A Case for Releasing Analog Recordings in High-resolution Audio

[gmgraves]

I was having a discussion with a fellow audiophile over the weekend. He pointed out that I had purchased a lot of titles in 24/96 from places like HDTracks and others that were from analog tape masters. He further mentioned that I was probably wasting my money because, as he put it, "The one thing an old analog tape recording from the days before digital doesn't need is bandwidth above 20 KHz."

 

Well, he certainly has a point. An experiment that I used to run for my own amusement was was to test the usable high frequency response of various tape formulations that I employed to record the local municipal Symphony Orchestra, the San Jose (CA) Symphony under Maestro Georg Cleve. 

 

I had a ritual in those days; the evening before the performance I would sit down at my two Otari MX5050 tape decks (and before them, my modified Ampex 350s with custom record/play electronics) and prepare each for the next night. I would start by checking the head alignment on both machines and tweaking it if necessary (which I would do once more, after moving them to the Center of the Performing Arts the next night). Then I would set the EQ of each machine to match whatever brand and type of tape I was using for that recording. To assure myself that both machines were perfect, I would then spot-check the frequency response of both decks at different frequencies to make sure that both decks were not only flat, but the same across the board.

 

One of the frequencies that I would use was 21 KHz. I would take a 21 KHz tone from my oscillator, and record it at -20 dB at 15 ips (38 cm/sec) and feed the output of the playback heads to my oscilloscope. 

 

Irrespective of the tape-type being used, whether 10.5" reels of Maxell Chromium Dioxide tape, Ampex 407 Grand Master, or Sony FerroChrome tape, the results were always just about the same. I would record the tone at -20 dB and the 'scope would show it coming off the tape at -80 dB and so buried in the tape hiss as to be useless. Before I could get the output at 21 KHz to match the input in level, I had to reduce the record level to less than -50 dB. Then the record and playback level at 21 Khz were the same. 

 

No, analog tape recordings, do not need the extended frequency response of 24-bit audio whether 48 KHz, 88.2, or 96 KHz  sampling rate because there is nothing useful up there. These were fresh recordings, mind you. These old analog tapes made on older tape formulations and equipment weren't as good as the tape and equipment that I was using, and long term storage took it's toll as well. Does that mean that there is absolutely no advantage to transferring these older performances to high-resolution digital? No, I don't think it does. 

 

In spite of the fact that these old recordings are limited in frequency response and roll off pretty steeply above 15 KHz (the highest frequency to which pro decks could/can be aligned) and the dynamic range of these old recordings falls far short of what 24-bit is capable, I think the 24-bit quantization has a big advantage in that the dynamic range window afforded by 24 (or 32) bit transfer makes for recordings that have much better lower level resolution that they can have when transferred to 16-bit Red Book. I have compared CD transfers of many of these older analog recordings with their 24/96 "descendants" and find the 24/96 versions to be better in every case. The high-res versions seem much more "effortless" and much cleaner than the CDs despite the fact that these high-res transfers are made from master tapes that are now 10 or even 20 years older than they were when the CD (or even the SACD) transfers were made. While, in practicality, I hear little to choose between a 24/96 (or 192) transfer and a SACD transfer of this older material, what difference that I do hear are more down to how the transfers were made rather than the format. Most of my SACD transfers of analog material were made by Sony from the Columbia archives and consist of titles from Miles Davis, Dave Brubeck, George Szell, Leonard Bernstein, Bruno Walter, etc. These works were transferred to DSD in the late 1990s and early 2000s and thus the masters were about 20 years younger than the LPCM transfers being made now (unless Sony is merely converting the DSD masters made then to high-res LPCM, which is likely).

 

Thoughts, comments?

[AJ Soundfield]

19 minutes ago, gmgraves said:

I think the 24-bit quantization has a big advantage in that the dynamic range window afforded by 24 (or 32) bit transfer makes for recordings that have much better lower level resolution that they can have when transferred to 16-bit Red Book.

No

 

If I'm reading this correct, it appears to have around 68db max SNR

http://lcweb2.loc.gov/master/mbrs/recording_preservation/manuals/Otari MX-5050 Professional Recorder (manual).pdf

[Ralf11]

what about bit depth?

[gmgraves]

35 minutes ago, AJ Soundfield said:

No

 

If I'm reading this correct, it appears to have around 68db max SNR

http://lcweb2.loc.gov/master/mbrs/recording_preservation/manuals/Otari MX-5050 Professional Recorder (manual).pdf

Well, yes, actually. with a 24-bit wide window, a program with a less than 70 dB dynamic range can be positioned in that window to have the 4 LSBs not toggled by the program content. This improves distortion and digital noise added to the original program. 

 

Half-Track, 15 ips pro tape Decks do have a ~65 dB max SNR, but the actual SN is largely a function of the tape formulations used. Older formulations like Scotch 203, have up to a 5 dB worse SNR (~-61 dB relative to 185 nW of fluxivity) than say, Maxell or TDK CrO2 and about 7 dB worse than Sony FECr formulations of the late '70's. These latter tapes have a peaked high-frequency retentivity and therefore must be EQ'd to roll-off more radically on the high-end to achieve a flat frequency response. This improves SNR by rolling-off the tape noise along with the frequency response. Add Dolby 'A' noise reduction and you can increase the practical round-trip SNR by another 6 or so dB. It is possible to have an analog tape recording with a round trip SNR approaching 75 dB.  

[gmgraves]

7 minutes ago, Ralf11 said:

what about bit depth?

 

With regard to what? I need a little more context than that!

[Ralf11]

I am just wondering if the "interpolation" or smoothing of amplitudes with a higher bit depth might give a detectable SQ increment.

[esldude]

If your choice is 24/96 or 16/44 I would go with the 24/96.  If they have remastered it then you are closer to that.  They have to use a downsampler to get to 44.1 khz and they don't always use a really good one.

 

If your choice is 24/96 and 24/48 then the downsampling you avoid might still matter, but probably matters less.   Unless I knew it was already a stellar recording I would go with 24/48. 

 

If done well, a much bigger if than it should be, 16/44 should be just fine for analog tape sources.  The envelope of the signal possible on 16/44 simply exceeds analog tape in every way. 

 

Now I have not measured it, but I thought 15 ips tape at -20 db could manage something like -3 db at 20 khz on some machines.  These were listed specs for the Revox A77. 

 

15 ips - 30 Hz - 20 kHz ± 1.5 dB
7.5 ips - 30 Hz - 20 kHz + 2 / - 3 dB
3.75 ips - 30 Hz - 16 kHz + 2 / - 3 dB

 

http://www.endino.com/graphs/

 

Here are some measurements of some later 1/2 inch, 1 inch and 2 inch tapes which look better than your experience with the 21 khz signals. 

[Don Hills]

22 minutes ago, Ralf11 said:

I am just wondering if the "interpolation" or smoothing of amplitudes with a higher bit depth might give a detectable SQ increment.

 

It shouldn't. It's a common misunderstanding. Provided that the original signal fits within the "bit window", the resolution is the same regardless of the maximum bit depth available. A typical analog recording might have up to 75 dB or so between maximum level, depending on how hard the tape is pushed, and where the signal starts to disappear into the noise. With care to avoid clipping, and properly dithered, the signal from the tape will fit into 16 bits with room to spare. Digitising it at 24 bit depth will still only use 16 bits, the lowest 8 bits will be filled with random noise. It seems counter-intuitive - surely the lowest level signals won't be recorded with the same resolution as higher level signals? - but that's the way it works when properly dithered. Remember Monty's show and tell video? It clearly shows a very low level signal being digitised. Without dither there is distortion due to very few quantising levels available. Properly dithered, all of the distortion disappears.

[AJ Soundfield]

29 minutes ago, Don Hills said:

properly dithered, the signal from the tape will fit into 16 bits with room to spare.

Yep

[The Computer Audiophile]

10 minutes ago, AJ Soundfield said:

Yep

I guess this is how one gets to over 600 posts in a little over a month :~)

[d_elm]

IMHO the case for 88.1k and higher sampling rates is for a relaxed and higher frequency low pass filter.  Some material. eg coral works, also benefits  from a higher sampling frequency.  Check your downloads for scams with a spectrogram generatd by spek.exe, from spek.cc.

[Lighthouse]

I believe one of HA mods claimed that it is possible to achieve effective 120+ dB with proper dither with 16/44.1.

 

Basically, the reason why 24/96 is used for analog transfer is simply convenience.

 

Which I understand, since I also use 24/96 for my Vinyl digital transfer works due to convenience as well.

 

 

But, that's not the main issue. The issue is that music labels trying to sell these analog transfer as "High-Res" format despite it is not even close to being High-Res. Companies abuse the fact that people do not understand these '24/96' stuffs only indicates the size of the bucket. It does not tell whether water is actually filled that bucket or not.

[Lighthouse]

3 hours ago, The Computer Audiophile said:

I guess this is how one gets to over 600 posts in a little over a month :~)

 

I guess that's ok though. I am still a "Freshman Member" despite the fact that I signed up this forum in 2011. At least... he is not fresh anymore.

[esldude]

3 hours ago, d_elm said:

IMHO the case for 88.1k and higher sampling rates is for a relaxed and higher frequency low pass filter.  Some material. eg coral works, also benefits  from a higher sampling frequency.  Check your downloads for scams with a spectrogram generatd by spek.exe, from spek.cc.

You know I hear this a lot.  I would even think it a good use of the extra spectrum.  Yet few indeed are the ADCs or DACs that work like that.  Most have a 2050 hz transition zone at 44.1 khz and 4100 hz transition zone at 88.2 khz.  So and so forth.    The flat band should end at 25 khz or no more than 30 khz imo.  Lavry gear does this, but few others.

[gmgraves]

6 hours ago, esldude said:

If your choice is 24/96 or 16/44 I would go with the 24/96.  If they have remastered it then you are closer to that.  They have to use a downsampler to get to 44.1 khz and they don't always use a really good one.

 

If your choice is 24/96 and 24/48 then the downsampling you avoid might still matter, but probably matters less.   Unless I knew it was already a stellar recording I would go with 24/48. 

 

If done well, a much bigger if than it should be, 16/44 should be just fine for analog tape sources.  The envelope of the signal possible on 16/44 simply exceeds analog tape in every way. 

 

Now I have not measured it, but I thought 15 ips tape at -20 db could manage something like -3 db at 20 khz on some machines.  These were listed specs for the Revox A77. 

 

15 ips - 30 Hz - 20 kHz ± 1.5 dB
7.5 ips - 30 Hz - 20 kHz + 2 / - 3 dB
3.75 ips - 30 Hz - 16 kHz + 2 / - 3 dB

 

http://www.endino.com/graphs/

 

Here are some measurements of some later 1/2 inch, 1 inch and 2 inch tapes which look better than your experience with the 21 khz signals. 

 

Maybe theoretically. And perhaps playback only, but I seriously doubt these figures for round trip. Especially for an A77. Actually, I've never measured a pro deck that had anywhere near this kind of performance in the real world, and I've set-up Ampex R100's, Studers, etc. Multi-tracks (16, 24, 48 tracks) are even worse because the tracks are so narrow. 48 tracks on 2-inch tape, and 24 tracks on 1-inch tape have individual tracks the width of a compact cassette tape track!

[gmgraves]

6 hours ago, Ralf11 said:

I am just wondering if the "interpolation" or smoothing of amplitudes with a higher bit depth might give a detectable SQ increment.

 

Please quote the post to which you are responding. Making comments in a vacuum, doesn't really advance the conversation. Thanks.

 

 

 

 

[mansr]

5 hours ago, esldude said:

You know I hear this a lot.  I would even think it a good use of the extra spectrum.  Yet few indeed are the ADCs or DACs that work like that.  Most have a 2050 hz transition zone at 44.1 khz and 4100 hz transition zone at 88.2 khz.  So and so forth.    The flat band should end at 25 khz or no more than 30 khz imo.  Lavry gear does this, but few others.

A sharp filter at 20 kHz just might make an audible difference. The same filter scaled to 40 kHz will be a lot less audible since any nasty effects (ringing, passband ripple, etc) are well removed from the audible range and there is generally less content at those high frequencies in the first place.

[elcorso]

11 hours ago, Lighthouse said:

 

I guess that's ok though. I am still a "Freshman Member" despite the fact that I signed up this forum in 2011. At least... he is not fresh anymore.

 

We children need "candy" as a reward. 

 

The quantity is sometimes awarded over quality. I'm not down to disqualifying the quality of AJ Soundfield (but other members of the forum), only time will tell regarding AJ.

 

Anyway, you can always change your automatic status of "Freshman Member" for whatever you want.

 

Roch

 

PS/ Sorry for the out of thread in this interesting (to me) discussion.

[elcorso]

9 hours ago, gmgraves said:

 

Maybe theoretically. And perhaps playback only, but I seriously doubt these figures for round trip. Especially for an A77. Actually, I've never measured a pro deck that had anywhere near this kind of performance in the real world, and I've set-up Ampex R100's, Studers, etc. Multi-tracks (16, 24, 48 tracks) are even worse because the tracks are so narrow. 48 tracks on 2-inch tape, and 24 tracks on 1-inch tape have individual tracks the width of a compact cassette tape track!

 

Pardon my ignorance.  There are 2, 1, or 1/2 inch analogue tape machines for two tracks only?   If they are, the track width is wider?

 

Thanks,

 

Roch

 

PS/ I owned, a long time ago, an Japan made, two tracks tape machine, but I can't remember. Or was out of my interest but that time and the SQ was extraordinary.

[firedog]

Maybe it is simply that the hi-res versions tend to have better mastering than CD masters of the same tape.

I personally have a bunch of tape to hi-res remasters that I think are better sounding than CD versions of the same album. 

I also tend to think that DSD makes a better sounding result from tape than PCM. More like analog. And I'm not one of those people that thinks DSD is superior, I just think it sounds different, and I tend to like tape to DSD remasters. 

[gmgraves]

37 minutes ago, elcorso said:

 

Pardon my ignorance.  There are 2, 1, or 1/2 inch analogue tape machines for two tracks only?   If they are, the track width is wider?

 

Thanks,

 

Roch

 

PS/ I owned, a long time ago, an Japan made, two tracks tape machine, but I can't remember. Or was out of my interest but that time and the SQ was extraordinary.

 

Well. there certainly were 1/2", 1", and 2" tape machines, but they were mostly more than two-track. Usually, 1/2" machines were four or eight track, and the 1" machines were generally sixteen track and so-on. These are not hard and fast rules, however as there were 1/4" tape machines with four and eight tracks, 1/2" ,machines with 16 tracks, etc. Recorder companies could configure machines about any way a customer might want. After all, it's just a matter of fitting the proper heads and the required number of record/play electronics.  The essential truth here is that the more tracks for any given tape width, the narrower the tracks and narrow tracks have a poorer SNR (due to the reduced number of magnetic domains passing the head gap at any given time).

 

My edition of Howard Tremane's Audio Cyclopedia shows that there is an NAB spec for a two-track 1/2" tape format, but I've never seen one. The closest I've ever seen is to that are two-track 35mm film recorders which used magnetic film with sprocket holes along the edges just like 35mm cine film. The difference is that instead of a photographic emulsion, this film stock was coated with magnetic oxide, just like recording tape. All of the 35mm film recorders that I've ever seen were branded RCA, and were made in Italy, but I suspect that Ampex made them too. They were used from the late '50's through the 1960's by such recording companies as Command Records, Verve Records, Vanguard Records Etc. Command used to record the the Pittsburg Symphony under William Stienburg with 35mm recorders. I watched Verve record Stan Getz and Charlie Byrd's Jazz Samba album (the original Bossa Nova release) in a church in Washington DC with two 35mm RCA "portable" recorders. At the risk of digressing, this has always somewhat puzzled me, because apparently the album was comprised entirely of takes recorded on a 1/4" machine and the 35mm takes were not used to cut the album! (???)

[gmgraves]

4 minutes ago, firedog said:

Maybe it is simply that the hi-res versions tend to have better mastering than CD masters of the same tape.

I personally have a bunch of tape to hi-res remasters that I think are better sounding than CD versions of the same album. 

I also tend to think that DSD makes a better sounding result from tape than PCM. More like analog. And I'm not one of those people that thinks DSD is superior, I just think it sounds different, and I tend to like tape to DSD remasters. 

 As do I. I have a number of hi-res versions of analog recordings that I earlier bought on CD. The hi-res (24/96) versions sound much better than the CD versions. OTOH, I have not noticed that SACD versions of analog productions sound better than LPCM hi-res. Just the other day, I compared a version of Miles Davis' Kind of Blue with the SACD of same. Frankly I heard no differences that I couldn't associate with playback equipment. The SACD was played on a Sony XA777ES SACD player and the LPCM version was streamed from my Mac Mini to my Logitech Squeezebox Touch and decoded by a Schiit Yggdrasil DAC. 

[elcorso]

5 hours ago, gmgraves said:

 

Well. there certainly were 1/2", 1", and 2" tape machines, but they were mostly more than two-track. Usually, 1/2" machines were four or eight track, and the 1" machines were generally sixteen track and so-on. These are not hard and fast rules, however as there were 1/4" tape machines with four and eight tracks, 1/2" ,machines with 16 tracks, etc. Recorder companies could configure machines about any way a customer might want. After all, it's just a matter of fitting the proper heads and the required number of record/play electronics.  The essential truth here is that the more tracks for any given tape width, the narrower the tracks and narrow tracks have a poorer SNR (due to the reduced number of magnetic domains passing the head gap at any given time).

 

My edition of Howard Tremane's Audio Cyclopedia shows that there is an NAB spec for a two-track 1/2" tape format, but I've never seen one. The closest I've ever seen is to that are two-track 35mm film recorders which used magnetic film with sprocket holes along the edges just like 35mm cine film. The difference is that instead of a photographic emulsion, this film stock was coated with magnetic oxide, just like recording tape. All of the 35mm film recorders that I've ever seen were branded RCA, and were made in Italy, but I suspect that Ampex made them too. They were used from the late '50's through the 1960's by such recording companies as Command Records, Verve Records, Vanguard Records Etc. Command used to record the the Pittsburg Symphony under William Stienburg with 35mm recorders. I watched Verve record Stan Getz and Charlie Byrd's Jazz Samba album (the original Bossa Nova release) in a church in Washington DC with two 35mm RCA "portable" recorders. At the risk of digressing, this has always somewhat puzzled me, because apparently the album was comprised entirely of takes recorded on a 1/4" machine and the 35mm takes were not used to cut the album! (???)

 

Thanks George!

 

I own some releases on SACD from Everest Recordings, that state on the album art, recorded on 35mm magnetic-film tape. The transfers are not extraordinary to my taste, but I appreciate that they have made the digital release attempt.

 

Here is some literature I found recently on the web:

 

http://www.cool.conservation-us.org/byform/mailing-lists/arsclist/2008/12/msg00076.html

 

Roch

[elcorso]

5 hours ago, gmgraves said:

 As do I. I have a number of hi-res versions of analog recordings that I earlier bought on CD. The hi-res (24/96) versions sound much better than the CD versions. OTOH, I have not noticed that SACD versions of analog productions sound better than LPCM hi-res. Just the other day, I compared a version of Miles Davis' Kind of Blue with the SACD of same. Frankly I heard no differences that I couldn't associate with playback equipment. The SACD was played on a Sony XA777ES SACD player and the LPCM version was streamed from my Mac Mini to my Logitech Squeezebox Touch and decoded by a Schiit Yggdrasil DAC. 

 

To my taste it is not the same to play and listen DSD recordings from an SACD player that streamed from a PC to a good DSD DAC, that please you.

 

But I will not discuss about tastes, because it is something very personal.

 

Roch

[gmgraves]

2 hours ago, elcorso said:

 

Thanks George!

 

I own some releases on SACD from Everest Recordings, that state on the album art, recorded on 35mm magnetic-film tape. The transfers are not extraordinary to my taste, but I appreciate that they have made the digital release attempt.

 

Here is some literature I found recently on the web:

 

http://www.cool.conservation-us.org/byform/mailing-lists/arsclist/2008/12/msg00076.html

 

Roch

 

Yes, Bert Whyte did make some everest recordings (Bert WAS Everest. Harry Belock of Block Instruments just provided the capital, Bert did all the work) on 35mm film. That the transfers (which, mostly,  I haven't heard) are not very good is the fault of the transfers, not the original recordings which, I can assure you, were mostly audiophile quality on LP. The one that I do have is Eugene Goosens and the London Philharmonic in works by Villa Lobos and Alberto Ginastera. I bought it  on 24/96 from HDTracks and while it would sound OK if it were a decent transfer, it has so much flutter in it that it is unlistenable. (HDTracks offered me 15% off on my next purchase. But believe me, with that kind of customer service attitude, it was a number of years before I bought anything from them again!).