hello,
Maybe someone can tell me if there is more benefit in recording, mixing and mastering in 32 bit.(the manual of cubase reccommends to record in 32 bit mode)
I ask this because of a quote of a mixing engineer who said that windows is working in 32 bit mode and therefore you must use 32 bit mode in recording and so on.
This would then prevent windows from recalulating every single step in the process to a lower bitrate with glitches and distortion as a result.
The advantage would be that your signal could be louder without the distortion.
negative on the other hand is the bigger wavefiles you get as a result of the higher bitrates.
I can also imagine that the quallity of your sound is better because of the greater wordlength i.e. more samples in a given signal.
But since a lot of people use 24 bit as their recording bitrate and this is what often is talked about, what should i do if processor, memory and harddisk space are no restriction to do recordings in 32 bit.
Anyone?
Greetz Hans
24 bit versus 32 bit
Re: 24 bit versus 32 bit
Unless you are recording at very low levels, there is no technical reason of any great justification to say that you should use 32 bit over 24 bit. You are very unlikely to hear any difference as the noise floor in 24 bit is likely to be much lower than the inherent noise in the material you are recording. For instance, with anything recorded using a microphone, you'd be lucky to get a signal to noise of better than 75dB-80dB whereas the noise floor in 24 bit is 120dB down.
Sure, you'll have people swearing black and blue that they can hear the difference, like they can hear the difference between playback from CD's compared to the same over hard-disk (according to some, hard-disk playback has less jitter than CD, though there's no technical justification for that, apart from one clock source being more stable than another). Ask any of these folks to reliably choose between the two in any of these cases in double blind tests and they will fail, as they have done in the past.
My advice is to take all this exaggerated talk of superiority with a grain of salt. What sets apart a good recording from a bad one, particularly so in this day and age, is recording technique. It's what you do rather than the gear. The history of recording is proof enough to me, as there are ample examples of outstanding recordings done decades ago on equipment with rather inadequate performance, yet they still stand up as great recordings. The irony is, that people attribute, as always, the great sound to the gear they were using, rather than the talent of the people involved, so you have lots of people running around buying up classic analog gear to try and obtain that classic analog sound, or worse still, plugins purportedly emulating analog gear sounds. Its the understanding of the process and not the pieces in the process, that makes the real difference.
Getting back to the point, there may be CPU usage reasons for using 32 bit over 24 bit, in as much as if the DAW you are using uses 32 bit floats internally and needs to convert 24 bit fixed point to 32 bit floating point, it will use more CPU than merely reading 32 bit floating point off disk. However, the difference might not be that great, so I'd suggest you do some tests with your DAW and see what difference there is in CPU usage between the two approaches, and then decide if you want to trade off disk usage with CPU usage or visa-versa.
regards,
Paavo.
Sure, you'll have people swearing black and blue that they can hear the difference, like they can hear the difference between playback from CD's compared to the same over hard-disk (according to some, hard-disk playback has less jitter than CD, though there's no technical justification for that, apart from one clock source being more stable than another). Ask any of these folks to reliably choose between the two in any of these cases in double blind tests and they will fail, as they have done in the past.
My advice is to take all this exaggerated talk of superiority with a grain of salt. What sets apart a good recording from a bad one, particularly so in this day and age, is recording technique. It's what you do rather than the gear. The history of recording is proof enough to me, as there are ample examples of outstanding recordings done decades ago on equipment with rather inadequate performance, yet they still stand up as great recordings. The irony is, that people attribute, as always, the great sound to the gear they were using, rather than the talent of the people involved, so you have lots of people running around buying up classic analog gear to try and obtain that classic analog sound, or worse still, plugins purportedly emulating analog gear sounds. Its the understanding of the process and not the pieces in the process, that makes the real difference.
Getting back to the point, there may be CPU usage reasons for using 32 bit over 24 bit, in as much as if the DAW you are using uses 32 bit floats internally and needs to convert 24 bit fixed point to 32 bit floating point, it will use more CPU than merely reading 32 bit floating point off disk. However, the difference might not be that great, so I'd suggest you do some tests with your DAW and see what difference there is in CPU usage between the two approaches, and then decide if you want to trade off disk usage with CPU usage or visa-versa.
regards,
Paavo.
Re: 24 bit versus 32 bit
Ok ,Thank you Paavo,
I read your comments and i agree for the recording techniques.
Maybe superior recording techniques technically do exist but that does'nt say anything about good recording or a well recorded song.
The thing is, besides the above mentioned, if you have the possibility to do so (32 bit and upcoming 64 bit) why shouldn't you use it ?
The people who recorded their stuff in the sixties, seventies and so on used the best equipment in their opinion they could get their hands on and was available at that time and it didn't stop them from making fantastic albums.(and people still do nowadays)
I do agree with you if you say that fancy stuff does'nt make better recordings or give you an advantage soundwise if you don't understand what you are doing, it has always been that way, no matter what you are doing and it always will be that way.
For example, your piece of software does make a difference in the mixing and mastering process, (maybe even in the recording process) if you know what you are doing.
No understanding, no good result.( with understanding i don't mean twittering around with numbers, i mean understanding the use of your equipment)
So i talk mostly technically instead of soundwise ( at least i don't hear a difference soundwise because of my age
and the damage taken from playing around a little bit).
Does 32 bit have an advantage technically and hopefully soundwise above 24 bit, although i most probably won't hear it.
I still don't know.
Greetz hans
I read your comments and i agree for the recording techniques.
Maybe superior recording techniques technically do exist but that does'nt say anything about good recording or a well recorded song.
The thing is, besides the above mentioned, if you have the possibility to do so (32 bit and upcoming 64 bit) why shouldn't you use it ?
The people who recorded their stuff in the sixties, seventies and so on used the best equipment in their opinion they could get their hands on and was available at that time and it didn't stop them from making fantastic albums.(and people still do nowadays)
I do agree with you if you say that fancy stuff does'nt make better recordings or give you an advantage soundwise if you don't understand what you are doing, it has always been that way, no matter what you are doing and it always will be that way.
For example, your piece of software does make a difference in the mixing and mastering process, (maybe even in the recording process) if you know what you are doing.
No understanding, no good result.( with understanding i don't mean twittering around with numbers, i mean understanding the use of your equipment)
So i talk mostly technically instead of soundwise ( at least i don't hear a difference soundwise because of my age
![Wink :wink:](./images/smilies/icon_wink.gif)
Does 32 bit have an advantage technically and hopefully soundwise above 24 bit, although i most probably won't hear it.
I still don't know.
Greetz hans
Re: 24 bit versus 32 bit
Hello Hans,
I'm not sure if you quite followed my line of explanation. I'm not poo-pooing higher resolutions but just looking at the total picture which includes the weakest link.
The impression I got from the original post was "should I be using 32 bit when I can get a more recording done on my workstation with lower resolutions" (ie. disk space consuption). To that end, there is no reason why you shouldn't use 24 bit over 32 bit if you want that extra disk space. Similarly, there is no technical reason why you shouldn't use lower sampling rates either, for example, using 44.1kHz if you are targeting CD format (though I can hear a deafening protestation from high sampling rate protagonists). My viewpoint is a scientific one based on the mathematics of the process. Though you hear many people espousing high sample rates, I've never seen any credible evidence to that effect (ie. double blind listening tests).
Now, getting back to the point, 32 bit is technically superior to 24 bit. It has 8 bits of extra resolution so it's noise floor and distortion will be lower, but the difference is betweem 120dB down and 160dB. Both are extremely low so you won't hear it. Your speakers will be producing far more distortion than that. Then, consider your final target. If you are recording to produce a CD with 16 bit resolution then all those resolution differences will be essentially lost by the time you produce your final 16 bit master, even if you are using some wonderful noise shaping resolution enhancing dither. The requirement one needs to produce a good CD (or music DVD for that matter) is enough resolution to avoid excessive additive noise in mixing your material together. For that purpose 16 bit is inadequate but 24 bit is more than adequate. You basically have to mix 65 thousand 24 bit channels together before your noise floor is raised to 16 bit levels.
Again, 32 bit is technically superior to 24 bit so you can use it if you wish, but don't expect it to make a huge difference in sound as the weakest link is elsewhere. Take note of your final target, the weaknesses of the components in the chain, and concentrate your efforts on the weakest links. 32 bit makes for a good slogan, but in reality, its a pretty inconsequential difference.
regards,
Paavo.
I'm not sure if you quite followed my line of explanation. I'm not poo-pooing higher resolutions but just looking at the total picture which includes the weakest link.
The impression I got from the original post was "should I be using 32 bit when I can get a more recording done on my workstation with lower resolutions" (ie. disk space consuption). To that end, there is no reason why you shouldn't use 24 bit over 32 bit if you want that extra disk space. Similarly, there is no technical reason why you shouldn't use lower sampling rates either, for example, using 44.1kHz if you are targeting CD format (though I can hear a deafening protestation from high sampling rate protagonists). My viewpoint is a scientific one based on the mathematics of the process. Though you hear many people espousing high sample rates, I've never seen any credible evidence to that effect (ie. double blind listening tests).
Now, getting back to the point, 32 bit is technically superior to 24 bit. It has 8 bits of extra resolution so it's noise floor and distortion will be lower, but the difference is betweem 120dB down and 160dB. Both are extremely low so you won't hear it. Your speakers will be producing far more distortion than that. Then, consider your final target. If you are recording to produce a CD with 16 bit resolution then all those resolution differences will be essentially lost by the time you produce your final 16 bit master, even if you are using some wonderful noise shaping resolution enhancing dither. The requirement one needs to produce a good CD (or music DVD for that matter) is enough resolution to avoid excessive additive noise in mixing your material together. For that purpose 16 bit is inadequate but 24 bit is more than adequate. You basically have to mix 65 thousand 24 bit channels together before your noise floor is raised to 16 bit levels.
Again, 32 bit is technically superior to 24 bit so you can use it if you wish, but don't expect it to make a huge difference in sound as the weakest link is elsewhere. Take note of your final target, the weaknesses of the components in the chain, and concentrate your efforts on the weakest links. 32 bit makes for a good slogan, but in reality, its a pretty inconsequential difference.
regards,
Paavo.
Re: 24 bit versus 32 bit
Actually, just to circumvent any possible criticism, I'll further qualify my statement "You basically have to mix 65 thousand 24 bit channels together before your noise floor is raised to 16 bit levels". That statement is based on the assumption that the additive noise in each channel is statistically independent from each other channel (ie. incoherent) and that the mixing process adds no extra noise. It could be seen as the best possible case.
The worst possible case would be the channel noise being coherent (ie. the exact same noise signal in all channels) in which case you'd need to mix 256 channels together before the noise floor reaches 16 bit levels. The reality is that the worst possible case is highly unlikely as it requires that each channel has exactly the same signal in it (ie. mixing together copies of the same track aligned in time).
This will never happen in practice, so the quantisation noise in each channel will be pretty close to statistically independent of any other track, so the main limitation in my original statement lies in how much noise the mixing process itself would add. That is likely to be in the same order of magnitude as the quantisation noise itself, so a realistic figure woud still require mixing thousands of channels together to obtain a 16 bit noise floor (say 10,000 or so).
regards,
Paavo.
The worst possible case would be the channel noise being coherent (ie. the exact same noise signal in all channels) in which case you'd need to mix 256 channels together before the noise floor reaches 16 bit levels. The reality is that the worst possible case is highly unlikely as it requires that each channel has exactly the same signal in it (ie. mixing together copies of the same track aligned in time).
This will never happen in practice, so the quantisation noise in each channel will be pretty close to statistically independent of any other track, so the main limitation in my original statement lies in how much noise the mixing process itself would add. That is likely to be in the same order of magnitude as the quantisation noise itself, so a realistic figure woud still require mixing thousands of channels together to obtain a 16 bit noise floor (say 10,000 or so).
regards,
Paavo.
-
- Posts: 40
- Joined: Sat Dec 04, 2004 7:18 pm
- Location: Arkansas
- Contact:
Re: 24 bit versus 32 bit
Hello Hans,
We should make a distinction here.
There are different storage formats for PCM data...
24-bit (fixed format integer)
32-bit (fixed format integer)
32-bit (float)
Most DAW software works in either 24-bit fixed point or 32-bit float format. Both are relatively equivalent. Float uses a 24-bit mantissa, so the dynamic range is about the same as 24-bit fixed point. So it is all a matter of whether or not the calculations done are in fixed format or float.
Floating point calculations can take advantage of a cpu's floating point processor.
Current cpu's use a word size of 32-bits and newer cpu's are now coming about that use a 64-bit word size. But, operating systems and application software both have to be designed to take advantage of this.
Twenty-four bit fixed point files and 32-bit float files are about the same in terms of the dynamic range that can be represented. So using one over the other is a matter of preference or convenience.
We should make a distinction here.
There are different storage formats for PCM data...
24-bit (fixed format integer)
32-bit (fixed format integer)
32-bit (float)
Most DAW software works in either 24-bit fixed point or 32-bit float format. Both are relatively equivalent. Float uses a 24-bit mantissa, so the dynamic range is about the same as 24-bit fixed point. So it is all a matter of whether or not the calculations done are in fixed format or float.
Floating point calculations can take advantage of a cpu's floating point processor.
Current cpu's use a word size of 32-bits and newer cpu's are now coming about that use a 64-bit word size. But, operating systems and application software both have to be designed to take advantage of this.
Twenty-four bit fixed point files and 32-bit float files are about the same in terms of the dynamic range that can be represented. So using one over the other is a matter of preference or convenience.