more headroom with 24 bit

  • Thread starter Thread starter wes480
  • Start date Start date
alright - i see what the concensus is now..makes perfect since. I think I am getting it...however...I am not sure that everything is exactly right...so, i need someone to prove me wrong on part of my way of thinking.

16 bit has 96db range, 0 to -96
24 bit has 144db range, 0 to -144

ok, that coincides with what everyone is saying..you can go softer with 24bit.

so, at 0db, in 16bit..its gonna look like this

1111 1111 1111 1111

at 0db in 24bit, it will look like this...

1111 1111 1111 1111 1111 1111

^--- and you are saying that those two are the exact same sound, at 0db?

I'm still hung up on the "0 is a reference point" thing. something in my logic is blocking me from thinking that the actual dB level of 0 wouldn't change with change in bit depth.

Does anyone at least see *why* i am thinking that? Maybe its just me...links to any good articles? hehe. I'm in trouble! :)
 
With digital 0db is different than analog. In analog 0 is considered the sweet spot. I am not too technical but I imagine this has something to do with sending out the best signal quality for line level. You can push it higher and depending on the quality of the equipment you could still get usable and even more pleasing results. That is where the term 'headroom' comes from. It was the ability to push the unit beyond 0. Because analog is a physical medium the better the tape or equipment the more you could push it. This applies to tape recorders and mixers.

In digital there is a hard coded limit. Once you pass that limit you get that absolutely useless digital distortion. They had to make that limit have some value so they made it 0db. You cant go past 0db in digital.

Now 0db is not a real world volume level. The real volume depends on what the user sets his volume at. 0db represents the best or highest possible signal and lowest self induced noise.

Now since 0db is the ABSOLUTE maximum in digital. The only place to go from there is down. It's kind of like how newer video camera chips have the ability to get more usable pictures from lower light sources. They can practically see in the dark. That is what 24bit does to audio. It can record much quieter sources and they can still be heard. This means that the much more subtle sounds of an instruments decay can now actually be recorded when before it would have been lost in the noise floor.

Did that make any sense?
 
Quote:- pchorman, I first used Pro Tools Free @ 24 bits, dithered that down to 16 bits and then did some editing with Cool Edit. It was in Cool Edit that I noticed all the peaks should have been overmodulated, but they weren't. Am I being misled by my VU meters? I then tried 32bit floating point recording with Cool Edit and confirmed that I can't trust my signal level meters any longer. I get no clipping when I go way beyond the 0dB threshold. Can anyone explain this?


Yes you can record beyond 0 db in cool edit. I do it al the time. Its the only software thats like recording to tape. You can push the meters into the red occasionally with out it clipping. The software lets you do it. Just make sure you final mix is below 0db.

Scott...
 
yes Tex, makes sense.

Let me sum all this up with one final question-

Is the loudest that a 16bit sound can be, the same loudness as a 24bit sound?

(16 1's, vs. 24 1's)
 
Is the loudest that a 16bit sound can be, the same loudness as a 24bit sound?
Click to expand...
Yes. 0dB, or full scale on your system, is as loud as anything can get. If you push the fader up when a continuous tone sound is already peaking at 0db, you will get distortion, and nothing more. Even if it's 8-bit, this is as loud as it gets.
 
They are both going to be equally loud. The main difference is that the 24bit recording will get the loud sound and more quiet stuff at the same preamp level.
 
I hope this doesn't confuse the matter...

First off, all digital audio tech nuts (like me) really should read Julian Dunn's "Measurement Techniques for Digital Audio" It's available from Audio Precision (they make a really cool audio test system) and it tells you pretty much everything there is to know about testing ADC's and DAC's.

That said, among various other parameters, one of things that he provides a number of concise methods for testing is "Input for Full-Scale Amplitude." Basically, it's a test to figure out what level (referenced to dbV) generates a full-code output (both positive- and negative-going.) One of the things he points out is that, if a converter has a nonzero DC-offset (in other words, if "zero" is actually a positive or negative output code), then either the positive or negative rail will be hit first, thus you cannot get the other maximum output code for a symmetrical signal (like a sine wave.) In this case, the resolution of each bit is constant.

While he does not explicitly state this, there are other aspects of the AtoD design that can alter what signal level (again, referenced to a real-world dbV) results in a full-code output. All ADC's have some kind of input buffer before actually driving the conversion circuitry. Besides potentially (or intentionally) bandwidth-limiting the signal, this can introduce gain errors as well. If the gain is greater than one-to-one, then a lower signal level will generate a full-code output. If the gain is less than one-to-one, a higher signal can be fed in before hitting full code. In both of these cases, the actual resolution of each bit is changing.

What that says is that you cannot direcly compare any two ADC chips, because internal variations can cause varying input levels to all give the same full-code output.

But we're not comparing ADC CHIPs, we're comparing full-blown coverter systems, which likely introduce their own gain stages, or some other kind of level control. This is where it gets really fuzzy. The only way you can tell what level will produce a full-code output is if the manufacturer specifies (ie +4dBU = full scale digital). Then you know that when the trim knob is at some nominal setting, a +4dBU signal WILL give you a digital 0dbFS. Then you can apply the theoretical dynamic range of either a 24bit or 16bit (or 20bit) converter to tell you the theoretical lowest-level signal that will turn on the least-significant bit on the converter. Of course, at this point system noise will limit that actual lowest-level signal you can digitize, so it will not be -144dBFS for a 24bit converter, but once you get above the noise floor, each bit will have whatever resolution results from its 0dBFS=xdBV equivalence.

Put differently, it's up to each manufacturer what INPUT level gives you 0dBFS, and it is again up to the manufacturer what 0dbFS becomes after DtoA conversion. But if you look at your specs, you can figure it all out and determine what input level will give you your desired output level WITHOUT getting digital clipping in between.
 
Thanks, that helped fill in some of my technical blanks.

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