Bit-depth info / question

Just to point out that even if the stair steps in that diagram did actually exist as voltage, that part of the wave would be at the sample frequncy, far above what anyone claims to be able to hear, above nyquist and filtered out. So it isn't there. Adding more samples doesn't get you anything, it won't make the reconstruction of a wave below the nyquist frequency any more accurate because there is no more accuracy to have.
 
snow lizard said:
This is PCM audio. I think there was a thread a few years back where we figured out that dBv values properly logged out correspond pretty closely but not exactly to the 6 dB per bit thing. Essentially the bit depth slices the volt into equal steps. More bits, smaller slices. The actual voltage change results in a 6 dB increase at approximately twice the voltage from where you started at.
Click to expand...
If each bit represents 6db, then 16 bit audio would have no way to express a level between -96dbfs and infinity. 24 bit audio has no way to express levels between -144 dbfs and infinity. The only gain with more bits is between the digital noise floor and no level at all.

You get more steps with 24 bit, but you are also covering more ground because you are pushing the noise floor down farther.
 
Ethan said:
It might seem that way,
Click to expand...

It is.

Ethan said:
but all that's affected is the noise floor.
Click to expand...

Perhaps not.

Ethan said:
What the "resolution" arguments miss is the reconstruction filter built into every D/A converter. This smooths the output waveform thus removing the steps. The attached figure is from my Audio Expert book, and it shows how the output waveform is identical to the input, within the frequency and noise limits of the current sample rate and bit depth.
Click to expand...

Cool! We agree that there are steps.

D/A isn't the only step in the process. Hang on, I'm going to post some screenshots of 8 bit, 440 Hz sine waves at different levels. The results are interesting. These are sine waves mind you, not complex waveforms like reverb tails or cymbal fades. Not quite sure what this has to do with real world applications but it might be worth noting.
 
Farview said:
I know this is meant for Ethan...But the steps never actually exist. There is no audio anywhere in the system that contains those 'steps'. The voltage changes are never stepped.
Click to expand...

Er, no "audio" containing steps but there are certainly discrete samples, each with a binary number representing such a step. Yes, these samples are re-drawn into a continuous waveform again before you can listen to the audio--but the samples DO exist in the system.
 
Yes, but it doesn't matter because it is never part of what you can hear. AND, the steps would be at the sampling frequency, so you still wouldn't hear them as they would be likely filtered out by the amp and or speaker system you were using to monitor the signal...and of course your ears...
 
Well, it does matter to some extent because any processing in your mix is math being performed on these samples, not on the processed audio version.
 
Bobbsy said:
Well, it does matter to some extent because any processing in your mix is math being performed on these samples, not on the processed audio version.
Click to expand...
Wouldn't you imagine that the processing would take that into account? A lot of plugins upsample to process, then downsample...

How did we get on a sample rate discussion when the original question was about bit depth?
 
Okay, let's start slow. This is a screenshot of a 440 Hz sine wave generated in Cool Edit, rendered to 8 bit PCM at -46.6 dBfs.

Notice anything odd?46.jpg
 
"How did we get on a sample rate discussion when the original question was about bit depth? "

Sorry Jay, c'est moir!
I started that off I think and went on to mention tape speed. Regarding which I would like to say I said "analogous to" that means "like" not "exactly the same as".

I was well aware of the bass woodles etc.

Dave.
 
Well, you can't really look at bit depth and sample rate in isolation since they are totally entwined. In a 16 bit/44.1 kHz situation you are getting 44,100 samples comprised of 16 zeroes and ones each every second.

Of course the processing takes into account the format of the samples it's dealing with. My point is that the samples themselves (i.e. what some are calling steps) are important in their own right since the only places you're working with those nice round waveforms are on the input and output. I don't know about your DAW but, on mine, (Audition) if I zoom far enough into the waveform view, I can see individual samples superimposed on the nicely curved line representing the waveform.
 
Here's a sine wave at -42 dBfs, 8 bit.

42sine.jpg

Here's -30dBfs.

30sine.jpg

-18 dBfs. Starting to look and sound more like a sine wave but it still isn't coming back at -18 and the distortion is quite audible and way above the noise floor.

18sine.jpg
 
Bobbsy said:
Please can you give details of your methodology (including sample rates). I can't duplicate this in the current version of Audition.
Click to expand...

I just opened the wave editor and created a new file. Mono, 44.1 sample rate, 8 bit audio. Click on generate>tones. The box allows you to select sine wave, frequency and level. It's also handy for generating a sine sweep for measuring room acoustics but I just had it at 440 Hz. No modulations or overtones were used, it's just a sine wave.
 
Here's a 16 bit sine wave at -48dBfs. Looks good, sounds like a sine wave and it's coming back at -48.

16bitsine.jpg


-60. Already starting to distort.


16-60.jpg


I'm pretty sure what we're seeing here is quantization errors.
 
snow lizard said:
I just opened the wave editor and created a new file. Mono, 44.1 sample rate, 8 bit audio. Click on generate>tones. The box allows you to select sine wave, frequency and level. It's also handy for generating a sine sweep for measuring room acoustics but I just had it at 440 Hz. No modulations or overtones were used, it's just a sine wave.
Click to expand...
So you didn't start with a sine wave and sample it at 8 bit, you told your program to generate one and this is what it gave you?
 
You must log in or register to reply here.
Back
Top