I take some issue with their presentation of spatial reproduction, especially when they have a visualizer that uses a human head. It's one thing to get raw audio, but to then suggest or hint that coincident capsules can accurately produce binaural content is one of my famous sticking points 'round these parts and that doesn't give me a lot of confidence in the company's technical know how.
Don't all ambisonic decoders do binaural? Even the inexpensive Zoom all in one device does this. What's to prevent software from being able to reproduce binaural? I know that for it to be actual binaural (vs HRTF) the mics have to be located in ears, and everyone's ear is different, but it seems like software should be able to get close enough.
Not sure about "all" but I'll take issue with any that do, including Zoom. The fact I've not heard of this Voyage Audio company and the fact they added in a binaural reproducer is not surprising if they're just following a market trend, but it also shows they don't understand the core tech underneath it.
Software cannot reproduce a correct binaural sound field from an ambisonic mic. This is because the ambisonic microphone capsules are effectively co-located in space, whereas a binaural hearing system - two ears on either side of the head - are not; scientific literature suggests an average of about 17 cm between the ears (the foundation for ORTF, for what it's worth). Ambisonic microphones rely on spherical harmonics to encode and/or decode directional information, which itself relies on the capsules to be co-located in space. That's not really practically possible, so the caps generally sit on a virtual sphere with some radius. This means that there's phase distortion as a "steering" function is applied to an ambisonic decoding; this happens at a relatively low frequency depending on the radius the capsules sit on, it is a linear phase distortion up to a certain frequency limit but it can rarely be undone in a manner that keeps the frequency response of the system flat.
Binaural filters, generally speaking, are encoded and assume a difference between the ears. They are usually captured (and/or modeled) with two receivers that are physically separated in space. Most of the binaural encoding schemes I see use a head-related transfer function, which can be thought of as a filter which models the system from a speaker in a free acoustic field captured by a binaural dummy head (two mics embedded inside of something that sort of looks like the upper 1/3 of a crash test dummy, with specific acoustical properties to model a real human).
And therein lies the rub. The wave field at the ambisonic microphone simply is not located at the same point in space as two human ears. Not just that, but trying to apply a binaural encoding on top of an ambisonic recording introduces a highly nonlinear phase distortion from the fact that solving for the wave equation, the wave field is different at the two points in space. These phase errors compound and multiply in the frequency domain. This has the practical effect of heavily distorting the sonic image and potentially introducing catastrophic phase cancellation all throughout the hearing range.
One good way of thinking about this problem is: mid-side (which itself can be thought of as a 1st order ambisonic system consisting of the 0th and 3rd harmonics) is NOT the same as AB. You can't mix XY ormid-side to be AB, and vice versa. One captures pure level cues, one captures pure time cues. The standard fare near-coincident patterns we rely on such as ORTF, DIN, NOS, etc all rely on some blend of a level cue and a time cue, which is closer to what a human hearing system does compared to AB or XY. Trying to convert one to the other causes similar errors as what I've mentioned above.
One potential solution to this problem is to make the ambisonic microphone have a similar radius to the human head. But then the upper limit at which the phase response is linear is VERY low - I'd need to double-check my math, but I believe that puts the phase distortion starting at 20 Hz, acgtually - and it's impractically large, even more so than something like a Jecklin disc. So you lose all the benefits of what an ambisonic system might be trying to encode.
The only company that's suggested anything to me that they understand this problem and have a valid solution is Core Sound, where Len suggests bilateral ambisonics - you can view the Core Sound ambisonics thread in this sub forum for Len's and my discussion on the idea. I think it's a great idea, but it requires two ambisonic microphones and an equally large number of preamps/channels on our decks, and in the present I think there are more effective ways to caputre a great sound field than bilateral ambisonics (needless to say, I'm waiting for the tech to catch up...).
Topic: Voyage Audio Spatial Mic (Read 3326 times)