24 bit dynamic level difference

  • Thread starter Thread starter Big Mike B
  • Start date Start date
Farview said:
I never understood intersample peaks. If anyhing happens between the samples, it would be happening at a frequency above nyquist and would be filtered out anyway. What am I missing?
Click to expand...

As you say the low pass filter takes out the higher frequencies. Sharp corners on waveforms indicate high frequencies, like if there were straight lines drawn from sample to sample. So the low pass filter rounds off those corners. If two of those corners are at or very near 0dBFS then the filter may not allow the waveform's line to turn that corner sharply enough to avoid going over.

You don't get this with a simple A/D/A path as the output voltage = the input voltage. But as soon as you start processing in the digital domain you can create digital audio that can produce analog peaks higher than what 0dBFS represents. Some converters have headroom for this and others don't
 
Ok, got the intesample peak thing. It makes perfect sense now.

As far as the dynamic range difference between 16 and 24 bit goes, if all bits at zero is negative infinity, lowest bit at 1 is -96dbfs in 16 bit and -144dbfs in 24 bit. In other words, the jump from silence to the quietest signal it can store is different depending on how many bits you are using.
 
Farview said:
As far as the dynamic range difference between 16 and 24 bit goes, if all bits at zero is negative infinity, lowest bit at 1 is -96dbfs in 16 bit and -144dbfs in 24 bit. In other words, the jump from silence to the quietest signal it can store is different depending on how many bits you are using.
Click to expand...

From what I've read it's about like that. That "jump" is probably the noise floor. Really, 16 bit is "enough", but it's really nice to have the extra room.
 
Wow a wee bit beyond newbie answers. Esoteric maybe LOL.
 
On thing that no one has mentioned is that when you convert the 24 bit files to 16 bit for CD's, you actually loose the noise floor to large degree as the noise is the 1st to go. This is a good reason to work in 24 bit for as long as possible before conversion.

Alan.
 
YOu lose the 24 bit noise floor, but replace it with a 16 bit noise floor, which is louder.
 
I was taking about the noise that existed in the 24 bit recording, over the multiple tracks, when the tracks are mixed down that noise is summed. When the 24 bit is converted to 16 bit the first thing to go is the noise as the dynamic range is decreased. I have actually heard this happen.

Alan.
 
It is instructive to consider the input and output amplifiers to the A/D-D/A converters themselves?

Douglas Self in his Small Signal Amplifiers give some useful figures and circuits.

The input drive need for 0dBFS for many converters is only about 5V peak to peak or some 1.7V rms and many are less than this.

That implies that for a noise floor of a typical, (not brilliant!) -100dB the noise of the drive circuitry must be less than 17microvolts, some 37dB BELOW the puny level specc' of the SM7b!

Then the FS output is the same so the output amplifier needs to be as quiet but also have a gain of some 20dB in pro gear if 0dBFS = ~+24dBu.

I think this shows that the converters themselves are essentially noiseless for 24 bit op' and all the design problems are analogue!

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