The Virtual Microphone

[okgb]

I suppose you could , " add " phase problems to an existing signal to make it sound different ?

 

[Steven Slate]

I have to be careful about replying to this without coming across as dismissive, however this forum is a more focused on the technical aspects than most, and the last reply was very vague... -I 'get' not revealing proprietary information, but -to quote "Scotty" (James Doohan) from Star Trek:

"Ye cannae change the laws of physics"

Not even with sales-talk.

Again, I don't doubt the cleverness of the product, or the people involved... but the laws of physics suggest that the information collected by these microphones is insufficient to execute a more complete simulation... no matter HOW immensely clever the people involved.

..unless someone has repealed the laws, and I somehow missed the memo.

Hi, thanks for the post.

I do not understand what laws of physics we claim to be changing or breaking.  A microphone reproduces sound.  We can hear the sound that it captures.  We can use the microphone to capture many different sources, at different amplitudes, at different axises, and in different environments and acoustic spaces...  We can get a clear picture of what the mic is capable of reproducing.  I don't think that is deniable.

Our job, is to take that vast information and represent it in an algorithm that when applied to our ML-1 microphone, recreates the sound of the microphone in a way that is convincing enough so that the human ear cannot decipher a difference between the model and the real thing.

There is no magic in microphones.  Despite the fact that some are quite complex in their harmonics and dynamic frequency response, it can all be measured, and it can be recreated. 

For what it's worth, Fabrice feels that in at least the case of the large diaphragm models, it is less complex than modeling the Studer tape machines that we did for our VTM plugin.  Ironically, modeling the dynamic mics has proven to be more difficult since we start with a vastly different condenser.  But even so, we are able to get very close to some of the famous dynamic mics as well.

Cheers,
Steven

 

[SSLtech]

We can use the microphone to capture many different sources, at different amplitudes, at different axises, and in different environments and acoustic spaces...  We can get a clear picture of what the mic is capable of reproducing.  I don't think that is deniable.

But you can't.

Not with either of those mics. You cannot have a complete picture of the soundfield, and that's my point.

You absolutely CANNOT discriminate between on-axis and off-axis pickup. Without that, it doesn't matter what your associates tell you, I'm afraid what your suggesting is not completely possible.

I bow to Ricardo on these matters, and its entirely possible that my thinking may be flawed... I welcome some correction. -But the Tetramic would indeed be better capable at making these discriminations, as would a Nimbus array, and -though the latter is large and cumbersome and somewhat tedious to set up- it may be slightly more accurate still than a tetramic. -Again, I defer to Ricardo's vastly greater experience in this area: I myself only had my Soundfield A/B formats epiphany almost exactly a decade ago, and when I began researching into the subject, I quickly discovered that he has probably forgotten more about the subject than I'll possibly ever know.

Again, not wishing to pooh-pooh the concept, and I'm certainly not saying that something like this cannot be made much more accurately... -it certainly can. -But the current approach would require more than one capsule. -Per microphone iteration. -In fact, since we're only talking modeling SINGLE-AXIS microphones, I would suspect that a design like the Sennheiser MKH-800 'twin' might be usable, since it can fully discriminate between pressure and velocity (along a single axis) waves... But I haven't experimented or analyzed it to a degree where I'm yet prepared to bet nor deny that it could... Again, Ricardo has far more experience in the field than I.

 

[Steven Slate]

With respect, you are telling me I can't do something that i've already done.  Our capsule's polar pattern is able to mimic the off axis and proximity response of the classic mics (in cardioid) in a way that is close enough to fool the human ear. 

I'll have to defer to actual audio from here on out.  I see no other way to prove my point.

Cheers,
Steven

 

[SSLtech]

...But we only have your say-so, that you have "done it".

On the other hand, I absolutely oppose any notion that it is physically possible to create a microphone which will simulate the behaviour of the Royer microphone as simulated in the video.

This assertion is based on physics -not opinions, and I'm afraid I don't believe that you have "done it". -Nor can you, with the microphones shown.

 

[JohnRoberts]

You guys are arguing apples vs. oranges.  SSL is talking about the actual physics involved, and you are talking about "fooling" human perception.

You can both be correct within reason, while I will have to take your word for the success of the perceptual charade.

I looked at this decades ago and dismissed it as impractical (for back then). Tools are better now and you don't have to fool all the people all the time to have a merchantable product. This being a technical forum you will get push back about over reaching claims.

You may be able to mimic the dominant sonic cues of a given mic configuration without completely recreating the physical  transfer function. As I probably already posted, the success of this may vary with the specific sound sources being mic'd. Being able to successfully cover some popular mics in some popular applications seems easier to accept.

JR

 

[Steven Slate]

...But we only have your say-so, that you have "done it".

On the other hand, I absolutely oppose any notion that it is physically possible to create a microphone which will simulate the behaviour of the Royer microphone as simulated in the video.

This assertion is based on physics -not opinions, and I'm afraid I don't believe that you have "done it". -Nor can you, with the microphones shown.

Here you are correct.  The Royer is a figure 8 mic and our ML-2 is a cardioid. Having said that, we are able to create a model with similar sounding transient and frequency response, and have found that using it in applications where one would use the ribbon mic has proven to sound great.  At the end of the day that is the goal.. make stuff sound good. 

However, in our large diaphragm models, we are able to very very closely replicate the sound both on and off axis of other large diaphragm cardioid mics.

Cheers,
Steven

 

[SSLtech]

Thanks John,

And I don't doubt that things can be mimicked for a given circumstance, and I'm sure nobody doubts that transferring frequency response and transient response curves is a fairly trivial matter... but once the simulated circumstance alters or differs in any way, then the simulation will begin to fall short. -As mentioned earlier, micing at the center of a cone versus at the cone edge, and 'up close' versus further away will all produce different pressure-to-velocity ratios across the spectrum, and since a Royer will almost completely ignore the pressure component and respond almost exclusively to the velocity component, these changes will result in variations from the simulation.

Expanding the technique to use something like -as mentioned earlier- something like an MKH-800 Twin would make things a LOT closer... but as it is, there are significant limitations upon how possible the simulation can be.

Seriously, I think you should consider this... It doesn't HAVE to be an MKH-800; it could be another mic entirely... or a couple of mics; a larger-diameter version using a Von Braunmuhl-Weber type capsule, but with separate signals from BOTH sides recorded would ALSO allow you to do this. -I believe that the Swedish manufacturer Pearl microphones also does a mic which could be used.

Without doing so, you may indeed be able to persuade people that you can "similarize" a microphone... but if you used a model more along the lines I'm suggesting, you could do an awful lot more cool stuff... and get closer. -You could change mic patterns after the fact. You could genuinely simulate the Royer a lot more closely...

-but then there's even more that COULD be done. -Automatic blast detection and canceling, for example. -Bearing in mind that a blast plosive will be opposing polarity on both capsules, it would be a trivial matter to seek them out and cancel them... kind of like a De-Esser, but for pops. -without the intrusive issues associated with a pop filter. -I read on another forum that JJ Blair, whom I think you may know (?) mentioned that he really wished there'd be more simulation of what he vaguely called 'phase response' . -He was referring -I suspect- issues arising from the path length difference between front and rear... and this too could be better simulated and accounted for, once you're provided by independent signals from front and rear capsules.

The possibilities would be MUCH less restricted, and in all likelihood you could achieve a rather closer 'simulation' by starting from a rather more physically relevant start point than a single cardioid microphone signal.

I'm genuinely hoping that Stephen can see the possibilities and benefits in this, rather than saying "well, we're sure we've simulated it very closely." -I appreciate that may well be what the company believes, but I'm too focused on the physical limitations which have to be acknowledged with regards to any claims.

I don't doubt that as it stands it may "fool people" (not my own words, please note), but I'd be much more interested in more closely replicating a system than merely "fooling" people.

Getting people to 'believe' something is sometimes also a trivial matter... Especially so when they really WANT to believe it.

 

[JohnRoberts]

Not to vote one way or the other, but an old theme of mine is that stereophony is inherently flawed as we try to capture and reproduce a more dimensional physical sound field with less playback dimensions in a different space. While stereo may be flawed in theory, it has been wildly successful for decades because the additional spatial cues from using two sound sources for playback is more than 2x better than mono for a perception of sound space. Unfortunately there are real but quickly diminishing returns from adding additional channels beyond 2.

There is a place for precision and accuracy, and place for psycho-acoustics to coexist in recording. 

JR

 

[SSLtech]

Ah, now you're entering the territory of the evangelists, John!

-As I'm sure Ricardo knows, Michael Gerzon was passionate to the point, I understand, of irritating a good number of people on this subject... -Are you familiar with Gerzon's work?

http://www.michaelgerzonphotos.org.uk/

For tasters; a shot of the earliest tetrahedral prototype:

 

[Holger]

deleted. wasn't my best post.

 

[JohnRoberts]

Ah, now you're entering the territory of the evangelists, John!

-As I'm sure Ricardo knows, Michael Gerzon was passionate to the point, I understand, of irritating a good number of people on this subject... -Are you familiar with Gerzon's work?

http://www.michaelgerzonphotos.org.uk/

Familiar in general but not chapter and verse.  I used to design and sell surround sound decoders in the late '70s.

It still seems flawed to me, even if you can accurately collect the three dimensional sound field at a single point, what do you then do with that information?

One exotic approach that does hold water IMO is dummy head recording, but only for playback through ear buds or maybe cans, so the playback does not have to deal with the different playback room acoustics. The only obvious error in that approach is that the dummy head outer ear, pinnae combing will not match up perfectly wih the recorded combing  but probably good enough.  Perhaps you could embed tiny microphones in your own ears to capture a performance you want to experience again later... while you can't experience it properly that first time while tracking with mics stuffed in your inner ears. 

For tasters; a shot of the earliest tetrahedral prototype:

My larger point is that there is no perfect way (that I know of) to accurately capture and reproduce complex sound fields in disparate spaces. This doesn't really matter since the human brain discards most of the actual information coming in, to interpret the sound space by focussing on significant aural cues. Hitting us with an adequate amount of believable sound cues can deliver a pleasant listening experience (like good stereo).  Since I don't believe there is a perfect solution to realize, I do not lose any sleep over chasing after one.

More importantly do no harm by adding artifacts that ruin the illusion, and make stuff that sounds good.

JR

 

[ricardo]

There's a number of topics that are being discussed here.

Firstly, I think its highly possible that the Slate LD mike could sound similar to U47, C12 etc in the environment it was designed for.  This environment is the Recording Booth in a multitrack studio or multi-miking a drum kit.  But it wouldn't be a good substitute for the engineers involved in some of the best recordings ever made http://davidpickett.wordpress.com/2009/07/26/the-golden-age-of-stereo/  These famous mikes were in fact designed to suit these people.

Steven, I think if you recorded the back diaphragm signal too, you might do a better job at these Jurassic tasks which I realised are NOT your market.

I shouldn't be fazed that Stock, Aitken & Waterman used my Mk4 Sounfield to make more Top 10 hits in the 80's than any other Pop Producers .. but I still regard this use as heretical. 
______________

On Gerzon, what he did was to turn the question of reproduction around.  Instead of asking, "How do we make it perfect?", he asked "What CAN we make perfect?"

Ambisonics combines Soundfield Reproduction (making things perfect when we can) with Psychoacoustics (making sure the remaining 'errors' are at least plausible to the Mk1 Human Head)

A very simple and important demo of this is that a simple 4 speaker Ambisonic system treats all horizontal directions equally.  It's impossible to distinguish the speaker directions if a signal is panned horizontally.

This is NOT the case with stereo.  Sounds panned CF sound a LOT different from sounds panned to each speaker or at half left or right .. the stability of these sources with small head movements is different too.

There is a bigger problem with dummy head recording than matching HRTFs.  ALL studies show that moving head cues are far more important than fixed head cues.

The simple 4 speaker Ambisonic system attempts to correlate fixed head & moving head cues too.

There's actually a lot of work at present on binaural playback of Ambisonic recordings.

 

[SSLtech]

I did some rumination overnight, and I did realize that I'd narrowed my thinking too far in one respect. -Yes, I believe that you would indeed still need a full three-axis array (eg. Tetramic) to be able to model off-axis anomalies, and since that runs things MUCH further up the cost ladder, (and adds considerably to setup complexity) I doubt that's really a reasonable goal.

On reading Ricardo's last post, I think that sums it up best insofar as taking target use etc. into consideration.  -I wish I could be more succinct sometimes (and I bet many others wish I could, too!)

As for SAW; At the time, I professed to dislike their 'cookie-cutter' product, but -truth be told- I'd have happily enjoyed a share in that success! -Pete Waterman of course was a great 'salesman' for the enterprise, and his promotional abilities should not be underestimated. A well-engineered product however, is a well-engineered product.

Likewise, Steven's promotion is a considerable factor in the likely success of this product. -But I feel that like the Kemper modeling amplifiers, Impulse-Response based reverberation systems, or any number of other 'simulation' products, when the prospective buyer WANTS to believe that the simulation is essentially 'as good as' the real thing, then the product is as good as "sold".

I'll restate that I suspect an 'advanced' version of this also recording the backplate signal has the potential to go much further. As to how the cost/benefit ratio may be perceived by users; that's not my field of expertise.

 

[JohnRoberts]

On Gerzon, what he did was to turn the question of reproduction around.  Instead of asking, "How do we make it perfect?", he asked "What CAN we make perfect?"

+1 it can't be perfect, period.

There is a bigger problem with dummy head recording than matching HRTFs.  ALL studies show that moving head cues are far more important than fixed head cues.

Good point... a major problem with IEMs too... no room cues.

There's actually a lot of work at present on binaural playback of Ambisonic recordings.

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.

JR

 

[sahib]

The best microphone of the 20th century, the Calrec Mk4 Soundfield, was also my baby 

Ricardo,

This prompted me to dig out my Jurassic time electronic magazines and yes Wireless World, October and November '79 issues have Ken Farrar's article on what I assume is the MK1.  Would I be correct?

 

[SSLtech]

This I think was the first iteration:

By the MkIV it was pretty slick.

My MkIV in the middle, with my two ST250's either side:

 

[iampoor1]

Trying to sell this product to anyone here isnt going to work Mr Slate....

...Unless you offer it in kit form in the white market thread. 

 

[SSLtech]

Well, let's not have any joking around that gets misunderstood...

What the hell though: here's a shot of the two U47s and their two more modern counterparts (because I'm a pic-whore once I get going!):

 

[jsteiger]

.....What the hell though: here's a shot of the two U47s and their two more modern counterparts (because I'm a pic-whore once I get going!):

And that sweet pic is taken on a REAL API console!  ;D