32 bit float audio

I think we all know by now what this section is for.
JamesD
Old Hand
Posts: 997
Joined: Tue Oct 09, 2007 4:26 pm
Location: North Yorkshire

#16

Post by JamesD »

Damn! I jusy typed a long reply and lost it due to finger trouble...

Briefly, At the wav bitstream head are the metadata chunks that define the bitstream parameters. One of those chunks is the FMT CHUNK - the format chunk and it contains a parameter called BlockAlign. This defines in bytes (8bits) the size of the data block used to house the essence PCM data. it is 2 for 2 bytes (16bits) for a normal 16 bit PCM stream and 3 (24 bits) for 18, 20 & 24 bit PCM streams and 4 for 32 bit PCM streams. This is the container size I was referring too - I know broadcast engineers and sound engineers do not always use the terminology the same way - this is common in broadcasting when discussing wrappers and containers for broadcast signals...

So when we have a BlockAlign of 4 but only have a 16bit bitstream held within it then we say we have a container size of 32bits with a 16 bit audio stream...

The reference to two 16 bit containers in one 32 bit container is purely a size comparison of the datablocks - a bit sloppy but something I use when explaining it to broadcast engineers as its meaningful to them - sorry if it confused you!.

Hope this helps explain what I mean!

James
gazjam
User
Posts: 103
Joined: Sat May 05, 2012 1:21 pm

#17

Post by gazjam »

Discovered something similar (as in 32bit audio being better) but from a different angle.

The new Dac does up to 32/384, anything I've had previously only went as far as 24bit.
I use a windows based server running Jriver, and have always set it to output the highest Bitdepth my Dac could take which was 24bit.

Can run it at 32bit output, though interestingly I have to output 32bit integer, not floating point.
Guess that's down to the Dac chip?

It's not really a case of "upsampling" to 32bit and improving things, more a case of less downsampling from Jriver's internal 64bit audio path.

It's a nice improvement, I wonder if there would be any real world difference between 32bit floating point and integer?
User avatar
Nick
Site Admin
Posts: 15707
Joined: Sun May 06, 2007 10:20 am
Location: West Yorkshire

#18

Post by Nick »

Not sure you would ever want to send floating point to a dac, I predict bad things would happen if it received IEEE floating format instead of integers.
Whenever an honest man discovers that he's mistaken, he will either cease to be mistaken or he will cease to be honest.
Post Reply