Conversion Part II
The record label has identified and secured the “best available” analog master tape of a wonderful album and are looking to transfer it to digital for release through one of the high-resolution digital music retailers. What should be a simple step of playing back the original master and digitizing the analog output at a high-resolution spec is often an unnecessarily convoluted process. I’m hearing more and more about the production paths used in creating the new digital “masters” and I’m dismayed. For example, here is a description of the conversion process associated with the recordings done by my friend Peter McGrath, who works for the high-end speaker manufacturer Wilson Audio.
The first step is to evaluate the condition of the analog master tapes. The tapes are slowly rewound and each tape splice checked and redone if necessary. Many older tapes suffer from terrible scrape flutter when played back due to the evaporation of the original tape lubricant. This lubricant allows the tape for smoothly glide over the heads and through the tape path. If none of the lubricant is left, the tape will shutter and squeal uncontrollably. The solution for this situation is to bake the tapes in a convection oven for several hours, which draws the last little bit of lubricant from the Mylar backing. Once baked the original tapes can be transferred to the digital domain for about 24 hours.
I read that one prominent engineer applies additional lubricant to any problematic tapes prior to the transfer. This is a very bad idea. The output level of high frequencies demands that the oxide layer of the tape be in intimate contact with the playback head. Any dust or substance that gets in between the tape and the head will cause significant loss of high-frequency information. So it’s ludicrous to deliberately pollute that gap with any foreign substance.
A very high quality analog tape machine is then used to playback the tape. A Studer A-80 two-channel deck equipped with state-of-the-art heads (Flux Magnetics Extended Response Repro head or a Wolke DIN Butterfly Repro head) is usually used. The tape handling characteristics of the Studer machines is near perfect. Although, I did meet a guy at the recent AES convention that is doing them one better. I’ll get back to him in another post.
Then a custom EMM Labs ADC-8 MK IV ADC modified by Andreas Koch converts the analog signals to DSD 64 for capture on a Sonoma System (at 2.8224 MHz). At this point, if there’s any work to be done such as gain changes, reverb, panning, or equalization, the DSD bitstreams are transferred back to analog and sent through a Neve analog mastering console (because remember you can’t do anything to a DSD 1-bit stream). The output of the mastering desk is then captured back to DSD using a Merging Technologies Pyramix DSD/DXD Masscore workstation before the systems’ Apodizing filters are applied during a final conversion to PCM…the ultimate sales format.
Holy cow! What are these people thinking? Do they really believe that multiple conversion stages and routing the signals through additional analog stages is going to result in a better master?
I don’t know any competent audio engineer that would choose a signal path that includes so many conversions and transformations. I count four conversion steps in the process described above and each time you convert from one form to another you lose fidelity. But it’s being done and we’re the ones suffering from this conversion nightmare.
Wouldn’t it better to simply transfer the analog master tapes to PCM using 192 kHz/24-bits using a really spectacular AD convertor? If any additional remastering needs to be done…do it in the digital domain using the widely available tools available to those studios working with PCM (which is virtually all of them). Following this procedure, the original analog tapes will only be subjected to one conversion. Keep it simple.
The guys at Wilson Audio got some very bad advice and the delivered files will be less than they could be as a result. The obsession with all things DSD is ensuring that future generations of listeners will be hearing diminished versions of the masters.
If you must use DSD because you prefer the sound, then do it at the final stage in your personal playback system. Don’t make me suffer from compromises that you accept because some designer or magazine reviewer told you DSD is a great thing. It’s not.
I’m so glad that I’m sitting out this round of the high resolution audio wars.
What ever happened to the old axiom, KISS?
Baking a magnetic tape. Magnetism and heat. I see something mutually exclusive here.
It’s actually standard operating procedure for retrieving old tapes…done it many times.
Hi Dr. AIX,
You mention that each time you convert from one form to another you lose fidelity. Why does conversion cause a loss, exactly what is that loss, and is it always a given?
I am asking not to challenge but rather to learn something about how the professionals convert music from analog masters to high-resolution digital formats and its consequences.
Phil C
Transferring from an analog tape master increases the noise floor by 6 dB or 1-bit in PCM terms. Capturing an analog playback from a source tape to high-resolution digital requires very high-end tape transports equipped with special heads, custom playback electronics, and highly accurate EQ curves. There are invariably albeit small losses caused by changes in tape speed etc. Converting from DSD to DXD or high-resolution PCM is not a seamless conversion…flaws in clocks, rounding errors etc occur. The best thing to do is to start in PCM and stay in PCM without any conversions.
Baking tape: Yes, very common cure, but the problem isn’t specifically “scrape flutter”, which is a rather high frequency longitudinal speed variation that happens when long sections of unsupported tape pass over a stationary element in the tape path. The result is a high frequency, somewhat broadband FM component added to the signal, which when bad enough comes out as noise wrapped around signal, disappearing when signal is not present. Baking has no effect on scrape flutter, but using a machine will roller guides rather than stationary ones is the fix.
The squeal that prompts baking tape is caused by a phenomenon called “binder breakdown”. There was a chemical incompatibility between the static dissipative back-coating and the organic oxide binder materials used to hold the magnetic oxide onto the tape base. An oxide layer is stored laying on top of a back-coat layer, and over time the binders break down. When the tape is played, any stationary tape path element causes oxide shed. When there’s enough oxide shed built up, and it doesn’t take a lot, the result is a form of “stiction”, where the tape resonates at a mid-band frequency, impressing this tone onto the audio. If it’s bad enough, the squeal is permanently impressed on the tape. Baking this tape for 8 hours at 130 degrees in a convection oven temporarily re-establishes the binder, and the tape can be played for about 24 hours, then the binder breaks down again.
The problem was confined to early back-coated tapes like Scotch 206 and 207, Ampex 406, 407, 456 and 457, and others. Once the issue was discovered in the mid 1980s, back-coating and binder formulae changed, and the issue was eliminated. Even so, there are plenty of late 1980s recordings made on older tape stock that suffer binder-breakdown. The mechanism and “cure” was covered in an Ampex publication in the mid 1980s, though other manufacturers had the same issue.
Thanks very much for the additional information.
I find it odd when the analog transfer process is mentioned that there is no mention of an effort to speed-match through pitch analysis, no mention of precision play head azimuth alignment to the tape, no mention of an attempt to match the original recording EQ with whatever tones may have been provided. I’d spend a lot of attention on head azimuth, even over the others, as a head out of az with the tape will cause rapid and severe HF loss, and a timing error between the two channels. There are many tools that can be used to establish proper azimuth, including an X-Y scope, phase meter, and both mono sum and L-R sum. Studer made a head block with a play azimuth knob with calibration marks, though that’s going to be a rare bird now. As for EQ, you’d start with a standard tape, hopefully new-ish, and a log sweep, get it as close as possible, then make small adjustments to match the (hopefully included) set-up tones on a master – post azimuth adjustment of course.
I concur with the use of 24 bit PCM, not sure there’s any point to 192KHz for analog tape though. Proper bias and record EQ would pretty much kill any significant response above 25KHz, more likely 22KHz, apart from distortion products.
Again, the proper physical alignment and electrical calibration steps much be taken prior to any transfer. Mastering engineers know this stuff quite well…many tracking and mixing engineers have never learned it.