Yes, in a lab. A CD, DVD or Blu-ray stores its data as a spiral of microscopic marks, and a 3D laser scanning confocal microscope can photograph and measure those marks, even on a disc that has been snapped in half. Researchers have then turned the pictures back into bits, doing by hand and in software what a disc drive does in real time. It’s genuinely clever, and genuinely limited: none of the published work we found shows a complete, error-corrected file recovered from a shattered disc, and imaging a whole CD takes hundreds of hours with a light microscope and years with an electron microscope. It isn’t a service you can buy. For a disc that’s scratched, fading or was never finalised, a drive and an image-first rescue come first.
You really can photograph data
Everything on an optical disc is physical. A pressed CD carries its data as one long spiral of tiny pits moulded into the plastic and coated with a thin metal layer, usually aluminium. That spiral runs from about 25 mm out to 58 mm from the centre, makes about 20,600 turns and is about 5.4 km long (our arithmetic from the CD standard, ECMA-130). Our lesson on how a CD stores data explains what the pits mean. The point here is simpler: they are shapes, and shapes can be photographed.
A drive reads those marks with a laser as the disc spins past at more than a metre a second. A microscope reads the same marks slowly, as pictures, from a disc that doesn’t have to spin at all. That’s the whole appeal: a disc in pieces can’t go in a drive, but every piece can go under a microscope.
Why a broken disc isn’t empty
When a disc breaks, the damage is a narrow line, not the whole disc. A US National Science Foundation project at Ferris State University and George Mason University broke, cut and bent discs and then examined them with a laser microscope. Cutting and bending did similar, limited damage, and nearly all of the data outside the break was intact. Gogolin, the project’s lead researcher, estimated that a disc snapped in half keeps about 99 per cent of its data, apart from a zone near the cut. The catch he pointed out: the bigger a file, the more likely it is to cross the destroyed zone.
Where the data sits decides how you get at it:
| Disc | Where the data layer is | What that means under a microscope |
|---|---|---|
| CD | On top of 1.2 mm of clear plastic, very close to the label side | Imaging up through 1.2 mm of plastic blurs the picture, so researchers often image from the label side once the lacquer is off |
| DVD | In the middle, between two bonded 0.6 mm halves (a dual-layer disc has two layers about 55 µm apart) | The bonded halves have to be split apart to reach it |
| Blu-ray | Under a 0.1 mm cover layer | Needs optics corrected for the cover, or the cover removed |
This is also why the official advice on destroying discs is so blunt. NIST’s media-sanitisation guideline says CDs, DVDs and Blu-rays can’t be wiped; they have to be destroyed, by grinding off a CD’s layers, incinerating them, or shredding them into particles no bigger than 0.5 mm along an edge and 0.25 mm² in area. The September 2025 revision of the guideline still uses an optical-disc shredder as an example, and notes that pieces 50% too large mean the job wasn’t done. A disc snapped in half hasn’t been destroyed. It’s been turned into two pieces of readable plastic.
What a 3D laser scanning confocal microscope is
An ordinary microscope takes a picture. A laser scanning confocal microscope builds one, point by point: it scans a tightly focused laser spot across the surface, and a pinhole in front of the detector lets through only the light coming back from the plane that’s in focus. Step the focus up and down, and the height at which each point reflects most brightly is the height of the surface there, so what comes out is a 3D map of the surface rather than a flat photo. A pit on a disc shows up as a dip in that map.
.jpg){kind=link}
The instruments used on discs are industrial “laser microscopes” made for measuring surfaces. A current example is the Evident (Olympus) LEXT OLS5100: a 405 nm violet laser, objectives with a numerical aperture (NA) of 0.95 at 50× and 100×, magnification from 54× to 17,280×, up to 4,096 × 4,096 points per measurement, and a motorised 100 × 100 mm stage that stitches many fields into one big image. It displays height in steps of 0.5 nm. Evident quotes a lateral resolving power of 0.12 µm for the earlier OLS5000 and OLS4500 (it doesn’t quote one for the OLS5100). Keyence’s VK-X3100 head (since succeeded by the VK-X4100) is similar: a 404 nm laser and 42× to 28,800×. The published disc project used the older Olympus LEXT OLS4000 and OLS4100.
None of this is cheap. In 2015 the University of Sheffield tendered for a “3D laser scanning microscope with sub-micron resolution” with a contract value of £90,000 to £110,000. We couldn’t find an Australian price, but a six-figure lab instrument is the right picture.
Can it see the pits?
Yes, comfortably for CDs and DVDs. The question is how the smallest marks compare with the finest detail a microscope can separate, which for light is roughly the wavelength divided by twice the NA.
| Disc | Track spacing | Shortest mark | Its drive reads with |
|---|---|---|---|
| CD | 1.6 µm | 0.83–0.97 µm long, about 0.5 µm wide | 780 nm laser, NA 0.45 |
| DVD | 0.74 µm | 0.40 µm (0.44 µm dual layer) | 650 nm laser, NA 0.6 |
| Blu-ray (25 GB) | 0.32 µm | 0.149 µm | 405 nm laser, NA 0.85 |
| Instrument | Finest detail | CD | DVD | Blu-ray | The catch |
|---|---|---|---|---|---|
| Laser confocal (405 nm, NA 0.95) | ≈ 0.21 µm | Yes | Yes | Tracks yes; shortest marks below the limit | Buried layers; Blu-ray needs cover-layer correction |
| Ordinary microscope, white light (NA 0.4) | ≈ 0.69 µm | Yes | Marginal (our estimate) | Only special research optics | Blurry through 1.2 mm of plastic |
| Atomic force microscope (AFM) | Better than 0.00015 µm | Yes | Yes | Yes | Layer must be exposed; a tiny field (30 × 30 µm on one instrument) |
| Scanning electron microscope (SEM) | Fine enough for CD and DVD marks | Yes | Yes | Not tested in the sources we found | Layer must be exposed; very slow |
A correction to a common claim: Blu-ray is not beyond light. A Blu-ray drive itself reads with a 405 nm laser through an NA 0.85 lens, and a 405 nm confocal microscope with an NA 0.95 objective resolves finer detail than the drive does. Blu-ray’s shortest marks are below both limits, but the drive doesn’t see them as separate shapes either: it recovers them from the shape of its signal, and decoding from images would need the same kind of processing. The University of Arizona built solid-immersion-lens microscopes (one with an NA of 1.4) to image Blu-ray marks through the 0.1 mm cover layer, and AFM resolves them easily once the layers are exposed; undergraduate chemistry students do it in a teaching lab. With DVD and Blu-ray, the hard part is reaching the buried layer without distorting the picture. Resolution isn’t the problem.
From picture to bytes
The picture is the easy half. A disc drive’s chips turn the laser signal into data in real time, and they never show you the steps. Working from images, every one of those steps has to be rebuilt in software, and in the published projects some were done by hand.
- Reach the data layer. For a CD, from the label side once the lacquer is removed, or through 1.2 mm of plastic and accept a blurrier image. For a DVD, split the bonded halves. For Blu-ray, get through or remove the 0.1 mm cover.
- Image overlapping tiles and stitch them. At high magnification each picture covers a tiny patch. The hobbyist Peter Monta took 125 overlapping frames, 60 µm apart, and stitched them with the free panorama program Hugin. The Brno thesis used FEI’s Maps software.
- Follow the spiral. Monta sampled along a fitted circle and noted that bigger areas need adaptive tracking: the picture-based version of the tracking servo that keeps a player’s laser on the track.
- Measure each pit and land, and recover the clock. Decide what is pit and what is land, then work out the length of each run in channel bits (written T). On a CD every run is between 3T and 11T. Monta corrected the drift in his clock with a hand-fitted curve.
- Turn edges into 1s. Every change between pit and land is a 1; every channel bit without a change is a 0.
- Find the frame sync. A CD frame is 588 channel bits, starting with a 24-bit sync pattern: an 11T run followed by another 11T run, which stands out clearly once you plot the data.
- Demodulate. A CD stores each byte as a 14-bit pattern (EFM) chosen so there are always 2 to 10 zeros between ones; a lookup table turns them back into bytes. A DVD uses 16-bit patterns (EFMPlus), with data runs of 3T to 11T and a 14T run only in its sync codes. Blu-ray uses yet another code, 1-7PP, with runs of 2T to 8T. A pattern that isn’t in the table becomes an erasure: a byte marked as known-bad, which error correction handles better than an unknown one.
- Correct the errors. A CD uses two Reed–Solomon codes, C1 and C2, interleaved so a scratch’s damage is spread thin (our lesson on how computers check data explains the idea). C1 works one frame at a time. C2 needs bytes spread over up to 108 frames, about 17.6 to 20.6 mm of continuous track. A DVD groups 16 sectors (32 KiB of data) into one error-correction block of 37,856 bytes, about 82.5 mm of track. Its row code works on short stretches; its column code needs the whole block.
- Assemble sectors, then files. A CD-ROM sector is 98 frames carrying 2,352 bytes; in the usual Mode 1, 2,048 of them are your data, protected by three layers of checking and correction (EDC, ECC and CIRC). Once sectors decode, the disc’s file system (ISO 9660 or UDF) or file-carving tools take over.
How big does a piece need to be?
A fragment of disc doesn’t hold a neat chunk of the data. Because the track is a spiral, a fragment holds thousands of short, separate arcs, one for each turn of the spiral that crosses it: about 625 arcs for every millimetre of the piece, measured outwards from the centre. Each arc’s neighbours are a whole revolution away in the data.
That’s where the interleaving bites. To decode one CD sector all the way through C2, you need roughly 98 + 108 frames on one unbroken arc: about 33 mm of track. That’s about 76° of the disc at the inner edge of the data and about 33° at the rim (our arithmetic from ECMA-130). Smaller pieces still give frame-level bytes through C1, as Monta showed, but not error-corrected sectors. And data within about 17 to 20 mm of a break, measured along the track, loses its C2 partners. A DVD’s full error-correction block needs about 82.5 mm of arc, close to half a turn at the inner edge. What’s lost at a break is the crushed zone itself plus every codeword that straddles it, and since a file needs the right arcs from every piece, a missing piece means missing sectors.
Who has actually done it
Not many people, and mostly in universities. These are the projects we could confirm from their own papers, records and write-ups:
| Who, when | Method | How far they got |
|---|---|---|
| University of Arizona, 2004 (Kasanavesi, Milster, Felix, Choi) | High-power microscope frames turned into a drive-like signal, then signal processing | Recovered the pre-groove timing information (ATIP) from a recordable CD fragment, and set a benchmark for the minimum track length needed |
| University of Arizona, 2005 (Milster and colleagues) | A spin stand and signal processing, not a microscope | Recovery shown on broken fragments and on microwaved, knurled and sanded CDs |
| Texas State University with Lucere Data, 2007 | A conference paper on recovering data from damaged CDs and DVDs | We could only see the citation, not the paper, so we don’t report its results |
| Ferris State University with George Mason University, 2011–2014 (NSF grant, US$356,318) | Olympus LEXT OLS4000/OLS4100 laser confocal microscopes: measure each mark, reverse the encoding | Read pits and lands and recovered encoded data as a proof of concept: some characters, decoded by hand, without full files. Automation was described as under way |
| Brno University of Technology, 2016 (bachelor’s thesis, Jarolím) | Scanning electron microscope on data layers peeled apart with steam | A correct raw bitstream from fragments that met the image-quality bar; full decoding left as future work. Preparing samples sometimes destroyed almost all the data |
| Peter Monta (hobbyist), 2022–23 | A Nikon 20× NA 0.4 objective, a Logitech C270 webcam and a 3D printer as the stage | About 40 frames (20,000 channel bits): frame sync found and the inner C1 code decoded |
| Tech Tangent, reported 2026 | A budget digital microscope | On an analogue LaserDisc, film credits were legible straight from the pit pattern. On a CD: nothing recognisable |
Monta’s is the one to read if you like this sort of thing. A webcam, a microscope objective and a 3D printer got him from a stitched picture to Reed–Solomon-checked bytes, and he stopped only because the CD’s outer error-correction code is spread about 100 frames away from the inner one, further than his image reached. His stitched images and plots are on his blog: part 1 and part 2. The Ferris State team’s laser-microscope images of damaged discs are in Evident’s application note. The Tech Tangent result is a useful reminder: on a digital disc, looking isn’t reading. The pictures only mean something once they’re decoded.
We looked for papers in the main digital-forensics journals and conferences on decoding optical discs from images and didn’t find any. We also found no paper since 2020 using machine learning to decode disc marks from images, and no automated “disc reader” product built on a confocal microscope.
The analogue cousin: IRENE
The success story is on records, not CDs. In 2000, Carl Haber, a particle physicist at Lawrence Berkeley National Laboratory, heard a radio report about fragile recordings at the Library of Congress and adapted particle-detector imaging to photograph sound grooves. The system, IRENE, was developed at Berkeley Lab from 2000 to 2003 with support from the Library of Congress, and named after a scan of the Weavers’ 1950 “Goodnight Irene”. It images grooves in 2D with a camera, or in 3D with a confocal probe, then works out the sound from the shape. In 2007 a 3D scan took about 20 hours per minute of sound.
It has saved recordings nobody could play. A 1944 glass-based lacquer radio disc in Columbia University’s collection had shattered, with small pieces missing. NEDCC, the conservation centre that runs IRENE as a service, imaged each surviving fragment separately and reassembled the recording in software; you can still hear clicks where lacquer was lost. Columbia’s write-up shows the pieces. More recently Berkeley Lab scanned a crack-riddled 1880s disc from Alexander Graham Bell’s Volta Laboratory, and used the speech-recognition tool Whisper and ChatGPT to identify the noisy recitation as Charles Wolfe’s 1817 poem The Burial of Sir John Moore.
The logic is exactly the one a broken CD needs: image every piece, put them back together in the computer. Records are easier for two reasons. A groove is 10 to 100 times larger than a CD pit, and it’s analogue, so a gap is just a click. A digital disc needs exact runs of bits to get through its error correction, which is why the CD projects above stop where they do.
Burnt discs, fading dye and peeling layers
- Burnt discs can be imaged too. On a CD-R or DVD-R the marks are written into a dye layer rather than pressed into plastic. NIST explains that the written spots absorb the reading laser, which gives a contrast much like pits, and the Ferris State work covered home-burnt discs as well as pressed ones. Writing also leaves real bumps: atomic force microscope studies of CD-Rs measured bumps 41 nm high in the plastic at 7 mW of writing power, appearing from 3.5 mW. So a burnt CD-R carries physical relief, not just a colour change. The Arizona fragment paper recovered timing information from a recordable disc’s pre-groove.
- Rewritable discs are an open question. RW discs store data in a phase-change metal alloy, which NIST notes degrades faster than dye. We found no published work imaging RW marks to recover data.
- Faded dye is gone. The organic dye in recordable discs breaks down over time, faster with heat and ultraviolet light. Where it has failed, there’s no contrast left to photograph.
- Holes in the metal are gone. On a CD, a pinhole or scratch through the metal on the label side destroys the reflection there, and nothing repairs it. Under a microscope that patch is simply missing. Scratches on the data layer itself destroy marks too: small damage takes out a huge number of pits and lands.
- Peeling layers cut both ways. The NSF team found the layers of broken discs flaked off the plastic quickly. That loses data, but it also made it easier to image the data layer from the top than through the plastic.
Where forensics fits
Who does it. Publicly, university research groups: Arizona, Texas State with Lucere Data, Ferris State with George Mason, and Brno. For analogue recordings, NEDCC’s IRENE service. We couldn’t confirm any government lab or commercial data-recovery company that publicly describes microscope-based recovery of digital optical discs.
When it’s worth it. When the data matters more than the time and cost: evidence or intelligence, where the Ferris State team also framed it as a security question; heritage audio, where IRENE works; and research.
No rulebook yet. The Scientific Working Group on Digital Evidence (SWGDE) publishes best practice for handling damaged storage devices; its 2025 edition covers hard drives, SSDs, NVMe and flash memory, and says it doesn’t cover optical media. Handling and chain of custody for a broken disc would follow general digital-evidence practice. We found no procedure written specifically for optical discs.
The mirror image is destruction. If labs can read a snapped disc, a snapped disc isn’t safe. Hence NIST’s 0.25 mm² particle size, and hence the University of Arizona building a high-power laser system specifically to make CD-R and CD-RW marks unrecoverable.
How long, and how much
| Method | Time | Source |
|---|---|---|
| Microscope images, decoded the conventional way | About 500 hours for a CD-sized area | Background section of a University of Arizona patent |
| Scanning electron microscope | 240 minutes per mm²: about 1,500 days for a whole CD | Brno thesis |
| Laser confocal at 50×, whole CD | About 131,000 tiles, 10 to 57 days of scanning and about 4.4 TB of height data | Our own rough estimate, not a published figure |
| Webcam and microscope objective | 125 images for about 40 frames, roughly 7 mm of track | Peter Monta |
| IRENE in 3D (records, 2007) | About 20 hours per minute of sound | Vision Systems Design |
Our estimate assumes a 50× field about 256 µm across and 5 to 30 seconds per tile with 25% overlap. It’s in the same range as the 500-hour figure, which is the point: imaging alone is weeks of machine time per disc, before anyone decodes anything.
On money: the microscope is a six-figure instrument (that £90,000–£110,000 tender), the Ferris State project was a three-year US$356,318 research grant, and we found no published price for recovering one disc this way.
What it means for you
- Scratched, fading or never finalised? This isn’t the tool. A different drive, an image-first rescue with GNU ddrescue and, for the shiny side, careful resurfacing get far more back, far faster. Our guide to saving a damaged or unfinalised DVD walks through it, and our client DVD rescue shows what that looks like in practice.
- Cracked but in one piece? A drive can often still read most of it. Our DVD guide covers reading a cracked disc without it breaking up in the drive.
- In pieces? Keep every piece, and don’t glue them: labs image each fragment separately and reassemble the data in software, so gluing gains nothing. Most of the data is physically still there. But we don’t know of anyone offering microscope reading of CDs or DVDs as a service, so treat any promise of it with caution.
- Getting rid of discs? Snapping isn’t enough. For sensitive data, NIST’s answer is to shred discs to particles of 0.25 mm² or smaller, or incinerate them.
Got a disc that won’t read?
Our media transfer service in Sydney includes a careful read attempt on scratched, ageing and unfinalised discs: image first, different drives for the difficult ones, and an honest report of anything that couldn’t be saved. If the disc is cracked or in pieces, tell us when you get in touch, and we’ll say plainly what is and isn’t realistic.
Common questions
Can data be recovered from a CD that’s snapped in half?
What is a 3D laser scanning confocal microscope?
Can a microscope read a Blu-ray disc?
Is there a company that will do this for me?
How long does it take?
Does it work on burnt CD-Rs and DVD-Rs?
Should I glue a broken disc back together?
How do I destroy a disc so this can’t be done?
Sources
All checked 13/10/2026. Figures marked as our arithmetic or estimate are ours, worked from the cited standards and specifications.
Standards and government guidance
- ECMA-130, Data interchange on read-only 120 mm optical data disks (CD-ROM), 2nd edition (June 1996), Ecma International. Track pitch, scanning speed and channel bit rate, EFM, 588-bit frames, CIRC and its 108-frame delay, sectors, data radii. PDF, checked 13/10/2026
- ECMA-267, 120 mm DVD – Read-only disk, 3rd edition (April 2001), Ecma International. Track pitch, channel bit length, EFMPlus, 0.6 mm substrates, dual-layer spacer, ECC block. PDF, checked 13/10/2026
- F. R. Byers, Care and Handling of CDs and DVDs: A Guide for Librarians and Archivists, NIST SP 500-252 (October 2003). Where the layers sit, dye and phase-change discs, label-side damage, dye fade. PDF, checked 13/10/2026
- NIST SP 800-88 Rev. 1, Guidelines for Media Sanitization (December 2014, withdrawn 26/09/2025), Table A-7, optical media. PDF, checked 13/10/2026
- R. Chandramouli, E. A. Hibbard, NIST SP 800-88 Rev. 2, Guidelines for Media Sanitization (September 2025). PDF, checked 13/10/2026
- SWGDE, Best Practices for Handling Damaged Digital Storage Devices, 14-F-002-2.0 (04/03/2025). PDF, checked 13/10/2026
- Blu-ray Disc founders’ format announcement, 19/02/2002 (Hitachi news release). 405 nm, NA 0.85, 0.1 mm cover layer, 0.32 µm track pitch, minimum mark lengths. link, checked 13/10/2026
Research papers, theses and grants
- S. K. Kasanavesi, T. D. Milster, D. Felix, T. Choi, “Data recovery from a compact disc fragment”, Proc. SPIE 5380, 116–127 (2004). University of Arizona record, checked 13/10/2026
- T. D. Milster et al., “Dynamic Data Recovery from Damaged CD Media”, ISOM/ODS 2005, paper WC5, Optica. link, checked 13/10/2026
- D. E. Tamir, W. Davis, L. Wolfe, R. McNiece, “Recovery of data from damaged CD/DVD”, Proc. SPIE 6696, 66962I (2007). Citation only. DOI, checked 13/10/2026
- T. Choi et al., “Optical data destruction system: device design and test results”, Proc. SPIE 6879 (2008). University of Arizona record, checked 13/10/2026
- J. Zhang et al., “Microscope system for Blu-ray disc samples”, Applied Optics 49(36), 6878–6887 (2010). University of Arizona record, checked 13/10/2026
- V. Densmore et al., “A high-NA solid immersion objective for imaging a Blue-ray disc and investigating subsurface damage”, Proc. SPIE 9201 (2014). University of Arizona record, checked 13/10/2026
- NSF award CNS-1116268, “Digital Laser Microscopy Analysis of Active, Altered, and Damaged Storage Media for Security and Digital Forensic Recovery” (Ferris State University, 2011–2014), with project outcomes report. NSF award page, checked 13/10/2026
- J. Jarolím (supervisor M. Drahanský), Reconstruction of Data from Damaged CD/DVD, bachelor’s thesis, Brno University of Technology, 2016. link, checked 13/10/2026
- S. F. Bancroft, M. A. Ali, P. Kohli, “An Inquiry-Based Introduction to Atomic Force Microscopy Techniques through Optical Storage Disc Surface Imaging”, J. Chem. Educ. 99, 3030–3038 (2022). link, checked 13/10/2026
- J. Huh, J. Kim, J. Kim, “Deformation Effects and Recording Characteristics of Compact Disc-Recordables”, Jpn. J. Appl. Phys. 36, 403 (1997), and J. Huh, J. Kim, “Thermal Decomposition and Deformation of Dye and Polycarbonate in Compact Disc-Recordables”, Jpn. J. Appl. Phys. 36, 7233 (1997). DOI · DOI, checked 13/10/2026
- W. G. Lutz, G. F. Meyers, E. Kolinski, Applications of Atomic Force Microscopy in Optical Disc Technology, Digital Instruments/Veeco application note AN18. CD and DVD pit sizes, drive optics. PDF, checked 13/10/2026
Patents
- T. D. Milster, S. Kannan, US 7,974,170 B2 (Arizona Board of Regents, 2011). Background: about 500 hours for a CD-sized area by the microscope-image method. link, checked 13/10/2026
- Sharp and Memory-Tech, US 9,697,863 B2 (2017). Blu-ray 1-7PP modulation, 2T to 8T marks. link, checked 13/10/2026
Project write-ups and news
- R&D World, “Restoring Lost Data: 3-D Digital Laser Microscopy Creates Visual Roadmap” (06/04/2015). The 99% estimate and the larger-files caveat. link, checked 13/10/2026
- G. Gogolin, J. Jones, D. Brower, “Maximizing data recovery”, application note, Evident Scientific. link, checked 13/10/2026
- “Imaging Damaged Optical Media Using Digital Laser Microscopy”, AZoM (01/08/2014). link, checked 13/10/2026
- “Olympus microscope recovers data from damaged optical media”, Laser Focus World (04/03/2015). link, checked 13/10/2026
- P. Monta, “Compact Disc microscopy”, parts 1 and 2, Signal and noise (22/12/2022 and 14/01/2023). part 1 · part 2, checked 13/10/2026
- J. List, “Reading Data From A CD, With A Microscope”, Hackaday (22/01/2023). link, checked 13/10/2026
- carrotIndustries, “Reading the Red Book: decoding Compact Disc Digital Audio” (GitHub, 2023). NRZI and CD decoding steps. link, checked 13/10/2026
- R. Speed, “Retro tech fan views LaserDisc movie data with a budget microscope”, The Register (10/03/2026). link, checked 13/10/2026
Instruments and optics
- Evident, LEXT OLS5100 laser scanning microscope, specifications. link, checked 13/10/2026
- Evident, LEXT OLS4500 product page, and “LEXT principles: features”. The 0.12 µm lateral figure for the OLS4500/OLS5000. OLS4500 · features, checked 13/10/2026
- Keyence, VK-X3100 measurement head, specifications. link, checked 13/10/2026
- Nikon MicroscopyU, “Resolution”. link, checked 13/10/2026
- University of Sheffield, “High Definition Laser Scanning Microscope” tender, UK Contracts Finder (21/10/2015). link, checked 13/10/2026
IRENE
- V. Fadeyev, C. Haber et al., “Reconstruction of Recorded Sound from an Edison Cylinder using Three-Dimensional Non-Contact Optical Surface Metrology”, LBNL-54927 (2004). PDF, checked 13/10/2026
- A. Cohn, “Play It Again, IRENE”, Berkeley Engineering (02/01/2008). link, checked 13/10/2026
- W. C. Holton, “Rescuing recorded sound from silence”, Vision Systems Design (01/10/2007). link, checked 13/10/2026
- Berkeley Lab, “Old sounds revealed” (14/12/2011) and “Voices from the past get an AI assist”. link · link, checked 13/10/2026
- AUTM Better World Project, “IRENE restores sound from old or damaged recordings”. link, checked 13/10/2026
- A. P. Han, “A Physicist Conjures Sounds From the Past”, Science Friday (24/01/2014). link, checked 13/10/2026
- J. Gertz, “Hearing Voices from a Broken Disc”, The Long View, Columbia University Libraries (09/01/2017). link, checked 13/10/2026
- R. Rodgers (NEDCC), “Putting the Band Back Together!”, Birthplace of Country Music (06/01/2020). link, checked 13/10/2026
- W. Atkins, “On the Road with the Frank C. Brown Collection”, Duke University Libraries (19/07/2016). link, checked 13/10/2026
Photographs (all via Wikimedia Commons, used under the licence shown)
- ZEISS Microscopy, “ZEISS LSM 800 for Materials”, Wikimedia Commons, CC BY 2.0 (resized). file page, checked 13/10/2026
- Stephan Sprinz, “AFM image of a DVD”, Wikimedia Commons, CC BY 4.0. file page, checked 13/10/2026
- Adaptron (English Wikipedia), “Piece of exploded CD”, Wikimedia Commons, Public domain (resized, converted to JPEG). file page, checked 13/10/2026