32 bit float audio

[Cressy Snr]

I was messing around with the Audio Midi setup on my Mac Mini when I came across an option to output digital audio at 32 bit floating bit depth. I clicked it to see what would happen.
I wasn't expecting the M1 DAC to keep playing but it did and the sound seemed to gain better rhythm and dynamics. switch back to 24 bit (which is what iTunes files are) and it went flatter and less dynamic.

I've left Audio Midi set at 32 bit float for now as it makes music sound more organic and dare I say it more like vinyl.

What is 32 bit float output? Some sort of studio thing for pro recording?
Whatever it is, it is not half bad. Presumably you must have 32 bit recordings to get the best out of it?

[Neal]

Its normally of use Steve when recording and manipulating the data in something like Audacity...go to the end of this for more details:

http://wiki.audacityteam.org/wiki/Bit_Depth

[ed]

there's something very scary going on if your midi set up is affecting your audio stream.......

the midi should be routed firmly to your on-board soft-synth or your DAW software.....shirley?

[Cressy Snr]

Cheers Neal, I'll have a look.

there's something very scary going on if your midi set up is affecting your audio stream.......

the midi should be routed firmly to your on-board soft-synth or your DAW software.....shirley?

Hi Ed
There is no MIDI involved.
The control panel is a dual thing that you use to set digital output from the Mac Mini. It is where you choose the output device eg the optical digital, USB etc.
The MIDI is a seperate part of the control panel for pros to use with synths and other bits as you say (I think)
I touched it once and all it did was put up a message saying there were no devices connected, which there wouldn't be, if you know what I mean.

[Nick]

Not sure there is a 32 bit OUTPUT format, it normally refers to the internal storage and how any internal processing is done. Something like PROTools and Audacity will store tracks in 32 bit float format so you don't suffer from lost of resolution in processing.

[Cressy Snr]

Not sure there is a 32 bit OUTPUT format, it normally refers to the internal storage and how any internal processing is done. Something like PROTools and Audacity will store tracks in 32 bit float format so you don't suffer from lost of resolution in processing.

Hmmm.. that's an odd thing then.
Wonder what selecting 32 bit float option in the popup menu actually does then. Perhaps the audio is treated differently by the CPU, than it is when 24 or 16 bit integer is selected.
It is definitely repeatable as you can switch on the fly between bit depths and the 32 bit option sounds different to the other two.

[Nick]

Not sure there is a 32 bit OUTPUT format, it normally refers to the internal storage and how any internal processing is done. Something like PROTools and Audacity will store tracks in 32 bit float format so you don't suffer from lost of resolution in processing.

Hmmm.. that's an odd thing then.
Wonder what selecting 32 bit float option in the popup menu actually does then. Perhaps the audio is treated differently by the CPU, than it is when 24 or 16 bit integer is selected.
It is definitely repeatable as you can switch on the fly between bit depths and the 32 bit option sounds different to the other two.

It could be altering the format that the data is being passed in between the various layers of the audio system. For it to make a difference infers to me that there is some processing of the data going on in the system somewhere.

For example, reducing the gain of a 24 bit signal by 12db And then increasing it again by 12dB again, will remove the lower two bits if 24 bit integer is used, but if 32 bit float is used the output would equal the input once it was converted back to a 24 bit stream for the DAC.

I can';t imagine it would alter the latency in the system, we are way past the days of external maths coprocessors (showing my age).

Of course if this was a open system, we could just look at the code and know for certain (it may be available for all I know)

[JamesD]

Just noticed this thread.

There is a 32bit version of WAV. The bitstream is made up of 32bit (actually two 16 bit) containers and it can have 16, 20, 24 or 32 bit audio signals held inside the 32 bit container.

J.

[Nick]

Yep, not saying you could not have 32 bit sources, as you say the WAV header will cover 32 bit resolution. But you wont find many. Easy enough to generate though I guess by taking a 16bit 44.1k source and extrapolating up to 32/386k. No extra information would be generated by the process though.

[JamesD]

I've used the 32bit containers quite a lot but the only time I used it with 32 bit audio was moving the signal between digital processors and a mixer (digidesign 48 track beastie)

J

[ed]

I'm a tad confused here....not sure what the original question refers to now, and I'm wondering why the end user/consumer would want to switch....

I only ask because I thought the players read the header and switched accordingly...and....if the source material is 16 bit and you ask your s/w to process 32 bits then the software is doing a lot of unnecessary work, shirley, cos it can't magic another 16 bits out of thin air...........(well it can but they're all zero)

I might be missing the point completely

I know I've stuck my nose in where it doesn't belong but I was prompted because my latest DAW defaults to 32 bit float waves, which take up 60 odd mb for a 3 minute song....and I keep having to go back after a mix and reset the thing to 16 bits and start again.....so far I havn't been able to tell the difference.

[JamesD]

I'm not aware of 32bit containers used in a domestic setting. We use them in broadcasting a lot as we can have one bitstream and a known and set bitrate (and therefore known filesize per minute) but have any of the bitrates inside the container so the transport is audio bitrate independent. Sometimes that is important - paticularly when used in a wrapper holding multiple signals both video and/or audio...

I suspect the audio is still 16bit inside the 32 bit containers that the DAW is using so it won't sound any different...

J.

[ed]

Hi James

it's all got well trixy now....

The bottom of this list is what I was referring to, but I'm not sure how this relates to your container....

http://www.cakewalk.com/Documentation/d ... 2.146.html

[JamesD]

32 bit float is the 32 bit container format and it can carry anything from 16bit to 32 bit audio in it...

32 bit audio really refers to the arithmetic processing bandwidth of the audio chain - it uses 32 bit numbers and so 32 bit float was originated to hold these numbers between processing blocks in the audio chain...

Now you can record (and publish if you want) in 32 bit float... Its a bit pointless as ADC precision is still limited to about 22 bits only so 24 bits gives all the precision one can theoretically record... but the format has it use in production and post production.

Now I see 64 bit float... simply because some modern processors use 64 bit arithmetic rather than 32 bit arithmetic... same comments apply but more so!!!

James

[ed]

Hi James

I don't want to labour this so don't feel obliged to answer, it's way off Steve's original post anyway, but........

I understand from years in the industry what 32bit floating point numbers are, and I understand how wave samples can be stored as 32bit fp numbers in a wav file format(format code 0x0003), and I understand how this can work:



and I understand how it is very useful to recover a clipped sample, and I understand how containers work in, for example, an AVI/MPEG media context..

but I don't understand your use of container in wav file format, or how you arrive at two 16 bit integers in the 32 bits? If you have a mo, or can be bothered, I'd ge grateful..

I do understand and agree about the value(or lack of) at the consumer end, but that's another story.
I'm now going to try and receate the pro-tools recovery example, I have to admit I was impressed.