32 bit audio

[demt]

I've just got vsdc free video edited and its got 32 bit audio and I've drowned in the pros and cons on the google .Personally I think 32 bits a good idea and I'm wondering if anybody else backs the idea or not.

[avasopht]

I've just got vsdc free video edited and its got 32 bit audio and I've drowned in the pros and cons on the google .Personally I think 32 bits a good idea and I'm wondering if anybody else backs the idea or not.

Reason stores audio as 32bit floats in project files.

Whether it's a good idea depends on what your needs are.

For simple playback there is no real benefit to floating point, since you are targeting hardware with a hard output limit (both upper and lower). For storing in-production audio you do have provision for floating point as your needs will vary greatly.

32bit integer for playback is pointless. 20 bits alone provides enough dynamic range to differentiate between the quietest audible sound and the threshold of pain.

[ejanuska]

I've just got vsdc free video edited and its got 32 bit audio and I've drowned in the pros and cons on the google .Personally I think 32 bits a good idea and I'm wondering if anybody else backs the idea or not.

Reason stores audio as 32bit floats in project files.

Whether it's a good idea depends on what your needs are.

For simple playback there is no real benefit to floating point, since you are targeting hardware with a hard output limit (both upper and lower). For storing in-production audio you do have provision for floating point as your needs will vary greatly.

32bit integer for playback is pointless. 20 bits alone provides enough dynamic range to differentiate between the quietest audible sound and the threshold of pain.

Why push only 20 bits when you can push 32 at no extra cost?

[avasopht]

Why push only 20 bits when you can push 32 at no extra cost?

Firstly I'm not suggesting using 20 bits. My main point is being crystal clear about what is actually going on rather than encouraging a cargo-cult like use of 32 bits (floats or integer).

Just FYI, 32 bit floats are 25 bit integers (for PCM) scaled exponentially by 8 bits (it's 24 bits of precision with a +/- sign, which is effectively a 25 bit "signed" integer). The extra bit comes for free due to the representation.

So with a 32 bit float, if you left the 8-bit exponent at a fixed value you still get 150dB of dynamic range just from the integer component.

My point is that whether 32-bits is a good idea depends on the use case. There are also other considerations (which is why Reason has a 64-bit integer mixer).

[jonheal]

There is a cost in performance, yes?

[Oquasec]

bitdepth? hell no. some even say it runs faster than 24bitdepth despite having an even more infinite amount of headroom.

[avasopht]

Performance wise everything depends on exactly what the application is doing.

Internally your plugins will be mostly using 32-bit float (some 64-bit float).

In terms of a good idea; a good idea for what? Bear in mind that Reason already does this.

[ejanuska]

Why push only 20 bits when you can push 32 at no extra cost?

Firstly I'm not suggesting using 20 bits. My main point is being crystal clear about what is actually going on rather than encouraging a cargo-cult like use of 32 bits (floats or integer).

Just FYI, 32 bit floats are 25 bit integers (for PCM) scaled exponentially by 8 bits (it's 24 bits of precision with a +/- sign, which is effectively a 25 bit "signed" integer). The extra bit comes for free due to the representation.

So with a 32 bit float, if you left the 8-bit exponent at a fixed value you still get 150dB of dynamic range just from the integer component.

My point is that whether 32-bits is a good idea depends on the use case. There are also other considerations (which is why Reason has a 64-bit integer mixer).

and my point is give us the best resolution possible and let the user decide their needs.

[avasopht]

and my point is give us the best resolution possible and let the user decide their needs.

You have to draw the line somewhere. The best resolution "possible" is far too wide of a net and includes bit-depths that are 100% pointless. For instance apart from a 1 in a billion edge case, nobody has a need that cannot be catered for with 32-bits, so giving the option of 65536-bits doesn't have a point.

If anyone has a practical need for more than 32-bits I'd like to hear it.

[pushedbutton]

If you ever need to record a conversation between an ant and blue whale 32Bit has you covered.

[Oquasec]

vst bridge.