Audio Gurus: Please help me move from 16 to 24 bits.

[Migu_lez]

.
Hello, everyone.

I've been reading/researching a lot on this topic, but I still have a few unanswered questions in my mind. PLEASE bear with me. I hope at least a few other users find this thread helpful.

I'll be as clear as possible with my questions to be able to get good answers from you. If you really mean to help me out, I'd REALLY appreciate if you can advise me and answer my questions one by one.

By the way, I read this wonderful article, but I'm still unclear about a few things. Here we go... (http://www.digido.com/modules.php?name=News&file=article&sid=14">http://www.digido.com/modules.php?name=News&file=article&sid=14)

1.- I understand now what dither is, what it does, and how it works. What I really want to know is, what happens when you make the bit depth HIGHER from a LOWER source? For instance, what happens to the audio if you go from 16 BITS to 24 BITS? Do you need to dither as well?

1a.- Does that going up in the bit depth do nasty things to the audio? Or does it just adds meaningless numbers, preserving the exact sound from the original sound?

The reason I ask this is because my main sample librar (VSL Pro Edition) is recorded at 16 bits. Most of the loops I own are also recorded at 16 bits:
2.- So, if I convert the 16 bit loops to 24 bits in a 24 bit DP audio project, will I mess in anyway up the audio loops?

My setup mostly involves slaved PCs using Gigastudio. I have TDIFF audio cards going from the PCs to my Tascam DM-24 mixer. From there, the TDIFFs go to the 2408 MK3, then to the MAster Mac G5.

3.- Ok, so how does the fact that the VSL samples are provided at 16 bits affect me if I want to start working on 24 bits?

3a.- The way I understand the signal chain, and PLEASE correct me if I'm wrong, is that I would open a new 24 bit DP project and set the clock to 24 bits internal. Make sure the DM 24 is set to that to, and also the 2408 mk3, right? Maybe also the TDIFF audio cards from the PCs.

So,
4.- if I record the VSL output (via busses) of the DM24 to a DP audio track, will it be the correct way to do it?

4a.- Am I missing something from point 3a or 4? Any other things I must be aware of, or even avoid?

5.- Am I correct in the assumption that (providing points 3a and 4 are ok) eventhough the samples are originally provided at 16 bits, I will get much better resolution in my final mix because I'm working at 24 bits, and the processing from plugs, gain changes, reverbs, etc, are processed at 24?

6.- Will it really be better doing that than just satying at 16 bits, as in my case, I've been doing so far?

7.- Fom what I get from tha above-mentioned article, is that in theory my sound would improve A LOT, but it mostly referred to 24 bit source files, and staying at 24 until mastering down to 16 for CD. But will I really hear a big improvement if the source files were provided at 16 bits, then "cheat" and record them at 24 in DP when I capture the MIDI to audio?

By the way, I'm planning on staying at 44.100 at the moment. Maybe after I get this concepts solidly, i will then venture to 96 KHz.

You guys have NO idea of how much I will appreciate your help on this matters. I'm just a composer trying to understand these non-musical technical details.
If you ever come down to Mexico, I'll make sure I get you drunk on Tequila or Mezcal (the one with the worm). Maybe we'll even go down-town and jam with Mariachis [:P]

Man, it seems that the more I read and research, the more questions I get... it can be frustrating sometimes. Especially because there are bad, good, better, and best ways to do things.


By the way, please don't be annoyed if you see this same post over at Unicornation... I'm just trying to get as much info and advise as possible.

So again, thank you for your help.

Hasta pronto!

[Conquer]

Hi Miguélez. I think you'll probably get feedback on this from people who know a lot more about sampling than me, but it seems to me that if a file was originally recorded at 16-bit depth you won't gain anything by trying to convert it to 24-bit. The 'real' audio information is already contained in the 16-bit file; the extra 8 bits would just be 'imaginary' extra data artifically created by the conversion program (assuming such a thing exists).

That's my theory. I'm no technical expert, so everyone please forgive me if I've overlooked something vital here!

[Dietz]

This is true. No damage will happen to your audio when converting from 16 to 24 bit. Quite on the contrary, all processing will most likely sound better (although almost all modern audio programs work with 32bit floating-point internally anyway). - You don't need dither or any other additional processing for this step.

When you go back from 24 to 16 bit, take care to dither the signal correctly. Dither is nothing else than a certain kind of added noise that masks the truncation errors which appear when "cutting" the lowest 8 bit of dynamic resolution. The human ear is more foregiving in respect to this random noise than to signal-related artifacts.

Converting audio to a different _sample-rate_ is a bit more demanding. Even _up_sampling a signal from 44.1 to 96 kHz can introduce small artefacts if not done properly. _Down_sampling is the most difficult task of all mentioned in this thread, as severe filtering has to take place to avoid aliasing ("ghost-tones"). Only the really good algorithms should be used in this case.

****

This leads me to the common misunderstanding that you express in question 3a: Bit depth has nothing to do with the clock. The clock is responsible for the samplingrate (read: the resolution of the signal's frequency content), while the bit-depth determins the resolution of a signals dynamic content. It's like a simple coordinate system where you draw a wave into: x=f(s), y=n(bit).

HTH,

[Migu_lez]

@Another User said:

This leads me to the common misunderstanding that you express in question 3a: Bit depth has nothing to do with the clock. The clock is responsible for the samplingrate (read: the resolution of the signal's frequency content), while the bit-depth determins the resolution of a signals dynamic content. It's like a simple coordinate system where you draw a wave into: x=f(s), y=n(bit).

Thank you for clarifying that, Dietz.
A lot of my new desire to work in 24 bits now was aroused by the article i mentioned above. Basically, Katz states that one of the bigger benefits of 24 is that every time you move a mixer fader in the program, insert a plug-in, or whatever, then the audio will be "damaged" little by little. Working with 24 bits, since it is a longer word,minimizes truncation, less distortion, truer digital representations of the real sound, etc.

Please keep the knowledge and advice comming. I love this stuff. I am feeling a bit more confident about this already [H]

Thank you all!

[Dietz]

Dear Tanuj,

if you're talking about Cubase SX, you could use the UV22 dither plug-in (from SX' "Other Plugins"-section) as the _last_ plugin (in the last slot - this is after the fader!) of your master-channel. Set this plugin to 16 bit and export your mixdown as usually.

If you plan to do a mix that goes to a professional mastering studio, you should export to 24 bit (or even 32 bit floating, if the mastering studio gives its OK for this format). The same is true when you do (sub-)mixes of your music that will be used again in another mixdown lateron.

Higher sample rates than 44.1 kHz don't make much sense as long as you work solely with Vienna Instruments. As soon as you whish to incorporate your own recordings of delicate acoustic material (e.g. a steel-string acoustic guitar), you _might_ notice some sonic improvement, but to be honest, the difference may be just enough to be audible in a direct A/B-comparison.

HTH,

[Peter Roos]

My approach is to render/bounce the mix in my sequencer to a 24 or 32 bit file and then to do pre-mastering in SoundForge.

This separates the mixing stage (with all the plugins and thus CPU power required) from the pre-mastering stage.

In SoundForge I convert the mix to 32 bit floating point, apply some subtle global EQ-ing and/or multiband compression, followed by subtle limiting and then making a few different "saves". I keep the raw mix, and make 24 and 16 bit masters, by doing the dithering during or after the limiting.

I can use the 16 bit version for making Mp3's and if I want to check the mix carefully, I use a 24 or even 32 bit version.

Making (time-stamped) different versions can be really handy if you ever want to re-do some pre-mastering. And if you ever need to convert to 38 Khz, you can use the 32 bit file. Upsampling a 16 bit file is a no-no in my opinion.

Dithering comes in different flavors, if your music is to be processed again by others, make sure to select a subtle algorithm, otherwise the reprocessing may suffer from the noise-shaping and dithering.

Cheers,

Peter