Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: 32 bit Float or Linear? (Read 11315 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

32 bit Float or Linear?

What is the technical difference between 32 bit float and linear?

32 bit Float or Linear?

Reply #1 –
That's like comparing apples and oranges. Please re-formulate your question in a more understandable way, or at least in a context.

32 bit Float or Linear?

Reply #2 –
What makes it either 'linear' or 'float.'
Is 32 bit linear mathematically the same as 32 bit float?
Is linear different to float when it comes to dsp?

32 bit Float or Linear?

Reply #3 –
32bit linear = 32bit integer, it contains whole-number values within a certain fixed range, just like 16 bit linear contains integer values from -32767 to 32768.  And then 32bit floating point contains, well, floating point values, fractional number values.  I believe, and I could be wrong, it is a 24bit fractional number with an 8 bit exponent or modifier component.  Not sure I have the exponent/modifier description correct, but I think that's how it works.  Almost 3am here and I'm just about to hit the hay so I'm a little hazy at the moment

<edit> and please correct me if I'm wrong (as I most possibly could be)

32 bit Float or Linear?

Reply #4 –
Quote
What makes it either 'linear' or 'float.'
Is 32 bit linear mathematically the same as 32 bit float?
Is linear different to float when it comes to dsp?

The right word would be "integer" or better fixed point, instead of "linear". So, 32 bit float gives more flexibility when comes to DSP, as opposed to 32 bit fixed point.

As to the exact meaning of the terms  floating point vs. fixed point, it's a bit complicated for me to explain now, I'm somewhat in a hurry, but just do a Google search.

32 bit Float or Linear?

Reply #5 –
is there an audible difference which can be heared by the human ear? Does it affect output quality if using floating point or integer?
Celeron 1300 Tualatin @ 1589, 512 MB Micronchip CL2, 0,5 TB HD, Audigy 2 ZX Platinum Pro, Sony MDS-JE530 DAC, Sony STR-GX 311 Ampifier, Sennheiser HD 570 Headphones, Creative Inspire T7700 7.1

32 bit Float or Linear?

Reply #6 –
If you mean at Foobar, 32 bit float. point output won't give better quality than 24 fixed point. bit output, due to real-world limitations of hardware. If you don't have a 24 bit card, output at anything higher than 16 bit will give worse quality.

32 bit Float or Linear?

Reply #8 –
I liked the 1500dB remark, but it made me think.  Would it not be possible to use the bits to give you better volume resolution within the same dynamic range?  Granted, we can only hear so much difference in volume, but you'd think it could still make a difference in certain situations.

32 bit Float or Linear?

Reply #9 –
I guess our soundcards end up being the 'bottleneck' in quality, being only 16 - 24 bit integer.

32 bit Float or Linear?

Reply #10 –
Compatibility?
I wonder whether 32 bit audio is constant across different programs like linear is.
E.G. would a 32 bit float WAV from Foobar2000 work the same if played back in 32 bit Cubase?

32 bit Float or Linear?

Reply #11 –
32 bit float is a standard wav format. I've used Foobar to convert an mp3 to 32 bit float for editing in Steinberg WaveLab 4 and Cool Edit Pro. I have no doubt Cubase will also work, although I've never tried it.
That's so plausible, I can't believe it.

32 bit Float or Linear?

Reply #12 –
Quote
I guess our soundcards end up being the 'bottleneck' in quality, being only 16 - 24 bit integer.

true, but for intermediate editing etc was what I was thinking.  I guess it'd be largely if not wholy imperceptible anyways.  I was just thinking that a similar concept in graphics might apply.  While 24bit colour should be enough for a given picture, when blending a lot of things together to get the finished picture, as they do in games, you sometimes get an accumulation of errors which results in visible banding in the finished image.  And now video card manufacturers are starting to use 64bit and 128bit colour to combat that, among other things.

32 bit Float or Linear?

Reply #13 –
Quote
true, but for intermediate editing etc was what I was thinking.

You're certainly right about that. It's best to work in the highest bitrate that your software/hardware will allow, then master down to your target -- for example, 16 bits with dithering. Foobar works the same way.
That's so plausible, I can't believe it.

32 bit Float or Linear?

Reply #14 –
Quote
I guess our soundcards end up being the 'bottleneck' in quality, being only 16 - 24 bit integer.

Naw, I'd say it's more like our ears being the 'bottleneck' in quality, only perceiving a maximum of 17/18 bit quality, in very, very rare circumstances.

Note: numbers are somewhat subjective, but apparently it is very, very difficult to notice bit-difference at the 16-bit level. Thus, would it be correct to say that our ears can perceive 17 or 18 bit max?

32 bit Float or Linear?

Reply #15 –
Quote
I liked the 1500dB remark, but it made me think.  Would it not be possible to use the bits to give you better volume resolution within the same dynamic range?  Granted, we can only hear so much difference in volume, but you'd think it could still make a difference in certain situations.

Floating point allows one or the other, or anything inbetween.

32 bit Float or Linear?

Reply #16 –
Quote
would it be correct to say that our ears can perceive 17 or 18 bit max?

The max. theorical dynamic range of human ear is around 120 dB (equivalent to 20 bit). 0 dB is around the lowest we can hear (IIRC in practice it's a little bit less, <0 dB can be audible), and at the threshold of pain is around 120 dB.

But this is all under ideal conditions. In practice all listening rooms have a noise floor of around 30 dB and more, and I guess nobody wants to reach the 120 dB pain limit of his ears 

I believe that under real world conditions, 16 bit is above the usual range of operation of our ears.

32 bit Float or Linear?

Reply #17 –
As 32 bit float has a resolution well beyond (according to the creators of Cubase) 32 bit linear, could 32 bit float be likened to a form of lossless compression?

32 bit Float or Linear?

Reply #18 –
Quote
As 32 bit float has a resolution well beyond (according to the creators of Cubase) 32 bit linear


They're both 32 bits - they both have exactly the same 'resolution'.

You could say that 32 bits linear has more accuracy and 32 bits float has more range.

There is no relation to lossless compression.

32 bit Float or Linear?

Reply #19 –
Ok... now i am so confused.  I have a M-Audio Revolution... which should I use?  24bit, 32bit linear, or 32 float? 

32 bit Float or Linear?

Reply #20 –
Quote
Ok... now i am so confused.  I have a M-Audio Revolution... which should I use?  24bit, 32bit linear, or 32 float?  :blink:

You should use 24 bit dithered.

And in the future, please search the forums before asking.

32 bit Float or Linear?

Reply #21 –
I did. I searched for m-audio revolution and found nothing saying if it supported 32bit. 

So my Revo card doesn't support either of those 32bit settings?  If so, what the hell am I supposed to use in Cool Edit Pro?  I want to take advantage of my cards 24bit recording capability, but Cool Edit only allows 16 or 32bit recording.  So I've just been using 32bit floating point, and it seemed to work fine.

P.S. In Foobar, should I keep dither on?

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

Now that I think about it, why do we even have to choose which resolution to have all audio played at?  Why doesn't foobar just play a file at whatever the resolution of that file is?  Like, if u load and play a 16bit mp3 or wave file then it playes back at 16bits, unchanged.  Then in the same playlist, u play a 24bit wave file and it is plays back at 24bits, unchanged.  And the same for 32bits. 

32 bit Float or Linear?

Reply #22 –
because if you have a 16bit mp3 and it's using a replaygain value then you are losing data if you play it back as 16bit vs 24bit

32 bit Float or Linear?

Reply #23 –
and if u play back a 24bit file using replay gain at 24bits it would loose data?

Is that why 32bit was brought about?  for things to be done with the 24bit files without loosing data?  like headroom?

so thats why cooledit does all its mixing in 32 bit even on 16bit tracks.... oooooohh

32 bit Float or Linear?

Reply #24 –
Quote
because if you have a 16bit mp3 and it's using a replaygain value then you are losing data if you play it back as 16bit vs 24bit

There is no 16 bit mp3.