32 bit

When, how and why is used?

 

32 bit, What?

 

The PlayStation was 32 bits

 

I have a 32-bit socket set, I use it often.

 

Pfft
8 bits for me

 

I heard 32 bit before, something about studio

 

Float or fixed? 99% of all DACs are 24 bit fixed.

 

I dont know

 

Research digital audio on Wikipedia. If you want to go deeper than that, then get Ken Pohlmann's excellent text, "The Principles Of Digital Audio".

 

Well studios can save the recording to 32 bit file, some do. It really seems it may be overkill but IMO why not, probably does not add much cost. Although I suspect most of these, if not all, get downsampled to 24bit anyways in the mastering process to prepare for vinyl pressing and SACD or for hi-rez downloading.

 

The advantages mostly come into play when editing and processing files:

Pro Tools 10: The Benefits of 32 Bit Floating Point Audio

 

Presuming the topic is digital audio, the sampling rate and bits are used to represent the volume of the waveform. Sampling rate is simply how often you measure the volume of the sound you are recording. For CDs, samples are taken 44,100 times per second or 44.1 kilohertz (kHz). The bits are how many numbers are used to represent the volume each time the sample is taken. More bits let you represent more values for the volume, meaning the samples are closer to the original waveform, and the noise/distortion will be less. 16 bits is used for CD, and IF the recording is done skillfully can sound really great. "High resolution" digital uses more samples (48, 96, or even 192 thousand times per second) and/or more bits (usually 24).

I've seen 32 bits in two audio contexts. The first is Digital-to-Analog Converters (DACs) where since it is so many more bits than the recordings themselves it is probably meaningless. The other context is within digital sound processing, in studios or laptops or recording equipment. There, the extra bits can be useful because any kind of processing like equalization or reverberation create more bits, which if you just kept 16 or whatever would cause an accumulation of rounding errors, increasing noise and distortion. 32 bits gives wiggle room to do all that stuff and the still put out 16 or 24 "good" bits at the end.

 

So master tapes can be 32 bit?

 

No. Tapes are not a digital format.

 

I dont understand if people had go, to make 24 bit dacs and music why, dont they go all the way?

Is 32 bit the maximun or the limit?

 

So you mean there are in fact 32bits dacs out there? I will search this info now.

 

But they can transfer the tape sound to 32 bit?

 

Keep in mind, there is also digital tape.

 

There is no such thing as digital tape. All tape is analog. Period. But some tapes might store 1's and 0's instead of a waveform. Decades ago my Commodore computer stored programs on a cassette. But you could play it.

 

Yes, there are. My DAC goes up to 32bit, so I have set my Foobar output to 32bit. Not sure if I gain something with it, since I only have maybe one album in 32bit (a vinyl rip).

32bit DACs seem to be growing in numbers nowadays, just like DSD DACs.

 

While we're at it: I heard that Foobar's volume engine works with 32 Bit floating-point. Since I usually use Foobar with ReplayGain, would setting my output to 32bit be a way to avoid the theoretical degradation of SQ by decreasing the volume?

 

This thread = This forum in a nutshell.

 

Working with 32 bits can be computationally advantageous. Consider 24 bit audio: when it is being transferred between memory and CPU for operations, it doesn't fit evenly in a 32 bit or 64 bit data bus. There are packed 24 bit formats for video codecs that are hardware accelerated (and your video card might expand 24 bit color internally to 32 bit to work on it), but not so much explicitly for audio.

A normal CPU will do integer operations on only 16/32/64 bit - and because of the need for audio-compatible dithering, we can't just let CPU math instructions round or truncate operations the way they normally would.

However, 32 bit is a floating point format. It has 24 bits of significant precision, the equivalent of a calculator with an 8 digit display. However it also includes an 8 bit exponent, that can scale the significand from 1 to a 38-digit number. This is a particularly good format for manipulating audio, you can turn down the volume 720dB, and still have 24 bit audio quality and dynamic range (the "significant number of digits" if you track the precision of your physics measurements).

What makes this format good is that we don't have to do the dithering as a second step like we would do when doing operations on 16 bit audio, or as a fourth step if step 2 and 3 were to upconvert for processing in the filter, then downconvert. We can just do math. Since it is floating point, we can take advantage of floating point hardware, especially SSE-type hardware-accelerated instructions (Streaming Single instruction, multiple data Extensions), where we can perform the same operation on large amounts of data instead of one word at a time.

Will it sound better that integer 20 bits? Probably no, but we can process and manipulate the audio digitally any number of times in audio workstation software and know there is absolutely no loss from it's "bittyness".

 

Now now , play nicely with the other children.

 

Files can be big