32 Bit Float Explained

[antcollinet]

Thank you all. So in terms of the product linked at the start of the discussion (and because I have no interest in accurately representing noises that would kill me), the main advantage is that the person recording just doesn't have to worry too much about gain staging. This mostly due to the multiple ADCs. But 32 float simplifies the process. So if you are going into a situation where you want record the frogs and Starship close up this tool simplifies the job. Or if you (like me) just don't want to be arsed about gain staging when recording, this system might work for you. If so, cool! I'm in the market for a new interface....

Bingo - spot on.

Person recording - who might not be an engineer - doesn't have to give a rats ass about the device he is recording from, or what levels he needs to set.

Can we stop now?

 

[pkane]

Can we stop now?

Yes, please?

 

[DonH56]

One more attempt to explain, with pictures: https://www.audiosciencereview.com/...floating-point-adc-system.48021/#post-1721130

 

[earlevel]

Eh? Consider me daft, but I mean it's delta-sigma. You can basically have as many bits as you want. As far as I can tell, 32-bit data goes into the DACs and 32-bit data comes out of the ADCs. Good enough for me!

I think we both got our points across fine, I just wanted to circle back on this one point (having been on vacation for the past 2+ weeks)...

What I was getting at—I alluded to this when I mentioned successive approximation ADCs of old—was that delta-sigma churns out those bits by referencing with a comparator. If the noise level of not only the input (which would be sufficient on its own) but the internal electronics doing the reference and comparison as well is greater than the bit level being resolved...then no, you can't churn out as many bits as you want, not with any more meaning than hard-wiring the bits to zero. Not good enough for me.

 

[AnalogSteph]

We were talking about different things then.

What you get out of the delta-sigma converter itself is roughly looking like this, typically at 64fs or 128fs:

(I borrowed this one from a DAC running at 32fs, but fundamentally it's the same. Also, you need to imagine the actual analog noise floor.)

Thereafter the signal is being filtered and decimated multiple times, typically 128fs / 64fs --> 8fs in 1-2 stages and then in 3 stages from 8fs to 1fs in the main digital filter. Decimation inevitably involves requantization.

The minimum number of bits required at each decimation stage will be a matter of how much of the shaped noise is still within the noise post-filtering and would be an interesting discussion within itself. More importantly in the context of our discussion, however, there is no inherent maximum. So you basically can throw as many bits at it as you want, whether they still add anything useful or not. That's all I was getting at. This is how you end up with a 32-bit ADC with 90 dB(A) worth of dynamic range.

 

[earlevel]

We were talking about different things then.

What you get out of the delta-sigma converter itself is roughly looking like this, typically at 64fs or 128fs:
View attachment 316090
(I borrowed this one from a DAC running at 32fs, but fundamentally it's the same. Also, you need to imagine the actual analog noise floor.)

Thereafter the signal is being filtered and decimated multiple times, typically 128fs / 64fs --> 8fs in 1-2 stages and then in 3 stages from 8fs to 1fs in the main digital filter. Decimation inevitably involves requantization.

The minimum number of bits required at each decimation stage will be a matter of how much of the shaped noise is still within the noise post-filtering and would be an interesting discussion within itself. More importantly in the context of our discussion, however, there is no inherent maximum. So you basically can throw as many bits at it as you want, whether they still add anything useful or not. That's all I was getting at. This is how you end up with a 32-bit ADC with 90 dB(A) worth of dynamic range.

Sure, it's the "whether they still add anything useful or not". In high school, I learned that you can't get more precision out of a calculation than you put in. If you have two measurements that are accurate to within a ten of an inch, and you add them together, it's not correct to come up with a number like "4.3000000 inches". Like I said, if you can claim true 32 bits converters, why not 64, etc. What favor is this doing anyone? (rhetorically)

 

[svyet]

I sent my MixPre3-II to @amirm for testing. As a Mac user, I am not familiar with using ASIO drivers in Windows. I have a $150 Windows11 laptop that I tried to install and test the MixPre before sending it to Amir, but could not get the ASIO drivers working, so I suggested that he just use it in plug-and-play mode, which limits it to 24bit/96Khz. It is very likely that it was my unfamiliarity with Windows that was why I couldn't get it to work.

Given the discussion in this thread, it seems that it would be good if Amir could use the ASIO drivers to enable 32-bit capture for his testing?

https://www.sounddevices.com/asio-driver-for-mixpre-ii-series/

Longtime reader here, bought the Topping DX3 Pro+ a couple years ago based off the ASR review.
Thank you, @amirm and everyone else.

That said, I can't wait for the MixPre results. I have the original MixPre-3 and a MixPre-6 II. Likely will sell the 6 II for lack of need for 32-bit recording and extra channel anymore (and I'll have to live without SD to USB copy function). Nonetheless, I believe both generations share DAC and headphone amp hardware. For that reason, @amirm in your test of the MixPre ADC, could you please also test and post measurements for its DAC performance (there is a 3.5mm stereo line out on the MixPre-3 and MixPre-6), as well as headphone amp performance? The headphone amp is simply spec'd at "300 mW + 300 mW, for use with any impedence headphones."

Those will all be very useful numbers to determine strengths and limitations of the MixPre for various use cases.

Thanks again--
svyet

 

[Andysu]

My contention is that if Sound Devices and Zoom are reducing high end distortion they're doing it through AGC. Has absolutely nothing to do with 32-bit float. In order to test this we'd have to see if they squish, say, a sine-tonish wave as the amplitudes increase. Since I can't seem to get anyone to learn enough 32-bit float to see why it doesn't increase fidelity what point that test? HAHA! Still, if you've seen something like that done I'd like to see it. Or, if I was going to test, what would you test the Zoom F6 against? Thanks!

PS. I just contacted Zoom, I submitted this support request:

In your audio description on Amazon you claim "32-bit float technology for distortion-free dynamic range" Since 32-bit float, in my understanding, is inherently distorted (imprecise), I'd like to know how this is possible before I buy your unit to test. I don't want to throw money away. Thanks!

zoom h4n pro

 

[Bergante]

I sent my MixPre3-II to @amirm for testing. As a Mac user, I am not familiar with using ASIO drivers in Windows. I have a $150 Windows11 laptop that I tried to install and test the MixPre before sending it to Amir, but could not get the ASIO drivers working, so I suggested that he just use it in plug-and-play mode, which limits it to 24bit/96Khz. It is very likely that it was my unfamiliarity with Windows that was why I couldn't get it to work.

Given the discussion in this thread, it seems that it would be good if Amir could use the ASIO drivers to enable 32-bit capture for his testing?

https://www.sounddevices.com/asio-driver-for-mixpre-ii-series/

The MixPre is a class compliant multitrack USB audio device. So even an iPhone will recognize it without drivers.

In order to accomodate lobotomysed(*) operating systems that in 2023 still don't support anything beyond stereo natively, MixPres can be configured in stereo only mode. Of course you lose functionality.

Anyway I am not sure about the 32 bit situation as an audio interface. You need software capable of supporting that. 32 bit sampling is most useful give you a lot of latitude in order to set the microphone inpu gain.

(*) Yes, you need those "ASIO" drivers. The funny thing is, ASIO is not officially part of the OS but a third party add-on, so a second class citizen.