Taking Pictures with a DRAM Chip

I am the original author of this writeup.

Did this hack together with my brother twenty years ago (look at the file dates)!

Strange to see my 1994 hack at #1 on hacker news twenty years later..

It was done with 64kBit DRAMs in a ceramic package. Descrambling the physical chip layout was a pain..

Wow that brings back the memories. DRAM cameras, or "Rameras" were popular with robotics experimenters in the 80's and 90's. From "Android Design" by Martin Bradley Weinstein:

"The ramera was first developed by the Robotics group at Case Western Reserve University in Cleveland Ohio, in 1978."

Apparently they wrote up their work in the Sensor Group Journal. They had used a 4008 which was only 64 x 64 pixels.

Not surprisingly this work was the basis for early optical mice which use what is a DRAM circuit as an imager to detect motion of the mouse in two directions without using a rolling ball. This led to a number of roboticists hacking their optical mice into simple cameras (which was much easier than popping the lid on an increasingly hard to find DRAM chip). That practice was then replaced by using CMOS imagers that had been pretty cheap by the early 2000's and of course these days you can get a camera module for a phone very inexpensively (see the PiCAM for $29 for example)

Almost all DRAM has some amount of address scrambling--logically adjacent cells are not physically adjacent. It can get quite complicated. Here's a PDF that describes some of the why and how: http://ce-publications.et.tudelft.nl/publications/1162_addre...

I have been told that a quick way to reverse engineer the scrambling is to shine a circular pattern onto the physical dram, then read out the data and test common scramblings until the logical data shows the same circle.

This was an amusing read. I'd not heard of using DRAMs as image sensors, but I had discovered that EPROMs sort-of worked as image sensors. Back in the 80s I was writing video poker machine games. 6502 running at 700KHz clock rate, very simple boards using interleaved video/CPU memory access, and all graphics done via 8x8pixel sprites, that were programmed in EPROMs. One day I noticed that bright sunlight on a sprite EPROM (no cover label) produced on-screen garbage. With a bit of experimenting, I found I could get a fairly decent image into the screen. Wrote a program to generate an index mapping into the game screen RAM, so the sprite EPROM was mapping to the screen 'about right', then found that by programming all 1 (or 0, I forget now) into the EPROM then partially erasing it, I could get it quite sensitive to light. I only tried hand holding assorted crappy lenses in front of the EPROM window, and there were all sorts of optical and electrical and geometrical artifacts. But it kind of worked. Pity I have no photographs of those experiments. Didn't consider it anything more than a silly amusement. Especially since I already had a nice B&W PAL video camera.

One of my teachers did this in the 1980s with eproms. You could get a reasonable 1bpp silhouette out of a 2716 and a couple of lenses out of a broken pair of binoculars. You had to program it first then use the sun to erase the picture onto it which meant exposure times in the order of 20 mins on a sunny day to flip enough bits plus some figuring out of how the cells were organised. He had one wired to a BBC micro user port with a couple of shift registers and a BASIC program that copied the current state into video memory. I think this was around 1986 and was probably the coolest thing I'd ever seen at the time.

Edit: just remembered - he worked out that if you used a high power camera flash and fired it 20-30 times it had the same effect.

From what it looks like, if anyone wants to build one they'll need:

- An IC (the one he used seems to be http://www.ebay.com/itm/5-Rare-Vintage-GOLD-NEC-044-D4164D-4...)

- A small lens to place over the IC (http://www.ebay.com/itm/MTV-6MM-CCTV-IR-Lens-For-Security-IP...)

- A parallel to USB cord if you don't have a computer with a parallel port (http://www.ebay.com/itm/USB-to-PRINTER-DB25-25-Pin-Parallel-...)

Please correct me if I'm wrong about any of these parts - I haven't tried to build one myself.

Here's my rudimentary understanding of the pin connection: (to connect it yourself, use https://upload.wikimedia.org/wikipedia/commons/b/b3/Pin_numb..., https://upload.wikimedia.org/wikipedia/commons/e/e0/Parallel..., and https://upload.wikimedia.org/wikipedia/commons/e/e1/25_Pin_D... as guides)

IC-Parallel

2-1

3-14

4-16

5-8

6-4

7-3

8-18

9-9

10-7

11-6

12-5

13-2

14-12

15-17

16-25

Watch out - the source and binary file (http://www.kurzschluss.com/kuckuck/kuckuck.zip) is a tarbomb.

Of course these days there are only few 4164 with the metal cap remaining, and they can be as expensive as a cheap webcam: http://www.ebay.com/bhp/4164-ram

Found the main claim page: http://dramclaims.com

"You are included in the Settlements if:

You are a person or business that purchased DRAM or a device containing DRAM;Your purchase was made anytime from January 1, 1998 through December 31, 2002;Your purchase was made in the United States including Puerto Rico, the Northern Mariana Islands, and the territories or from a seller located in the United States, including Puerto Rico, the Northern Mariana Islands, and the territories and;Your DRAM purchase was not made directly from any of the DRAM manufacturers."

Guess this is the reason some of you mentioned why the price was so high? well I'll be damned, submit your claims!

Small, cheap, and high-volume eats the world, and it's hard to beat DRAM. For any silicon hacker, this is disruption-in-waiting: make DRAM compute. Doesn't matter how, just do it well. Whether it be naive CMOS sensor substitutes like this, Micron's Automata processor - http://www.micron.com/about/innovations/automata-processing - (their IEEE paper is surprisingly good), or whatever you find fit, once you hit volume production you've probably won. It's hard for standard logic processes to hit cost-parity with this beast.

I'm also impressed by google translator!

The technique described takes a low quality sensor (a DRAM chip with an open window onto it) and uses it to produce reasonable quality grayscale images.

Interesting to look at the sourcecode of the Kuckoo program (before the widespread adoption of UTF-8!) and how it uses High Dynamic Range tricks to read off the image at different exposures.

With much faster RAM and smaller cells nowadays maybe one could do something cool with very fast photography, but of course the faster/smaller the RAM the more trouble interfacing with it and calibrating the light intake.

so I just googled 'DRAM chip' and this class action lawsuit popped up if anyone is interested: http://blog.seattlepi.com/boomerconsumer/2014/03/04/dram-chi...

" DRAM is a widely used form of computer memory that is found in many products including personal computers, servers, and MP3 players. Regardless of the brand name, most of these types of products produced between 1998 and 2002 contained DRAM chips.

“This price-fixing wrongly forced consumers and state agencies to pay more for computers and other high-tech products containing DRAM chips,” Attorney General Bob Ferguson said Tuesday. “This settlement will provide some monetary relief for affected consumers.”

The total multi-state settlement is $310 million. The amount of money returned to Washington state consumers depends on how many claims are filed and the types of products they purchased, Ferguson said."

price fixing?! interesting.

I was going to comment that Steve Ciarcia did this in the 80s, and then I see he is credited on the page.

Byte September / October 1983 Steve Ciarcia: "Build the Micro D-Cam Solid-State Video Camera"

ftp://164.41.45.4/pub/electronics/circuit_cellar/BYTE_Projects/D-Cam2.pdf

Looks like its US Patent No. 4441125

https://www.google.com/patents/US4441125