White Glove: Fairytales – page 2

Forensics on Early Releases

The first LP releases of Fairytales (1982, 2005) were made by transferring the analogue output of an MX-80 to cutting lathe. Our expectation is that replaying on the recording machine would result in the correct pitch for the LPs.
Digital releases (1986, 2004 & 2007) were made by feeding the analogue into a Sony PCM-F1 and it seems that at least the early CDs used a digital process to make the CD master. Our pitch investigation suggests that this PCM-F1 was in fact based on 44.056 kHz, with the result that the early CDs were high in pitch and quite brittle in sound due to the cascade of PCM-F1 interleaved ADC and the digital de-emphasis applied to the CD master.

2015 abandoned release

When the master tape was rediscovered in 2010, there was a concerted world-wide search to find an MX-80 machine that could play it back. The original machine had been modified to 48 kHz operation by then and was not in service.

A few remaining MX-80 machines were found in New York, Berkeley and Nashville; the majority of which were set up for 48 kHz, but none of these machines were able to lock to the tape, with the exception of one machine in Battery Studios (which had originally been owned by Tom Jung of DMP). Morten Lindberg (2L) had made some 88.2 kHz transfers of the outtake tapes in 2005 which have given us valuable references, but unfortunately, this machine broke down before the main master could be attempted. Attempts in 2016 to find working machines in the USA and Japan were not successful.
In a collaboration between Ringve Music Museum and the National Library in Mo i Rana there were several attempts to transfer the original master into a modern digital form. This project involved both repair and servo adjustments to the original machine.
In 2015 there was an attempt to recover an analogue version from the repaired machine. The audio path was from the analogue output directly at 192 kHz into a Pyramix workstation. To persuade the machine to lock it was necessary to manually apply resistance to the rollers (slowing the tape speed below 15 ips while the digital electronics locked at a rate near to 48 kHz. This was a heroic but difficult process. The resulting audio was not at the correct pitch and had noticeable ‘wow’.

Norway Library work for the 2018 MQA Edition

One problem remains with the original machine: there is inadequate information on how to modify the hardware so that it can play the Fairytales masters at original rate, also bearing in mind that it is a museum artefact and would then be unable to deal with the 48 kHz masters of other recordings. The National library bought a second (48 kHz) machine to facilitate access to digital streams and some time was spent using varispeed to extract the data directly from the tape. This effort had modest success, sufficient to get the test tones and all but the first phrase of track 1. Unfortunately, it has subsequently proved impossible to extract the rest of the album.

From a reclamation point of view, getting this snapshot of the digital information on the tape has been very helpful because we can separate the ADC and DAC influences and characterise both converters. Any future work should concentrate on modifying and adjusting the original machine so that all the digital data can be obtained, particularly as we describe later how the ADC was characterised.

In early 2016 MQA asked Thomas to make some measurements of the ADC and DAC of the original machine using the pass-through monitoring path (which cascades the ADC and DAC). A complex proprietary chirp signal was provided that allowed us to extract a high-precision estimate of the amplitude and phase frequency responses as well as the impulse response of the two channels.
By using a more gentle method to establish a lock, a 192 kHz digital capture was made of the analogue output playing the 50.35 kHz master on the library machine. This analogue transfer ended up being the best starting place for 8 out of 10 of the songs on the album. Of course, the analogue transfer was at the wrong pitch, approximately in the ratio 48k: 50.35k.

Fairytales: page 3