The Fifth Element #89 Page 2

To demonstrate the musical significance of this, we have posted an MP3 consisting of two excerpts of fast solo passagework beginning at bar 102 of the first movement (see image): first from the fast-and-sharp SACD, followed by the same snippet speed-corrected to A=440. Each excerpt is preceded by a sinewave tone: first 446.4, then 440Hz after a repeat of 446.4Hz.

At 440Hz (excerpt 2), the passagework sounds to me like exceptionally fleet-fingered playing. In the speeded-up excerpt 1, I feel there is a disconnect between the apparent speed of the arpeggios and the quantity and quality of the articulation sounds and the tone of the violin. This is especially so at the two D's in rapid succession in bar 102: the first at the bottom of the last arpeggio, the second at the bottom of the next arpeggio, with a bowing change in between.

There is a limit to how rapidly violin notes can be played with a full tone, a limit imposed by the nature of the instrument itself—specifically, the violin's acoustical "transfer function," which takes time. As the vibration of the string excites the bridge, the bridge excites the bass bar and the sound post, which in turn excite the violin's belly and back, which excite the mass of air inside the instrument—and only when all of that has happened do you have a full tone.

I think the speed-corrected example remakes the connection, which was unmade by the remix tape, of the organic congruity of the nature of the sounds of making the music with the music-making's essential pace. An unbiased listener, Laura K.T. Stokes, Music Librarian at Brown University, remarked, "FWIW, I heard two completely different performances, so to speak, in these excerpts. The second excerpt is much warmer, and oddly, to me the contour of the phrasing works better."

Perhaps a workable analogy would be to an opera singer's singing an extremely difficult passage at half the indicated tempo, then having the engineers speed up the file without changing its pitch—something easily done with a digital recording. In such a scenario, I think the same kind of artificial disconnect between the music-making and the music might happen.

Brown University's much-played copy of the post-1973 LP.

Indiana Marks and the Temple of the Lost Tapes
I can only speculate as to how the speed/pitch error in the "Re-Mix" replacement "Master" tape arose. However, I'm certain that, at the recording session, the Brahms Concerto was played with Oistrakh and the orchestra tuned to A=440. I'm equally certain that pre-1973 US LPs were released at A=440.

Over and above the assurances from the archivist and musicians of the Cleveland Orchestra, using Amadeus Pro II, I performed spectrum analyses of more than a dozen live and studio recordings. The key findings are:

• Szell/Cleveland: Schubert, Symphonies 8 & 9; March 1960 & November 1957 (CD, Sony Classical SBK 48268). Cumulate all, 439Hz.

• Szell/Cleveland: Dvorák, Slavonic Dances, Opp. 46 & 72; January 1963–January 1965 (CD, Sony Classical SBK 48161). Cumulate all, 440.2Hz.

• Oistrakh/Lev Oborin: Beethoven, Violin Sonata 9 in A, Op.47 ("Kreutzer"); Paris, 1962 (studio recording, Le Chant du Monde) (CD, Philips 412 570-2). Solo-piano intro only, 440Hz. C#/A double-stop in solo-violin intro, bar 3 only, 440Hz. Cumulate entire first movement, 441Hz.

• Oistrakh/Otmar Nussio, Swiss Italian Radio Orchestra: Mozart, Violin Concerto 3 in G, K.216; Lugano, 1961 (CD, Dynamic CDS 389). Cumulate all, 440.6Hz.

• Oistrakh/Richter: Brahms, Violin Sonata 3, third movement; TV broadcast from Alice Tully Hall, March 18, 1970 (DVD, VAI 4369). Cumulate all, 440Hz.

• Oistrakh/Szell/Cleveland: Brahms, Violin Concerto, first movement, solo cadenza only; May 1969 (LP, Angel SFO 36033, pre-1973 pressing in collection of Brown University). Cumulate all, 440.3Hz.

• Oistrakh/Szell/Cleveland: Brahms, Violin Concerto, first movement; May 1969 (LP, Angel SFO 36033, pressing purchased by John Marks in Nashville, Tennessee, 1977 or later; ripped by Michael Fremer, to confirm unexpected result). Cumulate all, 444.1Hz.

• Oistrakh//Szell/Cleveland: Brahms, Violin Concerto, entire work; May 1969/2012 (CD layer of SACD/CD, EMI 9 55978 2). Cumulate all, 446.5Hz.

Spectral analysis of the Cadenza from the pre-1973 Brown University LP.

Spectral analysis of the Concerto from the entire SACD.

Spectral analysis of the first movement of the Brahms Violin Concerto from JM's post-1972 LP.

Back in the day, if someone I liked had enjoyed an LP I'd played for him or her, I would sometimes give that person the record, and replace it with a new, unplayed one for myself. At the time, I didn't realize that the earlier pressings would almost always sound better. Ah, me. I'm sure that, before 1977, I had given away two or three BOSC LPs. I have no idea today whether the earliest one I bought had been at proper speed and pitch.

That said, thanks to Michael Fremer, we now know that the last LP of this recording that I bought, ca 1977, was from a corrupted source. We also know that the speed-and-pitch error was not limited to the UK. Post-1972 US LPs were bad, too. It's surprising that, more than 10 years before the BOSC recording was first transferred to CD, the LPs on sale in the US were nearly 1.5% sharp and fast (footnote 1).

This unexpected finding somewhat reduces my chagrin that I didn't hear the speed and pitch discrepancies in the SACD—after all, both my Nashville LP and early-'90s CD were sharp and fast.

I tremble to think how little reason there is to hope that the Esoteric SACD of the BOSC recording from Japan, used copies of which now trade for over $250 on eBay, is from an uncorrupted source. However, I have not been able to get anyone to make a rip of that disc's CD layer for me. Unsurprisingly, Richard Lane, who invited me down this rabbit hole, reports that Testament's reissue of the recording on LP is—sharp and fast.

"The Future Lies Ahead!"
I take no joy in bearing bad news, and less at eating crow in public. We do know that, at some time, an EMI engineer heard the remix replacement tape as playing sharp, and noted it in the box log. It now seems that the LP that was pulled out to check against was a post-master-tape substitution LP whose stamper had been cut from the same corrupt remix tape it was being used to check against, and so the wrong conclusion was reached. The question is, what can we do now?

EMI no longer exists; its constituent parts were sold off here and there, with the Warner Music Group getting the classical department. I have so far been unable to interest anyone at Warner or Capitol Studios in looking for the BOSC session tapes. However, if those tapes could be found, I think that fixing this problem would be a perfect project for crowdfunding. Kickstart it!

The perfect wine bottle with the imperfect, bookbound SACD.

Photo credit: John Marks w/Thanks to ENO Wines, Providence

Locating the master tapes, finding out if they're playable (see sidebar), and then going forward with remastering and repressing, would be a risky proposition for the traditional reissue-label business model. But if 1000 hardcore Oistrakh fans promised to donate x dollars for feasibility studies, and a further y dollars to make a limited-edition SACD and downloads, the proposition becomes much less flaky. One of the well-known commercial reissue firms might take a shot at it.

And with all of this, I guess I've just launched the prices of pre-1973 US pressings of the Angel LP of the BOSC recording into geosynchronous orbit.

I think the ideal wine for baked-crow pie would be a legendary Rhône, such as a Vieux Télégraphe. However, given the realities, any wine that can stand up to game should suffice.

If you can't hear a significant pitch difference in the test tones or the music samples posted at the top of this page, consider that when the speeded-up SACD version of this recording has reached the end of the movement, the original version still has the measures above to play.—John Atkinson.

Footnote 1: I recently bought on eBay, from the UK, an LP that the seller claimed was the first UK pressing. It turned out to be a Dutch pressing from the fourth stamper. But after ripping by MF and analysis by me and Amadeus Pro II, this LP, too, turned out to be sharp and fast. I informed the seller of this by e-mail but never received a response.

woodford's picture

I have a copy of this catalog number- how does one tell whether it's a pre or post 1977 pressing?

and I agree about the Vieux Telegraph- i have a '98 in my cellar.

John Marks's picture

I believe that if your pressing has a scribing in the dead wax area something like S 1 36033 A 1 or S 1 36033 B 1 or S 1 36033 C 1, it was cut from the original edited stereo mixdown true master tape.

The Angel, "Mastered By Capitol" (a stamping in the dead wax area) pressing I bought in Nashville in 1977 is scribed S 1 36033 F 13 #2, from which I infer but it is only an inference that its source was an acetate that was the sixth ("F") US cutting. Because it too is fast, I conclude that at or about the same time the UK master was replaced the US master was replaced too.

So now we just have to hope that the original US master was not thrown out or recorded over.


woodford's picture

is what mine says- what does the 4, or 13 #2 signify?

in any event, it sounds the same to me as the Classic reissue. i Heifetz SACD, and Perlman/Giulini CD, and Grumiaux/Davis on vinyl. as a performance my preference has always been Perlman/Giulini, which i prefer to his later recording w/ barenboim.

John Marks's picture

The LP replication process goes (1) cut acetate, (2) negative master, (3) matrix or mother, and (4) stamper.

As far as I know, a cut acetate makes one negative master which makes one matrix which can make several stampers before it wears out and a new cutting is required.

So I believe that the stamper that stamped your LP was the fourth stamper made from the sixth acetate. Seeing as that acetate is the same as my LP's, that should mean that your LP is fast and sharp too.


deckeda's picture

That might provide the answer as to what happened to the original. Producers might recall these kinds of things where the labels and label archivists do not.

As for getting the music properly re-done (assuming the better sources can indeed be found) the only person I've read about who's had success where others have not is Chad Kassem. My understanding is that he shows the labels a combination of cash and perseverance other people don't. And that gets answers, at least.

That said, apart from pitch, the overloading is the main sonic problem? Let's say overloading isn't due to the 2nd or 3rd gen ('73) tape. Pitch and speed can be corrected with software but of course that's not the sort of homework a buyer should be expected to do for a R2DF candidate.

John Marks's picture

The first thing that you must understand is that the company EMI no longer exists. It was broken up and the classical division went to Warner; I have had no luck getting a response from anyone at Warner or at Capitol Studios. I have also reached out to some former employees who either ignored me or had no information.

Two, as a general rule, in a big label (unlike a small label) once the recording is in manufacturing, the recording engineer moves on and the producer moves on. In this case the recording engineer died some time ago. Tom Fine shared with me much information about Carson Taylor's work methods at the time, and the Cleveland Orchestra scanned copies of a logistics and scheduling document, but that was all 1969, NOT at the time of the master tape substitution.

In theory, there might be a paper trail in dead storage wherever EMI ancient history is stored, but in fact all that might have been sent to the landfill or recycling. The session producer died some time ago, again, not surprising, as recording Soviet musicians with an American orchestra was the most important thing EMI had going on at the time, so they used their top people, whose careers had started in the 1950s. The in-machine Cleveland master tape, if it still exists, is more than 45 years old. Visualize the final scene of the first Indiana Jones movie, and you will not be far from what my experience has been.

Over and above that the trail is cold, the reason why there was a "remix" that resulted in a master tape replacement ultimately is only academic because it was wrong. The troubling word is "re-mix." To me, that indicates that there is a strong possibility that someone somewhere went back to the in-machine Cleveland master and did a new stereo mixdown. So, to get at the sonic truth, we need either (1) the in-machine Cleveland master and either a working Dynatrack playback setup or a means to effect those "Boosted-to-NAB (and back)" track changeovers in the digital domain, or (2) Carson Taylor's original, correct-speed stereo mixdown razor-blade-edited master tape. In an ideal world. I'd prefer the first option because that would allow a more natural balancing between soloist and orchestra.

I have discussed this recording with Chad Kassem several times. He is not interested, and I can understand why. Chad is a Blues guy. More importantly, this is not a sonic blockbuster like Scheherazade or Symphonic Dances. It is core classical repertory. There are at least 80 violinists represented on non-historical discs, with many, like Oistrakh, having more than one performance (4 studio, as many as 10 live). So lists 195 Brahms Violin Concertos. Now, what an entrepreneur could do is Kickstart the project and not pull the trigger and collect the money until the One True Master Tape is found. But that is not Chad's business model and I have no hard feelings.

I don't know why you say "Let's say overloading isn't due to the 2nd or 3rd gen ('73) tape." I have not heard a pre-1973 pristine LP on a good system. I rather doubt that Carson Taylor would have approved an overloaded stereo master--and he, I am told, did his own editing, which, back then, was razor blade and splicing tape. My assumption (yes, it is an inference from established facts) is that in 1973 they played the Cleveland in-machine back not realizing that it required special playback electronics, which is why there is a notation somewhere (I was told by Abbey Road) that there was a tape box in the archives marked "Unplayable." Whether they kludged from that or had some other Plan B, I do not claim to have a clue.

One would think that with all the audiophile-reissue formats this performance has appeared in, there would be remaining demand. But the conventional business model has a break-even at or over 4,000 units, which is why I believe, based on 30 years in the recording business, that Kickstarting is the only solution. As far as I can see, only that gives you a handful of cash to wave in the face of Warner and say, we are ready to prepay the license fees so you don't have to trust us or chase us.



Wineguy1's picture

Great article. What was the vintage of this Chateauneuf?

John Marks's picture

Hi-I was really not paying attention to the vintages. I went to a wine store and asked if I could pose a bottle with the SACD book and take some pictures. I liked the dramatic high-contrast label and the designation "La Crau," which was as close as one can get to the crow I had to eat.


Wineguy1's picture

Ah I see. It is a terrific wine nonetheless and indeed if a crow pie were served up it would pair perfectly. I see one reader below with some 1998 in his cellar which is drinking beautifully.

davip's picture

I'm just getting into SACD down here in the deepest, darkest Rhone Valley (Department de l'Ain), but Chateauneuf? A €5 wine on a €20 budget one might say, whatever the domaine. It's ok if you don't mind your wine tasting like molasses (but then it is a molasse of 13 seppages), but a bog-standard mis-en-bouteille 'Vieux Papes' shows it a clean pair of heels for 1/10th of the cost (I doubt that you'll get it over the pond, mind). Still, in a milieu where people justify paying 10,000s for an ethernet cable or a mains lead, Each To Their Own...

John Marks's picture

HI. Thanks for posting. The web version omits some aspects of the print version, in this case, the headline "Which Wine Goes Best With Crow?" Although that thought recurs at the very end of the text too.

So, Eating Crow was on my mind when I stopped in at a fancy wine shop downtown and asked to pose a bottle with the SACD book and take some pictures. I chose that bottle because of the designation "La Crau." Given the major expense of that bottle, and the fact that I don't have Sam Tellig's credit card number memorized, I did not buy it.

For my next act I will build and market a mains lead made from worn-out Pirastro Eudoxa Olive Rigid violin G and D strings that were used only on pre-1800 Cremonese violins. Harmonic richness such that you can imagine you are smelling all 13 (or 14) distinct grape varieties in a great CTNDP!!! A bargain at USD10,000 per 2-meter mains lead! Send me your money before someone else does!

Tee hee.


wozwoz's picture

Congratulations on a splendid article and brilliant detective work combined with academic rigour. And yes - the 440Hz snippet sounds much more natural.

Next, you will be telling us that Richter actually recorded his famous 1970/19S73 Schloß Klesheim Melodiya recording of the Bach Preuldes and Fugues at 440Hz too ... even though they sound almost a semitone sharp ! (something I have never understood).

It would be disappointing if it transpires that some engineers back in 1973 thought it 'cool' to tweak recordings to make artists appear, say, more virtuosic ... the way people today might photoshop a photo today to tweak it. Ooh - look what we can do with this fancy-shmancy tape deck machine!!! ....

Separately, as I commented on your original article, I found the SACD transfer by EMI of the Brahms to be highly distorted ... for whatever reason. Unfortunately, the EMI engineers did not simply do a tape to DSD transfer, but ran a whole bunch of artificial digital filters, cleaners and disinfectants over the source file to de-hiss it, ... which rather kills the whole point of transferring it to SACD (which is to keep things au naturel).

John Marks's picture

I really doubt an engineer took into his head to replace an LP cutting tape prepared by Carson Taylor or under his direction with a speeded up "remix" just because he thought it was a good idea.

One possible explanation is that because Oistrakh had recorded the Brahms in Paris for EMI in 1960, and in 1969 also recorded the Beethoven Triple in Berlin for EMI, the Cleveland tapes were the Odd Men Out in terms of pitch, and some executive decided to speed up the Cleveland tapes so they would sound more similar in pitch to the European tapes, but in so doing they truly distorted the pacing and phrasing.

I am keeping at this; I am trying to get Warner Brothers to do a vault search in California and in England. If a non-corrupt stereo mixdown master is located, or even the multitrack session tapes, then at least a reissue outfit like Mobile Fidelity can evaluate it as a business proposition, and if not, some ambitious person can Kickstart it.


dspandrc's picture

I fail to see why a recording that was classified as R2D4 should be downgraded because it might have been speeded-up. I would have thought that the R2D4 was awarded for what was heard on the recording - and that has not changed.

It is always tricky to determine the pitch at which something was played when the original recording was on analogue tape. This is because there was seldom a good absolute frequency reference, such as we now have with digital recordings.

John Marks makes an assumption that because it is thought that the Cleveland orchestra always tuned to A440 then that must have been the actual pitch used. I have analyzed a few of Oistrakh’s recordings and they are mostly pitched higher than A440.

The recording I have of the Brahms concerto featuring Oistrakh with Kondrashin is at A447, the recording I have of the Brahms concerto featuring Oistrakh with Klemperer is at A445. The Brahms double concerto I have featuring Oistrakh and Rostopovich with the Cleveland orchestra under Szell is at A447. I also looked at Oistrakh playing Mozart and found it to be at A447 as well. The fundamental question is does anybody know with absolute certainty which pitch Oistrakh asked conductors/orchestras to tune to when accompanying him.

It is, of course, possible that most of the Oistrakh recordings that I have (no matter what the label) have been speeded-up by playing the tape a little faster and it is, of course, possible that the original release of the Brahms concerto had been slowed-down by somebody who thought that the Cleveland always tuned to A440. It is also possible that the directive to keep the higher pitch on re-issue had come from performer input after having heard the original release.

The spread of Oistrakh's movement timings, from his various recordings, is significantly greater than 1% (if I remember correctly his spread was greater than 7%).

regnaD kciN's picture

While I don't have the recordings or software on hand to check these, the fact that Oistrakh's pitch seems to be higher than 440 on several other noted recordings would suggest that this was a habit for him. I would also note that, while the first movement's length of 22:35 would appear to be potentially 20 seconds shorter than it would be if the tape were re-pitched to 440, it barely differs from the 22:36 of his recording with Klemperer, and is slightly slower than the 22:30 of his version with Kondrashin (all timings courtesy of Tidal). Now, maybe every version of Oistrakh's Brahms concerto (along with, as noted above, several other of his recordings), all coming from different companies over the course of several decades, happened to be coincidentally and accidentally sped up by about the same amount. On the other hand, I would have to wonder if, perhaps, the "no speed adjustment" directive, rather than coming from some apathetic EMI "suit" with no expertise in the matter, might indeed have come from a source that had knowledge that the original recording was, in fact, made at 445?

I mention this open question because I notice that you've paired with the folks at HDTT to produce a "pitch-corrected" version of this recording -- a laudable enterprise if the original is, indeed, running too fast, but an unnecessary version if it isn't. Pity HDTT didn't release an album -- simple to do with a download-only company -- offering both versions of their transfer, so that buyers could compare for themselves.

regnaD kciN's picture

My previous skeptiucal comments on this subject were based on the running time of the 1969 first movement compared with the similar timings on Oistrakh's previous recordings. This timing of 22:35 was taken from the EMI "Great Recordings of the Century" CD release of 2003. As JM commented above, he knew that the SACD was sourced from "a PCM transfer." In the interest of using a higher-res master, I always assumed that the PCM source in question was the 24/96 transfer available at HDtracks, among others. Since, as I understood it, the GRotC series usually made 24/96 transfers and converted them to Redbook for CD release, I always assumed that the GRotC and hi-res versions were from the same remaster. My mistake! Examining the hi-res download version, I see that it came from a different transfer made in 2011. But that was only the first surprise. While the GRotC CD clearly had a timing of 22:35 for the first movement, the same track on the 24/96 had a timing of...22:23. "Triple Yikes!" as JM might say.

This, fortunately, became an easy matter to research further. Ripping the CD, and converting the 24/96 track to Redbook resolution, it was a simple task to open the two together in Audacity. And there it was...from first non-zero sample to last, the CD track had a length of approximately 22:26.261. By comparison, the 24/96 length was 22:12.200...more than fourteen seconds shorter. If that was the source used for the SACD, no wonder it would sound rushed and unnatural! (And here's the kicker: if this issue was caused by something odd in the digital transfer itself, it likely wouldn't even manifest itself as being sharp, since digital manipulation would affect the speed but not the pitch.)

The bottom line here appears to be that, all other issues from the '70s aside, there is something drastically wrong with the 2011 24/96 transfer that constitutes a significant corruption over any previous version. Unfortunately, it would appear that the current "Warner Classics" CD is likely sourced from that corrupted transfer, leaving the EMI GRotC CD as the last "official" release without this problem...glad I own a copy! (IMHO, samples I've heard of the "speed-corrected" HDTT suggest that the reel-to-reel tape used is several steps away from the master tape, with noticeable degradation of sound quality.)

dspandrc's picture

I usually find track lengths to be somewhat unreliable in determining the lengths of the actual music. I always analyze the frequency content of tracks and determine the centre frequency of the "A".


regnaD kciN's picture doesn't change the fact that, when you have two transfers ostensibly from the same master tape, and the length of the music, from first note to last, on one is twelve seconds shorter than the other, it's pretty much proof-positive that at least one of those transfers has a serious flaw -- and that remains true no matter what the pitch used (as I pointed out, if the problem was in the digital transfer, the length -- and thus the apparent pace -- would be drastically different, but the pitch would be unchanged).

And note that I didn't just use "track lengths" (which can be deceptive due to extra silence at the beginning or end); I loaded both tracks into an audio editor and measured them from first non-zero sample to last, thus ensuring that any additional silence in one or the other had no bearing on the results.

dspandrc's picture

I have two recordings of Szell's Bruckner 3 which came from the same masters but one of which leaves out some of the music. Frequency analysis gets over this potential problem.


SleeperSmith's picture

Kickstart it, and let's spread the word.

Post it on the forums. 4000 records shouldn't be that big an ask?? Especially for such a legendary performance like this which will not happen again ever.