Safeguarding digital archives, Part 2 - A better way

Part 1 of this article ended with the suggestion that standard NAND flash memory products are marketed under specifications that are based on statistical measures of time limits on the robustness of the data held in the floating gate cells and the number of times the cells can be reliably overwritten with new data. How much risk you are willing to take between dumping data off camera cards and using a back-up system varies from person to person and the eventual use of the information. A professional photographer will need those files for their very livelihood while a few snaps of another day of little Johnny at the park on the swing might be less important, or perhaps more.

I recently had a chance to work directly with NAND flash memory chips and programming them. Although I had used good old EEPROM (electrically erasable programmable read only memory) technology before, this was an eye opener to say the least. For someone whose starting point was consumer flash memory cards, it is quite different to work at "bare metal" reading and writing directly to the memory chip. That's because all flash cards on the market come bundled with their own microcontroller that intermediates your camera or computer file system and the actual memory chips. The main reasons for this are the variation in memory cells across the memory as well as the need to account for changes to cells over time. A quick look at a high density device from any leading manufacturer will soon prove to you that it is software that really makes this hardware tick. In fact, the microcontroller would get pretty busy keeping track of bad memory sites if it wasn't for error correction codes (ECC). Without ECC, you would be getting hundreds rather than thousands of shots on your multi-gigabyte SD card. The success of NAND flash as a high density storage medium depends utterly on ECC to correct otherwise useless memory cells.

Obviously, this is why we hear so much these days about the importance of backing up. Getting back to the serious photographer, most have a discipline for multiple types and locations of back-ups that is second nature - as much a part of the process as composition and focus. But not many would disagree that it's still a hassle. Worse, there can often be some nagging doubt. "Will my online storage provider be in business 20 years from now if I lose my other disks?" So as I've said before, there must be a more convenient and perhaps a safer way to keep these memories alive for our old age and beyond.

Remember when you used to buy film, have it processed, review the prints and put your favourites into photo albums? When it was time for more pictures, you needed to buy more film. I'm not claiming that film is better than digital or even more robust, but somewhere in a closet or shoe box, you may have kept those old original negatives in case the paper prints were lost or damaged. I recently found my own treasure trove of old negatives collected by my mother from her own dearly departed relatives as well as shots she took herself.

So what if we gave up on the idea of re-usable digital memory cards and moved to the true digital analogy of the old film days? Flash card manufactures like SanDisk surely salivated once upon a time for the idea that people would pop into the drug store for a new photo card every time their previous one was filled. But consumers never adopted this approach for obvious reasons.

Could any semiconductor manufacturer produce a product that trades off re-writability for robustness?

SanDisk acquired Matrix Semiconductor and attempted to offer the extremely robust memory to police forces a few years later under the WORM SD brand. The main selling point was that chain of custody for evidence would be less of a target for defence lawyers at trial if police photographs were shot on a digital storage platform that could not be re-written. Of secondary importance seemed to be that the evidence would always be there - barring some sort of catastrophic event creating serious physical damage to the memory cards.

The core of the Matrix 3D memory was a one-time programmable anti-fuse cell. The simplicity and regularity of the Matrix technology allowed the semiconductor wafer manufacturer to push lithography limits beyond specifications and accomplish very small linewidths for the cells and wiring. Furthermore, the cells could be stacked layer upon layer creating the first true 3D memory chip.

(Original image source:  Matrix Semiconductor Inc.)

Passing a current through the anti-fuse at the intersection of individual row and column address lines fused the two sides of the cell bridging the electrical barrier between the two normally inactive sides. More specifically, the cells were activated (programmed) by rupturing a dielectric between two conductors. In this elegant approach, a simple, potentially quite dense storage system was devised. More to the point, it was nearly indestructible - at least from the electrical point of view. Once the two conductors bridged the intervening dielectric barrier, it is difficult to imagine anything short of hammer or power saw that would break the bond that programmed these bits.

So will I ever live to see my dream archival digital film? It seems unlikely. Sandisk is probably sticking with standard NAND flash chips in its Memory Vault product. It would be hard to fault the business strategy or repackaging existing products to address a niche of consumers that might be willing to pay a premium (retail price of the 8 GB Memory Vault is $49.99 - compare that to the thumb drives next to checkouts at your local drug store or supermarket). But even if Sandisk did decide to use the true archival memory in their current product, it misses the whole point of building in the ultimate convenience of a one-time use digital film into the photo archiving process.