I find myself on Internet Archive a lot during these dog days of summer. Delving into classic texts like Edgar Rice Burroughs A Princess of Mars or Aldous Huxley's The Perennial Philosophy. Discovering a forgotten H. P. Lovecraft story in the Weird Tales archive. Mining old time radio shows like Suspense for story inspiration. And using the Internet Arcade for screen grabs that can be used in retro-style game texture art. It makes me think I should do a better job of preserving my own output. You never know what future generations may find useful!
It's like the Project Gutenberg of audiobooks, recorded by volunteers. The recordings are new, but many of them are of old books, now out of copyright.
The quality of their readers varies, but there are some surprisingly good readers on there, such as my favorite so far, David Clarke, who did a superb reading of The Count of Monte Cristo.
 - https://librivox.org/
 - https://librivox.org/the-count-of-monte-cristo-version-3-by-...
Some guy just wanted to tell everyone some neat little facts about this thing he apparently knows a lot about. I find it fascinated how much people care to know about things like this.
EDIT: whoever this "arc-alison" character is, they're prolific - I'm finding their informational reviews all over this archive.
I feel really rude for saying that because I would hate to discourage people from contributing on the internet, especially to a project like this. And on that note, this project is awesome! As a crate digger, I can see myself spending a lot of time trawling through these tracks.
So many great shows too. Such an awesome resource for fans.
Digitized from a shellac record, at 78 revolutions per minute. Four stylii were used to transfer this record. They are 3.8mm truncated conical, 2.3mm truncated conical, 2.8mm truncated conical, 3.3mm truncated conical. These were recorded flat and then also equalized with NAB.
The preferred version suggested by an audio engineer at George Blood, L.P. is the equalized version recorded with the 2.3mm truncated conical stylus, and has been copied to have the more friendly filename.
I'm trying to guess but can't imagine what the reasoning for this is. I've tried A/B/C/D testing a few tracks on some crappy speakers and can't discern any difference.
While it's certainly admirable to try and digitize it as thoroughly as possible, I just can't see how a difference of 0.5mm in the stylus width is worth increasing your work load 4x times over (having to record each record 4 times rather than just once).
"During the 78rpm era there are no standards for speed, stylus size, or record/playback equalization. Within the trade there is broad agreement that optimizing playback requires both knowledge of the documentation that’s available on these parameters for each label over time, and some amount of judgment. There are many reasons why judgment is necessary. One reason is that the disc may be worn from being played many times with the correct stylus size. Better results may come from using a different (“the wrong”) size stylus because it sits in a portion of the groove that is in better condition. But there’s no free lunch. Using a smaller size may mean a noisier transfer as it plays a less cleanly molded part of the disc. Using a larger size may increase tracing distortion that is the result of the larger size not fitting all the way to the bottom of the smaller grooves of higher frequencies. [...]"
1. "friendly filename" sounds good, little static/pop, etc
2. Super loud squeal thing in the background, yuck. Voices sound poorly equalized
3. Quieter than (2) but way more noise than (1), voices causing weird audio artifacts in my headphones (as if they're blowing their available range) and are radically changing volume in the middle of a line
4. Weird squirrely noises on high-volume peaks, sounds like crap on the loudest parts. Right channel is like, totally f'd in the A man.
5. Seems to be same as (1), but with the standardized filename. I gotta agree with mr audio engineer, it just sounds the best.
(I would note that due to the way the recording is made, it may be the case that a real needle would jerk around a bit more. On the other hand, it wouldn't have to jerk around much before this entirely stops working, so I'd guess in the end it's probably pretty accurate.)
You might want to try with a half-decent usb dac and a set of good headphones if the goal was an a/b test?
You can see a picture of one of the four-armed turntables here: http://great78.archive.org/preservation/
The surface noise works well with the song.
The Laser turntable was a really bad idea, one that attempts to solve a problem that doesn't really exist in practice: record wear.
A stylus rides on the physical surface of the record. The information (audio) is on the physical height and x-position of the groove, not on the image of the groove. Thus the laser turntable also reads dirt and even damage that wouldn't be read if the stylus was used.
On a proper turntable with a proper (reasonable quality, not-worn) stylus, a vinyl record can be played in excess of 1000 times before noticing audio degradation, according to AES tests done in the '60s. Under an electron microscope it was reported that the vinyl surface appears to "flow" or "compress" under the action of the stylus.
78rpm "shellac" records are designed so the needle wears (!). For this, the osmium needles of the era should be discarded after two record sides. What happened in real life was that the needles weren't discarded, so they developed cutting edges that did damage the records.
I don't think I'd agree that the laser turn table was a bad idea necessarily though - I mean if the laser turntable worked as you'd want it to, it could be really nice, because you wouldn't need to have any moving parts potentially, or fewer moving parts (series of mirrors to position the laser or something). And fewer moving parts -> machine will last longer.
Normal turntable, moving parts:
1. turntable spindle
(2). Motor (sometimes integrated into the spindle assembly)
3. tonearm bearings
1. turntable spindle
(3). tray loader
4. tangential tonearm positioning motor
5. laser servo mechanism for precise positioning
6. laser focusing servo
The laser spot size covered about three grooves. But for playing three grooves at once, it sounded good enough to be interesting. Didn't follow up, however.
For example this one from 1902: https://archive.org/details/78_medley-of-emmetts-yodles_yodl...
I'm sure Izotope would give the RX license for free in exchange for a blog post (or any other audio software company).
I currently listen to "A Duke Ellington Panorama", just nice!
Thanks for that and keep up the awesome work!
"A consumer will never need 24-bit. Ever."
"24 bit audio is as useless as 192kHz sampling"
I would love to hear a good explanation so I can decide about the future of my audio library.
However, if you want to prepare audio (or any media) yourself, it helps to have input material that's much higher quality than your output format so that you can mess around with it and still have something that's at least a little higher quality than your output format.
I guess archive.org hopes that these recordings will be remixed and re-used and incorporated into future creations, as well as preserving their original form.
> Finally, the digital effects used in studios to mix music benefit from the higher 24-bit resolution file for microscopic processing duties.
Using 24-bit over 16-bit keeps the door open to any post-processing the user wants to do (e.g. restoration/enhancement). I very much doubt that they consider it necessary, or even useful, for general listening.
When the source is analog (like these old records), I can see the argument for preserving them at the highest reasonable depth. The fundamental purpose of the archive is to preserve, regardless of whether or not people subjectively prefer the sound.
If the subjective experience of 24-bit is unpleasant to some people, you can always go down, but you can't add back bits that were never captured, so it's my opinion that 24 is the appropriate format here.
Of course, from a philosophical perspective, we're still talking about two methods for reaching infinity and which gets "closer"
I really like some of the audio here but it needs some post processing. The only thing I can find to do it is audacity and it doesn't look very friendly to scripting.
To outline the starting assumption and desired ending points, what we have are four tracks with different needle sizes. The smaller needles tend to "wobble" in the track more (will pop/hiss more) but the larger sizes may miss fine detail (treble) that the smaller needles pick up. The assertion of the Internet Archive is that the needle that is closest in size to the one used to record the track will produce the best output, but again we are not limited to just a single track, we can programmatically combine them to produce a better output. The desired end-goal is a "clean" track with maximal spectral quality and minimal pop/hiss.
I think there are two distinct tasks here. Maximizing spectral quality and de-popping the track.
For the first task, my layman's description of "maximizing spectral quality" would be that we combine the frequency ranges that each track is "best" at. In other words the finest needle has the best treble, while the biggest needle has the cleanest bass. That might be implemented by some kind of averaging, or a weighted average (eg weight tracks that are "most different" from the average track, or from the cleanest track).
Then you de-pop the resulting track. In terms of machine learning, this should be something that is amenable to deep learning. If you train a net to identify what a "pop" or "hiss" is then you can have it directly produce a clean output, or produce a "pop/hiss track" that you can then subtract from the input waveform (same thing).
If you want something more programmatic, you could again play around with generating a "noise track" by subtracting the "clean" signal (biggest needle) from the "most detailed" signal (finest needle), perhaps also repeating this with each other track as well. The "noise track" would still have some signal inside it and you would need to apply some other method to further separate that out, but you would be working only on a portion of the total signal so in theory you would lose less detail than working on the whole signal.
If anything, let me tell you that "clicks&pops" can be greatly minimized simply by using a better "turntable+tonearm+cartridge" combination!!
Just at a glance, I'm seeing The Light Crust Doughboys, basically a string band supergroup. Multiple members would go on to found famous western swing bands (Bob Wills, Milton Brown). Very proto-rock-and-roll -- listen to that electric guitar -- Elvis would cover some Western Swing numbers in his early days.
Also seeing some older stuff, including a few recordings by the (arguable) best banjo player of all time, Vess L. Ossman (from 1907). Pretty cool to listen to these march numbers and then hear them evolve into jazz/ragtime only a couple years later (this is a recording by Fred Van Eps, the second best banjo player of all time, from 1914).
EDITS: seeing some other personal favorites:
Hank Penny, a favorite western swing singer of mine. He usually does it hot/upbeat/fun.
Blind Blake, a guitarist who could play the fretboard like a ragtime piano!
Oh, and here's the WWII-era Bob Wills I was waiting for. Got that classic Leon McAuliffe pedal steel playing. No Tommy Duncan vocals, unfortunately.
Neat! An old solo Art Tatum! Widely considered the best pianist of all time... And another, a whole album!
Really classic early electric guitar playing on a jump blues number by T-Bone Walker. I actually believe he's one of the first to use the electric guitar in blues.
Great steel guitar playing on this Gene Autry cowboy number.
Looks like there's a lot of Django for all you gypsy jazz fans. Never heard this take on Avalon before, I dig it.
Lot more to dig through and lot of obscure stuff I'd like to give a shot, but I'm out of time for now...
(Though these are songs you can already find on CD or on Spotify.)
The Archive is taking a fair use stance here, but as with some of their other projects there's a significant "ask for forgiveness, not permission" element at work as well.
A lot more here: https://archive.org/details/78rpm
My eventual life goal is to do something similar with my Brazilian record collection... have the skeleton of such catalog at: https://www.novedos.com/collection.
Cab Calloway, The Man from Harlem
*IANAL and this may not be the case for all the material but I'm sure that there is mountains of inspiration to be mined.
Probably to allow for better post-processing in case one groove wall is in better condition than the other. They seem to be going for completeness rather than hard drive efficiency.
> This means we deliver both groove walls of 4 different stylus sizes with and without EQ for a total of 16 channels of audio. The most comprehensive presentation of 78rpm discs ever!