On 2022-Mar-31, at 11:56 PM, Mark Huffstutter via cctalk wrote: ------------------------ Great document! Yup, internal reflection.

Wasn't some of this glass delay line memory used in early raster-scanned computer video displays? --Chuck

On 4/1/22 10:27, Paul Koning wrote: I know that the CRT display controller on the CDC 200 series terminal (INTERCOM, Export/Import 200 software) used a 10 msec magnetostrictive delay line.for image storage. Glass would seem to be a more mechanically robust storage medium. See:page 1-5 http://bitsavers.org/pdf/cdc/terminal/82128000_200_User_Terminal_Hardware_R… Later raster terminals used MOS shift register memory. The STAR-100 stations used a track on the station microdrum for video refresh. --Chuck

Hey, I've got one of those somewhere (the delay line, not the terminal ;-) )! I do still use the cabinet as a desk, as well as a few parts here and there; to think that today something like an Arduino nano can replace that desk-sized cabinet containing a substantial power supply and a card cage with at least a dozen cards IIRC... I'm still amazed by how far we've come in less than my lifetime... m On Fri, Apr 1, 2022 at 2:30 PM Chuck Guzis via cctalk <cctalk at classiccmp.org> wrote:

On Sat, 2 Apr 2022 at 00:34, Bill Gunshannon via cctalk <cctalk at classiccmp.org> wrote: The comparisons of size, power, storage, cost, power usage, heat output and so on are often made. What is less often observed are the facts that a machine that takes multiple trailers can be repaired with spare parts. Anything made from ICs basically can't, except by replacing the ICs. What if you can't make ICs any more? Or rather, what level of IC fabrication would it be possible to construct from scratch? And if the war were long (sticking to the military context) or the conditions extreme (say, radical rapid global warming and a retreat to polar regions) and so all the factories and infrastructure were lost or had to be abandoned... I'm thinking of the current global chip shortages, and the floods in Thailand that screwed up supplies of hard disks a decade ago. What could be constructed from scratch, given copious amounts of scrap and waste for source materials? And aside from the hardware: Yes, modern computers have vast amounts of storage and power, but we use them all on OSes and apps that need millions of lines of code in dozens of languages, and gigabytes of storage. As a result, although they are vastly quicker, latency is arguably _worse_ now: https://danluu.com/input-lag/ What if... we had to reconstruct entire OSes and software stacks from scratch? Maybe because the authors were dead. Maybe because the computers they needed didn't work any more and couldn't be repaired. Maybe because we had to make new computers and they were smaller and slower. How much could be rebuilt? What could be learned from the mistakes of old? I recently wrote these pieces, which were fun to research: https://www.theregister.com/2022/03/29/non_c_operating_systems/ https://www.theregister.com/2022/03/31/serenityos/ There are so many UNIX-like kernels in C on Github I couldn't count them all. In the many dozens, maybe hundreds. This is doable. But there are fewer in C++. Fewer still in Rust or Go. Very few in anything Wirthian or with garbage-collection. Well indeed! http://collapseos.org/ is relevant to this. But is an 8-bit the biggest we could realistically construct if all the IC fabs were destroyed? -- Liam Proven ~ Profile: https://about.me/liamproven Email: lproven at cix.co.uk ~ gMail/gTalk/FB: lproven at gmail.com Twitter/LinkedIn: lproven ~ Skype: liamproven UK: (+44) 7939-087884 ~ Czech [+ WhatsApp/Telegram/Signal]: (+420) 702-829-053

But that true for earlier machines, too. Replacing broken transistors requires having replacement transistors suitable for the circuit in question. For a 1960s era machine that may be quite hard; while transistors are easy to find, transistors with suitable characteristics might not be. And what are they? I'd love to know what the specs of the transistors in CDC's 6000 series "cordwood" modules are. Other than the stage delay (5 ns) I have no idea. That's a fun question and a full answer would probably make a good book. For transistors the answer is only marginally simpler. For tubes, quite a lot simpler. (There's a nice Youtube video of someone making tubes, in his basement. You need glass blowing equipment, a spot welder, vacuum pumps, and an inductive heating system. That's about it for tools, as I recall. Materials: pyrex glass, kovar feed through wires, tungsten filaments, not sure what the electrodes are made of.) For semiconductors, you'd start with machinery to make ultra-pure materials (silicon, I'd assume). A Czochralski crystal growing machine to make the cylinders of pure mono-crystal silicon from which wafers are sliced. Polishing machinery. Wafer coating machines. Wafer steppers. Etching, metal coating, diffusion, etc......... most of which also require very pure and often exotic ingredients. (I remember being amazed to read that chlorine trifluoride is used as a cleaner in the semiconductor industry. Look up the properties of that compound, it will blow your mind.) Reading the specs of the latest generation wafer steppers from ASML (the only company in the world with the technology) boggles the mind, especially if you have some understanding of precision machinery design. And even earlier generation steppers are not easy devices to make. I'm not sure what's involved in doing one precise enough for, say, an 1980s era microprocessor. Or even an SN7400. Transistors are basically the same as small ICs, unless you go to really ancient types (point contact, mesa, alloy diffusion). If I had to build a simple computer starting from a pile of rubble I'd seriously consider building it from tubes. paul

From: "steven--- via cctech" <cctech at classiccmp.org> Paul and others said What if you can't make ICs any more? Or rather, what level of IC fabrication would it be possible to construct from scratch? For semiconductors, you'd start with machinery to make ultra-pure materials (silicon, I'd assume). A Czochralski crystal growing machine to make the cylinders of pure mono-crystal silicon from which wafers are sliced. Polishing machinery. Wafer coating machines. Wafer steppers. Etching, metal coating, diffusion, etc......... most of which also require very pure and often exotic ingredients. (I remember being amazed to read that chlorine trifluoride is used as a cleaner in the semiconductor industry. Look up the properties of that compound, it will blow your mind.) Which brings to mind the amazing work of Sam Zeloof: https://www.youtube.com/results?search_query=zeloof+z1+chip <https://www.youtube.com/results?search_query=zeloof+z1+chip> It really depends on how "from scratch" we have to go. In a more real world scenario where the ability to make large, low density MSI chips in the micrometer scale is somehow lost through scrapping and ignorance, recreating the machines required to make them on an industrial scale shouldn't cost a huge amount. It's proof, that if someone can make these chips in his garage on a shoestring budget, a few hundred grand should be able to recreate the technology in order to reproduce them industrially. Obviously there'd be an iterative process, much like happened in history, to shrink the die process if you were indeed trying to get smaller. However, if we had some sort of massive scale Carrington event (or larger) that fried every IC ever made, things would definitely take a step back. I imagine that the large crystal-growing tanks themselves would be fine, but the control gear etc would be truly sent back to the dark ages. It might take 5 years or more to re-engineer all the control gear to work in an electricity-free (or at least IC free) world, with many other things like food, water and fuel supplies taking precedence. Then there's the task of re-engineering the IC production from paper copies of research papers and documentation (no internet, remember), which could prove rather difficult to duplicate and distribute without photocopiers. We'd essentially be taken back to the early 60's when it came to IC production, having to re-engineer everything from scratch. Much of the non-electronic machinery and R&D would still be available, so i imagine getting back to where we are now would only take 20 or 30 years, instead of 60, but it would still be a long and iterative process which would be likely hampered by the whole "end-of-the-world Armageddon" thing the Carrington-scale event would have caused. I imagine most people would be worried about putting food in their bellies and keeping warm, rather than worrying about getting their Exchange server working. I think i may have overthought this. Josh Rice