16-bit... It really is enough!

[danadam]

http://www.2l.no/hires/index.html

I am not sure how indicative it is for requiring more bitdepth or not, but one of the things that SoX stats command shows is a minimum RMS per 50 ms window:

-108.03 dB  2L-038_MQA2016-88k-24b_01.flac            MOZART Violin Concerto in D major KV 218, I. Allegro / MOZART Violin Concertos (MQA remix 2016)
 -71.14 dB  2L-053_04_stereo-96kHz-24b.flac           Haydn String Quartet In D, Op. 76, No. 5 - Finale - Presto / STRING QUARTETS vol. I Haydn - Solberg - Grieg
 -69.62 dB  2L-056_03_stereo_96kHz.flac               VIVALDI Cantata Rv 679: Che Giova Il Sospirar, Povero Core - Recitativo II / Bellezza Crudel - VIVALDI
 -70.07 dB  2L-056_04_stereo_96kHz.flac               VIVALDI Cantata Rv 679: Che Giova Il Sospirar, Povero Core - Aria: Cupido, Tu Vedi / Bellezza Crudel - VIVALDI
 -71.02 dB  2L-064_stereo96kHz_01_08.flac             From Let Us Garlands Bring, Op.18: Come Away, Death / Come Away, Death
 -62.07 dB  2L-077-stereo-96kHz_21.flac               Undring / harmOrgan
 -76.79 dB  2L-082_stereo-96kHz_01.flac               Ubi Caritas / Piano Improvisations
 -67.94 dB  2L-084_stereo-96kHz_11.flac               Eat! Drink! (Uriah, David) / David and Bathsheba, an opera-oratorio by Ståle Kleiberg
 -86.68 dB  2L-086_stereo-96kHz_15.flac               Bozza: Children’s Overture / La Voie Triomphale
 -60.13 dB  2L-087_stereo-96kHz_06.flac               Blågutten / Quiet Winter Night
 -75.95 dB  2L-092_stereo-96kHz_01.flac               Living / LIVING
 -60.75 dB  2L-093_stereo-96kHz_03.flac               Schubert String Quartet No. 14 in D minor D. 810: III. Scherzo. Allegro molto / The Schubert Connection
 -67.38 dB  2L-106_stereo_PCM-96k_MAGNIFICAT_04.flac  Arnesen: MAGNIFICAT 4. Et misericordia / MAGNIFICAT
 -67.53 dB  2L-125_stereo-88k-24b_04.flac             Frank Bridge Variations: 4. Romance / REFLECTIONS
 -83.63 dB  2L38_01_96kHz.flac                        Violin Concerto no. 4 in D major KV 218 - Allegro / MOZART
 -73.56 dB  2L48SACD_14_stereo_96k.flac               NORTH COUNTRY II / Stone Rose
 -85.51 dB  2L50SACD_tr01_stereo_96.flac              Simple Symphony  op. 4 - Boisterous Bourree / DIVERTIMENTI

Top two is -108.03 and -86.68 dB. After trimming 1 sec from starts and ends to reduce impact of fade ins/outs, these change:

 -83.54 dB  2L-038_MQA2016-88k-24b_01.flac            MOZART Violin Concerto in D major KV 218, I. Allegro / MOZART Violin Concertos (MQA remix 2016)
 -62.87 dB  2L-056_03_stereo_96kHz.flac               VIVALDI Cantata Rv 679: Che Giova Il Sospirar, Povero Core - Recitativo II / Bellezza Crudel - VIVALDI
 -68.91 dB  2L-056_04_stereo_96kHz.flac               VIVALDI Cantata Rv 679: Che Giova Il Sospirar, Povero Core - Aria: Cupido, Tu Vedi / Bellezza Crudel - VIVALDI
 -71.30 dB  2L-082_stereo-96kHz_01.flac               Ubi Caritas / Piano Improvisations
 -65.72 dB  2L-084_stereo-96kHz_11.flac               Eat! Drink! (Uriah, David) / David and Bathsheba, an opera-oratorio by Ståle Kleiberg
 -77.46 dB  2L-086_stereo-96kHz_15.flac               Bozza: Children’s Overture / La Voie Triomphale
 -68.55 dB  2L-092_stereo-96kHz_01.flac               Living / LIVING
 -56.14 dB  2L-093_stereo-96kHz_03.flac               Schubert String Quartet No. 14 in D minor D. 810: III. Scherzo. Allegro molto / The Schubert Connection
 -77.98 dB  2L38_01_96kHz.flac                        Violin Concerto no. 4 in D major KV 218 - Allegro / MOZART
 -73.25 dB  2L48SACD_14_stereo_96k.flac               NORTH COUNTRY II / Stone Rose
 -79.86 dB  2L50SACD_tr01_stereo_96.flac              Simple Symphony  op. 4 - Boisterous Bourree / DIVERTIMENTI

New top two is -83.54 and -79.86 dB. That's better than most, I guess, but I'd say still plenty of room left.

 

[KSTR]

The minimum RMS level of course is governed be the analog noise floor of the recording plus any dithering noise. When the bandwidth is higher than 20kHz one should filter to that bandwidth before calculating the RMS levels.
We also have to account for that we can "hear into" the noise some 10dB at least... assuming the level is sufficient. With zero-compressed/-limited classical and jazz recording the level in the quiet section is often low enough that this might be relevant.

In general I also find that 16bits, when properly done (competent production, including dithering etc), is sufficient for all but some very few corner cases, as the final output format. On the other hand, except for exotics, all DACs handle 24bit data and many of the higher sample rates, disk storage and streaming throughput is no issue, so why not use the higher resultion if it is available.

I provided the 2L links because I think it is a nice resource for having the guaranteed exact same material in just different formats, ideal for comparisions. Also, the production quality is high so we may assume all the down-sampling and dithering from the original DSD/DXD masters is competent.

 

[watchnerd]

On the other hand, except for exotics, all DACs handle 24bit data and many of the higher sample rates, disk storage and streaming throughput is no issue, so why not use the higher resultion if it is available.

Sure, I do this all the time, because my NAS has GBs to spare.

But I don't think it's necessary nor sonically superior.

 

[Sal1950]

Joking apart my guess is it is the extra reverb from mechanical (mainly) and airborne (maybe) feedback.
I base this on measuring how substantial this is, particularly at LF, deciding to site my record player outside the listening room to avoid it and being disappointed by the sound once I did.

LOL

On the other hand, except for exotics, all DACs handle 24bit data and many of the higher sample rates, disk storage and streaming throughput is no issue, so why not use the higher resultion if it is available.

For live recordings absolutely. But as a upsample of an existing source there's nothing to be gained except wasted space.

 

[earlevel]

try record a 24-track blank loop 16-bit master, every track added is a +3dB in noise floor

You mean that you need a doubling of (uncorrelated) tracks to add +3 dB. But, any addition in noise noise due to the number of tracks mixed is accompanied by a similar boost in signal gain. The SNR stays the same, basically, and for any overall gain increase, the person doing the mixing and/or mastering has to pull the gain back down the master fader or individual tracks to compensate, so the noise floor stays basically the same no matter how many tracks you mix. Of course, we mix with more than 16 bits for other reasons, but the number of tracks doesn't matter.

 

[MakeMineVinyl]

For live recordings absolutely. But as a upsample of an existing source there's nothing to be gained except wasted space.

I would bet there are exactly zero masters being recorded today which are not at least 24 bit / 96kHz. Storage is indeed cheap, and there is zero justification for NOT having the higher resolution if the cost is the same, which in most cases it is. I remember that guy I knew who insisted on naming all computer files in the DOS format of 8 characters - He really needed to get with the 20th century, at least. It was just 'my grandaddy did it this way' stubbornness

 

[Blumlein 88]

I would bet there are exactly zero masters being recorded today which are not at least 24 bit / 96kHz. Storage is indeed cheap, and there is zero justification for NOT having the higher resolution if the cost is the same, which in most cases it is. I remember that guy I knew who insisted on naming all computer files in the DOS format of 8 characters - He really needed to get with the 20th century, at least. It was just 'my grandaddy did it this way' stubbornness

No, the overwhelming majority of current recording is 44.1/24 and 48/24. 96 khz is done, but still a rarity. At one time DSP had an advantage done at 96 khz, but even that is no longer the case.

 

[MakeMineVinyl]

No, the overwhelming majority of current recording is 44.1/24 and 48/24. 96 khz is done, but still a rarity. At one time DSP had an advantage done at 96 khz, but even that is no longer the case.

That's still higher than 16 bit/44.1kHz, isn't it? I've been using 24/48 in ProTools since the 90s, when they first came out with it.

 

[Prep74]

The minimum RMS level of course is governed be the analog noise floor of the recording plus any dithering noise. When the bandwidth is higher than 20kHz one should filter to that bandwidth before calculating the RMS levels.
We also have to account for that we can "hear into" the noise some 10dB at least... assuming the level is sufficient. With zero-compressed/-limited classical and jazz recording the level in the quiet section is often low enough that this might be relevant.

In general I also find that 16bits, when properly done (competent production, including dithering etc), is sufficient for all but some very few corner cases, as the final output format. On the other hand, except for exotics, all DACs handle 24bit data and many of the higher sample rates, disk storage and streaming throughput is no issue, so why not use the higher resultion if it is available.

I provided the 2L links because I think it is a nice resource for having the guaranteed exact same material in just different formats, ideal for comparisions. Also, the production quality is high so we may assume all the down-sampling and dithering from the original DSD/DXD masters is competent.

Could you specify specific these classical/jazz tracks or corner cases where a difference can be heard? It would go against the grain of the accumulated evidence over the past 30 years or so. I'd be particularly interested in whether these tracks came from an analog source as 16 bits (done properly) is already overkill to capture all the information in detail - 13 bits would be sufficient.

Lastly, rather than using files from 2L, isn't it better and easier just to use your own hi res files in foobar (with the ABX plugin) or alternatively download the test tracks from Waldrep's Real HD Audio? At least one can be confident of provenance. If I recall correctly, one of the audio engineer members on the Head Fi Sound Science site found some flaws with the down conversion of 2L test files - and those on the Linn website.

I sort of agree with your sentiment though, about why not use the hi res files given storage and playback are no longer issues. The issue is that often the hi res remasters do not sound as good as the original CD releases. The hi res files that I do have are the ones that where the remaster has improved the sound quality (Neil Diamond is one of many examples), others I deleted and kept the CD version, eg ELO, Pink Floyd and others I have a mix depending on the album (eg Led Zeppelin).

 

[Prep74]

That's still higher than 16 bit/44.1kHz, isn't it? I've been using 24/48 in ProTools since the 90s, when they first came out with it.

It is interesting though that while 24 bit (and higher these days) studio equipment has resulted in the ability to make higher quality recordings/productions, there is another school of thought that it has generally gone the other way, due to the ability to further process and squash the sound in a way that was not possible with analog or early digital. It is the main reason why some LP records and some (many?) early CDs sound better than modern remasters. Take for example Dire Straits Brothers in Arms. It is widely considered to be audiophile quality. It has also been remastered several times, including half speed LP records. Although these remasters have been done well, it is a 16/44 recording and the early CD issues were a bit perfect copy of the master. It is a matter of subjective taste whether the remasters sound better or not.

 

[Sal1950]

No, the overwhelming majority of current recording is 44.1/24 and 48/24. 96 khz is done, but still a rarity. At one time DSP had an advantage done at 96 khz, but even that is no longer the case.

WOW, I'm shocked. I've got no knowledge of the pro recording world but if asked, I would have been willing to bet money the overwhelming number of modern recordings was done at 24/96 or better.

 

[Blumlein 88]

WOW, I'm shocked. I've got no knowledge of the pro recording world but if asked, I would have been willing to bet money the overwhelming number of modern recordings was done at 24/96 or better.

According to people in the business no. Some do record at 96. Some fair number think it is better than lower rates. Bulk of recording is at the lower rates.

 

[xykreinov]

Just here to say this is a good thread. Lots of handy, detailed replies. I like having succinct things to link to when talking to buddies about audio.

 

[Don Hills]

WOW, I'm shocked. I've got no knowledge of the pro recording world but if asked, I would have been willing to bet money the overwhelming number of modern recordings was done at 24/96 or better.

I believe you're both right. It depends on how many tracks you need to record at once. Classical tends to need fewer tracks than rock / pop and thus is more likely to be 24/96.

 

[Prep74]

I believe you're both right. It depends on how many tracks you need to record at once. Classical tends to need fewer tracks than rock / pop and thus is more likely to be 24/96.

Yes, it varies a bit. Many are 24/48 but 24/96 is common too. Hardly any studios that I know of do 24/192, which makes one wonder about the provenance of the hi res 24/192 tracks. The worst example though is DSD. Typically it is the higher end studios that can afford the equipment and staff for true DSD recording, though many use it as an archival format. The problem is that vary few consumer DSD releases are true DSD all the way through and these tend to be classical live recordings. The more general releases are either from a PCM or analog source or converted to PCM and back for mixing and mastering which kinda defeats its purpose. Probably irrelevant in a sense as there is no credible evidence that humans can distinguish between true DSD and PCM (with it at least being 16/44), but marketing prevails.

 

[Blumlein 88]

Yes, it varies a bit. Many are 24/48 but 24/96 is common too. Hardly any studios that I know of do 24/192, which makes one wonder about the provenance of the hi res 24/192 tracks. The worst example though is DSD. Typically it is the higher end studios that can afford the equipment and staff for true DSD recording, though many use it as an archival format. The problem is that vary few consumer DSD releases are true DSD all the way through and these tend to be classical live recordings. The more general releases are either from a PCM or analog source or converted to PCM and back for mixing and mastering which kinda defeats its purpose. Probably irrelevant in a sense as there is no credible evidence that humans can distinguish between true DSD and PCM (with it at least being 16/44), but marketing prevails.

Yeah, there pretty much has never been real DSD. So how did it get a rep for having a sound good or bad?

And boy studios doing 192 are really rare. So then ask yourself about DACs that do 384 or 768 khz??

 

[earlevel]

No, the overwhelming majority of current recording is 44.1/24 and 48/24. 96 khz is done, but still a rarity. At one time DSP had an advantage done at 96 khz, but even that is no longer the case.

OK, note that I'm not making this argument, someone else made it (I'm a little skeptical); on gearsl**z, industry old timer Bob Ohlsson said:

"Today, most major label recording is done at 96k. Back in the '90s, most recording was done at 48k while after the mid '90s some was done at 88.2k. This was because the extremely popular Pacific Microsonics mastering converter could only operate at 44 and 88. By 2010, high quality conversions between all sample rates no longer took hours when they weren't 2 to 1. At that point, most of the recording industry moved to 96k."

But...I've seen other people say that 44.1 kHz still dominates, and many services/library accept only that rate. I find that true in the smaller segment I have dealt with. But Bob should know, regarding "major label recording". Personally, I track at 44.1k still because I save the horsepower for DSP. I would upsample to make someone happy who needed 96 KHz.

 

[Blumlein 88]

OK, note that I'm not making this argument, someone else made it (I'm a little skeptical); on gearsl**z, industry old timer Bob Ohlsson said:

"Today, most major label recording is done at 96k. Back in the '90s, most recording was done at 48k while after the mid '90s some was done at 88.2k. This was because the extremely popular Pacific Microsonics mastering converter could only operate at 44 and 88. By 2010, high quality conversions between all sample rates no longer took hours when they weren't 2 to 1. At that point, most of the recording industry moved to 96k."

But...I've seen other people say that 44.1 kHz still dominates, and many services/library accept only that rate. I find that true in the smaller segment I have dealt with. But Bob should know, regarding "major label recording". Personally, I track at 44.1k still because I save the horsepower for DSP. I would upsample to make someone happy who needed 96 KHz.

It is a regularly recurring topic at Gearslutz. Everytime the preponderance is low sample rates. The Pacific Microsonics gear likely had little to do with setting rates either way as there has never been very many of them out there. So there is some 96k out there. Probably almost any studio would accommodate you if you ask for 96 khz. In general not much music is recorded at that overall I don't think.

 

[Sal1950]

Yeah, there pretty much has never been real DSD. So how did it get a rep for having a sound good or bad?

I've had that debate on the true believer websites for years. The ones who claim that DSD sounds SO much better, even when they take a plain ole 16/44 Redbook and resample it to 2x or 4x DSD and make the "thousand veils disappear, even my wife hears it" claims.

So there is some 96k out there

Only one I know for sure is Mark Waldrep at AIX Records who records everything at 24/96 or better, and then sells 24/96 files in stereo or multich.
I have a number of his recordings and they always sound excellent but that is the result of great care in making the recording and nothing to do with data rate, even he admits it.

 

[blse59]

I guess when they said "perfect sound, forever" they weren't kidding.