Hacker Newsnew | past | comments | ask | show | jobs | submitlogin
Smithsonian will restore hundreds of the oldest sound recordings (smithsonianmag.com)
82 points by 8bitsrule on Jan 18, 2023 | hide | past | favorite | 13 comments


They didn't mention the technology used, which is called IRENE.[1] Both cylinder and disc recordings can be imaged. This results in a huge 3D model of the recording medium, which can then be "played" in simulation. Even broken records can be recovered.

[1] https://irene.lbl.gov/


This is very cool. I've been pondering this for a while. With all the very expensive audiophile turntables out there it seems that the true goal for an audiophile should be listening without degrading the vinyl, this.

It also makes me think. This decoding system can pull out essentially height data, height of the groove at moment in time.

With that, it's not outside the realm of possibility for us to create a small actuator that takes the place of then stylus, manually moves the turntable needle, recreating the exact sound as if the record were being traditionally played.


Non-contact record players been built.[1] The trouble is, non-contact record players track every tiny bit of dust. Elaborate record cleaning is required. So they're not useful for routine use. Also, they're overly expensive low volume products.

The IRENE people are trying machine learning to recognize and remove dust and scratches. That sort of thing is for reading archival items once and generating a archival digital copy at some very high sampling rate and resolution. That's then down-converted for listening.

Many early cylinder records sound quite good when played back with modern equipment. It turns out that most of the distortion came from the mechanical playback process, where huge stylus pressures were used to get enough motion to directly drive a speaker membrane. The recording process was making reasonably good recordings.

(History of electronics: "We need more volts." "We need more gain." "We need more frequency." "We need more stability." "We need more power capacity." "We need more gates". "We need less heat.")

CDs, DVDs, etc. have error-correcting codes, so they don't have this problem.

[1] https://en.wikipedia.org/wiki/Laser_turntable


There exists such a thing as a laser turntable. They're probably not as popular because they came out on the tail end of vinyl being popular, they didn't deal well with dust, and could only play opaque records.


There are some experimental sound recordings made a few decades earlier by a Frenchman [0] , although they were not intended for playback.

[0]: https://en.m.wikipedia.org/wiki/%C3%89douard-L%C3%A9on_Scott...


What is a recording if not to allow for playback?


"Unlike Thomas Edison's later invention of 1877, the phonograph, the phonautograph created only visual images of the sound and did not have the ability to play back its recordings."

In my mind this seems like the difference between the visual representation of the waveform an audio editor shows vs the literal bits encoding the actual audio.

In practice one can approximate the original sound from the image of the waveform, but in most cases it will likely lack in resolution relative to a medium engineered for playback.


Oh please I hope they release the raw data from the scan in addition to the restored audio as PCM.


I think it was back in the 70's when people used lasers as a "needle" to play records without touching them.

I seem to recall 30 years ago or so that someone took photos of records and managed to reproduce the sound from them.


I also seem to recall photos-sound, but have yet to rediscover it via web search.


Laser turntable, 1970s - https://en.wikipedia.org/wiki/Laser_turntable , which goes on:

> The [1980s production] prototype revealed an interesting flaw of laser turntables: they are so accurate that they "play" every particle of dirt and dust on the record, instead of pushing them aside as a conventional stylus would.

In the same Wikipedia entry:

> A similar technology is to scan or photograph the grooves of the record, and then reconstruct the sound from the modulation of the groove revealed by the image. Research groups that developed this technology include:

IRENE is the one used by the Smithsonian. https://irene.lbl.gov/links/audio-files-for-june-24-2021-pre... has some of the recordings recovered by IRENE.

I've wondered if regular cameras would get good enough that you could take a picture of a record with your phone and have it play the sound. IRENE and VisualAudio links show that a microscope is needed, so no, it doesn't seem "phone"-o-graphic playback will be any time soon.


https://doc.rero.ch/record/7958/files/StotzerS.pdf is a PhD thesis for VisualAudio. It says "To get a good audio quality, a resolution of about 1 μm is needed, which would mean 25000 dpi."

Yeah, it's going to be a while before phones get there!

Also, here's Heine's 1977 laser scanning phonograph paper - https://www.aes.org/e-lib/browse.cfm?elib=3098 .

Here's a Saphir paper - https://hal.science/hal-01885324/ .

Here's an approach using an off-the-shelf scanner - https://www.aes.org/e-lib/browse.cfm?elib=16071 (the VisualAudio PhD thesis mentions these as having lower quality).


Thank you. I'm hoping IRENE gets discussed in detail here. We will eek out someone qualified to elaborate on Laser Turntables.

https://www.bl.uk/events/unlocking-our-sound-heritage-inspir...




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: