24 bit audio question

yes when recording, use 24 bit. you will often not use every available bit at the loudest moment. when youuse every available bit at the loudest moment in a tune 16 bit 44.1 can accurately reproduce everything a human can hear upto 96 deciibels.

Rob.Art

Hi
At SAE the teaching was the higher you start the better it will compress down.
I still start @ 4800Hz and 24 bit , I have run @ 196 / 64 the audio files are huge
Regards
Rob

Uh ...

Yes, record at 24 bits.  Even the average, prosumer convertors that we all know and love outspeck the theortical 96 dB range of 16 bit audio.  Unless you prefer a higher noise floor, then use 16 bit.

@

Don't bother.

There are no 32 bit recording interfaces available.

24 bit is more than sufficient

I can tell a difference in 48k over 44.1k in listening to vocal details.

Or I was able to some years back but always went with 24bits for headroom and mixing. You can mix low and bring things up with digital with no problems, where in analog you had to get every track pretty hot.

You start a project at a certain level and when everything is added it will be too hot unless you start low enough, kind of thinking.

A lot of people want to work at higher sample rates like 88.2 and 96 and above but there is a substantial cost in performance that you don't have with 48k.

Unless you have the 64 bit audio engine on. 

I switched from 16 to 24 bits a while back - I think whether the difference is audible to the naked ear or not will likely come down to a variety of other factors - including other components in your audio chain, how well the source is recorded, the overall noise level of the recording, how attuned your ears are, etc. 

However, regardless of whether you can hear the difference or not, I believe one advantage to recording 24 bit is that it provides increased headroom, which is particularly handy when recording vocals - you can lower the input gain down a bit, so you don't have to worry about compression to protect from peaks clipping and still have plenty of headroom to increase the level in the final mix.

I view 24 bits as an easy way to record quality audio without all the angst that 16 bit audio has. As a poster mentioned 24 bits gives added headroom. That may not be technically accurate but it is a fair characterization in the sense that one can be less attentive of ones levels with it. 

On the other hand 16 bits is the CD standard and no one can say that CDs don't sound good.

As to dithering, it takes place when we go from a higher bit depth to a lower one because of truncation errors. Its a way to mask the artifacts that may occur when bits are cut off. The funny thing is that often there is sometimes little to no degradation of the audio because of the cut off of the bits. It really depends on whether there is useful data being truncated or not. This is one reason we have various types of dither.  

It is a good idea to test which dither works best with the material being processed.

I have maintained that we get no perceivable benefit from using bit depths greater than 24. Sonar allows 32 and 64 bit files but the true usefulness of those bit depths has yet to be demonstrated to me. The thing people often forget is that you can never hear either a 32 bit file or a 64 bit file. No D/A converter can process those bit depths thus at best you will hear 24 bit audio only.

So to sum up record at 24 bits if you can and when exporting use dither if needed and burn to CD and be happy. However, keep in mind dither is a form of distortion that is cumulative so try to dither only once. 

   

FastBikerBoy

Correct, there is definitely confusion here.

If you choose 32 in the record bit depth the rest of the file will be padded out with zeros taking more disk space for no benefit, not that that is much of an issue nowadays.

I was always under the impression that if we set Sonar to record in 32bit, then live dithering of 32bit FX during playback never came in to play?

Seems to me, it would be very beneficial to set Sonar to record in 32bit if that would eliminate dithering during playback, especially with FX such as reverb ... etc.

Am I thinking wrong on this?

Are you a recording studio for hire whose work, as it is recorded, will be commercially mastered and released? If you're like me and you're a "pro-sumer" (a nut recording in his basement) instead of the aforementioned recording studio than these details will probably not have any affect on whether your final product is artistically and creatively (or even commercially) viable. This is a question I ask myself when I'm trying to finish a song: Am I an artist or an engineer? If I say I'm an engineer (or I get slick and I say both) I ask my self do I have the equipment in my basement the the big recording or mastering houses have in their facility? Their expensive boxes DO make a BIG difference in SOUND quality. If I'm being honest I'll then try to make the best song or song idea I can with the best mix I can with MY realistic circumstances. Sound quality wise I can only go as far as my resources but creatively the skies the limit.

Well,... the arguement will go on but I'll say it anyway. 24 bit recordings have more detail in them than 16 bit or MP3's. That's what my ears tell me, and that's what the meters say on the way out. Go ahead, flame me, I can take it. 

Ah thanks John, that makes perfect sense, always nice to learn something in a not so technical sense, much easier to understand. 
So assuming all this noise 'adds up' over many tracks, is there really a discernible difference between 16/24bit in the average pop song if recorded well, or is this something more noticeable in a more 'naked' production? 
My reason behind this thinking is that for instance a signal path can introduce much more noise ie a guitar amp buzz or maybe a slight ground hum from another piece of equipment than the noise introduced from the average soundcard whether at 16 or 24 bits, so for the average guy recording a song 'at home' does it really matter? 

Bibs

Wow, I'd never thought about it like that. Deciding if added noise is an artistic decision or a technical no. I have truly never thought about it in the context of ambience.
I may be slightly clueless here but I'm guessing that noise, or at least the right type, adds a warmth that the human ear finds pleasing for whatever reason. Which to me almost makes the whole noise floor argument slightly redundant if it is favourable to add it, although im sure that there are different types of noise that introduce pleasing or not so pleasing harmonics. Was this a big issue in the days of tape? 
Thanks for your posts John, very intriguing and thought provoking, so much to learn, so little time :-)

Bibs

Bub

FastBikerBoy

Correct, there is definitely confusion here.

If you choose 32 in the record bit depth the rest of the file will be padded out with zeros taking more disk space for no benefit, not that that is much of an issue nowadays.

I was always under the impression that if we set Sonar to record in 32bit, then live dithering of 32bit FX during playback never came in to play?

Seems to me, it would be very beneficial to set Sonar to record in 32bit if that would eliminate dithering during playback, especially with FX such as reverb ... etc.

Am I thinking wrong on this?

 
 
 
Hi Bub

In case anyone else knows a definitive answer, the question again is ...

Does dithering come in to play during playback of the project when we set Sonar to record 32bit?

 
 
The definitive answer to the question is no.
 
Explanation of why follows............
 
This is how I understand it and is correct as far as I'm aware but I'm no DAW programming expert.............
 
Where much of the confusion comes in is the confusion between file bit depth and internal bit depth used for calculations. Then there is the integer and floating point to consider.
 
Integer file size first.
 
The AD converters are converting the audio into digital and presenting Sonar with a mathematical representation of that audio. Currently the highest resolution that will be at is 24 bit. If you then tell Sonar to store it as a 32 bit file it will but it can't invent digital information so once the 24 bits that the AD converter has been stored in the 32 bit file the other bits are padded out with zeros.
 
The files use integer resolution (For want of a better word) and internal bit calculations are floating point, allowing for less than whole numbers when performing the calculations and hence greater accuracy.
 
The audio engine then performs the calculations using floating point notation either 32 or 64 bits on these files (regardless of the file resolution) 
 
Note that the bit depth used at the calculation stage bares no relation and has nothing to do with the file bit depth. Of course if the material is more accurate to start with the calculations will also be more accurate but at file bit depths higher than 16 bits accuracy isn't really an issue anyway. As bitflipper indicated earlier it's only if file bit depth is less than 16 bits that the stored information isn't going to be an accurate representation of the original sound.
 
 
There is no dithering performed on calculations because there is no need for it. Dithering is just the process of smoothing out any truncation of data which only happens when reducing bit depth. Think about it, as I said earlier if 24 bit data is written to a 32 bit file the extra bits are just zeros. If it goes the other way though and a 24 bit data gets written to a 16 bit file then all that happens is that the last 8 bits are truncated, resulting in distortion. Hence the need for dithering when reducing bit depth.
 
At 64 bit or 32 bit floating point resolution there is plenty of "room" to perform the calculations needed without chopping out any important information, hence no dithering required.
 
I guess some may be wondering that if the calculation on the files creates results using more than 24 bits where does the extra data get written? The simple answer is it doesn't, remember that all calculations are in real time performed internally and aren't actually written anywhere until a bounce or export happens, that's where the "Render bit depth" in Preferences-->File--->Audio data comes in. I'd advise that is set to 32 bits or higher. (32 bits is "almost lossless").
 
The only time dithering should be done is when reducing bit depth, and then preferably only once. If you are exporting a project for further post mix processing such as mastering you should leave the export file bit depth at the project's bit depth settings assuming that the program being used for mastering can handle it of course.
 
Only when it's being exported to a format that can't handle the existing project bit depth such as a 32 bit project to CD should the bit depth be reduced and dither applied.
 
Of course if any of the cleverer than me cookies consider any of the above wrong I'm more than happy to be corrected but that's how I understand it.
 
HTH

FastBikerBoy

Bub

FastBikerBoy

Correct, there is definitely confusion here.

If you choose 32 in the record bit depth the rest of the file will be padded out with zeros taking more disk space for no benefit, not that that is much of an issue nowadays.

I was always under the impression that if we set Sonar to record in 32bit, then live dithering of 32bit FX during playback never came in to play?

Seems to me, it would be very beneficial to set Sonar to record in 32bit if that would eliminate dithering during playback, especially with FX such as reverb ... etc.

Am I thinking wrong on this?

 
 
 
Hi Bub

In case anyone else knows a definitive answer, the question again is ...

Does dithering come in to play during playback of the project when we set Sonar to record 32bit?

 
 
The definitive answer to the question is no.
 
Explanation of why follows............
 
This is how I understand it and is correct as far as I'm aware but I'm no DAW programming expert.............
 
Where much of the confusion comes in is the confusion between file bit depth and internal bit depth used for calculations. Then there is the integer and floating point to consider.

Yes, this is what I'm talking about. I was always under the impression that if you had a 24bit or lower file you were working with in Sonar, and used a 32bit Float effect, Sonar was converting the file on-the-fly and dithering was applied during playback and you had the option to select which dithering method you wanted to use in the 'Playback and Record' options screen. But it looks like that is not the case.

I was just reading the manual trying to find more info about it and it says the 'Dither' setting in 'Playback and Record' options is for when you bounce, freeze, or apply effects. It says nothing about what goes on during playback like I was thinking.

Thanks for clarifying that.

:-)