SoylentNews is people
SoylentNews
The Great 78 Project over at the Internet Archive has been professionally digitizing old 78 RPM records for a while now. These records were all made between 1898 and sometime in the 1950s. Over 20 collections have been selected for digital access and physical preservation with the help of George Blood, L.P. and the Archive of Contemporary Music. So far about 26,000 of the 78s have been added to the Internet Archive. Each disc has about 3 minutes of audio per side. Most of the discs are made from shellac and really quite brittle, perhaps even more brittle than today's digital formats.
This discussion has been archived.
No new comments can be posted.
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.
Life is to you a dashing and bold adventure.
(Score: 2) by wonkey_monkey on Monday August 14 2017, @03:26PM (3 children)
You might have got a slight improvement by averaging them.
systemd is Roko's Basilisk
(Score: 1, Interesting) by Anonymous Coward on Monday August 14 2017, @04:31PM (2 children)
> You might have got a slight improvement by averaging them.
Or not, if they were slightly out of phase (esp. at high frequencies)...
Getting a good mono signal from a true stereo pair is not trivial, adding them together often kills off a lot of the high frequency information as the distance from a "point source" may be slightly different for the two mics.
(Score: 0) by Anonymous Coward on Monday August 14 2017, @05:31PM
One could use frequency filters to average the lower-frequency end of the channels together, and use only one channel for higher frequencies to avoid the "alignment" problems you mentioned. Of course, that complicates things. But if you want professional or high-quality results, experimentation is warranted.
Post-recording phase adjusters could also be used to align higher frequencies. It's also possible that if dust or a scratch is on one stereo channel but not the other, then it could be subtracted out from the final result altogether rather than averaged in. One could set a "difference threshold" such that if the wave-form difference between each channel differs too much, then only include the channel with more "normal looking" wave set, based on surrounding wave patterns.
Laser-scanning of the grooves to recreate a 3D model of them could even be better than 2 channels, if it has enough resolution. It would be easier to identify (estimate) bad spots. One figures the original wave bumps took up most of the groove's shape. Scratches and dust would either take up just a portion of the "valley", or not fit the typical grove profile. A lump of dust would be convex instead of concave, for example. And, scratches would either mostly be on the upper parts, and/or make diagonal patterns, which are unlikely in the original. It's kind of like digitally patching up a beat-up statue: you know mostly what it's supposed to look like based on common sense and similar statues of the time, and thus can make pretty good guesses about where to put the stucco. Of course, some guesses will be wrong, but a having a clean statue may be worth it.
(Score: 2) by wonkey_monkey on Monday August 14 2017, @06:24PM
That's not the case here though. The original recording was also mono.
systemd is Roko's Basilisk