32 bit "Float"

[JonUrban]

Can someone explain to me what a "32-bit Float" bit-rate is?

I understand that there's 8bit, 16bit, 24bit, and even 64bit and more, but what is the significance of the "float"?

I used a demuxer to check out the audio on the "A Hard Days Night" Blu-Ray, and the demuxer program output 32bit "Float" and 48kHz. Sound Forge had no problem with it, but I don't think that you can go into Chrome with a 32bit Float wav file. (PS - I know you can resample, I'm just curious as to what the float is)

 

[ingresman]

If I understand your question correctly......
When you get something digital e.g a CD, the sample size is an indication of many bits you need to store 1 sample. E.g with a CD you get 16 bit and it is held as an integer. (You get 6db of signal/noise ratio which is why CD's specifiy 16*6 = 96db). Now because you only have 16 bits the range of values you can have is 32768 to -32767, (or 65536 discreet values) which for most uses is ok. However if you have a high definition you really need to have more resoution. So if you had a 32 bit integer you would get 4294967296 discreet values or 2147483648 to -2147483647 (or there abouts), so you get in theory a much smoother digital representation. The reason why you would hold this as a float is to get an even bigger range at the loss of a bit of accuracy and if the sample needs manipulating (e.g. multipying by .707 in a SQ decode) you get a much miore accurate answer if you hold the value as floating point or simply a float. I would imagine that tools like Audition use 32 bit float internaly.

As ever wikipedia is a good place to look http://en.wikipedia.org/wiki/Single_precision and http://www-mmsp.ece.mcgill.ca/Documents/AudioFormats/WAVE/WAVE.html is quite a good little guide to wav.
Hope this helps !

 

[Ge Someone]

I'll give it a try to explain, for a full explanation see Wikipedia or try utah.edu.

Float is computer speak for floating point number.
In floating point representation, a value is split into a mantisse and an exponent (forget the terms).
The first part determines the precision, the second part the range of the values. For example a 32bit float is usually a 24bit mantisse and 8 bit exponent.

Now a simplified example, I will use the decimal system (not the binary)
consider a floating point with 8 digits mantisse (precission) and 2 digits exponent.

the number 12345678 would be represented like 12345678 x 101 (sorry superscript not available, I mean to the power of 1)

If you multiply by, let's say, 10,000 it would become: 123456780000
represented like 12345678 x 104

if you divide the original number by 100,000 it would be: 123.45678
represented like 12345678 x 10-5

As you can see the precision stays 8 digits alway whereas a fixed 8 digits would have given you an overflow in the first example and 123 (losing precision) in the second. Similar with 32bit floats you keep the 24bit precision even if you went off the scale during editing.

(I wrote this at the same time as ingresman and I hope it adds to the understanding)

 

[Key]

But would the files on a Blu-ray really be Floating Point? I don't think it should be anything except 24-bit int. Except maybe 16-bit

 

[ingresman]

http://www.blu-ray.com/faq/#bluray_vs_dvd_comparison has a side by side comprason of dvd vs blu-ray. I would imagine that the audio is 24 bit as the codecs are similar to DVD. A not more research should throw up some more info.

 

[Ge Someone]

But would the files on a Blu-ray really be Floating Point? I don't think it should be anything except 24-bit int. Except maybe 16-bit

I agree, the floating point format is only used as an in between format. It is often used when editing/mixing, even mastering. The end result will then be converted back to 24bit or 16bit (integers).
Looks like that demuxer coverts the data to 32bit floating point assuming you're gonna work on it, it might be a setting though.

 

[Key]

That's what I was thinking. What are you using to Demux? I haven't had any trouble with TSMuxeR. I personally don't trust floating point files haha I just don't understand them. I prefer 32-bit int if I do go above 24-bit. And most of the programs I use list 32-bit Int as being higher precision than 32-bit float - although everytime I say this someone argues the opposite. I think the reason is you would have to go above 0dB in order for 32-bit float to be a higher precision? (not sure on that) which can pose problems down the line I would think.

 

[JonUrban]

Thanks guys. Wow. Nothing like a "techie" post or two!
I think I got it. You just edit your files with the "float", then resample back to a usable rate. I am sure that the default output from the program I used was the 32bit float, that's why I ended up with that rate. The files on the disc are listed as 24bit.

--------------------------------------------------------------------

Yes, I used tsMuxerGUI to demux the Blu-Ray. I found the program on the net and it was shareware, so I tried it. It seems to do the job. Actually, I just wanted to get a look at the wav files. They are not worth doing anything with.

Is there something better for demuxing the m2ts files? As long as the program price is reasonable (not some pro thing costing thousands of bucks), I'd be interested in buying it, or at least checking it out.

Below is the info in the tsMuxerGUI "About" screen. I have no idea what a lot of this means, and have less of an idea how to use it all, but for what I needed to do it worked fine, at least I thought it did!?

SmartLabs tsMuxeR – the software utility to create TS and M2TS files for IP broadcasting as well as for viewing at hardware video players (i.e., Dune HD Ultra, Sony Playstation3 and others).

SmartLabs tsMuxeR is a freeware.

Supported incoming formats:
TS;
M2TS;
Blu-ray;
Demux option.

Supported videocodecs:
H.264
Microsoft VC-1;
MPEG-2.

Supported audiocodecs:
AAC;
AC3 / E-AC3(DD+);
Dolby True HD (for streams with AC3 core only);
DTS/ DTS-HD;
LPCM.

Supported subtitle types:
M2TS Presentation graphic stream.
SRT text subtitles
Supported containers and formats:
Elementary stream;
Transport stream TS and M2TS;
Program stream EVO/VOB/MPG;
Matroska MKV/MKA.
MP4/MOV.

Main features:
Automatic or manual fps adjustment while mixing;
Level changing as well as SEI, SPS/PPS elements and NAL unit delimiter cycle insertion while mixing H.264;
Audio tracks and subtitles time shifting;
Ability to extract DTS core from DTS-HD;
Ability to extract AC3 core from True-HD;
Ability to join files;
Ability to adjust fps for subtitles;
Ability to convert LPCM streams into WAVE and vice versa;
Track language information injection into blu-ray structure and TS header;
Ability to cut source files;
Ability to split output file;
Ability to detect audio delay for TS/M2TS/MPG/VOB/EVO sources;
Ability to remove pulldown info from stream;
Ability to open Blu-ray playlist (MPLS) files;
Ability to convert SRT subtitles to PGS;
Tags for SRT subtitles support - tags for changing font, color, size, etc.; tag's syntax is similar to HTML;
United cross-platform GUI - Windows, Linux, MacOS.