The Virtual Microphone

[ricardo]

Keith, the trepanned Soundfield you show is a Mk2.  I believe only 1 exists and is in the estate of the late Prof. Peter Fellgett.

The Mk1 was done with 4 goosenecks and Calrec CM1050 cardioid stick mikes.

The Control Box is for a Mk3 (or probably a Mk3A), the first commercial version and the first I was involved with.

Mk3s, 3As & Mk4s are of course the Holy Hand Grenades of Ambisonia (with apologies to Monty P).  No others have this Sacred status.

Yes, binaural playback eliminates a major variable. Perhaps at some point an accelerometer on the listeners head could tweak for and simulate expected combing changes from head movement, while that would probably be weird too.

Current experiments involve an Android phone taped to your headphones.

There is a 'feature' in the present software that requires afficionados to face North to reset CF.  I've suggested to the developer, that while he sorts this out, he incorporates an option to face Mecca for those of the Muslim faith. 

The thread is on the Sursound forum.
__________________

Keith, have you really got 2 'real' U47's that you believe have the vintage magic?  Are they matched?

I've never had my paws on one.

I know vintage C12s sound very close to at least one modern version of the 414.

When the 'new' transistorized 414 came out, the BBC asked us to measure them cos some of their engineers thought the C12 sounded better.  I turned it into a 'measure every mike in sight' exercise and begged & borrowed loadsa mikes.  No U47.  I think the Neumann I looked at was the stereo SM69 & maybe a U67 or U87.  Circa 1980 so my memory is hazy.

 

[SSLtech]

Ricardo, yep; only the background console in that photo belongs to work:

As for 'got the Mojo'... I'm not comfortable with making bold claims. -For one thing, there's more than one type of capsule out there: the PVC-skinned M7 is worshiped by many as the 'grail' (again, we're inevitably back to Python!) but both of these particular examples came from the factory equipped with K47 capsules, which marks them as later versions, which would also correspond to their serial numbers.

-The serial numbers are close, but not consecutive. -They're a few tens apart. The insides on one are as came from the factory, while the second one has a replaced output transformer (AMI-Oliver Archut) so they can't be described as 'matched' at all. -Oddly the AMI-equipped one sounds 'stronger' (I hate subjective desctiptions!) so sometime the second one may get an xfmr replacement, too.

 

[abbey road d enfer]

A friend who works in the robotics field is about to publish a paper on his company's artificially intelligent nanobot processor that is the size of a dime.  To him, making a digital model of frequency, harmonics, saturation, and phase is laughably simplistic.

It's not the first time I hear from a guy who's full of his own s..t importance claiming he can outdesign any piece of audio equipment while he's watching Oprah and picking his nose. Most cases it turns into a major disappointment.
The Magic Alex syndrome.

 

[gyraf]

Current experiments involve an Android phone taped to your headphones.

Aah - very interesting indeed. We've discussed this quite a lot at the neuro-cognitive research group meetings at my local university: Lack of head-movement-cues in audio while wearing headphones as a major source of listener fatigue. Think of your experiences in anechoic chambers.

Detecting the movements is easy enough these days - but the delivered audio needs to be mapped into a "virtual room" simulating some kind of a natural acoustic environment where the audio is poured out, and in which the "pickup" placement is moved to generate the needed auditory micro-cues. And macro cues - this is where it gets fun.

Turning your head will alter not only perspective, but also place you elsewhere on the local comb-filter. And sources will be louder as you move towards them.

The really interesting application (for people like us) will probably be that this technique will enable us to map/downmix surround sound into LR headphone streams at last

General consensus around here is that phone manufacturers will be quite happy to have something as complex as this to spend the new model's processing power on....

Jakob E.

 

[ricardo]

Detecting the movements is easy enough these days - but the delivered audio needs to be mapped into a "virtual room" simulating some kind of a natural acoustic environment where the audio is poured out, and in which the "pickup" placement is moved to generate the needed auditory micro-cues. And macro cues - this is where it gets fun.

Turning your head will alter not only perspective, but also place you elsewhere on the local comb-filter. And sources will be louder as you move towards them.

Ambisonics allows the creation of this 'natural acoustic environment' because an Ambisonic Recording defines (at least to 1st order though there are many higher order experimental recordings) the Soundfield at a 'point'.

These binaural experiments create a virtual Ambisonic speaker rig and allow head movements within this re-created soundfield.

A limited amount of moving towards the source is also possible though this is con as the soundfield is only captured at a point.  But it can be very effective for the correct source.  Sources do other things beside get louder as you move closer.

Even the 'comb filter' can be made innocuous by picking the right Ambisonic order and the number of speakers.

Lack of head-movement-cues in audio while wearing headphones as a major source of listener fatigue

I wouldn't describe it as listener fatigue.

But head-movement cues are BY FAR the most important in determining directions.  ALL localisation experiments show this.  Even if ALL other cues are destroyed or conflicting, the Mk1HH quickly learns to use head-movement cues to re-establish its aural view of the world.

However, a whole generation has grown up with headphone as their main music source.  I'm still not sure what the repercussions of this are except ... 2 channel sound is more important than ever making surround sound efforts even less commercially viable.

Hence the efforts with Android phones ... 

 

[SSLtech]

But head-movement cues are BY FAR the most important in determining directions.  ALL localisation experiments show this.  Even if ALL other cues are destroyed or conflicting, the Mk1HH quickly learns to use head-movement cues to re-establish its aural view of the world.

The head doesn't even have to be human.

Perhaps 15 years ago I watched how our dog determined the angle of elevation of a noisy flying insect that he couldn't SEE... through a slightly-opened garage door.

To do this, he cocked his head somewhat sideways... and when you LOOK for it, it's plain that it's an instinctive trick for any dog who is trying to localize something. (there's videos of dogs looking at laptop videos of other dogs/people and puzzling as to how they come to be on the screen... in every instance, they cock their heads at a high angle, and 'prick' their ears...)

Beyond vertical-axis rotation (to reinforce localization on the horizontal plane) the HUMAN head also tilts slightly -and instinctively- to locate sounds vertically... though since -historically- relatively little of survival interest comes from below the floor or above our heads, the angle of tilt is probably limited to about 15 degrees. -I suspect that this is why relatively small vertical axis spacing is needed in 'periphonic' arrays to suggest verticality to the Mk1HH...

I should specify that much of this is personal observation and conjecture rather than established fact, and I welcome corrective input...

Damn... we've strayed a LONG way off the topic. -Sorry. -But I do find this stuff so bloody fascinating. -Thanks to Gyraf, Ricardo, JR, Abbey and all. -THIS is a superb example of the "richness" of this forum which I've missed for a while!

 

[JohnRoberts]

The shape of the outer ear controls the pinnae (?) transforms that control the comb filtering that occurs when sounds bounce around  transiting the outer ear on the way to our eardrum. The different elevations result in different combs and sound character. This has been well studied, and I don't recall the woman's name but IIRC she had multiple papers published in the AES journals back in the '70s.

While my recollection is vague I think sounds from above and below create similar transforms, but we naturally assume above since we are generally standing on a hard boundary. Sounds from front and back experience serious frequency response differences due to shading from the outer ear naturally pointing forward.

I am sensitive to this phenomena while out jogging on roads. I can tell from the sound character of the tire noise without even seeing the car whether it is coming from in front of me or from behind. 

JR

 

[SSLtech]

I've always been 'distrustful' of the pinnae's filtration as being in any way "complete" in terms of confirming position, or solving the fore/aft duality... Not to eliminate it, but I've always seen it as 'contributive' rather than in any way complete.

The spectral modification is easily observable. We have a nice fountain outside work -presumably to mask the traffic noise from the busy road nearby- and it's a wonderfully useful broad-spectrum noise source to listen to and demonstrate the spectral differences as you turn. -Similarly, car tire noise (specially on a wet road) is nice and broad, and sufficiently decorrelated for the ear/brain to focus on generalities of spectrum, without being distracted by having to 'decode' content or meaning.

But surely it requires a prior knowledge of the spectral signature (or at least an expected spectral signature) to decide whether the received/observed spectrum is likely fore or aft... And -again- involuntary head movement is a handy and instinctive way to support/confirm/disprove any localization ambiguities.

I DO believe that imposing spectral (pinna-related) modification in addition to other cues such as amplitude is absolutely valid... specially since the use of headphones removes this effect. -But while I agree it's easy to observe or demonstrate, I'm not completely convinced that it's a dominant cue. -Again, this is my own suspicion/observation/surmising, and I thoroughly welcome further meat to chew on.

 

[Dylan W]

My understanding is that the filtration contributed by the pinnae is limited to vertical localization, as JR said, since interaural time and amplitude variations, which are the primary cues for stereo localization, don't mean much in the vertical plane.

Interesting what ricardo was saying about head movement. There have been studies done where they've taken the pinnae out of the picture and lesioned the superior olivary complex (which compares the signal from both ears and localizes the sound horizontally), and subjects still have no problem localizing as long as they're able to move their heads. I'm don't know much about what role head movement plays in a broader sense--certainly it's sufficient in the absence of other cues.

They've also done work where they come up with a new prosthetic pinnae shape, like an alien or something, and subjects adapt to the new system of filtration very quickly. Once they've adapted, they take the prosthetic on and off and can use either to localize vertically without a problem (implying that the brain creates two sets of "impulse responses" and can load them separately...)

 

[abbey road d enfer]

The horizontal and vertical localization processes are very different.
Horizontal, it's mainly the time difference between left and right ear. Vertically it's the spectral notches that count.
Apparently (and instinctively I tend to believe it), these two processes involve different "computations". It seems the vertical process is not as accurate and, more importantly, not biunivocal (different directions may give similar notches).
In order to eliminate the ambiguity, it is then instinctive to use the horizontal measurement tool and adapt it to vertical measurement, by rotating it.
Since we have lost the ability to move our ears, we must move our heads. The only exception is Chuck Norris; he stays in place and moves the universe.

 

[ricardo]

There are thought to be 5 mechanisms for Auditory Localisation

- LF ITD or 'phase' : This is the basis of Stereo as we know it and first described by Blumlein and is important up to about 700Hz
- 'HF' ILD : important over a large band probably from 300Hz to about 5kHz.  There are some strange anomalies in this as the mechanism crosses over at important mid frequencies.
- HF ITD or time of arrival : here the envelope of 'HF' is used.
- Pinnae (or Response spectrum) cues : Varies with individuals.  The ONLY cue available for vertical localisation in the median plane but requires a priori knowledge of the sound.

The above 4 can be considered 'HRTF' cues cos each are a subset of da dreaded HRTFs.  The individual size of the Mk1HH affects all 4 but seems to give problems only for the last.

- Lastly, moving head cues.  There are moving head versions for most of the above mechanisms though most are not widely known.

The Mk1HH appears to use a combination of cues to determine localisation.  If most cues agree, you have low listening fatigue and good sound.  Early surround systems often had one strong cue which wasn't confirmed by the others.  Though localisation might have been 'good' in a formal test, this results in listening fatigue.
_____________________

Gerzon's http://www.aes.org/e-lib/browse.cfm?elib=6827 has references and shows how Ambisonics attempts to reconcile these cues in a surround sound system.  It is highly mathematical but also has useful practical examples showing how the theories predict eg known stereo phenomena.

Blauert is a compilation that appeared when I was in the bush and shows much new work on the subject(s).  I only managed to glance through a copy this Millenium.
_____________________

I've been involved with experiments to determine the importance of some these cues, notably the Greene-Lee Neckbrace and the Malham-Van Gogh Cap. 

The later involves a Latex Swimming Cap.  It confirms work on prisoners that were de-pinnaed for the crime of being pseudo Golden Pinnae.

 

[12afael]

Interesting. maybe  this is part of the reason why metal fans move their heads up and down, to have a "bigger" experience.  lucky us Chuck Norris does not need to shake his head.

Keith have the most amazing history of buying vintage microphones I `ve heard.

 

[SSLtech]

Ricardo...

Maybe 'Chopper Reid' would be worth including in any future pinna-function study... As a 'control group'.

(Someone might have to help him with the bigger words though... -anything longer than "fish" for example!)

 

[milestone]

I came across this thread and thought I'd post you this picture.
If it's the only one in existence then here it is…

I bought it several years ago from an old gent called Peter. It was an eBay transaction so I didn't get his surname, but I guess it must be the prof. you mention.

 

[SSLtech]

Wow!

Whereabouts are you located? -I'm presuming in the UK, but just wondering whether in the north or south...

 

[Harpo]

The only exception is Chuck Norris; he stays in place and moves the universe.