Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[mark4man]

Crew...

Is there anything wrong with bouncing 16-Bit comps or sets to a 24-Bit file; & then opening them in a new 24-Bit project for mastering? (in much the same way that one would bounce to stems to take to a mastering session)?

I know digital is all about numbers; & there shouldn't be any degradation in sound quality concerning transfers...but someone told me once it's better to bounce directly from the original 16-Bit project (to a 24-Bit stereo .wav); & even tho my instincts tell me different, well...I'm paranoid.

Anyway, here's what's happening:  8 out of 9 songs on my first CD were recorded & mixed as 16-Bit projects.  In those projects, I've applied numerous UAD-1 plug-ins to tracks & sub-busses.  I intend on upgrading to a new, fast, powerful workstation once the CD is released...but right now, after about a dozen & a half plug-ins applied within a 25 - 40 track project, I bog down & drop out.

To stay in the original project & keep working, I obviously have to bounce comps & sets to new tracks & archive the source tracks.

I'd like to instead bounce them to 24-Bit .wav file stems & open them all in a 24-Bit master project for a few last minute tweaks before bouncing off to my final stereo .wav for the CD.  In this way, I can take advantage of the full ambience added by the plug-ins; & also accommodate the extra resolution added to the signal due to the DSP.

But in bouncing from 16-Bit to 24-Bit, I have no way of performing a null test to make sure they sound the same.

Am I being nuts here?  Should I be worried, or is there some validity to staying in the original project (& if there is...I'm not seeing it.)

Thanks,

mark4man


[BTW - I remembered...the concern on the part of the person advising me was that the cumulative affect of successive DSP summing would erode the signal quality.  Is that relative to what I want to do?]

[Adam Dempsey]

mark4man wrote on Thu, 19 January 2006 13:03

I'd like to instead bounce them to 24-Bit .wav file stems & open them all in a 24-Bit master project for a few last minute tweaks before bouncing off to my final stereo .wav for the CD.  In this way, I can take advantage of the full ambience added by the plug-ins; & also accommodate the extra resolution added to the signal due to the DSP. [BTW - I remembered...the concern on the part of the person advising me was that the cumulative affect of successive DSP summing would erode the signal quality.  Is that relative to what I want to do?]

Yep, that would be the way to go. No need to do a null test as this is a case of minimising your resolution losses.

So , yes it's relative (relevant) to what you want to do. Bounce any dsp/gain/panning/mixing to 24 bit (or better still, 32 bit native in the DAW, then dither to 24 bit mix) - regardless of the source files' res (16 bit or other).

[cerberus]

increasing bit depth should only pad the data with zeroes. if you are concerned about the integrity of your algorithm: truncate a bumped-up file back to 16 bits (without dither) and it should null with the original.

most daws have a working bit depth of 24 or 32 bits, so their processing will be returned to their busses at this higher bit depth regardless of the bit depth of the source files.  so as you described your situation, i think there is really no need to bump up the bit depth of your files, all your processing has already been done at high bit depths, if you have bounced  within the project to 16 bit output, you were reducing the bit depth of the signal each time.  

I prefer to start at a high sample rate, so for that conversion, I find it necessary to bump up the bit depth first before src, which can use the extra bits, like any dsp process.

jeff dinces

[Ronny]

Both Adam and Jeff are correct, IMHO. Many processors add bits as it allows more accurate processing, some of the 32 bit hardware consoles add bits at the eq with a 56 bit accumulator and some DAW programs recommend that you convert 16 bit to 32 float for processing and than quantize back to 16 bit for the redbook final. I've been doing it that way since I bought my first DAW and have no complaints, contrary to what others say that it's better to process the 16 bit files without having to go through 2 dither process. I believe that the second dither is much less harmful than the benefits you gain by adding bits for processing, eq in particular.

[Ronny]

BTW, about the degradation and nulling. The files will null, in fact a 24 bit file dithered and quantized to 16 bit will retain all information above -93dB and will null with the file before quantization to the dithered noise floor of the final bit depth. This is likely why 16 bit cd's have not become obsolete after 10 years of 24 bit becoming popular. It's likely that the importance of higher bit depths is in the capture and processing, not so much in the final playback quality.

[mark4man]

Jeff...

Quote:

so as you described your situation, i think there is really no need to bump up the bit depth of your files, all your processing has already been done at high bit depths, if you have bounced within the project to 16 bit output, you were reducing the bit depth of the signal each time.

Please bear in mind...this is something I have to do to keep working, so I'm not sure I understand the advice fully.  You're telling me there's no benefit to bouncing up to 24, but don't bounce to 16 in the same project 'cause I'll loose resolution.

I have to mixdown certain track groups to keep going...otherwise I'm dropping out.

So if what you're telling me is correct, then I should mixdown to 24, right?

I just thought that...since I'm obligated to do this, I may as well create a 24-Bit master project & continue working (& then mixdown to my final redbook stereo .wav) from there.  This means that ultimately I'll be summing twice, which worried me a bit (pardon the pun.)


RoMo...

Quote:

BTW, about the degradation and nulling. The files will null, in fact a 24 bit file dithered and quantized to 16 bit will retain all information above -93dB and will null with the file before quantization to the dithered noise floor of the final bit depth.

Now...this I find interesting.  I would have thought that the dithering process would change the character of the audio, making any null impossible.  Now I can't wait to get back to the studio & try it...to hear how it sounds.


Adam...

Quote:

So , yes it's relative (relevant) to what you want to do. Bounce any dsp/gain/panning/mixing to 24 bit (or better still, 32 bit native in the DAW, then dither to 24 bit mix) - regardless of the source files' res (16 bit or other).

RoMo mentioned this also...32-Bit native within the DAW.  This is also interesting...not sure I can do this in SONAR...can't wait to get back to the studio & try.

Thank you friends...all good stuff...will digest, implement; & get back.

mark4man

[bobkatz]

mark4man wrote on Wed, 18 January 2006 21:03

But in bouncing from 16-Bit to 24-Bit, I have no way of performing a null test to make sure they sound the same.

You wouldn't have this issue if you used a real DAW! The wordlength restrictions in Pro Tools and Performer are entirely artificial. Other DAWs will let you intermingle 16 and 24 bit files with no problem. The good news, however, is that Pro Tools translates 16 bit files to 24 with absolute accuracy, no problem, just a bunch of extra stupid zeros taking up space.

As for your other questions, saving them for another reply when I have time.


BK

[mark4man]

Quote:

BTW, about the degradation and nulling. The files will null, in fact a 24 bit file dithered and quantized to 16 bit will retain all information above -93dB and will null with the file before quantization to the dithered noise floor of the final bit depth.

RoMo...(or Bob or anyone)...

I tried this test; & the results weren't as I expected.

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
I had been listening to hype suggesting that subsequent DAW versions sported audio engines that were nothing short of phenomenal (in terms of advancement)...

so I had created a project in SONAR2 of 9 tracks w/ 6 UAD-1 plug-ins (w/ 2 tracks panned hard left & right...since the discussion had centered on the possibility of newer software versions sounding different due to panning laws); & bounced the tracks to a stereo .wav, which I opened in the same project in SONAR4...

& it nulled perfectly...as it also did the other way around (S4 > S2)...

...[telling me that these newer versions may be re-written to provide greater node depth, fatten the code, or streamline the linkage within the architecture...but they can't change the fundamental structure of what they draw from the converters & in turn send to the hard disc & back again. They can't make PCM digital audio sound any better, except in the way they handle plug-ins.]
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

But...in the process I thought I'd try your test.

I bounced the 16-Bit tracks to a 24-Bit file; & opened it up in WaveLab...where I first dithered it back down to 16-Bit utilizing Apogee UV22HR; & then just saved it as a 16-Bit file, which simply truncates the file.

I imported them both back into the original (16-Bit) project; & they both nulled.

So I thought...I hear nothing in the monitors...but is it a perfect null?  I put the headphones on, cranked the volume all the way up; & listened.

The dithered file was a perfect null...dead silence.

But in the truncated file, I could hear white noise & a faint piano !!!

So I had to come back to the brains at PSW & ask:

shouldn't it be the other way around ?*?*?

It's dither that modulates the LSB & adds noise, right?  I'm certain the files were labeled correctly 'cause I bounced & processed again with the same results.

Is that weird, or does truncation change the very bottom of the bit scale in an audible way...but then...dithering doesn't?

mark4man

[Ronny]

Dithering lowers noise floor and increases dynamic range over a quantized signal that lops off the bits. Yes, you add noise but the quantization errors from the non-dithered truncation create more audible artifacts. If you had no audio playing and only the dither, it would be around -93dB on the 16 bit final. This is -12dB lower than your best Neumann mics self noise in cardioid patterns. Your 24 bit dithered to 16 bit file when one has the polarity reversed will null down to the floor of the dither. Any differences that you hear between the two files would be below -93dB.


Now you know why 16 bit redbook cd's haven't become obsolete when we've had quality 24 bit converters for a good 10 years now. It's the capture that's most important, degradation from reducing word from 24 to dithered 16 bit is not that big of a concern. IMHO.



From the Rane Audio Encyclopedia:

stochastic resonance - Communications. The science behind dither. A phenomenon of nonlinear systems where low-level input signals are amplified and optimized by adding noise, i.e., an increase in the input noise produces an improvement in the output signal-to-noise ratio "The effect requires three basic ingredients: (i) an energetic activation barrier or, more generally, a form of threshold; (ii) a weak coherent input (such as a periodic signal); (iii) a source of noise that is inherent in the system, or that adds to the coherent input. Given these features, the response of the system undergoes resonance-like behavior as a function of the noise level; hence the name stochastic resonance." [From Stochastic Resonance by L Gammaitoni, P Haenggi, P Jung, and F Marchesoni. Thanks S.H.!]

[mark4man]

Quote:

Dithering lowers noise floor and increases dynamic range over a quantized signal that lops off the bits. Yes, you add noise but the quantization errors from the non-dithered truncation create more audible artifacts.

Wow.

OK then...this could only mean one thing...that there is in fact a clear advantage to bouncing a 16-Bit project to a 24-Bit file (& then dithering it back down to 16-Bit.)

Because...the fact that it nulled perfectly means I captured all the plug-in ambience by using that process.

I can also only assume that truncation whacks off that part of the plug-in signal which appears below the 16-Bit floor (within the 32-Bit internal processing environment...since plug-in processing always expands the 16-Bit signal)...& that it would have been the same as if I'd simply just bounced the 16-Bit project to a 16-Bit .wav file & brought that back in & listened to the null at high headphone volume.  (I would have still heard the white noise & the faint piano.)

This is good news; & thanks for the technical info.

mark4man