alexis wrote:As an explanation for ONE of the factors for what he believes is inferior sound quality with today's recordings he states:
I was astonished to see that he had written that. Dr D is a very intelligent and knowledgable chap, but that is just misguided IMHO.
Narcoman and Elf have made all the sensible, relevant and pertinent points, but just to labour the point...
lower digital levels mean lower bit counts, and lower bit counts mean less detailed recordings ...
The first part of that statement is clearly true. For every 6dB below 0dBFS the audio will exercise one fewer bit in the 24-bit word that describes the amplitude.
However, to suggest that results in a less detailed recording is utter nonsense. It results in a noisier recording, certainly. The signal-to-noise ratio is degraded -- just as it is with any analogue system used in the same way. But the audio has no less 'resolution'.
I explained all this in an article called Digital Problems, Practical Solutions
four years ago.
The erroneus idea of reduced 'resolution' with lower wordlength usually comes from demos like this 8-bit (undithered) piano recording
in which you can clearly here distortions and artefacts and all manner of nasty things. Some would say that is proof of 'less resolution'...
However, that example is not representative of the complete digital process because it only demonstrates crude amplitude quantisation, which is inherently grossly non-linear. All digital audio systems worthy of the pro-audio industry use dithering to perfectly linearise the quantisation process. This results in no distortions, no artefacts, and only a smooth benign noise floor -- exactly like any analogue recording chain or device.
Here is the same piano recording
but properly dithered. You can hear a perfectly clean and detailed piano recording there, with no loss of resolution whatsoever. However, there is a constant noise floor...
Taking it to an extreme, here is a 3-bit recording of the same thing, but with 'noise-shaping' dither
which makes the noise floor slightly less objectionable through a spectral shaping process.
That's really just a three-bit recording. So if you set your recording levels to peak around -100dB that's pretty much what you'd be left with. Does that really sound like it's lacking resolution or detail? I can hear all the subtleties of the performance and the room acoustics buried in the noise.
I have seen and heard Prism Sound demonstrations where they have shown it perfectly possible to hear a clean, undistorted signal 15dB below the theoretical limit of a 16 bit system if proper dithering is employed with good quality converters.
So sorry... it's a common argument against digital audio, but it simply ain't true.
I grew up in this "new" generation of 24-bit sound cards, where "peak at -12dBFS or lower" seems to be the generally recommended way to record
It is the recommended way because it is the technically correct way, the same way as we evolved the use of analogue systems, and for exactly the same reasons.
The issue, as the Elf has said, is that people get scared when they seen unused space at the top of their meters, becaues they simply aren't used to seeing the headroom margin revealed in that way. Analogue meters just don't show the headroom margin -- but it's still there and for very important reasons.
And the 'warmth' thing is as Narco said -- a lack of masking harmonic distortions that make things sound cuddly and make mixing rather easier because the details aren't as exposed. It's not hard to introduce similar effects to a digital system if that's what you want, via plug-ins or external analogue hardware. But as far as I'm concerned having a clean and faithful recording chain is a good thing.... and I happily record with peaks under -10dBFS and average levels around -20dBFS without any concerns whatsoever.