> we built a [bigger] system called META, Mega Channel Extraterrestrial Assay. [...] this was amazing for [...] about 1985. [...] There's our META processor. It had 128 68000s with a pile of memory on each one [...] Here's a picture of the control room. It's got two racks here full of stuff, and you can't read that, which is good, because it would be embarrassing if you knew that it said META Supercomputer: 75 million instructions per second. You know, the Cray-1 wasn't much faster than that. This was a supercomputer. This had 7,000 64k DRAM chips. I soldered them all in. I know. And they cost $3 apiece. We spent $21,000 to get 60 megabytes of memory. [...] Here's a rack full of these that had nine of these racks. Each one of those has the memory on it. You can see that. And then it produced this stuff. This is, again, on the WICAT screen, and it gave you the whole spectrum.
> We need more channels. So we built a thing called BETA [...] It must have been Unix. Linux? I don't know. Something. So here's our rackets. We're down to one rack now, and we actually have lots of bytes. And here's a whole bunch of 90-megahertz Pentium PCs. That was state of the art back then, and we cobbled these things all together. And here's what one rack looked like. So it's now the billion-channel extraterrestrial assay. It no longer brags about being a supercomputer. We learned our lesson from that one. So that's what we did, and this thing was able to search the full water hole. [...] We were sort of proud of how fast this thing produces data. Two seconds of data would fill up a CD, which was state of the art optical storage in those days.
> Let's build an all sky optical SETI. [...] he came up with a really clever idea [...] if you take a fancy FPGA, one of these Xilinx parts, they have a whole bunch of inputs which are meant for digital signals, they're LVDS differential pairs. But they're very good comparators, so you can trick it into being a flash analog to digital converter, even though it doesn't know that's what you're doing. So you tie eight of these with a common signal, and put in a progression of biases, and you've got yourself an ADC. And you can run these things at 1 and 1/2 giga samples per second. And you can put 32 of these into one chip
the mastodon instance at cybre.space is retiring
see the end-of-life plan for details: https://cybre.space/~chr/cybre-space-eol