Basic question about 16 vs 24 bit

[Electricide]

in conclusion, then

16 bit and 24 bit both record low volume sounds, even those below -96, but the quantization noise in 16 bit interferes with our ability to hear that information below -96.

And yes, the dB scales confuses the voltage/amplitude discussion because it's not linear, but exponential.

[wayne kerr]

16 bit and 24 bit both record low volume sounds, even those below -96, but the quantization noise in 16 bit interferes with our ability to hear that information below -96.

Yes. Actually they both record the exact same thing, it's just a matter of the signal being at a higher level than the cumulative level of self-noise and circuit noise from the rest of the signal path. Somebody a while back actually proposed a theoretical maximum of 32 bits (4,294,967,296 discrete intervals) which would yield a theoretical dynamic range of 192 dB and likely a functional range of about 180 dB. But all the research at the time went into stabilizing the clock to yield a 192K sample rate.

And yes, the dB scales confuses the voltage/amplitude discussion because it's not linear, but exponential.

Well, actually the dB scale expresses an acoustic power ratio and is logarithmic (which IS exponential, but a specific KIND). Voltage is only used to measure the electrical potential of a circuit, and as such can represent the amplitude of a signal between input and output transducers.

[Piotr]

It would seem that the proper approach to take is to look at the problem in terms of noise reduction instead of trying to figure out where all those quantization levels reside.

The 65,536 levels in a 16 bit system may be the same 96 dB as the top (bottom?) 65,536 levels out of 16,777,216 in a 24 bit system. But ultimately the difference won't be overall volume per say. We already know that a 1-bit system can be as loud as a 24-bit system. What distinguishes the higher resolution gear is the ability to push the noise floor well below the threshold of human hearing.

I wonder what dogs think???

[Electricide]

and I think the benefits of a 180 dB dynamic range are overkill..how many rooms in the world are actually that quiet. What the hell's the dynamic range of human hearing? Of the monitors we use? The mics we record with?

It would be good to remove as much noise as possible, but not just for rhetorical purposes.

[wayne kerr]

and I think the benefits of a 180 dB dynamic range are overkill..how many rooms in the world are actually that quiet. What the hell's the dynamic range of human hearing? Of the monitors we use? The mics we record with?

Ahh C-man, you just had to din't ya! Remember, dynamic range is just ONE of the important functions of word length. Now we're gonna have to get into the highly theoretical (and headache inducing) world of time domain and spatiality...

But after lunch.

[MoreSpaceEcho]

Now we're gonna have to get into the highly theoretical (and headache inducing) world of time domain and spatiality...

Bring it!

just let me take a nap first.

s

[Electricide]

No, I know where this is going. As I stated earlier, ambient noise, harmonics, reverb tails, are all better off having the noise floor reduced. And, the top end of that full scale would sound even better than 24 bit.

And let's not forget that the world isn't even on a standard 24 bit consumer delivery format yet.

what was the point of this post?

Chris

[wayne kerr]

No, I know where this is going. As I stated earlier, ambient noise, harmonics, reverb tails, are all better off having the noise floor reduced.

Actually, time domains and time smear have nothing to do with noise floor. But seriously, after lunch

[Jesse Skeens]

Man, I just read my post, and yes, my head hurts too... but maybe someone will find this helpful.

-- edited for clarity (coulda fooled me!) --

this make sense to me. i think my problem was thinking that the extra bits were down below -96. now it seems that they are actually mostly near the top end of the db scale. but what counts is the less noise from not as much rounding off errors that gives us the s/n ratio. and that since half of the bits reside from 0 to -6db essentially that you need the extra resolution of 24bit to compensate for the nonlinear dispertion of steps per db which causes low level signals to be nasty since there are actually less bits down there compared to the top.

[Jesse Skeens]

Ok all this being said then why do the mastering guys over at prosoundweb make it sound like you dont need 24bit if you have a certain level of noise in your system? because it sounds like what we are saying here is that 24bit not only gives you lower noise and better resolution across the whole board. they seem to think that practically the only good thing about 24bit is capturing reverb tails. yet if it does give more steps even near full scale then youd think it would be good for the whole dynamic range not just the bottom.

[zoetrope]

yet if it does give more steps even near full scale then youd think it would be good for the whole dynamic range not just the bottom.

I don't think so. Let's say I'm noise, and you're signal, and we stand at opposite ends of a hallway. We measure the distance between us in yard.
3 yards.
That's too close!
Now we measure it in feet.
9 feet.
Ah, that's better, now I have a better signal to noise ratio, right?
Wrong. We're just measuring in smaller steps, but the distance between us remains the same.

Let's say you have a $2 sound card that lives inside your computer, where there is all kinds of electronic noise bouncing around at all kinds of frequencies; as someone on another board put it, everything from DC to daylight. I'm not sure what that means but it sounds cool.
Anyway, that's all more noise, so the noise level goes up without any improvement in signal, regardless of what bit rate or brand of ketchup you use. So to go back to the hallway, I walk closer to you. You can measure in feet, inches, millimeters, whatever you like, but the basic problem is still too much noise. All those subtle reverb tails and things live below the noise level caused by all the computer zips and zaps and radioactive spiders and whatnot. So there's no point trying to measure the reverb tail more accurately and make the steps finer below a certain level, that level being the one that benefits most from the increase in bit rate.

To use one more awful analogy, it doesn't matter if you're measuring in yards, feet, or inches if you're so deep in the fog that you can't see what you're measuring.

I think I'm actually getting this now.

If, like me, you've got your soundblaster live! living in your computer, there's no real advantage in going from 16 bits to whatever, because you're just doing a better job of measuring noise. Keep the signal between 0db and -6db (100% and 50%) when you record, and you will be doing the best you can with what you've got. If the guts of your ad/da converter live in one of those fancy breakout boxes, and you record in a quiet space with a quiet computer, no cats, no kids, no cars, then you might be happier with 24 bit.

Again, I invite all to point out the flaws in what I said.

[Roboburger]

yes, but think of it this way. Pretty much everyone here records their snare at a pretty good level, I assume. 16 vs 24 doesn't make a difference to signals that are loud. It would only matter if you plan on tracking your snare at 48dB below max.

BUT

most of your ambient snare sound, reverb trails, sub-harmonics, and all the stuff that contributes to a clean recording with "depth" and "air" probably happens in those bottom 12 bits.

And let's not forget that all of these individual tracks are gonna get summed up together, and maybe processed in some way (Comp, delay, 'verb, etc). A more accurate recording of each individual wave will make a more accurate summation.

Leaving headroom (mentioned earlier in the thread, talking 'bout Nashville) is important to do if the track will going through some "Math" stages, like rendering new tracks, so that the ten or whatever tracks don't add up to numbers above 100% which would meant yucky digital clipping. However, If you are gonna just run the track directly out of the system to your board through it's own seperate output, I would recommend getting as hot as safely possible.

Snare drums have always been my least favorite part of recording digitally. If ya get a good inital crack outta the thing, then the fade of the hit is unsatisfying, so then I tried hard limiting to stop some of the peaks and let the tail end of the snare have more resolution, but then the inital crack wasn't as good.

I am switching to 24/96 real soon (just got a Layla off eBay) and am looking forward to getting some comparisons under my belt.