[MLA] MLA web site/ Resolution Loss? No.....

[Kevin Gross]

I agree with what you're saying here. I think we are hung up on
terminology. We need something shorter than a paragraph. I used
"resolution". You suggest "loudness to noise ratio". The problem I have
with that is that people may take the noise to include any source of noise
(e.g. acoustic noise in the recording studio).

In any case, there are many professionals and audiophiles who obsess over
digital gain staging and any reduction in useful information in their
signals. Of course they should be looking at the weakest links first (e.g.
acoustic noise in the recording studio) but they tend to wave that approach
off claiming, "it all adds up" which is difficult to argue with.

Kevin Gross - AVA Networks

On Thu, Dec 21, 2017 at 5:17 AM, Bob Katz <bobkatz at digido.com> wrote:

> As far as resolution loss.... stop using that term. Here's why: You can be
> 100% sure that Bob Stuart properly dithers his processing. And since there
> is already attenuation or gain built into MQA, normalization does not
> contribute one single added calculation. It's completely trivial to include
> loudness normalization within MQA. So the first argument is there are no
> additional calculations. Since dithering removes all distortion and
> produces random noise at -141 dBFS, that's quieter than ANY analog noise
> floor EVER, in ANY analog system. Dithered digital calculations perform
> EXACTLY like analog, as the signal gets softer, it gets closer to a
> constant noise floor, only the digital noise floor is much quieter.
>
> So we need to start talking about the LOUDNESS and the NOISE LEVEL. The
> facts are that digital domain attenuation and analog domain attenuation are
> functionally equivalent, here's the proof:
>
> Stop thinking bits. Start talking about LOUDNESS to NOISE ratio, not
> SIGNAL TO NOISE RATIO. YOUR COMPARISON IS BETWEEN DIGITAL DOMAIN
> ATTENUATION VERSUS THE SO-CALLED "HOLY GRAIL" OF ANALOG DOMAIN ATTENUATION.
> That's supposedly your concern when you bring up the term "losing bits".
>
> Use your TARGET level to do a LOUDNESS TO NOISE RATIO calculation. Let's
> take the worst case scenario, using an extremely conservative TARGET, of,
> say -23 LUFS. If all music is normalized to -23 LUFS. Simplifying (a
> reasonable approximation), assuming a DAC noise floor of -120 dBFS RMS,
> then the signal-to-DAC-noise ratio is 120 minus 23, which is 97 dB. Now,
> consider the SPL, again worst case scenario. If you play music loudly, at
> 85 dB SPL. Then since -23 LUFS is 85 dB SPL, then the SPL of the DAC noise
> is 85 minus 97, which is MINUS 12 dB SPL. So the LOUDNESS to NOISE ratio
> (DAC noise) is 97 dB. That's 97 dB BELOW 85 dB SPL!!!!! If you play music
> softer, then the DAC noise is even quieter. This is with 0 dB attenuation,
> NO ATTENUATiON for the widest range, softest piece of classical music.
> Keep in mind that the 16 bit compact disc is, was and forever will be
> quiet enough for the widest range classical music, without any perceptible
> noise, except in rare cases of an extremely wide range piece of classical
> music in the quietest room. Even there, the 16-bit dither noise is quite
> tolerable. If the argument is that somehow  "old-fashioned" analog domain
> volume control is superior to digital domain, remember that the widest
> range piece of classical music would require NO digital attenuation at all.
> So with either method of atttenuation, you still end up with an inaudible
> noise floor of the DAC in the room at MINUS 12 dB SPL and a LOUDNESS TO
> NOISE ratio of 97 dB. Since it's dithered, there are no added distorion
> components.
>
> When do we have to attenuate? When the signal is LOUD. So let's compare
> the situation using a LOUD rock piece with a level of, say, -12 LUFS. We
> need 11 dB of attenuation to get to a target of -23 LUFS. With digital
> domain attenuation, the analog noise floor of the DAC remains constant, so
> you still end up with a DAC noise floor of MINUS 12 dB SPL, and a signal
> level of -23 LUFS, which is the same situation as the classical. In other
> words, if you were satisfied with the SNR of the classical music, then the
> rock performs with EXACTLY THE SAME LOUDNESS TO NOISE FLOOR after digital
> domain normalization. THE LOUDNESS TO NOISE FLOOR REMAINS CONSTANT.
>
> Gotcha?  Not really. Does the analog domain attenuation afford ANY audible
> improvement? With analog domain attenuation of 11 dB there is a theoretical
> "improvement" of 11 dB compared to the digital. So your LOUDNESS TO NOISE
> ratio "improves" from 97 dB to 108 dB. So the DAC noise goes down from -12
> dB SPL to MINUS 23 dB SPL. Basically your LOUDNESS to noise ratio
> "improves" from COMPLETELY INAUDIBLE to EVEN MORE COMPLETELY INAUDIBLE.
> Assuming you have a quiet analog line stage that's quieter than your DAC.
> The argument moves from ridiculous to even more completely ridiculous. Is
> this acoustically even measurable with any measurement gear in a real room,
> or even in an anechoic chamber?
>
>
> BK
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://grimmaudio.nl/pipermail/mla_grimmaudio.nl/attachments/20171229/5e609368/attachment.html>