Is Freeze & BTD reliable??

[MIDI Life Crisis]

Seeing your other post about DP acting strange when opening projects, I’d repeat the troubleshooting suggestions here. Please see other post. Lol

[Yiannis]

LOL. I ve allready seen your answer...thank you

[MIDI Life Crisis]

LOL. I ve allready seen your answer...thank you

That’s comforting. Lol.

[FMiguelez]

I'm glad someone explained this. I have been freezing and hearing slight differences for a while. I'm now exclusively in 32bitFP so I guess you recommend I turn off dither entirely from the pulldown audio menu select?

It depends on whether you prefer QD or dither noise in your frozen/merged 24 or 16-bit file... at around -140 dB

But I don't think those slight differences you mentioned are due to bit depth. (see below more details about this)

My personal recommendation is to actually NOT work at 32bFP but at normal 24 bits, especially if the individual 32bFP files are the result of a "fake" conversion from 24 bits (total waste of space for a "padded with zeroes" version --> It's still just a 32bFP representation of a 24 bit file --> Still 24 bit file).
Even if you record analog sources at 32bFP, it's a total waste, IMHO, since the ONLY difference between that and a 24 (or even 16) bit version would be the relative level of the noise floors. That's it. Nothing purer or "truer" to the original. Just distortion at around either, -192, -144 or -96 dB, respectively.
I'd stick to 24 bits> More than enough precision with more than enough headroom with more than enough low noise floor.

Like I've mentioned before, the only times I print at 32bFP is to get my final stems and final mix, and I admit it's tooooootally unnecessary, and almost a contradiction to what I've said before. But my excuses are>
1) that's only 8-10 tracks (as opposed to hundreds), 2) I avoid another unnecessary 24-32-24 round (the mastering engineer, seldomly used), 3) I don't have to deal with QD VS dither, 4) and it allows me to not torture myself every time I print my stems/final mix knowing I didn't leave some digital residue... at -144 dB
Still, probably unnecessary, and I couldn't tell one from the other in a DBT. But I do sleep better at night, for some idiotic reason


I have time and LOVE these topics. It's been a while since I haven't had a good little DSP rant, so in case you're interested in a little further discussion, here we go>

From what I've read and my own extensive tests, these are some of my conclusions>
The dither noise is a few dB louder than QD, but QD sounds more irritating and annoying.
I used to think that cumulative dither was a problem, but turns out, good dither (as random as possible and uncorrelated between LR and at the proper level and the proper kind) does not add as we'd expect. IIRC, it only does so after 2-3 times, but the more you add, the less they sum. It was an eye opener when I tried it. The QD, AFAIK, would sum as expected and it "pollutes" the timbres. This would be more of a problem with almost identical-sounding instruments playing in unison, I suppose.

Another eye-opener for me was when I actually heard what quantization distortion, on its own, sounds like (with signal-flow trickery, you can get to not hear the original signal and only hear the QD as you switch bit-depths). Very revealing. From listening to QD VS dither with the same test file (piano minus piano), I'd go for dither noise any day of the week. QD is so annoying!

But remember, these artefacts are so low in level that you just can't hear them. All things being equal, if you could tell a 32bFP apart from a 24 bit signal, it would mean you're listening to it at instant-hearing-damage levels for the rest of the signal (music).
There's also all the noise in your quiet studio (35-45 dB), there's analog-induced noise in your hardware, atmospheric noise, and even normal noise in your inner ear. And loud music on top of the mere -140 dB or so QD or dither...
Also, remember that you never actually hear a 32bFP signal, since your AD hardware converters are 20 bits, if you're lucky.

Just for fun, another revealing test, and it will definitely surprise you in a humbling way, is to see how loud you can have a track, any track, in the middle of your mix, FULL of QD before you even notice it (say, a piano track with a plugin that makes it 8 bits). Start with 8 and go up. By the time it's in 12-13 bits, you probably won't be able to tell the difference, at least not in the context of the mix. Hmmm....

[Timeline]

I seem to like the dither off and 32bit fp especially when mixing. Don't know why. Thanks for letting me know though. I'll leave it off for the time being.

[FMiguelez]

I seem to like the dither off and 32bit fp especially when mixing. Don't know why. Thanks for letting me know though. I'll leave it off for the time being.

My pleasure
Just remember that the Dither setting in the Audio menu is ONLY for freezing/merging files, offline plugin processing and that kind of stuff.

It does NOT do anything in the actual real-time mix, and for recording and bouncing it matters not if it's checked or unchecked, since these are different processes (you'd need a dither plugin for this if you want the noise, the usual way).

[Timeline]

How do you determine if a plug needs dither then. For instance the Motu limiter or the Precision UAD multi-band comp?

[FMiguelez]

How do you determine if a plug needs dither then. For instance the Motu limiter or the Precision UAD multi-band comp?

If the plugin developer is good, he will take care of all that properly himself (dithering if needed, handling up-sampling properly, etc.). But that's totally out of our control, and all we can do is trust the developer, or put it to the test with tools like Plugin Doctor.

If your question was meant more as when and where to dither ourselves, my suggestion is to only do it whenever you lower the bit depth of your music (i.e., when you master it and want to get a 24-bit print of DP's normal 32bFP handling). It should be the last step in the last plugin in your master fader chain, typically some kind of limiter, like DP's MW Limiter).
BTW, that one (and its little brother) are very cool because you can set any bit depth you want, not only 24 or 16 bits, which is wonderful for testing this stuff.

But this is almost splitting hairs, considering how un-hearably low the dither noise or QD levels are at 32-24 bits.

[Timeline]

Thanks. Great info.