Correcting a very odd pitch issue from a 1928 recording

I'm posting this purely for information - I hope it's of interest to some.

I've been working on remastering a series of recordings made by Mississippi John Hurt in 1928. Most of his 13 sides were cut at Okeh's New York studios in the December, but an earlier session took place at (most likely) a hotel, in Memphis, yielding two sides that were released as Okeh 8560.

When I started working on the first of these two I sensed something odd was going on with the pitch, and set about analysing the problem. Turns out the original recording machine had some significant speed issues. I checked the pitches of an F sharp harmonic that runs consistently through the song and used this to map out an inverse pitch change:



Because this is the inverse graph of the pitch inconsistencies when the record it played back at a constant speed, it also shows what the cutting lathe was doing - starting fast, winding down rapidly, and then more slowly.


A zoomed in FSE view of the recording, concentrating on a specific set of harmonics, looks like this:




After applying the sliding pitch change curve shown above, and then doing some very fine tuning using the visual FSE display and working at <0.1 semitone resolution in the pitch bender I was able to get this recording to look like this:




The other surviving track from these sessions exhibited precisely the same problem - by applying the same pitch-bend curve (adjusted so that the nodes hit the right timings - otherwise it stretches to the overall duration, which was shorter in the second song), I was able to almost entirely 'straighten out' the second song, with again only a very slight visual adjustment required once I was done.

I don't recall ever seeing exactly this problem with a 78rpm record before - has anyone else?


You can listen to the opening minute of the corrected and remastered version here


(The recordings, BTW, will be issued on Friday at www.pristineclassical.com - the seventh release in our historic Jazz and Blues series)

The only mechanism that I could envisage that could cause something similar would be if the motor driving the turntable was also driving the cutter head across the disc and that the load was varying as the cutter head moved from the outside to the inside.

A colleague of mine has suggested a poorly governed spring-driven motor was most likely to blame for this type of 'field' recording, given the date.

Recording companies had portable recording outfits that toured the "Territories" and places such as New Orleans, especially after the introduction of electrical recording in 1925.  Such outfits were invariably (lead-acid secondary) battery operated due to the variability (or even lack of) mains electricity supplies.    Speed regulation was often manual, and not very precise .   (Even the major studios used dc for this function to isolate themselves from eccentricities in the mains supply; some even used gravity feed to drive the lathe - notably EMI even into the '50s, until the arrival of the LP).   I am not aware of recording outfits that used spring drive for the lathe at this time.  (The BBC of course developed its miniature disc cutter during the war, which used spring drive for the turntable, and batteries only for the cutting amplifier - and there was that portable EMI tape recorder in the early '50s that used the same configuration).

I would suggest that the pitch behaviour observed suggests a partially discharged battery, where polarisation of the plates can occur during discharge, leading to increased voltage drop and hence lower speed.  During the break between takes, the cells recover.  I can't propose a scenario relating to increased cutter drag, because this should decrease as the surface speed under the cutter (and hence drag) decrease as the cutter moves from outer to inner diameter. 

Of course, one could also conjecture that with the intensity of playing, the guitar strings heat up and the pitch progessively drops during the song ... 

I wasn't thinking so much of cutter drag as  wear or stickyness of the helical drive shaft that presumably moved the cutter head across the disc.

The friction to traverse the head along the leadscrew is basically constant, and small (due to the high gearing ratio) in comparison to the effort of driving the turntable itself against the cutter head.   Unless the the leadscrew was damaged severely, I would tend to discount this as a cause of the observed behaviour.

But I should have said also that your solution was very effective and I will keep it in mind for archival transcription work that I do. 

I note that in some transcription of magnetic tapes, some work has been done on recovering the bias frequency and using this as a reference/pilot tone to corect long and short term speed variations eliminating wow, and to some degree flutter as well.   In future generations maybe they can use the high frequency 'dither' that DGG used for their Direct Metal Mastering process cutting directly onto copper blanks to eliminate wow and flutter when they transcribe them .....

Thanks for this, Panatrope - I've quoted you in full (which I hope is OK) on our website, and would like to credit you appropriately if you can supply a 'real' name!

You can see what I've written here:

http://www.pristineclassical.com/LargeWorks/Jazz/PABL004.php

The recording is now out:

Andrew

I'm happy to be quoted in the way you have done.  If nothing else, if I am wrong, somebody will correct it and I'll learn.  But I have an interest both in traditional jazz, and recording technology, and I can't immediately put my finger on the appropriate reference, but I have seen some photos of early portable recording outfits and this is what informed my response.

Of course the restoration and preservation of the recordings of early blues artists has exactly the same importance and early jazz recordings, and I will try to have a listen to some of the recordings in due course.

Meanwhile, I can be contacted at .   

(BTW, Panatrope was the name of the first electric phonograph produced by the Brunswick company in 1926.)

On the subject of early Jazz and Blues, perhaps before the two fully diverged into separate entities (as recognised by performers and public), a few weeks ago I was playing with Bessie Smith's 1925 acoustic recording of St. Louis Blues, which features Louis Armstrong's cornet playing. What astonished me more than anything was to find genuine harmonics from the cornet right up at 9kHz. As I said, this was from an acoustic 78...