View Single Post
Old 06-10-2019, 07:31 AM   #9
Human being with feelings
Join Date: Sep 2010
Posts: 9,003

Originally Posted by 3buddhas View Post
So upsampling SD (is that 44.1???) to HD (48? 96?) frees the mastering lab's converter from having to anti-alias, giving the lab a more faithful version of the original. No quality or resolution is added by the upsampling, but less quality is lost in the DA conversion. This then would be a good argument FOR upsampling a mixdown headed for mastering, yes?
You could make the argument that a SD (44.1k or 48k) master should be upsampled into a HD (88.2k or 96k) container.

This opens a can or worms a little...

Here's the deal as I understand things (and my experiments to the best of my ability seem to confirm):

Aliasing is a thing with SD sample rates. It generates noise in the audible band if left to run wild. So converter units use steep eq filters to truncate just below the Nyquist frequency.

These are analog eq's. This is the biggest factor in the 'sound quality' of a particular audio interface (specifically it's AD or DA stage).

HD passively gets around this by the sampling frequency being so far above the audio band it's all a moot point. Thus some converter units (in audio interfaces and home receivers) sound more transparent at HD.

This is important for recording and playback. Where the AD and DA conversions occur. In between we're just shuttling ones and zeros around. They either get to the other side of the road or you get an obvious error.

We could just use 24/96 as a format and shut up and move on. And that's kind of exactly what happened! 24 bit HD flac files are the consumer format of choice for downloads and the bluray is the physical disc format. These support multichannel surround as well.

So the worms...

That could be interpreted as a suggestion to record/mix/master at HD. Reduce to SD for the consumer. Make the consumer upsample if they wish.

Some people seem to claim that there IS some magic in the ultrasonic artifacts above the audio band... even though speakers don't reproduce that range. Nor do some amps.

I figure, might as well preserve HD content when possible/reasonable since the consumer format is 24/96 nowadays.

SD masters? Seems wrong to have SD masquerade as HD with an upsample. Let's leave that to the consumer to do if they wish here.

So what's the worst we can do today? Mp3 right. The lower bitrates of mp3 aside, this still raises the bar for audio fidelity next to older analog formats gone wild. (The malfuncting cassette decks or the toyish turntables.)

With some squashed pop music with just an OK mix, you'd be hard pressed to tell the difference between a 24/96 bluray and a 180k mp3! A dynamic classical recording or something artsy like a Pink Floyd album will reveal damage from not only the mp3, but reducing 24 bits to 16 bits.

OK, so that last bit sounds like an argument against HD!
If only the outliers (classical music and Pink Floyd) benefit from HD and nothing else is perceivable... what the hell?

I've heard the outliers. Mix prowess is still the elephant in the room. All these lesser formats (even mp3) are 1% next to the mix work. Proper mastering presentation can be a big deal. It's an opportunity to do damage if something goes wrong too! CD's that are volume war slammed and have the treble cranked up to 11 are degraded that way intentionally. It's an aesthetic that is apparently preferred by the portable listening device crowd with CDs. Or it's literally shit mastering engineers doing a shit job and somehow nobody is calling them out and returning discs en masse. (I honestly think it's the latter.)

Well that's a lot of technobabble!

I've never heard any generation loss of any kind with 24/96 format.
24/96 is the modern high falootin format with downloadable FLAC files and bluray discs.
The older format of CD still sounds pretty complete full fidelity as long as it isn't mastered by a shit engineer (or bs online service).
Consumer tip: Try upsampling any of your SD albums if you're curious.

That's what I land on currently.

Oh, and the world needs more surround sound mixes!

Originally Posted by 3buddhas View Post
Not sure I get this point yet, but would like to. I've heard "Don't worry about mid-chain overs because with floating point math they can be fixed end-chain" (Andrew Scheps, Kenny Gioia). David Gnozzi argues mid-chain overs miss out on plugin sweet spots. I wonder if I'm even close to getting YOUR point.
The point in floating point math is that you can't clip your digital mixing system internally. Everything is "fixed" as it were in analog and thus you had to watch your internal signals in analog boards. It's still a good workflow practice to gain stage internally, but the floating point math literally eliminates the hard requirement. You can go over zero with floating point. When you get to the final output that feeds DA converters, now you need to go back to fixed point and respect the zero limit.
serr is offline   Reply With Quote