Hello,

The new Sound Devices Mix Pre range is said to offer a 32bit float path. However the documentation is not very precise and doesn't tell whether the AD converter itself is 32bit float or if it's only at the output of some digital stage.

So, I would like to know more about the A/D converters of these devices. First, is it possible to have a floating point A/D converter? All the converters I knew about so far had an integer output. So, how do they achieve this? Is it a variation on the ubiquitous delta-sigma principle or do they use an alternative architecture?

I am interested both as a potential buyer and as a teacher in charge of A/D theory in a music school.

Thank you for helping.

32 bit float is the data stream, not the conversion resolution. The ADC chip numbers should be seen inside.

@ Jim Williams : do you have access to the device or to some picture of its circuit?

Sound Devices does not provide their schematics. Word on the street is that they use 2 A/D converters in their 32 bit float devices. Vague information is in their patents:

patents:
US9654134 B2 (USA),
CA2973142C (Canada),
EP3259845 (Europe)

No, I haven't read all of the patents, nor studied them. It seems to work.

They record wav files in 32bit float format as an option. Most DAW’s can import them without issue. Not sure at what stage the device translates the stream into 32bit float.

Thanks @ GLouie .

The US patent explains that several converters are used and allow to have a dynamic range close to 28bit / 168 dB.

The systems compares the output of two or more ADCs set to different gains. Some complex windowing is involved.

I have attached a diagram of the 3-stage version.

Quite interesting!

It would be interesting to compare to the method used by the Zoom F6, which also claims a dual A/D with different gains for 32 bit float. I don't know what the Sound Devices patent covers that Zoom does not do. I have not looked for details on the Zoom method, other than the user's manual.

There's actually a great article on Sound Devices own website which explains in rough terms how the 32bit is achieved

I seem to also remember somebody online saying somewhere something along the lines of 'someone patented the use of two ADs to achieve 32bit so Sound Devices used three and patented that instead'. But my memory is veeeeeery hazy and I didn't understand the comment when I read it so please don't quote any of this 😂. All I know is that 32bit works and sounds great on my unit!

you may wanna check out digico and merging (and possibly a few more manufacturers by now) as they both make 32-converters - and then there's stagetec which has been manufacturing their 28-bit 'true match' converters for ca. two decades...
(no idea on the innards, i'm no electrical engineer, designer or coder)

24 bits equals 140 db of resolution. We can't get there now due to thermal and current noise limitations of active devices. 32 bits is marketing. An FFT, THD+noise and IMD specs will tell more, if available.