avatar

Mike Rivers

Aqua Marine

Posts: 2,528 Member Since: 13/10/2012

Lead

May 17 17 6:24 PM

Tags : :

The Grammy P&E Wing has a new document out in which they reasonably intelligently set up some guidelies for what's "Hi resolution." And of course being who they are, they think everyone should record high resolution no matter how the listener hears it.

I didn't go through the whole thing with a fine tooth comb, but it does contain recommendations such as upgrade your software and equipment early and often. Maybe when I start getting Grammys, I can afford/justify that. What they do address is that the end users may not be able or willing to keep up with the higest resolution technology. And while, given good examples, most can hear that HiRez sounds better, they acknowledge that not everyone cares enough to make it exclusive.

Read it here:

=14pxhttps://www.grammy.org/files/pages/recommendations_for_hi_resolution_music_production_05_10_17.pdf


 



For a good time, call mikeriversaudio.wordpress.com

Quote    Reply   
avatar

zmix

Aqua Marine

Posts: 4,169 Member Since:20/01/2011

#3 [url]

May 18 17 8:48 AM

seth wrote:
Thanks Mike. I have to wait for some down time to dig into that, but I'm looking forward to reading it.

Enjoy, Seth, but bear in mind that it's written by people who don't particularly understand how digital audio works, and have an agenda to convince people who know even less.

In the section on "Bit Depth", for example, they state:

"The exponential expansion from 16 bit audio to 24 bit audio results in a ratio of 256 steps (at 24 bits) for every 1 step (at 16bits)."

In binary numbering systems, as used in PCM Audio,  bits are not equal sized "steps",  each additional bit that is added to the word is a least significant bit, and has 1/2 the value of the bit above it.

There is no difference between the upper 16 bits of a 24 bit signal and a 16 bit signal.

Each additional bit simply reduces the noise (and noise due to quatization error, which is eliminated by dither).

A 16 bit word  has a theoretical THD+noise of   0.0015849%.

Adding a 17th bit reduces the noise and distortion to 0.0007943%.

Adding 8 additional bits reduces noise and distortion (theoretically) to 0.0000063%, though due to physical constraints, 0.0001% or less is unheard of.

The rule of thumb in the "Hi-Fi" days was that anything less than 1% thd+noise was inaudible.











Last Edited By: zmix May 18 17 12:08 PM. Edited 5 times.

Quote    Reply   
avatar

seth

Ruby Baby

Posts: 5,634 Member Since:26/01/2011

#4 [url]

May 18 17 9:05 PM

I will take it with a grain of salt, Chuck. I'm interested to see what they say, and I have the feeling that sooner or later someone will make reference to this document in my day-to-day work. If nothing else I want to be able to say I've read it.

Since I got my AP Portable One I've been looking at all my gear. The lowest THD+N on any of my preamps is about .01% @1kHz. Nowhere near any of my digital gear, so it's academic. I go with what sounds good to me. BTW, I discovered one of my API's had a bad output transformer. Not horrible, but inconsistent with the others.

Quote    Reply   
avatar

Blue Note

Tin Man

Posts: 35 Member Since:22/01/2017

#5 [url]

May 19 17 1:08 AM

Thanks for the link, Mr. Rivers. Glad to see RL weigh in. Wondering about the omission of Dan Lavry's claim that going beyond ~ 60 kHz Fs introduces low end distortions because the settling time of the chips in ADCs can't actually keep up with broadband accuracy. He essentially claims it's distorting the signal if we go above 24/48 (since there is no 24/60 format). He allows for 24/96, since it's healing something else (the "cold" sounding effect of 1x filtering). I think his last stance on this matter was 5 years ago, however. Have chips' settling times gotten much better since then? How much is the low end getting distorted when we switch from 48 kHz to 96 kHz? Got pics?

Bløn Øte

Quote    Reply   
avatar

Blue Note

Tin Man

Posts: 35 Member Since:22/01/2017

#7 [url]

May 19 17 7:17 PM

I agree that the anti-alias filter is the worst offender in 24 bit LPCM audio. However, sometimes something better is also worse in a different way from which it is better, and the net result is later on recognized as a zero gain, or less (i.e., loss). Has anyone other than Dan Lavry measured distortion in the low end of audio that is the direct result of running a converter at 192 as opposed to something lower? Lavry was so convinced that it is deleterious to fidelity to exceed 2x Fs that he didn't include 4x rates in his converters, even though it cost him sales and would have been trivial to implement, since any prosumer manufacturer can provide it. There's a lot of excitement about largesse in numbers, generally. We abandoned the long form of counting, decades ago, probably since the influential milliardaires among us wanted to be known as billionaires. Please show the proof that his worries have been solved by advances in chip settling time since 2012 or at least let's see what he was worried about in the first place. Graphs? Displays? Numbers?

-B. Ø.

Quote    Reply   
avatar

zmix

Aqua Marine

Posts: 4,169 Member Since:20/01/2011

#8 [url]

May 19 17 7:52 PM

What's been explained to me, and confirmed by several manufacturers is that there are not enough clock cycles at higher sample rates to compute the filter properly, so aliasing is much worse at 2x and 4x rates.

Last Edited By: zmix May 19 17 8:21 PM. Edited 1 time.

Quote    Reply   
avatar

gtoledo3

Aqua Marine

Posts: 4,181 Member Since:23/10/2013

#9 [url]

May 20 17 9:00 AM

Very interesting... makes sense.

There can't really be an equivalent to "offline rendering" for capturing live music... but so much can be handled more eloquently in the visual realm because of the possibility of that.

Quote    Reply   
avatar

maarvold

Aqua Marine

Posts: 3,145 Member Since:23/01/2011

#10 [url]

May 20 17 9:18 AM

zmix wrote:
What's been explained to me, and confirmed by several manufacturers is that there are not enough clock cycles at higher sample rates to compute the filter properly, so aliasing is much worse at 2x and 4x rates.

 
I've been happily working at 24-48 almost all the time in recent years.  Thanks for giving me a possible explanation why.  

Quote    Reply   
avatar

zmix

Aqua Marine

Posts: 4,169 Member Since:20/01/2011

#11 [url]

May 20 17 11:14 AM

I'd love to know what professionals think about these NARAS guidelines, in the meetings it felt like there was great pushback from working professionals and sycophantic approval from dabblers and dilettantes...

Quote    Reply   
avatar

weedywet

Ruby Baby

Posts: 5,957 Member Since:20/01/2011

#12 [url]

May 20 17 11:35 AM

on the one had, I've ALWAYS had issues with the idea of NARAS "guidelines", going back to their Pro Tools session delivery guidelines

There isn't a need to standardize everything in life.

OTOH, I think 96k almost always sounds better in actual implementation and it's what I use the great majority of the time.

Quote    Reply   
avatar

barry hufker

Diamond Forever

Posts: 12,201 Member Since:26/01/2011

#13 [url]

May 20 17 4:38 PM

I always use 96k to record and master.  I've been pleased with the results over the years.  Of course I downsample to 44.1k when necessary but keep a 96 k master.

Quote    Reply   
avatar

weedywet

Ruby Baby

Posts: 5,957 Member Since:20/01/2011

#14 [url]

May 20 17 4:50 PM

I play the 96k file D-A at mastering.
where, after any eq and leveling, it's transferred to consumer formats as well as archived at 24/96

Quote    Reply   
Add Reply

Quick Reply

bbcode help