24 bit in and around, 16 bit out?

  • Thread starter Thread starter jedblue
  • Start date Start date
jedblue

jedblue

beep beep beep beep beep
In this study I found http://drewdaniels.com/audible.pdf the authors conclude that there is no audible difference between a high bit SACD and a 16 bit CD during playback as revealed in their long term double blind listening test.

However, they do make this interesting comment early on in the paper;

"The usefulness of the increased dynamic range afforded by longer word lengths for mixdown has never been in question."

If I accept their conclusions regarding 'no audible difference' stereo playback and if it's going to be a 16 bit 44.1 out cd final form, I wonder just what the "usefulness" of, say, 24 bit AD / DA, processing and mixing over 16 bit they refer to is?

My digital multitracker is fixed at 44.1 kHz 24 bit in and around and 16 bit cd / .wav out so I can't make any direct comparison. Any clarifications or experiences from those who've work with both 16 bit and 24 bit in AD / DA, processing and mixing would be appreciated.

G
 
It doesn't seem like they concluded that there was no difference between 16 bit and 24 bit, just that people tend not be be able to hear it.

When mixing, you will sum together the deficiencies of all the tracks. In other words, the difference between 16 and 24 bit doesn't matter much on a single track, but it does matter when you have a bunch of them and you mix them together.

CD is a final listening format, there is no further processing. When mixing, that extra bit depth helps keep things sounding good through the processing.
 
Farview said:
When mixing, you will sum together the deficiencies of all the tracks. In other words, the difference between 16 and 24 bit doesn't matter much on a single track, but it does matter when you have a bunch of them and you mix them together.
Click to expand...

+1

In other words, if you recorded a song at 16 bit from the get go, and do all the digital processing (DSP) at 16 bit, It would be easier to make the distinction between the audio quality of the final 16 bit mix of that session compared to if you did everything (recording dsp) at 24 bit and then dither down to 16 bit.

If you record everything at 24 bit and then a/b'ed the 24 bit and the 16 bit mixes from the 24 bit session, the distinction isn't as clear.
 
otoh, the end user ends up converting your masterpieces to an mp3 and listening to your art on $6 ratshack earbuds.
 
Another important point and distinction is that 24-bit gives the engineer a larger digital "canvas" to work with, with some 138dB or so of range to play with (vs. about 90 dB at 16-bit).

Both of these numbers are probably greater than the actual dynamic range of the analog signal coming in, and are almost certainly greater than the dynamic range of the final product for most popular genres (especially in today's sad environment of high compression). But just having that kind of blank canvas space to be able to capture and mix within is a major benefit for the mix engineer, if he knows how to take advantage of it.

G.
 
Think of it this way:

if you are working on recording all the original tracks and you can only record 16 bit, you have to pretty much nail the level and run close to over in order to avoid building up the noise floor to a noticeable level. And then you have to mix 16 bit tracks, so that ugly noise will be summed over the total number of tracks and further raise that ugly noise floor. This feels "cramped" and the process is a lot more stressful and difficult.

With 24 bit, you have a lot of room to work with, so you can leave a bit of buffer on the top end, never have to worry about "overs", mix a fair number of tracks and still have a result that is limited more by the noise level in your studio than by the recorder noise. Then when you dither down to 16 bit, you get the full performance of 16 bit in terms of noise.

Cheers,

Otto
 
jedblue said:
In this study I found http://drewdaniels.com/audible.pdf the authors conclude that there is no audible difference between a high bit SACD and a 16 bit CD during playback as revealed in their long term double blind listening test.

However, they do make this interesting comment early on in the paper;

"The usefulness of the increased dynamic range afforded by longer word lengths for mixdown has never been in question."

If I accept their conclusions regarding 'no audible difference' stereo playback and if it's going to be a 16 bit 44.1 out cd final form, I wonder just what the "usefulness" of, say, 24 bit AD / DA, processing and mixing over 16 bit they refer to is?

My digital multitracker is fixed at 44.1 kHz 24 bit in and around and 16 bit cd / .wav out so I can't make any direct comparison. Any clarifications or experiences from those who've work with both 16 bit and 24 bit in AD / DA, processing and mixing would be appreciated.

G
Click to expand...

The Meyer & Moran study is poorly done and was doomed from the outset. There is nothing there I would hang my hat on. They've really done a disservice to the recording community. I have a work in progress debunking their "research" that I will be making available very shortly. I began work on it last year, but had to put it on the back burner. It's nearly complete. I will post a link here once it's available.
 
Last edited:
Farview said:
It doesn't seem like they concluded that there was no difference between 16 bit and 24 bit, just that people tend not be be able to hear it.
Click to expand...

Yes, and this is true for lots of things in audio. For example, dither can be proven mathematically to be "better" than truncation. And it is better! But the improvement is 90 dB below the music, and thus inaudible. So in practice dither is little more than mental self-gratification.

the difference between 16 and 24 bit doesn't matter much on a single track, but it does matter when you have a bunch of them and you mix them together.
Click to expand...

This is another example where you can prove with math that using 24-bit tracks has lower noise and distortion. But I disagree that it makes an audible difference. At least for pop music recorded at sensible levels.

A few months ago there was multi-page "flame-fest" type thread at recordingreview.com where a fellow named Duggie insisted that 24 bits is better for tracking for the same reasons you state here. So I found this freeware +decimate VST bit-reducer plug-in:

http://www.soundhack.com/freeware.php

This plug-in is part of the freesound bundle, and it reduces the active bits on a track from 32 down to 1. Duggie was gracious enough to make two renders of his current tune - one with all tracks left at 24 bits, and another with this plug-in inserted on every track set to truncate at 16 bits. Using this plug-in is equivalent to having recorded at 16 bits originally. Not only did the two mix files null with only minor artifacts way down at the noise floor, Duggie actually changed his mind and ended up agreeing with me. Now that's a rare occurrence for an audio forum!

This plug-in is the best way I know to do a true apples-to-apples comparison of exactly what we're talking about. I urge everyone to try it for themselves.

--Ethan
 
Ethan Winer said:
... and another with this plug-in inserted on every track set to truncate at 16 bits. Using this plug-in is equivalent to having recorded at 16 bits originally. Not only did the two mix files null with only minor artifacts way down at the noise floor, Duggie actually changed his mind and ended up agreeing with me. Now that's a rare occurrence for an audio forum!

This plug-in is the best way I know to do a true apples-to-apples comparison of exactly what we're talking about. I urge everyone to try it for themselves.

--Ethan
Click to expand...
And truncated to boot. Interesting.
 
Ethan Winer said:
Yes, and this is true for lots of things in audio.
Click to expand...

Yes, everyone should record everything at 16 bit through a soundblaster card on a commodore 64 using packing blanket's for bass traps and egg crates for diffuser's in a domed mud hut.

It' all about the lowest common denominator.

Just kidding with you Ethan ; ) Cheers.
 
I should have also mentioned that I'm presenting this workshop at the upcoming AES show in New York City next month:

Audio Myths - Defining What Affects Audio Reproduction

I'll be demonstrating the +decimate plug-in, and many other such related topics. I realize not everyone can afford to pony up the $400 they charge to see all the workshops and demos. With luck, they'll let me make a video for everyone to see.

--Ethan
 
Hey Ethan, how you been. As you can see, I'm hanging here now at HR.... :)

I may hit the NYC AES this year. I haven't been to one since 2003.
Actually, I remember seeing you there in 2003...you were just starting to promote your new bass traps if I recall.

Not sure if I'll drop $400 for the workshop... :eek: ...but if I go, I'll stop by your booth to say hello.
icon14.gif
 
Waltz Mastering said:
Yes, everyone should record everything at 16 bit through a soundblaster card on a commodore 64 using packing blanket's for bass traps and egg crates for diffuser's in a domed mud hut.

It' all about the lowest common denominator.

Just kidding with you Ethan ; ) Cheers.
Click to expand...



domed mud hut.....hmmmm? studio....whats the dithering aspect of the mud absorption ratio....
 

Attachments

  • grasssod studio.webp
    grasssod studio.webp
    10.8 KB · Views: 149
COOLCAT said:
domed mud hut.....hmmmm? studio....whats the dithering aspect of the mud absorption ratio....
Click to expand...

No need for dither.

Petro for the generator is the biggest concern, unless your located next to a river where you can use water turbines.
 
Ethan Winer said:
This plug-in is the best way I know to do a true apples-to-apples comparison of exactly what we're talking about. I urge everyone to try it for themselves.
Click to expand...

You've got me intrigued, Ethan. I'm using a Yamaha AW1600 standalone. I suppose I could try out the plugin by tracking at 24-bit, porting tracks to the DAW, using +decimate and porting back. I can also just switch to 16-bit mode.

I assumed that it sounded marginally better in 24-bit mode, but I didn't really test that assumption very much. The difference is 8 tracks at 24-bit versus 16 tracks (8 mono plus 4 stereo at 16-bit). The specs say the overall dynamic range (AD-DA from mike or line input to stereo out) in 24-bit mode is at least 97 dB and typically 103 dB. One would hope that in 16-bit mode it would still be close to the theoretical limit of 96 dB. If that's the case, it doesn't seem I should be too stressed out about tracking at 16-bit versus 24-bit. Honestly, though, I tend to have mike or preamp noise at a higher level than 16-bit audio, already. One observation is that even in 16-bit mode, it strikes me that the noise is slightly less noticeable than back when I used a Tascam DA-38 16-bit machine. It may mainly be a difference in the character of the noise seeming less obnoxious.

At any rate, I'm going to do some recordings both ways and see if I really care about whatever difference there may be, if any!

Cheers,

Otto
 
Don't be confused. Go with 24 bit. There is much more headroom below 0 dbfs to work with.
 
Waltz Mastering said:
Don't be confused. Go with 24 bit. There is much more headroom below 0 dbfs to work with.
Click to expand...

Now I am confused by your terms. My headroom is the 24 dB I have between my standard operating level of -24 dBFS and 0 dBFS. That is determined by my level practices and is the same for either 16-bit or 24-bit.

The potential difference is a closer recorder noise floor in 16-bit. My recorder spec says the dynamic range is typically 103 dB and at least 97 dB (presumably in 24-bit mode). If the 16-bit performance is merely limited by the theoretical performance limit of 16 bit audio, the dynamic range should still be close to 96 dB, leaving me only a few dB fewer than at 24-bit. It's not clear to me that that difference is important.

Pardon my ignorance of digital recording, but since the recorder noise floor seems to be about the same either way, about the only possible benefit I see on this machine is if there is signal detail encoded below the recorder noise floor in 24-bit mode that would not be in 16-bit mode. I'm not sure if that is the case, and if it is, I'm not sure it would make a relevant difference in a final mix. Hence, the experiment...

Cheers,

Otto
 
You must log in or register to reply here.
Back
Top