparison but such mixes are exceptionally rare, so I present a short except below for your feedback.

So is it so that "good enough" is enough for multi-mic'ed recordings, but not for binaural recordings? Multi-mic'ed can shine on speakers, while binaural can shine on headphones. So why not record both ways? Have a double CD where disc one is the multi-mic'ed version for speakers (and headphones with crossfeed) and disc two is and extra disc for those who want to experience the music differently, binaurally with headphones? The dummy head would be just an extra pair of mics on the multi-mic setup and used on another version of the music. Since the binaural recording is pretty much what it is and you can't do much with it, it wouldn't mean much additional work. If the recording is 40 minutes or less, you can fit both versions on one disc.

If possible, I would like you to provide feedback on the following considerations:
1. Does it sound realistic, somewhat like "actually being there"?
2. Does the acoustic space sound coherent or is it a bit of a nonsensical mess? Does it sound like a real space or one artificially generated by a reverb unit?
3. What sort of space (real or artificial) does it sound like? IE. A studio, a cathedral, a concert hall or somewhere else?
4. Do you perceive the sound all between your ears (IE., a flat left/right stereo image) or does it feel more in front, above or behind you?
5. If you have the opportunity, what does it sound like when you play it on speakers and what differences did you notice with HP playback?
6. Any other observations, impressions or questions you may have.

1. On speakers the soundstage is deep and a bit narrow. I bit more width would add to the feeling of "actually being there." On headphones excessive ILD at lower frequencies destroys some of the realism compared to speakers, but crossfeed fixes this. However, at higher frequencies some additional width could help. The recordings has a general feeling, that only the front of the listener has acoustic space while the backside of the listener is completely damped.

2. Nonsensical mess in the sense that only frontside has "sonic activity", but that also means the recording has got good depth.

3. Like a cathedral which has been turned an anechoic chamber behind the listener. An anechoic chamber inside a cathedral the front wall removed and listener inside the anechoic chamber listening to echoic sounds coming from ahead only.

4. Headphones without crossfeed: In front of me, but VERY near my face! It's like a wall of sound and my nose is touching it. With crossfeed the sound jumps away from my face and is a few feet in front of me. With loudspeakers the sound come from between my speakers, but from a larger distance, even outside my room. Nothing comes from other directions. Depth is clearly the biggest strenght of this recording.

5. see above.

6. I analysed the recording in Audacity with my nyquist plugin. It analyses "D"-measure (channel correlation expressed in another way) on octave bands for low frequencies. D = (S/(S+M)) where M and S are mid and side channels. D = 0 means mono, D = 0.5 means equal amount in M and S channels and D = 1 means that left and right channels are identical, but have opposite polarity (antimono). The results are (recommendations for crossfeed level):

D (50 Hz) = 0.47 (-2.2 dB)
D (100 Hz) = 0.48 (-3.9 dB)
D (200 Hz) = 0.51 (-7.2 dB)
D (400 Hz) = 0.51
D (800 Hz) = 0.52

A good value for D (50 Hz) is about 0.25 and gradually larger for the higher frequency D measures if headphones listening is concerned. Above 800 Hz we can have D measures as large as 0.6 bringing width to the sound.

 

[gregorio]

Thanks for your response, interesting observations. I won't respond to them for now, I'll wait to see if there are any other responses first. This one though I will respond to:

Since the binaural recording is pretty much what it is and you can't do much with it, it wouldn't mean much additional work.

Unfortunately, it doesn't work both ways. There are certain things you can do, some editing processes for example. Those things we can do, require double the work, once for the multi-mic recording and once for the binaural recording. Those things we can't do means we have a version the musicians are not as happy with and don't want released.

G

 

[71 dB]

1. Thanks for your response, interesting observations. I won't respond to them for now, I'll wait to see if there are any other responses first. This one though I will respond to:


2. Unfortunately, it doesn't work both ways. There are certain things you can do, some editing processes for example. Those things we can do, require double the work, once for the multi-mic recording and once for the binaural recording. Those things we can't do means we have a version the musicians are not as happy with and don't want released.

G

1. You're welcome.
2. Too bad… …but this is normal for me, having other people insist my ideas don't work. That's why I have serious issues with my self-esteem.

 

[bigshot]

1. On speakers the soundstage is deep and a bit narrow. I bit more width would add to the feeling of "actually being there."

That's why 5.1 has a center channel speaker to fill in the phantom center.

 

[71 dB]

That's why 5.1 has a center channel speaker to fill in the phantom center.

That and to make movie dialogue clearer from other sounds. However, I fail to see the connection to my remark.

 

[bigshot]

Adding a center speaker allows you to double the width. For two speakers 8 feet is the maximum width. For 5.1 you can increase that to 16 feet. If you were going to increase the width between two speakers to 16 feet, you would have to sit so far away it would be lousy.

 

[james444]

1. Does it sound realistic, somewhat like "actually being there"?
2. Does the acoustic space sound coherent or is it a bit of a nonsensical mess? Does it sound like a real space or one artificially generated by a reverb unit?
3. What sort of space (real or artificial) does it sound like? IE. A studio, a cathedral, a concert hall or somewhere else?
4. Do you perceive the sound all between your ears (IE., a flat left/right stereo image) or does it feel more in front, above or behind you?
5. If you have the opportunity, what does it sound like when you play it on speakers and what differences did you notice with HP playback?
6. Any other observations, impressions or questions you may have.

Finally got the chance to listen to this on my IEMs:

1. Somewhat realistic, but as @71 dB said, part of the acoustic space seems to be sonically dead.

2. Horizontally, it sounds coherent to me, with seemingly realistic placement of singers. As far as depth is concerned, I get nice forward projection (out-of-head), which is rather rare when listening to multi-mic'ed recordings with IEMs. The distance from listener to all singers seems to be the same though (which may or may not correlate with how they were really placed). So far, so good. What makes the acoustic space feel somewhat odd in my book, is that I hear no reverb from the backside, and little if any reverb from floor and ceiling. The latter making it impossible for me to get a sense of the room's height.

3. Sounds like a small church to me, or a large empty room with no furniture. Just going by left-to-right info and depth, since there hardly seems to be any height or rear info.

4. Several feet in front of me, with pretty realistic stage width.

 

[bigshot]

I think room ambience alone is enough to get your brain to place things a couple of feet ahead. I haven't heard anything with binaural that sounds like speakers or a real performance on stage in front of me.

 

[gregorio]

OK, as that seems to be it, I'll explain the recording and address some of the points/observations made:

As I mentioned in the OP, this was a highly atypical recording for me, because in many respects it was not a professional recording, it was a favour for a friend and presented many constraints which would not exist if this were a professional recording. This is a recording of ancient ecclesiastical music but could not be recorded in a church, it had to be recorded in a particular school and only 1 hour 45 minutes in total was available for setup, soundcheck, the actual recording of 9 songs (29 minutes of music) and all overdubs. Although, I had done a recce the day before and decided on what setup I would use and where in the school to make the recording. All the medium classrooms available of course sounded like medium class rooms and nothing at all like a church but as the recording occurred in the evening, the main entrance hall was available. It wasn't actually a hall though, it was essentially a large corridor, about 20ft wide and 40ft long with a ceiling about 12ft high, all hard reflective surfaces. Across the back of this entrance hall/corridor was an open stairwell leading to the first floor, providing an area about 10ft deep roughly 40ft or so in height, again, hard reflective surfaces.

The choir consisted of 10 professional musicians and music teachers but all were amateur singers: 1 male bass and 9 females, split (effectively) into two groups of sopranos and a couple of altos. Not an ideal split of voices by any means but not entirely inappropriate for the genre. I placed the choir in a line (actually halfway between a line and a semi-circle) across the back of entrance hall, just in front of the stairs, facing the main entrance. I setup a stereo pair in the entrance hall about 15fit in front of the middle of choir, individual spot mics targetted at the bass and the altos and another a stereo pair halfway up the stairs, pointing upwards. The reasons for this setup were:

1. The stereo pair in the entrance hall would act as the main pair, although the acoustics/reverb of this corridor were rather poor, it was too "slappy" (not well diffused), had too short a pre-delay, too short a duration (RT) and was rather too bright.
2. The two altos and one bass were likely to be over balanced by the seven sopranos, mic'ing them individually provided the option during mixing of raising their level somewhat, independently of the sopranos and achieving a better balance.
3. The stereo pair on the stairs allowed the recording of the stairwell's reverb, which was far more diffuse, smoother, longer and less bright than that of the entrance hall. This provided the opportunity of mixing this stairwell reverb with the main stereo pair, reducing the problem of the slappy entrance hall acoustics and hopefully resulting in a reverb more representative of a church than a school entrance corridor. Hence why the choir was positioned next to the stairs, so more of the sound would spill into the stairwell and excite it's acoustics.

From the above then:

A. There was too much height, side and rear reverb recorded than would have existed in an actual church, because the ceiling, side and rear reflective boundaries were far closer to the main stereo pair than would be the case in a church. This is the opposite of the stated observations! Which leads me to speculate that the observations may be influenced by an expectation that the multi-mic'ing setup would ONLY be individual close mics and therefore, height and rear reflections would not be recorded at all or at a very low level?!

B. Neither of the responses suggested the recording venue was a school entrance corridor, both suggested a church or cathedral type location, albeit compromised in some way. This indicates that the manufactured illusion was at least somewhat successful, even for those who have expressed a prejudice against multi-mic'ing!

C. Depth perception: Almost all of what I personally perceive when using headphones to listen to music occurs inside my head between my ears, not in front or behind me. This is also true of binaural music recordings, although there is usually a greater sense of depth due to the additional reverb, however, I also get that greater sense of depth with just an ordinary (non-binaural) recording with additional reverb.

D. Width perception: The width perception on HPs is equivalent to standing a couple of feet or so in front of the mic position (closer to the choir), IE. Slightly wider than what would have been perceived had you stood where the mics were positioned. On speakers, it's somewhat narrower, equivalent to being several feet behind the mic position. Of course though, that would depend on the relative positioning of the listener's speakers. I would generally prioritise the speakers and create a somewhat wider mix but a greater consideration of HP playback was requested for this recording than is usual and therefore the intended/desired width is represented quite well with HPs.

F. I'm afraid the "excessive ILD" and crossfeed comments are indicative of an obsession with ILD and crossfeed, rather than the actual facts! The L/R level differences on this recording are, if anything slightly lower than would be experienced at the actual live event. FURTHERMORE, the recording venue was not ideal, it had thick external walls but was moderately close to a road, which resulted in a considerable amount of low frequency traffic rumble. As there was no musical content below about 120Hz, a high pass filter was applied during mixing at that frequency to reduce/remove that unwanted LF noise. Yet apparently, crossfeeding the remaining low level LF noise makes the recording sound more "real"? You can't mean a more "real" music performance because there is no music down there, you must mean a more "real" traffic rumble? I would definitely NOT recommend the use of crossfeed on this recording! ... Due to the addition of the stairwell reverb, the recording already has as much crossfeed on headphones as intended/desired.

G

 

[james444]

A. There was too much height, side and rear reverb recorded than would have existed in an actual church, because the ceiling, side and rear reflective boundaries were far closer to the main stereo pair than would be the case in a church. This is the opposite of the stated observations! Which leads me to speculate that the observations may be influenced by an expectation that the multi-mic'ing setup would ONLY be individual close mics and therefore, height and rear reflections would not be recorded at all or at a very low level?!

As far as I'm concerned, nope (to the bolded part). I simply lack the knowledge about multi-mic'ing techniques to even come up with such expectations. In fact, I was comparing your recording to one of @analogsurviver's JD recordings of a choir, and felt that the latter gave me a pretty good sense of the room's height, whereas your recording just didn't. An entirely subjective observation, of course. But no speculation involved in my case.

 

[gregorio]

[1] I simply lack the knowledge about multi-mic'ing techniques to even come up with such expectations.
[2] I was comparing your recording to one of @analogsurviver's JD recordings of a choir, and felt that the latter gave me a pretty good sense of the room's height, whereas your recording just didn't.
[3] An entirely subjective observation, of course.

1. Even if you don't know the specifics of multi-mic'ing, you have expressed the opinion (in the past) that binaural recordings typically provide you with much more "real life" spatiality than multi-mic'ed recordings, so there is an obvious "expectation", given that you knew from the outset that this was a multi-mic recording. How much that expectation influenced your observation is impossible to say but it's entirely reasonable to speculate that it had some influence. It's also possible that 71dB's comments had some influence. If you'd done your critical listening before you'd seen 71dB's comments and if you'd been told (and believed) it was a binaural recording, would your observations/impressions have been exactly the same? Probably not. Would they have been substantially different enough for you not to notice/realise that it wasn't an actual binaural recording? Possibly, but we'll never know for sure.

2. A JD recording would provide no more height information than the setup I used and would provide significantly less rear information. Of course though, there are several major differences between my recording and analogsurvivor's, not least of which is that I was not attempting to record/reproduce the performance acoustics, I was attempting to manufacture the illusion of a completely different acoustic to the one which actually existed, a feat which would be impossible with only a JD stereo (or a binaural) recording. So it's an apples to oranges comparison!

3. Ultimately, that is the problem. You heard the choir several feet in front of you, I don't. You thought it sounded somewhat like a small church, it isn't. It seems to you there isn't any height information but if anything there was too much. You didn't hear any rear reverb, neither do I but there is more than would exist in a typical actual small church recording and I don't hear much/any rear reverb on any binaural music recordings and I didn't hear any on the JD recording. Your observations are not particularly accurate but neither are they necessarily wrong either. Particularly when it comes to HP listening, tiny differences in the FR and other characteristics of the HPs, their position on your head, the size, shape and exact position of your pinna, your head width and absorption characteristics, all make a difference to how your brain interprets the information output by your HPs. That's why even actual binaural recordings only work for some people and different binaural recordings work better or worse than others. In other words, I'm not saying your observations on the example I posted are wrong, they are right for you but how many other people are they applicable to, especially if those other people were not aware about how it was recorded? This is what I mean by "ultimately that is the problem", as music engineers/producers we are always subject to our own subjective observations and trying to imagine what the subjective observations of the majority of our listeners will experience.

The example I posted wasn't trying to directly compete with binaural recording, just demonstrate that a multi-mic'ed mix could be created to produce a coherent soundstage on HPs. My example was hobbled by the fact that it still had to work well on speakers and by the rather extreme and highly atypical situation of trying to create a coherent soundstage that was deliberately and very significantly different to the one which actually existed at the live event. Obviously for you and 71dB it wasn't successful (I never thought it would be successful for 71dB!), it would have been interesting to see a few more responses though, and see if even despite it's hobbling factors it created a good enough illusion of a coherent soundstage for some/many. For me personally it doesn't but then my subjective observation is highly biased, I know for a fact that it's not a coherent soundstage or indeed even a representation of any actual/real soundstage.

G

 

[71 dB]

F. I'm afraid the "excessive ILD" and crossfeed comments are indicative of an obsession with ILD and crossfeed, rather than the actual facts! The L/R level differences on this recording are, if anything slightly lower than would be experienced at the actual live event. FURTHERMORE, the recording venue was not ideal, it had thick external walls but was moderately close to a road, which resulted in a considerable amount of low frequency traffic rumble. As there was no musical content below about 120Hz, a high pass filter was applied during mixing at that frequency to reduce/remove that unwanted LF noise. Yet apparently, crossfeeding the remaining low level LF noise makes the recording sound more "real"? You can't mean a more "real" music performance because there is no music down there, you must mean a more "real" traffic rumble? I would definitely NOT recommend the use of crossfeed on this recording! ... Due to the addition of the stairwell reverb, the recording already has as much crossfeed on headphones as intended/desired.

G

Indicative of obsession with natural spatiality with headphones. The ILD levels of this recording are totally ok for speakers, but not for headphones without crossfeed. I confirmed this both with my "ILD-trained" ears and technical analyse, D-measures. These ILD levels are totally typical for music recordings so there is no need to feel bad. Even if everything below 120 Hz is traffic noise, that noise should not have "excessive ILD" either and reverberation is not crossfeeding. In fact adding reverb can increase ILD.

People should try crossfeed themselves and decide. My own ears and technical analyse suggest crossfeed is needed for best results.

 

[gregorio]

[1] Even if everything below 120 Hz is traffic noise, that noise should not have "excessive ILD"
[2] I confirmed this both with my "ILD-trained" ears and technical analyse, D-measures.
[3] reverberation is not crossfeeding. In fact adding reverb can increase ILD.
[4] The ILD levels of this recording are totally ok for speakers, but not for headphones without crossfeed.
[5] People should try crossfeed themselves and decide. [5a] My own ears and technical analyse suggest crossfeed is needed for best results.

1. Everything below 120Hz has been filtered. Therefore at say 50Hz, the level is at least 40dB or so lower than the musical content and of course below 120Hz is an area of least sensitivity for our hearing anyway, so we're talking about a tiny, very low level and inconsequential ILD. "Tiny" and "inconsequential" are at the opposite end of the scale to "excessive" and therefore using the term "excessive" so inappropriately is a pretty good definition of an "obsession"!!

2. Am I understanding this correctly? You've trained your ears to detect a few dB of ILD at extremely low levels but those trained ears couldn't tell the difference between traffic noise and a choir singing? Training your ears for that level of ILD, while being insensitive to the actual sound/music you're listening to is about as indicative of an obsession with ILD as anything I can even imagine!!

3. Yes, some types of reverb can increase ILD, other types can reduce it and do in effect create crossfeed. In this case of course I know exactly the type of reverb AND how it's been added, and therefore that it is NOT increasing ILD but reducing it!

4. The amount of ILD in this recording is EXACTLY AS INTENDED for headphones AND very close to what actually existed at the live event! Therefore the ILD levels of this recording are TOTALLY ok for headphones!!

5. Yes, people should try crossfeed for themselves. Maybe there are others like you who are obsessed with crossfeeding, like reducing fidelity and getting further away from the artistic intentions of the recording?
5a. By "best results" you mean best results for someone obsessed with crossfeeding, regardless of whether in reality that's actually "worse results"! For "best results", in terms of artistic intent and fidelity, including the fidelity of the acoustics/"spatial information", then not only is crossfeed NOT "needed", it's actually detrimental!

All you've achieved with your response is indicated even more strongly that you are obsessed with ILD and crossfeeding!!

G

 

[james444]

Obviously for you and 71dB it wasn't successful...

As I said before, I'm glad you managed to arrange this test. To me, it provided a successful out-of-head experience, which I rarely get with IEMs. It was less successful in terms of realistic spatiality, which means I couldn't get a clear idea of the environment and placement of singers. Could be due to the recording / mixing technique, but could also be, because I never actually heard a real choir sing in a school entrance hall, so there was simply no information in my brain to pattern match this recording to.

As for those top / bottom / rear reflections and how they compare to @analogsurviver's JD recordings, his samples are on this page, so anyone interested can go listen and compare / judge the spatiality for themselves. I have the Mozart Requiem and the "VAL" CD, and all I can subjectively say, is that the spatiality on both recordings works for me (even though @analogsurviver himself insists it doesn't work with IEMs!). May be just plain luck with my HRTF, or the fact that I've heard several real choir concerts in churches, so the pattern matching might just work in that case. I honestly don't know, but I personally have no skin in the multi-mic vs. binaural game, so I try not to overthink things and rather just take them as they come.

So thanks again for the opportunity, regardless of the outcome, I felt it was an interesting experiment!