Slashdot Mirror
Making Old Sound Recordings Audible Again
orgelspieler writes "NPR is running a story on a safe way to reproduce sound from ancient phonographs that would otherwise be unplayable. The system, called IRENE, was installed in the Library of Congress last year. It can be used to replay records that are scratched, worn, broken, or just too fragile to play with a needle. It scans the groves optically and processes them into a sound file at speeds approaching real time. IRENE is great at removing pops and skips, but can add some hiss. Researchers are also working on a 3D model that is better at removing hiss."
And the National Library of Canada has had one of those units since 1992.
looks like 3d noise reduction is done in the image domain - not in the audio.
Can add some hiss to what? To the perfect Hi-Fi quality you are expected to get out of a century old phonograph?
To the level of his that the recording itself actually contains.
Old recordings actually did a very good job of making a record of the actual sound. But dust on and damage to the surface produced artifacts in the output signal when played with a needle.
Optical techniques can identify the actual flat surface of the groove and ignore the artifacts. But digital approaches to performing this scan and/or encoding the result add errors from quantization and digitizer nonlinearity, which appears as added hiss - the amount depending on the resolution and quality of the converter and/or scanner.
Bantam Dominique roosters crow a four-note song. Once you've heard it as "Happy BIRTHday" you can't NOT hear it that way
I've been a fan of ancient public domain music for a while now. I hope they are kind enough to post these on a website for our listening pleasure.
Give Kashyyyk back to the Wookies
I don't know about the Canadian system, but the Japanese system is different, and if I understand it correctly, much less capable. The Japanese system spins the disk and replaces the mechanical stylus with two lasers, if I read the description correctly. The IRENE system takes a picture of the surface and reconstructs the groove pattern from the image. The record surface does not move. This is why IRENE can scan a record even if it is broken. In the NPR article, they describe how they input an old recording that had a broken section. They just fit the two pieces together and scanned the surface. IRENE can also scan cylinders as well as disks, which the Japanese system cannot do for mechanical reasons.
Well at least ArchieBunker lived up to his pseudonym: ignorant and proud of it. When you are incapable of reading and understanding an article, I guess you have to compensate by trying to demean creative people who do worthwhile work. Instead of yawning, Archie should stay off Slashdot and go back to watching reruns of old TV shows, where no mental activity is required.
I don't know what slashdot is coming to, this is a total dupe! OK, so that story is from 2005, so what? OK, so I remember slashdot stories from two years ago. So what? Doesn't mean I don't have a life. Right? Right?
(Actually that other story is pretty cool, has some neat pictures and goes more in depth on the technology. And theres a nice thread talking about three-grooved records).
--
Looking for a C/C++ job in Silicon Valley?
Qxe4
the perfect Hi-Fi quality you are expected to get out of a century old phonograph?
Surprise, surprise, listen to the fine samples. The first collection sounds like it was recorded yesterday. The technique is unbelievably excellent. This is very good news for music preservation.
Friends don't help friends install M$ junk.
Because you didn't read the article! Laser record scanners have to spin the record around, this one scans it in place, so you can scan broken records or old Edison cylinders.
No, not improved. Just totally different.
This is not a laser-distance-based modeling system. That sort of system tracks along the grove mechanically (without touching, but still moves the lasers, much like a CD) and models the surface by reading the distance from the laser to the disk surface.
This system takes an image of the entire disk surface in one pass, with no moving parts. That image is then processed to construct a 3D model of the surface, and that model can then be processed to follow the groove track, much like the laser-based system physically scans the disk surface.
After the 3D model is constructed both systems work much the same way, but the construction of the model is significantly different. The laser-based system can only play flat disks (not say, wax cylinders), and cannot pre-process the disk to construct an accurate model from pieces of a disk. Also the image-based system could be used with any set of images of a disk sufficient to reconstruct the surface -- it would not be necessary to physically transport the disk in order to process it with such a system, so long as the necessary images can be produced at the disk's current location.
Coincidentally, I just spent the last weekend converting some old 78s using a modern (albeit not laser-based) record player. I wrote a little article about it here: http://www.ambor.com/public/78rpm/78rpm.html, including some sample audio clips that show what the raw recording sounds like and then shows what some open source audio restoration software can do.
While it's damn near impossible to get a hold of new music recorded in a high resolution, digital format, it's pretty safe to say that most music is widely available at a 44.1khz sample, 16 bit sound, with no compression.
Wrong sort of compression. All audio CDs are compressed heavily so that this week's Best Thing Ever sounds just that little bit louder than last week's Best Thing Ever.
I think it's partially a generational thing. Here in the UK, our national general music station, BBC Radio 2, long had a reputation of playing sixties music by day and jazz and big band music by night. Its target audience was generally the over 40s. These days it has pulled its audience back to the over 30s and has actually paid attention to what the over 30s listen to, and has become the best station in the country. But then again, I'm well over 30.
Even older and of great cultural importance are wax cylinder recordings.
From TFA : "When taking flat photographs, it can create a three-dimensional image of the groove on a record, or on an old wax cylinder. Haber been working with the University of California's Phoebe Hearst Museum of Anthropology, to reconstruct sound from field recordings, like one wax cylinder made around 1911 that features a Native American called Ishi."
This system takes an image of the entire disk surface in one pass, with no moving parts. That image is then processed to construct a 3D model of the surface, and that model can then be processed to follow the groove track, much like the laser-based system physically scans the disk surface.
The print article that TFA links to seems to suggest that, but I could swear that in the audio they aired yesterday morning the inventor said the current system uses 2D images. IIRC, the inventor said 3D images are planned for the next generation machine and are expected to help reduce the hiss. But IMNRC (I may not recall correctly)....
Music isn't all at the same volume. To make a piece stand out, you can use compression (dynamic compression, not like mp3 compression) to make the louder bits a bit quieter and the quiet bits a lot louder. This is why TV adverts seem so much louder than the rest of the programme material - the quiet bits have been turned up. Radio stations use a thing called an "Optimod" to get the maximum possible modulation without the signal distorting, which is essentially an extremely aggressive compressor. If you listen to certain dance music where the bass drum makes the rest of the track "pump" - fade up a bit after each beat - then you're hearing compression at work.
It's referred to as "the loudness war", the industry-wide effort to make every single and album sound louder than everyone else, at the expense of dynamic range.
Once again, the Wiki is your friend: http://en.wikipedia.org/wiki/Loudness_war
The sad part about it is that the kids I've tried to explain this to, actually like their music to be a dull wall-of-noise. And sadly, by the time they're mature enough to perhaps appreciate the subtleties of properly-recorded music, their hearing will be too damaged to do so.
(If only they'd GET OFF MY LAWN!)
Learn from the mistakes of others. You won't live long enough to make them all yourself.
Wow. I see what you mean. I didn't know that this was a common practice, but it certainly explains a lot. I have noticed that older tracks tend to be more listenable at high volumes in my car, and in particular, how newer tracks tend to have treble that hurts my ears at the same volumes. Either way, I guess it goes to show what a nerd thinks of when he reads "Compression." Thanks for the info.
Boot Windows, Linux, and ESX over the network for free.
The way I do my vinyl restoration is using Cool Edit Pro aka Adobe Audition. For the noise reduction pass, what you do is take a sample of the between-song space, which is normally empty, and then mathematically subtract this 'silence' from the entire sound file. Voila! The surface noise is eliminated from you recording, and you can then do any more de-clicking, etc. as you will.
It doesn't mean much now, it's built for the future.