There are all kinds of technology that appear through the ages that find immediate success, promise to revolutionize the world, but fade to obscurity almost as quickly. Things like the ZIP disk, RDRAM, the digital compact cassette, or even Nintendo’s VirtualBoy. Going even further back in time [smbaker] is taking a look a bubble memory, a technology that was so fast and cost-effective for its time that it could have been used as “universal” memory, combining storage and random-access memory into a single unit, but eventually other technological developments overshadowed its quirks.
[smbaker] is placing his magnetic bubble memory module to work in a Heathkit H8, an Intel 8080-based microcomputer from the the late 70s. The video goes into great detail on the theory of how these devices used moving “bubbles” of magnetism to store information and how these specific devices work before demonstrating the design and construction of a dedicated support card which hosts the module itself along with all of the necessary circuitry to allow it to communicate with the computer. From there he demonstrates booting the device using the bubble memory and performs several write and read actions using the module as a demonstration.
Eventually other technologies such as solid-state RAM and various hard disk drives caused the obsolescence of this technology, but it did hang on for a bit longer in industrial settings due to its ability to handle high vibrations and mechanical shocks, mostly thanks to the fact that they had no moving parts. Eventually things like Flash memory came around to put the final nail in the coffin for these types of memory modules, though. The Heathkit H8 is still a popular computer for retrocomputing enthusiasts nonetheless, and we’ve seen all kinds of different memory modules put to work in computers like these.
Heathkit – A once respected and revered company. Today [1] it is just another case of an old trademark being hijacked. That’s really a shame. With more innovation, community building, and a lot less greed, I think the Heathkit name-brand has potential.
1. Today’s Heathkit
https://shop.heathkit.com/shop
Never realised they still exist (somehow). I miss the old Heathkit.
50$ for an ATMega on a small board, not even Arduino compatible?
That’s a brave price… o.o
Took me back to discussions with my college physics tutor whose research field was improved magnetic materials for bubble memories. That’s a walk down Alzheimer’s alley :-).
I didn’t watch the whole 40 minute video, but I checked if it mentioned Intel, who was a BIG proponent for magnetic bubble memories “back in the day”. He did mention that he was using Intel bubbles. I recall how Intel pushed that in the media, in ads, and via their FAEs then, but then… it disappeared and I think they switched to flash and larger DRAM. The flash evolved from their UVPROMs that started from their humble PMOS 1702 256 x 8 bit beginnings that required a -46V programming pulse, so mostly not in-circuit programming! I have one or two still. Intel was also a early advocate and investor in Stanford R. Ovshinsky’s Ovonic phase-change memory development that lead to the printed transistors used in high-resolution LCDs that he predicted back in 1968. HP later used the concept to produce their version of a memristor (named by Leon Chua in 1970) that was touted as the ultimate memory, but then HP “devolved” into what its pieces are today.
I’m still waiting for someone to make that spinning holographic video memory “coin” I saw in the 1960 version of “The Time Machine” movie with Rod Taylor.
Wouldn’t bubble be better for high radiation environments? A modern “core”.
Bubble memory would have an advantage in rad hardened systems. Flash memory uses small capacitors to store info. Radiation strikes the dielectric material and ionizes it depleting the stored charge.
Back in the 70s and 80s, alpha particle radiation nearly destroyed the DRAM market, later found to be due to “bad Gold” (i.e. radioactive), passivation mechanisms (e.g. polyimide coating of the die) were first used to mitigate its effects. Then that and other rad-hardening techniques used for space and high-altitude electronics since then. Other mechanisms can and are used to “fix” the problem with memory correction HW and techniques like “memory scrubbing” to scan memory for “hits” and do corrections while they are still correctable (and detectable).
Wow great video. Fascinating combination of matching Bubble Memory to an H8 rather than a modern day PC. (Would be interesting to see that though). Bubble Memory was used in the 80s. One model of the NCR 2126 cash register had an optional Bubble Memory board that could hold the price look-up scanning file for small stores. I had one store with 6 lanes with laser scanners. Two of the registers were the Master and SubMaster both with Bubble Memory boards. Due to the 2126s design both machines memories were simultaneously updated as items were scanned. Should the Master go down, the SubMaster could be switched into Master mode and the store could continue scanning without loosing a single total. Scanning speed with all 6 lanes running was as good as today with the Bubble Memory handling lookup of bar codes and sending price and item detail info to each lane efficiently. This was in the day when there were clerks standing in each lane to check you out. NCR dropped Bubble Memory because of the expense of it and the cheaper memory that came out including early NVRAM.
I don’t recall bubble memory ever being considered “fast”, especially since it’s inherently serial access. Perhaps fast compared to a disk drive. The Intel 2102 was a 1 kbit static RAM with 350 ns access time, first available in 1972.
I designed made my own 2KB SRAM board for my 8008 based Mark 8 in 1977 using “surplus” memory that (I think) was 1.5 usec rated. I’m unsure if 350 nsec parts were available for a few more years and process improvement with finer lithography, but it they were, they had to be super pricey “picked” parts.