Attempt to Convert 45 RPM Record Scan to Audio
In 1997 Charlie Baker, who used to manage the electronic music studio at UCSB, Lee Worden, and I got the idea to convert a high-resolution scan of a vinyl record to sound. We weren't looking for high fidelity; we just wanted a proof-of-concept, and to have some fun.
I selected an old 45 of The Monkees "Steppin' Stone", because it was my sister's, it was heavy on bass and bass solo, and reproducing the bass from an image requires the least resolution. Calculations showed that we should be able to make a scan with sufficient resolution to reproduce low musical frequencies. We opted to use the graphics functions built-into the NeXT Computer library of NeXTstep to track and decode the groove.
Our first attempt failed miserably. We used an HP scanner, and the illumination source is mounted almost directly at the same 90˚ location from the surface of the scanning table as the sensor. The reflection from the 45˚ sides of the black vinyl record grooves was minimal. (An unfortunate example of the same stealth reflection principle as the F-117.) The contrast of the image of the grooves vs. that of the lands was too low to allow our software to track the grooves.
People we told about this thought that it would be a "great hack." We were told that we had an open invitation to the "Hacker's Convention" to show it whenever we got it to work.
I had a brainstorm: Why not increase the contrast by rubbing a bar of handsoap over the surface of the record, lodging white soap in the grooves, and therefore increasing the image contrast of the lands vs. grooves? We felt that we could follow the center of the white area, deducing where the center of the bottom of the groove would be.
This seemed to work a little better. Our program was able to track a groove for at least a revolution or two of the record before it "skipped." When we decoded the audio, we got a weird crackling modulated white-noise sound. At first we thought our audio translation S/W was broken. But then we noticed that the amplitude envelope of the noise followed the macroscopic swipes of the handsoap across the the record, which I had made radially from the center of the record outward. Apparently the offset of the center of the soap filling each groove was offset by depositing more soap on one side of the groove than the other. We somewhat successfully reproduced the soap swipes. But we couldn't tell for sure if we heard any bass guitar buried in the noise or not. Unfortunately we never got back to the project.
It would still be a great hack! Good luck to whomsoever manages to get it to work!
In 1997 Charlie Baker, who used to manage the electronic music studio at UCSB, Lee Worden, and I got the idea to convert a high-resolution scan of a vinyl record to sound. We weren't looking for high fidelity; we just wanted a proof-of-concept, and to have some fun.
I selected an old 45 of The Monkees "Steppin' Stone", because it was my sister's, it was heavy on bass and bass solo, and reproducing the bass from an image requires the least resolution. Calculations showed that we should be able to make a scan with sufficient resolution to reproduce low musical frequencies. We opted to use the graphics functions built-into the NeXT Computer library of NeXTstep to track and decode the groove.
Our first attempt failed miserably. We used an HP scanner, and the illumination source is mounted almost directly at the same 90˚ location from the surface of the scanning table as the sensor. The reflection from the 45˚ sides of the black vinyl record grooves was minimal. (An unfortunate example of the same stealth reflection principle as the F-117.) The contrast of the image of the grooves vs. that of the lands was too low to allow our software to track the grooves.
People we told about this thought that it would be a "great hack." We were told that we had an open invitation to the "Hacker's Convention" to show it whenever we got it to work.
I had a brainstorm: Why not increase the contrast by rubbing a bar of handsoap over the surface of the record, lodging white soap in the grooves, and therefore increasing the image contrast of the lands vs. grooves? We felt that we could follow the center of the white area, deducing where the center of the bottom of the groove would be.
This seemed to work a little better. Our program was able to track a groove for at least a revolution or two of the record before it "skipped." When we decoded the audio, we got a weird crackling modulated white-noise sound. At first we thought our audio translation S/W was broken. But then we noticed that the amplitude envelope of the noise followed the macroscopic swipes of the handsoap across the the record, which I had made radially from the center of the record outward. Apparently the offset of the center of the soap filling each groove was offset by depositing more soap on one side of the groove than the other. We somewhat successfully reproduced the soap swipes. But we couldn't tell for sure if we heard any bass guitar buried in the noise or not. Unfortunately we never got back to the project.
It would still be a great hack! Good luck to whomsoever manages to get it to work!
Very Cool project!
I think way back when, someone tried to read the grooves using a laser.
But the closes thing that ever became of it was using a laser to help with the tracking of the tonearm.
I had thought of your concept aswell but I am not a programer.
I think it is a great idea ,But my only question is can you excede the resolution using the analog to 24 bit conversoin format?
Back when I was much younger I had discoverd that by playing my LP's at 16 RPM's and recording them to tape and then playing back the tape at double speed I got a much cleaner recording with a much better dynamic range than the LP at normal speed.
When I start digitizing my LP's I will use the same method.
Just a few thoughts.
jer
I think way back when, someone tried to read the grooves using a laser.
But the closes thing that ever became of it was using a laser to help with the tracking of the tonearm.
I had thought of your concept aswell but I am not a programer.
I think it is a great idea ,But my only question is can you excede the resolution using the analog to 24 bit conversoin format?
Back when I was much younger I had discoverd that by playing my LP's at 16 RPM's and recording them to tape and then playing back the tape at double speed I got a much cleaner recording with a much better dynamic range than the LP at normal speed.
When I start digitizing my LP's I will use the same method.
Just a few thoughts.
jer
Last edited:
Laser turntable? Fantastique! Thanks, SS
Now all we need is a cheaper version.
As regards the scan decoding, you really want to see the undulations in both sides of the groove. If you had an opaque contrasty (white) liquid that would settle in the bottom of the 'V', then you would be able to distinguish the information more readily. There'd be some considerations of surface tension and wetting. Any materials technologists out there? Sounds like a nanotech application.
I just measured an LP, the innermost track is at a radius of 67mm, call it 65 for a margin. This equates to a circumference of 408mm. At 33.33RPM this is 13.6 metres/min, ~0.225 metres/sec. Taking an upper frequency of 20k this gives a minimum sinewave feature length of 11.3 microns or 0.44 thou. So the minimum scan resolution would need to be ~5000dpi, but more reasonably you'd want some oversampling, say 20,000dpi. Canon have a flatbed at 9600*9600.
OK, put it on the back burner for a couple of years.
Now all we need is a cheaper version.
As regards the scan decoding, you really want to see the undulations in both sides of the groove. If you had an opaque contrasty (white) liquid that would settle in the bottom of the 'V', then you would be able to distinguish the information more readily. There'd be some considerations of surface tension and wetting. Any materials technologists out there? Sounds like a nanotech application.
I just measured an LP, the innermost track is at a radius of 67mm, call it 65 for a margin. This equates to a circumference of 408mm. At 33.33RPM this is 13.6 metres/min, ~0.225 metres/sec. Taking an upper frequency of 20k this gives a minimum sinewave feature length of 11.3 microns or 0.44 thou. So the minimum scan resolution would need to be ~5000dpi, but more reasonably you'd want some oversampling, say 20,000dpi. Canon have a flatbed at 9600*9600.
OK, put it on the back burner for a couple of years.
That is absolute bunk.
At the minimum disc radius for a 33 1/3 RPM disc the groove speed is approximately 0.1 m /s. Allowing an upper bound of 50 kHz for things like quad LPs, the minimal wavelength is thus 2 um.
Nyquist theorem thus puts the minimal wavelength for capture of all data at 1 um which is well into the infrared.
Same argument works for amplitude excedpt here the limit is SNR.
Note that I'm being generous here - if you are scanning and converting to CD then obviously the minimal wavelength for capture is greater -around 5 um.
Last edited:
I was about to post a similar message. I think the Young's modulus of vinyl precludes any possibility of reading information at the 0.5 micron level. With a laser - perhaps.
There are two approaches to this to could first find out what wavelengths the vinyl reflects strongly at and use a laser of that wavelength to do the scan.
Or you could silver (or gold which is better at infrared) plate the record and scan the record like the op did.
Or you could do a shallow angle scan using a digital HD video camera and macro lens. This would have to be done twice from left and the right of the lands. The images would digitally combined to eliminate noise.
Or you could silver (or gold which is better at infrared) plate the record and scan the record like the op did.
Or you could do a shallow angle scan using a digital HD video camera and macro lens. This would have to be done twice from left and the right of the lands. The images would digitally combined to eliminate noise.
Wow, I didn't know about the ELP, or IRENE 2D and 3D scanner projects.
I really like the idea that the ELP laser does not ACCUMULATE dirt in use like a needle tends to do. A transcription (or just needle playing) that begins well with subsequent deteriorating qaulity as the needle cleans the groove and accumulates debris is very frustrating. Already a pretty neat product.
The current product is purposely analog and operating in real-time, with the understanding that subsequent digitization and processing is always still possible. But with development I would not be surprised if the ELP could be enhanced to run much faster, and perhaps the RIAA curve etc. also applied in digital post-processing.
It sure seems to me the laser ELP has room for improvement in the area of recognizing the difference between dirt or damage versus the cut signal of interest. I'm no expert on needle gem cut and groove contact theory, but it seems a needle suffers from deflection from the correct path by any small interfering dirt particle or additive damage located anywhere along the needle contact path, whether high or low in the groove (details dependent upon the particular needle cut). The ELP VSO feature allows the user to change the depth into the groove where the reading occurs. To state it simply (and slighlty incorreclty simplified), the needle is the worst case when dealing with dirt, where the entire needle is deviated in its path from any contact with even very small dirt regradless of how deep or shallow down the groove wall the particle is encountered. The laser system currently allows you to select ONE groove depth where the data is read, which is no doubt an improvement, especially when choosing a non-worn part of a worn groove.
But it seems logical that multile lasers reading at different groove depths could (with either analog processing or preferably digitization and digital post-processing) make intellegent decisions and do a better job. And that's just processing the data available at the same point along the groove. Just as cleaning up analog video involves processing across subsequent frames (across time), it should be possible to make huge improvements via specialized processing of the multiple-path (different-depth) data across time (using data before and after the point in the groove or point in time of interest) in post-processing (like scratch/pop processing software).
That link to the IRENE 2D and 3D readers is sure interesting! Look at those groove pictures! It's been years since I've looked at a groove in a microscope. It certainly seems to me that specialized procesisng along the general lines of the current "digital enhancement" processes incorporating specialized edge and line recognition (within the rules of one original cut) holds hope for pretty good recovery of even moldy vinyl. I'm going to have to read more about this. Some kind of digitization of very detailed groove data sure looks like it should lead to superior data recovery in sophisticated post-processing. Very cool stuff and food for thought and study. And with modern digital processing power, with devleopment this too could be much faster than real-time.
Too bad there is not sufficient truly mass market demand to make really economical products; something like a fast advanced ELP with digital procesing for prices only maybe a few times more than your PC's CD-reader instead of $20,000.
I really like the idea that the ELP laser does not ACCUMULATE dirt in use like a needle tends to do. A transcription (or just needle playing) that begins well with subsequent deteriorating qaulity as the needle cleans the groove and accumulates debris is very frustrating. Already a pretty neat product.
The current product is purposely analog and operating in real-time, with the understanding that subsequent digitization and processing is always still possible. But with development I would not be surprised if the ELP could be enhanced to run much faster, and perhaps the RIAA curve etc. also applied in digital post-processing.
It sure seems to me the laser ELP has room for improvement in the area of recognizing the difference between dirt or damage versus the cut signal of interest. I'm no expert on needle gem cut and groove contact theory, but it seems a needle suffers from deflection from the correct path by any small interfering dirt particle or additive damage located anywhere along the needle contact path, whether high or low in the groove (details dependent upon the particular needle cut). The ELP VSO feature allows the user to change the depth into the groove where the reading occurs. To state it simply (and slighlty incorreclty simplified), the needle is the worst case when dealing with dirt, where the entire needle is deviated in its path from any contact with even very small dirt regradless of how deep or shallow down the groove wall the particle is encountered. The laser system currently allows you to select ONE groove depth where the data is read, which is no doubt an improvement, especially when choosing a non-worn part of a worn groove.
But it seems logical that multile lasers reading at different groove depths could (with either analog processing or preferably digitization and digital post-processing) make intellegent decisions and do a better job. And that's just processing the data available at the same point along the groove. Just as cleaning up analog video involves processing across subsequent frames (across time), it should be possible to make huge improvements via specialized processing of the multiple-path (different-depth) data across time (using data before and after the point in the groove or point in time of interest) in post-processing (like scratch/pop processing software).
That link to the IRENE 2D and 3D readers is sure interesting! Look at those groove pictures! It's been years since I've looked at a groove in a microscope. It certainly seems to me that specialized procesisng along the general lines of the current "digital enhancement" processes incorporating specialized edge and line recognition (within the rules of one original cut) holds hope for pretty good recovery of even moldy vinyl. I'm going to have to read more about this. Some kind of digitization of very detailed groove data sure looks like it should lead to superior data recovery in sophisticated post-processing. Very cool stuff and food for thought and study. And with modern digital processing power, with devleopment this too could be much faster than real-time.
Too bad there is not sufficient truly mass market demand to make really economical products; something like a fast advanced ELP with digital procesing for prices only maybe a few times more than your PC's CD-reader instead of $20,000.
Last edited:
some of you are missing a physical limit of optics - look up Gaussian beam, diffraction limit, f number, Airy disc for optical systems
with any depth of field, high f number optics have a many wavelength Gaussian beam waist diameter
line contact styli can have 2 um radius - clearly giving smaller contact width than f-10 (ridiculously optimistic, requiring focal depth tracking at audio frequency) optics spot size with visible light
unfortunately for the op all of this speculation is pointless - if he just starts 1x transcription with good conventional phono setup he will be done a least a decade before anything else discussed here can give equivalent results at faster than real time - certainly not at a competitive cost if the tech is ever developed
with any depth of field, high f number optics have a many wavelength Gaussian beam waist diameter
line contact styli can have 2 um radius - clearly giving smaller contact width than f-10 (ridiculously optimistic, requiring focal depth tracking at audio frequency) optics spot size with visible light
unfortunately for the op all of this speculation is pointless - if he just starts 1x transcription with good conventional phono setup he will be done a least a decade before anything else discussed here can give equivalent results at faster than real time - certainly not at a competitive cost if the tech is ever developed
I thought that was a bunch of eggheads at a US lab
Research News : From Top Quarks to the Blues
Home Page
http://www-cdf.lbl.gov/~av/cylinder-paper-PVF.pdf
http://www-cdf.lbl.gov/~av/JAES-paper-LBNL.pdf
Edit: Sorry, missed the post where this was already mentioned. My Bad.
Last edited:
I know one pretty weird hi-fi guy, who occasionally recycles old Pioneer Laserdisc players into something capable of playing vinyl the same way ELP does. Can you believe it? It is not a joke.
As I can see, my post not taken seriously. As to me, I've discussed ordering one such LD to LP player conversion for myself. I'm still not sure if it makes a sense for me, because:
1) I'm pretty happy with my current heavy-weight Micro turntable, it's speed accuracy would be hardly surpassed by the LD conversion deck,
2) I'm in no need to digitize my humble collection of just 1000LPs,
3)Not sure how laser will behave on warped records ,
Besides that, it is still an interesting thing to do. I don't think there would be any patent infringement involved, with LD patent expired, and with different laser pick up design from the ELP own design.
1) I'm pretty happy with my current heavy-weight Micro turntable, it's speed accuracy would be hardly surpassed by the LD conversion deck,
2) I'm in no need to digitize my humble collection of just 1000LPs,
3)Not sure how laser will behave on warped records ,
Besides that, it is still an interesting thing to do. I don't think there would be any patent infringement involved, with LD patent expired, and with different laser pick up design from the ELP own design.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.