32 Bit Floating Point A/Ds, Mics

[mschnell]

There already are some, Reode does a big campaign on their USB Mic.

Such devices do away with manually setting the (analog) gain control, while preserving the same resolution of 24 Bit in the complete sensible input volume range.

Will this the the future standard in a foreseeable time frame ?

-Michael

[Glennbo]

Personally, I view USB microphones as podcaster mics. I would never use 7 USB mics on my acoustic drumkit, over the 7 standard XLR mics that plug into my 16 channel audio interface.

Keeping things in sync is pretty important, and using USB mics very likely sucks big time in that capacity. I used to use 2 M-Audio Delta 2496 cards, and even though they used the same clock, they were always a handful of samples off from each other, and I had to manually align tracks recorded on one card with the other.

Using USB mics would be a nightmare for me, but for someone who is only using one single mic, might be of some value.

[ScuzzyEye]

I agree with all that you said about the multi-USB mic setup. But there's no reason to think that there won't be 32-bit float interfaces that accept standard XLR mics. In fact Xoom and TASCAM already have portable recorders that do this. It's just an adjustment in form-factor to make it into a desktop interface. Drivers are a little iffy about 32-bit float PCM on Windows right now, but that's also just a chicken/egg problem. Get enough hardware with the feature, and the drivers will have to be there too.

[karbomusic]

^This.

I would place USB mics and interfaces in separate categories for the most part. I would think there are multitudes of creators who don't really know or care to know much about professional audio per se. So something like this for a podcaster etc. would be very valuable as they never really need to worry about getting a good signal to get the job done.

For DAWs et al, I'd imagine interfaces might become more of a thing however... Field recorders like the F6 (I'd imagine) exist because some field conditions contain dynamic range that may be difficult account for otherwise, from crickets to a jackhammer for example.

Also, for your basic bedroom producer, if someone tells them to use the gain knob to make sure it doesn't clip, that's really all they need. It would be nice to not have to even worry about that but what I'm saying is this might not be a super huge market since most interfaces can easily cover the circumstances so long as you don't set the gain too high. Possibly more "nice to have" than "mission critical" IMHO/YMMV. But still, there is the safety benefit of knowing you can't make that mistake and ruin "the take".

[DVDdoug]

From my VERY limited understand of "modern" 1-bit ADCs, successive approximation, and oversampling, I ASSUME it simply takes more processing power (in the chip) to make a floating-point chip.

There is still an upper voltage limit so it will clip way-before you get to the 32-bit floating-point limit. But it might go to +12 or +20dB or something.. That could be useful!

And "quiet side" is still limited by analog noise and inaccuracy/uncertanty.

From what I've read most 24-bit ADCs and DACs are only accurate to about 20-bits. I don't think 32-bit floating-point gets you any more absolute accuracy or usable dynamic range, just more headroom and numerical precision.

[mschnell]

Originally Posted by Glennbo

Personally, I view USB microphones as podcaster mics..

For now that is perfectly right. When using multiple A/D devices, you need to sync their Sample generation moments. This can't be done afterwards ("aggregation") in a lossless way.

So we feature two issues that are slightly interconnected.
(A) digital Format 24Bit Fixed point vs 32 Bit Floating point and
(B) Synchronization.

Re (A) this can be implemented and supposedly is useful with "standard" multi-Channel A/D Boxes connected to a PC via USB (or Thunderbolt, or other usual wiring). This would just have the advantage of no more bothering about analogue Gain setting. No new standard is necessary, as supposedly any DAW can handle 32 Bit Floating point audio streams.

Re (B) I think it should be possible to have devices exchange syncing information via USB. Maybe this is not possible by hooking all of them to a PC, but with dedicated "switches" I am pretty sure that can be done. Here a new standard "USB Word Clock" would be necessary to make that succeed. Maybe in the future this could become a standard with microphones, so we would need to have such (purely digital) "switches", instead of multi-channel AD boxes.

-Michael

[mschnell]

Originally Posted by DVDdoug

From what I've read most 24-bit ADCs and DACs are only accurate to about 20-bits. I don't think 32-bit floating-point gets you any more absolute accuracy or usable dynamic range, just more headroom and numerical precision.

Yep.

32 Bit Floating point provides 24 Bit resolution, when used in the same voltage range as a standard 24 Bit A/D. The quality (linearity and noise floor) supposedly depends on the price of the device hardware and may vary from 20 to close to 24 Bits for both transmission formats.

The floating point format just does something like automatically set the correct gain at any given "block" of time (and transfer the Gain setting to the DAW so there the potential long term resolution is higher that 24 Bit). (I recently did an estimation and found it in extreme cases to be more then 32 Bits, hence indeed 32 bit floating point is better than 32 bit fixed point.)

-Michael

[Glennbo]

Originally Posted by mschnell

Re (B) I think it should be possible to have devices exchange syncing information via USB. Maybe this is not possible by hooking all of them to a PC, but with dedicated "switches" I am pretty sure that can be done. Here a new standard "USB Word Clock" would be necessary to make that succeed. Maybe in the future this could become a standard with microphones, so we would need to have such (purely digital) "switches", instead of multi-channel AD boxes.

I would think a 32 bit floating point audio device with 8 or 16 XLR inputs would be the way to go, rather than tying to sync a bunch of single channel devices.

[mschnell]

Originally Posted by Glennbo

I would think a 32 bit floating point audio device with 8 or 16 XLR inputs would be the way to go, rather than tying to sync a bunch of single channel devices.

IMHO, in general it's a good idea to eliminate analoge cables. Up to now this was hardly possible due to technological limitations and price. But I feel the time has come that this might change. The current USB standard should be fast enough to allow for syncing, and it's proven that 32 Bit Floating point is doable (in Mics and boxes) without decreasing quality or steep increasing the price. Due to the available chips it's cheap and easy to manufacture advanced devices with USB hardware to adapt to cables.

-Michael

[Held]

If I remember correctly, the revolutionary part of those A/Ds is that they actually stack multiple A/Ds with different gain settings and then automatically select the best one for a given input volume.

Using floating point is just for marketing purposes. Using 32bit integers would actually retain more information, since the float's significand has only a precision of 24 bits compared to the 32 bits of an integer value.

[mschnell]

Originally Posted by Held

Using 32bit integers would actually retain more information, since the float's significand has only a precision of 24 bits compared to the 32 bits of an integer value.

32 Bit integer does feature a digital limit. The hardware designer need to set this above the analog limit of the circuity driving the A/D. By that the resolution at the low end gets very limited. Of course here, the noise floor gets more and more relevant, but I estimate that the lack of resolution with Integer is dominant over the noise.

OTOH you are correct that the resolution with 32 Bit Float is no more than 24 Bits. But as any affordable audio hardware will not provide linearity required for more than 24 Bit resolution this is not relevant.
I don't know how 32 Bit floating point devices hardware is designed. (I can imagine several approaches using digital and analogue circuity.)

-Michael