"24/192 Music Downloads
...and why they make no sense"
Full article:
http://people.xiph.org/~xiphmont/demo/neil-young.html
-
ArchaicRecords
- Posts: 56
- Joined: Fri Mar 03, 2006 10:36 pm
- Location: USA: Lexington, KY
24 bit / 192 KHz sampling rate - is it more desirable?
archaicrecords.com
Dan Lavry has explained that ~55 kHz sampling rate would be ideal for avoiding digital audio's notorious anti-aliasing filter artifacts (closed image, high end smear, digititus, etc...) and for maintaining accuracy (which goes down for the low end of the spectrum, as sampling rate goes up). He claims that the settling time between clock pulses needed for accurate encoding and decoding requires more of a rest than 2x and higher rates of sampling apparently allow for, with today's chips. Nevertheless, the deleterious effect of the steep low pass filtering at 22.05 kHz, used in CD and similar 44.1 kHz digital audio is detrimental in practice - even if not, in theory. Higher sampling rates are noticeably less colored by the antialiasing - regardless of who fails to distinguish between them in a double-blind A/B/X, but on a bad hair day...
The high end D/A converters (Lavry, Benchmark, Crane Song, etc...) routinely upsample to at least 2x before reconstruction. 1x is so closed sounding by comparison that 48 kHz is a breath of fresh air, since the modest improvement over 44.1 kHz is right in the critical zone between harmful and only mildly harmful. Recording at 24 bits and 48 kHz makes digital audio almost a pleasure to listen to... The song and mix has to be very good to sound great at only CD rates, if you ask me. And the impact of 20 bits at 48 kHz is quite illuminating compared to 16/44.1 such that the slight increase in storage and transfer capacity seems picayune.
So, I agree that 192 kHz is beyond audio and is wasteful of resources and even not accurate, compared to slower rates. But I also disagree that 44.1 kHz is high enough. Vinyl can easily outdo that, in spite of how the heavy hitters are filtering. The program should be filtered in band - not the medium!
16 bits is even theoretically more than 96 dB. The formula is something like 6.02N + 1.xxx + process gain, due to band limiting. So, even in theory there's over 98 dB. But, when you play a song through a requantizer which is turning the 24 bit source into a 16 bit destination, the 16 bit output is always a little _bit_ worse. (; You do the math.
Cheers,
Spike Volta
The high end D/A converters (Lavry, Benchmark, Crane Song, etc...) routinely upsample to at least 2x before reconstruction. 1x is so closed sounding by comparison that 48 kHz is a breath of fresh air, since the modest improvement over 44.1 kHz is right in the critical zone between harmful and only mildly harmful. Recording at 24 bits and 48 kHz makes digital audio almost a pleasure to listen to... The song and mix has to be very good to sound great at only CD rates, if you ask me. And the impact of 20 bits at 48 kHz is quite illuminating compared to 16/44.1 such that the slight increase in storage and transfer capacity seems picayune.
So, I agree that 192 kHz is beyond audio and is wasteful of resources and even not accurate, compared to slower rates. But I also disagree that 44.1 kHz is high enough. Vinyl can easily outdo that, in spite of how the heavy hitters are filtering. The program should be filtered in band - not the medium!
16 bits is even theoretically more than 96 dB. The formula is something like 6.02N + 1.xxx + process gain, due to band limiting. So, even in theory there's over 98 dB. But, when you play a song through a requantizer which is turning the 24 bit source into a 16 bit destination, the 16 bit output is always a little _bit_ worse. (; You do the math.
Cheers,
Spike Volta
-
montalbano
- Posts: 139
- Joined: Tue Aug 11, 2009 3:03 pm
- Location: Settala (MI), Italy
- Contact:
One band has supplied to us a 192Khz master some time ago
The record has sold very well in the indie rock scene here and the band was quite respectable FYI
They paid remarkable money to a recording studio for that wonderful job
When I got the master, I was astounded by the size of the files, and my mood was already bad that day; so the first thing I've done was to make a copy and convert it to 44,100 just to check
Then I counterphased the two versions (192 against 44,100) to check the differences
The difference was just a -60/-70dB quantizing noise
It is sad to see how people in the studios are getting money from the bands to sell nothing concrete
The record was cut with 44,100 files, 24 bits, and everything went OK.
In my view 24 bits have a real impact on the sound, while the sampling rate as this article says can go up to a certain limit, beyond which it is meaningless.
Anyway, nowadays that vinyls seem to be more a gadget or a "cool thing" to show and sell, people pretend to have 8 minutes cut on a 7" side or 30 minutes on a 12", they supply a master full of counterphased signals, and sometimes they even have the guts to complain about the quality of the sound or the playback skipping on some cheap turntables.
I start thinking that one should stick to RIAA standards, publish them on the site, and that's it, like Mossy said more than once.
I have recently dealt with a little complaint from a customer who told me that by mistake I'd cut his 7" @ 33 rather than 45. (he hadn't specified the speed in the order and the speed wasn't mentioned in the label as well).
The music was a drone-style music, very dark, with lots of basses. There were not high frequencies, so since the length was about 5:00 per side I decided to go 33 because that way I could have kept a decent output level and the trebles and the high freqs were so weak that even @ 33 rpm they had sounded OK. On the other side, to make the groove fit @ 45 I'd have to cut with a much lower output level and cut around 60Hz in order to limit the horizontal modulation.
It took me one hour to convince the customer but I am not sure he understood
The record has sold very well in the indie rock scene here and the band was quite respectable FYI
They paid remarkable money to a recording studio for that wonderful job
When I got the master, I was astounded by the size of the files, and my mood was already bad that day; so the first thing I've done was to make a copy and convert it to 44,100 just to check
Then I counterphased the two versions (192 against 44,100) to check the differences
The difference was just a -60/-70dB quantizing noise
It is sad to see how people in the studios are getting money from the bands to sell nothing concrete
The record was cut with 44,100 files, 24 bits, and everything went OK.
In my view 24 bits have a real impact on the sound, while the sampling rate as this article says can go up to a certain limit, beyond which it is meaningless.
Anyway, nowadays that vinyls seem to be more a gadget or a "cool thing" to show and sell, people pretend to have 8 minutes cut on a 7" side or 30 minutes on a 12", they supply a master full of counterphased signals, and sometimes they even have the guts to complain about the quality of the sound or the playback skipping on some cheap turntables.
I start thinking that one should stick to RIAA standards, publish them on the site, and that's it, like Mossy said more than once.
I have recently dealt with a little complaint from a customer who told me that by mistake I'd cut his 7" @ 33 rather than 45. (he hadn't specified the speed in the order and the speed wasn't mentioned in the label as well).
The music was a drone-style music, very dark, with lots of basses. There were not high frequencies, so since the length was about 5:00 per side I decided to go 33 because that way I could have kept a decent output level and the trebles and the high freqs were so weak that even @ 33 rpm they had sounded OK. On the other side, to make the groove fit @ 45 I'd have to cut with a much lower output level and cut around 60Hz in order to limit the horizontal modulation.
It took me one hour to convince the customer but I am not sure he understood
Phil from Phono Press, Milan, Italy
http://www.phonopress.it
http://www.phonopress.it
When would people will wake up and cut to well established rules as the RIAA standards
I have yet to find any damn article to tell me that human ears have evolved in the last 100 odd years since recorded sound come about
It is not surprising to see graphs from the 30's in so far as to the abilities of that part of human anatomy used today as nothing has changed
All these standards where put together by experts as well as a lot of input from a very wealthy and strong industry with millions of pressings already out there until such time as all agreed to them
Isn't this enough? why do people want to reinvent the wheel? it is beyond me
Loud cuts 7" at 33 rpm or 5+ minutes on a 7" at 45???
We just flatly refuse to do them We stick to the standard and that's it
Want to cook your cutting head so as to satisfy some 5 minute vinyl expert who is going to give you an order for 300 pressing and wants a "phat" cut Well there are people out there to do it they can go there Alternative! Go to it at your peril
The guys who rewind heads want to make a living as well
The point is: That All of this is is quite irrelevant in so far as vinyl is concerned cut in a professional environment sticking to well established standards standing for many years as proof
In so far as the loudness wars the sampling rate the bits and all of that is another kettle of fish which has no relevance whatsoever in what we do
A state of the art D to A converter is of course a necessity and that's where it all ends in so far as cutting a lacquer is concerned
Given that the mixing as well as the mastering has been attended to it becomes quite irrelevant in so far as vinyl is concerned as one still has to tweak the track here and there so as to get a satisfactory result
And of course that is the art of the cutting guy who has to know his machine like the back of his hand as well as its nuances as well as its peculiarities
Cheers
I have yet to find any damn article to tell me that human ears have evolved in the last 100 odd years since recorded sound come about
It is not surprising to see graphs from the 30's in so far as to the abilities of that part of human anatomy used today as nothing has changed
All these standards where put together by experts as well as a lot of input from a very wealthy and strong industry with millions of pressings already out there until such time as all agreed to them
Isn't this enough? why do people want to reinvent the wheel? it is beyond me
Loud cuts 7" at 33 rpm or 5+ minutes on a 7" at 45???
We just flatly refuse to do them We stick to the standard and that's it
Want to cook your cutting head so as to satisfy some 5 minute vinyl expert who is going to give you an order for 300 pressing and wants a "phat" cut Well there are people out there to do it they can go there Alternative! Go to it at your peril
The guys who rewind heads want to make a living as well
The point is: That All of this is is quite irrelevant in so far as vinyl is concerned cut in a professional environment sticking to well established standards standing for many years as proof
In so far as the loudness wars the sampling rate the bits and all of that is another kettle of fish which has no relevance whatsoever in what we do
A state of the art D to A converter is of course a necessity and that's where it all ends in so far as cutting a lacquer is concerned
Given that the mixing as well as the mastering has been attended to it becomes quite irrelevant in so far as vinyl is concerned as one still has to tweak the track here and there so as to get a satisfactory result
And of course that is the art of the cutting guy who has to know his machine like the back of his hand as well as its nuances as well as its peculiarities
Cheers
Chris
-
Aussie0zborn
- Posts: 1828
- Joined: Sat Mar 11, 2006 8:23 am
- Location: Australia
- Contact:
montalbano wrote:One band has supplied to us a 192Khz master some time ago
The record has sold very well in the indie rock scene here and the band was quite respectable FYI
They paid remarkable money to a recording studio for that wonderful job
When I got the master, I was astounded by the size of the files, and my mood was already bad that day; so the first thing I've done was to make a copy and convert it to 44,100 just to check
Then I counterphased the two versions (192 against 44,100) to check the differences
The difference was just a -60/-70dB quantizing noise
Curious as to how you did a Null Test on files that are at different sample rates?
montalbano wrote: It is sad to see how people in the studios are getting money from the bands to sell nothing concrete
The record was cut with 44,100 files, 24 bits, and everything went OK.
In my view 24 bits have a real impact on the sound, while the sampling rate as this article says can go up to a certain limit, beyond which it is meaningless.
Even worse - deleterious. Still, there is better-looking audio at higher than 44.1 kHz. This is not audiophile conjecture. As I mentioned, the no-nonsense designer Lavry has deliberately eschewed 192 kHz as market-driven and an overkill which interferes with input = output accuracy, ironically. But 2x is his defacto upsampling rate, since the slight imperfection in accuracy is less noticeable than CD rate anti-aliasing filters measured performance can deliver - on scope, MLSSA, or perhaps his decoder ring - Ziggy?.
I disagree with listening tests as prone to unnecessary complexity when methods exist for measuring performance, rather than merely judging it. As I mentioned in another forum, recently, the esteemed loudspeaker designer, Dr. Dunlavy, didn't use listening tests to create new products. Only writing things down and then looking at meters was his method. It is used by many other successful high end professional audio manufacturers, too. Most of these lab types are saying 44.1 was never enough. 55 kHz will never happen. But we already have 48 khz, and it's, in my opinion, a sweeter settting on all my ADCs than 44.1 is. My processing transfers are usually at 88.2, rather than 96k, fwiw, but that's cause it makes the signal processing of brick wall limiting sound smoother, and the lab coats are saying that it should. But the lab coat I like best doesn't say more is always best. Too much is too much. What we have already is close. Too bad it wasn't widely implemented outside of television production.
Sounds to me as if DSP is the best reason for high sampling rates. The anti-alias filter should be out of the way during analog modeling. If you have to upsample for DSP, at least try not to downsample any further than necessary - if at all. Bingo - another reason for high sample rate releases - no unnecessary DSP after the song is mixed and premastered. I realize we are still living in the past when it comes to the promises of the '90's. Remember when Mytek Digital converters were used in like 2000 (last Millenium/Century) to encode to 24/96 5.1 audio a jazz performance, recorded at McGill University, in Canada, and decoded in real (electrical propagation / convertion latency buffer speed) time, at UCLA, along with video, over Internet2? I thought we'd all be sharing complete Purple-Ray DVD's worth of content in seconds over the internets by now... 2012??? Come on!
Or modify RIAA to make room for synth treble energies... (: Big Al says it's from a different era of performing, miking, and sound crafting. Those 8+ dB SNR boost are welcome.montalbano wrote: Anyway, nowadays that vinyls seem to be more a gadget or a "cool thing" to show and sell, people pretend to have 8 minutes cut on a 7" side or 30 minutes on a 12", they supply a master full of counterphased signals, and sometimes they even have the guts to complain about the quality of the sound or the playback skipping on some cheap turntables.
I start thinking that one should stick to RIAA standards, publish them on the site, and that's it, like Mossy said more than once.
montalbano wrote:I have recently dealt with a little complaint from a customer who told me that by mistake I'd cut his 7" @ 33 rather than 45. (he hadn't specified the speed in the order and the speed wasn't mentioned in the label as well).
The music was a drone-style music, very dark, with lots of basses. There were not high frequencies, so since the length was about 5:00 per side I decided to go 33 because that way I could have kept a decent output level and the trebles and the high freqs were so weak that even @ 33 rpm they had sounded OK. On the other side, to make the groove fit @ 45 I'd have to cut with a much lower output level and cut around 60Hz in order to limit the horizontal modulation.
It took me one hour to convince the customer but I am not sure he understood
Tell them that if the mix had been right, it could have been a best-seller since "Sir Duke" was the last track on one of the LP sides of Songs in the Key of Life. That's 33 1/3 rpm - last song. No dullness to those horns. Probably some welcome self-erasure! (:
Cheers,
Spike Volta
-
montalbano
- Posts: 139
- Joined: Tue Aug 11, 2009 3:03 pm
- Location: Settala (MI), Italy
- Contact:
[quote="Serif"]Curious as to how you did a Null Test on files that are at different sample rates?[(/quote)]
I first converted the copy to 44,100 and left the original to 192. Then I routed the two stereo signals, original and copy (converted) to two different D/A converters, of course one set to 44,100 and the other to 192, same brand, same model, one is the master and the other is the "slave" (= spare) like most "fundamental" elements in our plant (from the water pump feeding the cooling circuit of the presses, to the Sx74 head. The reason is that if something gets damaged, you keep going with the spare, while the original is being fixed. We don't have a spare steam boiler though and there is not a spare Montalbano as well...)
OK, at that stage we were already on the analog domain coming from two identical sources so the two D/A converters outputs were routed to 4 mono channels on a Soundcraft TS24 mixer, L/R panpotted to wide open Stereo, and PRE buttons on to keep their original levels.
Every channel of the TS24 has the phase invert button so it was easy to check the differences.
I also tried with another way; the TS24 has two different circuits to route the signal, the MIX and the GROUP modes. The MIX mode skips the faders and the levels of each channel are adjustable by potentiometers which you can set exactly to the desired level because their movement is set by ticks, now if you set to 0dB all four channels and invert the phase of 2 of them, when you playback them together, their sum is routed directly to the control room.
Now being the two versions 100% counterphased, what you will hear is the real difference between the 44,100 and the 192Khz version. And you hear it already in the analog domain, I mean you hear the difference which would be routed to the SAL.
Now apparently I didn't hear nothing, just a very weak signal. It was not music but a digital noise. It is called quantizing noise. I measured it, very quickly, by using the VU meters of the master section, and it was around -70dB. That's it.
I first converted the copy to 44,100 and left the original to 192. Then I routed the two stereo signals, original and copy (converted) to two different D/A converters, of course one set to 44,100 and the other to 192, same brand, same model, one is the master and the other is the "slave" (= spare) like most "fundamental" elements in our plant (from the water pump feeding the cooling circuit of the presses, to the Sx74 head. The reason is that if something gets damaged, you keep going with the spare, while the original is being fixed. We don't have a spare steam boiler though and there is not a spare Montalbano as well...)
OK, at that stage we were already on the analog domain coming from two identical sources so the two D/A converters outputs were routed to 4 mono channels on a Soundcraft TS24 mixer, L/R panpotted to wide open Stereo, and PRE buttons on to keep their original levels.
Every channel of the TS24 has the phase invert button so it was easy to check the differences.
I also tried with another way; the TS24 has two different circuits to route the signal, the MIX and the GROUP modes. The MIX mode skips the faders and the levels of each channel are adjustable by potentiometers which you can set exactly to the desired level because their movement is set by ticks, now if you set to 0dB all four channels and invert the phase of 2 of them, when you playback them together, their sum is routed directly to the control room.
Now being the two versions 100% counterphased, what you will hear is the real difference between the 44,100 and the 192Khz version. And you hear it already in the analog domain, I mean you hear the difference which would be routed to the SAL.
Now apparently I didn't hear nothing, just a very weak signal. It was not music but a digital noise. It is called quantizing noise. I measured it, very quickly, by using the VU meters of the master section, and it was around -70dB. That's it.
Phil from Phono Press, Milan, Italy
http://www.phonopress.it
http://www.phonopress.it
-
montalbano
- Posts: 139
- Joined: Tue Aug 11, 2009 3:03 pm
- Location: Settala (MI), Italy
- Contact:
Sorry I can't edit my post
Just wanted to correct a little imperfection in my post, MIX mode of the Soundcraft actually doesn't get rid of the faders but rather sums the MIX signal whose level is set by the MIX potentiometers of every channel to their respective levels as set by faders. If you move faders to -oo you hear only the MIX signal
Anyway I am more a pressing man than a cutting man, as for this 192 kHz thing in my view one should ask Gengy who's the best and most famous cutter here or any other cutter with a loooong experience
But I have the feel they will tell you the same story
Just wanted to correct a little imperfection in my post, MIX mode of the Soundcraft actually doesn't get rid of the faders but rather sums the MIX signal whose level is set by the MIX potentiometers of every channel to their respective levels as set by faders. If you move faders to -oo you hear only the MIX signal
Anyway I am more a pressing man than a cutting man, as for this 192 kHz thing in my view one should ask Gengy who's the best and most famous cutter here or any other cutter with a loooong experience
But I have the feel they will tell you the same story
Phil from Phono Press, Milan, Italy
http://www.phonopress.it
http://www.phonopress.it
That's what happened to me. I think this forum is locked from user edits or something. I can still edit older posts that appear in different forums...
LOL
What D/A converter, may I ask, was playing at only 44.1 kHz?
Can't find one here that doesn't push up the 1x LPCM analog output to 2x bandwidth via automagic upsampling. Also, can't find an analog summing mixer around here with the noise floor I'd like for this test. Must build network...
Incidentally, not to drop names, but to illustrate the widespread belief among brilliant and learned designers, you can't play back at only 44,1 kHz anymore on the following DACs: Lavry (pick a model), Benchmark, CS Avocet, Mytek Digital, USA!, Weiss, etc... These are not companies who pander to audiophiles, either. The reason Lavry stands out as a good example, here, is that he, too, dismisses the logic of going beyond 2xFs for audio. Yet there's no way to trick his D/A not to upsample a 1x source at least to 88,2k.
Also, I wonder what the deal is. My mentor who is a legend but wishes to go nameless (lest he be associated with the likes of me!) says that theory is actually leaning to needing significantly more bandwidth than the highest partial of the highest fundamental for the following reason: Not because we have bat ears... But due to the need for realistic impulse response, essentially. When an orchestra plays a chord, rather than a single instrument, there are so many partials happening at different frequencies at once, that the wavefront approaches that of a square wave briefly - during the transients. If the DUT doesn't pass a 10 kHz square wave without droop, there can be no good sound out of it as it compromises those glistening instants of the leading edge of realistic sound propagation. I leave dissection of this morsel of thought to the floor as pizza awaits.
Until tomorrow,
- Chuck Spindle
LOL
What D/A converter, may I ask, was playing at only 44.1 kHz?
Can't find one here that doesn't push up the 1x LPCM analog output to 2x bandwidth via automagic upsampling. Also, can't find an analog summing mixer around here with the noise floor I'd like for this test. Must build network...
Incidentally, not to drop names, but to illustrate the widespread belief among brilliant and learned designers, you can't play back at only 44,1 kHz anymore on the following DACs: Lavry (pick a model), Benchmark, CS Avocet, Mytek Digital, USA!, Weiss, etc... These are not companies who pander to audiophiles, either. The reason Lavry stands out as a good example, here, is that he, too, dismisses the logic of going beyond 2xFs for audio. Yet there's no way to trick his D/A not to upsample a 1x source at least to 88,2k.
Also, I wonder what the deal is. My mentor who is a legend but wishes to go nameless (lest he be associated with the likes of me!) says that theory is actually leaning to needing significantly more bandwidth than the highest partial of the highest fundamental for the following reason: Not because we have bat ears... But due to the need for realistic impulse response, essentially. When an orchestra plays a chord, rather than a single instrument, there are so many partials happening at different frequencies at once, that the wavefront approaches that of a square wave briefly - during the transients. If the DUT doesn't pass a 10 kHz square wave without droop, there can be no good sound out of it as it compromises those glistening instants of the leading edge of realistic sound propagation. I leave dissection of this morsel of thought to the floor as pizza awaits.
Until tomorrow,
- Chuck Spindle
-
Steve E.
- Site Admin
- Posts: 1916
- Joined: Fri Jun 24, 2005 3:24 pm
- Location: Brooklyn, New York, USA
- Contact:
montalbano wrote:Sorry I can't edit my post
Weird....I checked the settings. You _should be able to edit posts, though you can't delete them. I dunno what's going on....Serif wrote:That's what happened to me. I think this forum is locked from user edits or something. I can still edit older posts that appear in different forums...
LOL
(end tangent)