[Outer World Apps] noted that there was no PDP-8/V made by DEC — a variant that used vacuum tubes. So he’s decided to make one using about 320 6J6A tubes. He’s got a plan and a few boards completed — we can’t wait to see it finished.
The logic is actually done by crystal rectifiers, but the tubes do inversion. To make an and/or/invert gate requires a single triode or half of a 6J6A. A D flip flop requires three tubes or two tubes for a latch. In addition to the “crystal diodes,” the memory and I/O are a Raspberry Pi, and there are transistors to do level conversion between the tube logic and the Pi.
The final product will run at a blistering 28 kHz or less. Possibly a lot less. So far, there is the sequencer with 80 tubes and the memory address register with 48 tubes. Remaining are the PC register, the AC register, and the next one slated for completion, the ALU.
What a labor of love! The PDP-8 had a simple instruction set, so it makes sense to pick it for a tube implementation. We get the concession to use some transistors and the Pi, although we would have loved to see this with core memory and all tubes. We have seen a PDP-8 transistorized. Of course, as long as you’ve got the Pi, you really have everything.
I recall seeing 5lb bags of memory cores – just the little ferrite rings, looked like a bag of chia seeds – dirt cheap at some surplus place, probably sometime in the 90s. Got a wild urge to wire up a few last summer, could not find even say 16 of them. (Getting gouged on flea bay doesn’t count)
So these are the sort of prices we’re talking about on eBay: https://www.ebay.com/itm/255494797059?hash=item3b7cad3703
Is 15 pounds for 5000 cores really gouging?
Well, that’s better than what I saw before, for certain.
Yeah, but they’re communist cores made in East German. That means that they may seem to work well in theory, but will not in reality.
I have a PDP-8e. No peripherals. Any programming needs to be done via the switches with the results in the lights. I do have a current loop interface card; I should rig up a USB adapter but I don’t have the time.
The right touch would be to show on the computer running it an ASR-33 teletype chugging away on the screen as data is transferred.
It took 3 boards for the CPU; a bus terminator board (which I think will be back); and 3 boards for 4K 12-bit words. One had the cores; the other two were drivers and addressing logic, with the drivers using discrete transistors. I have several spare boards of memory and some other stuff, plus a set of schematics. You can actually see the cores.
Ah, the days of wooden computers and iron programmers. And ferrite cores are largely iron, too.
For a while, I had a sign on the computer that “Real computers have front panels. And real programmers know how to use them.”
I have the pdp11/70 1/3 size from “Obsolescence Guaranteed” …. Point is I only boot strapped it from the front panel twice. That was enough for me :) . But I do like the ‘lights’ on it…. Seems real computers should show they are doing ‘something’ and the front panel makes it so :) . Glad I got it.
Labor of love above. I don’t have the patience/time for projects like that. My grand children would have to finish it…..
I knew someone who said, back in the 1970’s, that IBM sold a lot of computers because their front panels were so impressive. I’d love to find one somewhere. I had the chance to use a 360 something or other and read the manual about the front panel. The description for most of the lights was “For customer engineer [IBMese for Field Tech] use only”
11/70’s were large – several racks. Used at least one back in the day. I think a modern smartwatch has more compute power than even the 360’s of the era. And the Linux servers I use? Probably the few in the rack have combined more compute power than all the computers in the word in 1970.
I remember the IBM-1620 front panels in _War_Games_ (the movie), props for “The Whopper.” I did my first system programming on the 1620s at Purdue, adapting a card system FORTRAN to run under a Disk Operating System. They had a 407 printer on the set too.
the pdp-8/e systems i was involved with back in the day used a two-word bootstrap. the first word lit of a transfer that had been set up in the rk05 registers by the reset. the second word branched to itself; it would get overwritten by data coming in from the disk, at which point the system was off to the races.
Don’t see what so special, it’s just using MOSFETs. (Metal-Oxide-Space Field Effect Tubes)
I’d be impressed if he made it CMOS – using positron emission tubes.
Positronic brains like in Asimov’s I, Robot?
Why Asimov? Positronics are often mentioned in the Perry Rhodan universe. Guys, go get some culture and read proper Sci-Fi! Widen your horizon! 😃 Asimov.. Pfft. 🙄 What’s next? Lem? 😂
(Just kidding. 😜 Though sometimes I feel Asimov is a bit over-present here at hackaday.com 😒)
We do all know that positrons are the antimatter version of electrons, right? And therefore problematic? Not saying they don’t exist, but building a device based on them? Things that make you go “Hmm.”
Yeah, I know everyone’s joking around but sometimes stating the obvious is a form of joke.
I just finished an Industrial Robot course (I’m tutoring it) and Asimov & his “Three Laws” is mentioned in great detail in the first chapter. This is a serious class for industrial automation technicians.
What do you feed that thing? It must draw a hellatious amount of power. Hello PG&E, I’de like to buy a few hundred more KWh please.
Heater current for the 6J6A is 450mA so 320 times that is….
.. just a pinch more than 900 watts, about half what a respectable hair-dryer pulls. Of course, a hair dryer is usually employed for couple of minutes at most, whereas the PDP8 would probably need to run for a couple of hours at a time.
But that’s merely at 6,3v.
It’s drawing less than half an ampere (those 450mA), also.
In watts, it’s about 2,5w. Two and a half watts!
In tube dimensions, it’s almost nothing.
In relation, an ordinary 100Ah car battery has 13,8v and about 9000 ampere (cold start).
It will provide about 10A over 10 hours..
And if the tube characteristics are secondary (used as a switch rather than amp), the heater voltage could be reduced down to 5v or 4,5v, even.
That’s 2.835W per tube, this has 320 tubes.
But why LEDs ? We had a PDP-8 in the 1980s and it had incandescent lightbulbs as a status display. Another one I saw had LEDs but you had to kill the lights to see if they were von or off.
Power consumption, I suppose. A private home isn’t a laboratory.
There’s only so much current that can be handled by the mains.
But you’re right, I think. Incandescent lamps are superior and should be favored, whenever possible. They’re more eyefriendly, too and have a small delay that’s handy. Comparable to the afterglow effect on a green monitor (CRT) that reduces flicker.
Kind of spoils it by by using the Pi for IO and memory. Core memory and a terminal (preferably an ASR33) are called for.
Ug, yeah, I hate that too. I mean, don’t get me wrong, if I had a functioning vacuum tube computer I’d build a SBC-containing _peripheral_ for it to allow it to connect to the modern network and attach bulk memory, but the latter would be an emulation of an “imaginary magnetic drum farm” or something that could have been built with contemporary technology, but I didn’t have the physical room or resources to build.
Memory is the most interesting and challenging part of a vacuum tube computer! After the really wonky stuff with pulse-counting based logic was over (e.g. ENIAC in the 1940s), the real remaining challenge to scaling first generation computers – apart from tube reliability – was how to do memory. Williams tubes, delay lines, magnetic drums as not only main memory but registers, etc — core was the solution to these problems but it is a relatively poor fit for vacuum tube switching capabilities and so requires interesting electronics. Really, memory is the challenge.
LEDs are also a little tacky. Neons are much cooler. But if you use round orange LEDs and they’re behind a panel or something anyway, it might be okay aesthetically. To each their own. Usagi Electric is using VFDs in his low-voltage vacuum tube computer (not that it I’d hold it out as a model for aesthetics at all, but it is pretty successful as a project.)
Why an ASR33? It’s an ugly duck.
And a teletype typewriter, not a real terminal, even.
Just use a classic glass terminal from the late 70s. Say DEC VT-100. Looks more professional, too.
Second, it’s easy to complain about projects not being period-correct.
It happens all the time in the retro gaming scene, too. 🙂
But this project here, at its core, was a lot of work already.
The use of an Raspberry Pi is no sign of weakness or laziness.
It’s rather a practical decision and good in the early phase to “get things running”.
At a later stage, when the tube circuits are known to work fine, nothing prevents the creator to complete the computer by creating a matching drum memory, core memory.
If I understand correctly, by reading between the lines, that’s what the creator essentially meant on the site linked.
For example, he mentioned the replacement of the LEDs by neon lamps or incandescent lamps in the future. So he’s hinting that he’s not “done” with this project yet. 🙂
The PDP-8/e was sold with an ASR-33. I remember it well.
Yes. In many cases, the paper tape reader and punch of the ASR-33 was the only storage medium the computer had other than the nonvolatile core (which could get overwritten by a wayward program). In fact, typically a short loader program was toggled in via the switches, and then run to load off paper tape the loader that could handle larger programs. IIRC, the first toggled-in loader was the RIM (Read-In-Mode) loader and the paper tape one was the BIN loader, but I just found my manual’s pages have deteriorated to the point I don’t want to look it up. I remember using a Data General Nova that had the paper tape loader on Mylar tape so it wouldn’t wear out.
How backwards. Couldn’t they just have made the punch card/punch tape reader/writer an external/separate device? That noisy typewriter would be a better match for relay computer. Thank god I never had to operate a telex machine! Otherwise I would have lost my hearing. A noisy computer can be locked a way in a sound-proof room, but not a “terminal”.
If the ASR-33 had been a modern ball-head typewriter, I wouldn’t complain, even.
These were cool and had a nice, bearable sound.
But as it looks it’s really just a hacked telex machine that does ASCII.
There were external punched tape readers and writers, but they were more expensive (and vaster) and needed their own interface board.
As for teletypes, yes, they were derived from the earlier 5-bit machines. Amazing mechanical devices. The keyboard used a mechanical encoder based on bars running from the front of the keyboard to the back, one per key, which had tabs and slots to push down the 7 bars running lengthwise that operated microswitches, serialized by a sort of distributor working backwards. The received bits operated solenoids that moved the type cylinder up and rotated it to the correct position.
Computers had to delay after sending a carriage return because it took so long; many systems would send DEL characters that did nothing for timing. If real characters were sent, they would be typed wherever the printhead was at the time.
pdp 8/e is the only computer poster hated.
Hired to get the Illiac 3 working in 1972/3 at the University of Illinois at Urbana-Champaign.
8/e used to help get hardware working.
Crashed LOTS!
No boot strap loader.
Loader had to be toggled in by hand. :(–
Illiac 2 project leader announced to 3 group of hardware engineers/computer programmers in late fall
that the Illiac 3 project had failed.
All of us to seek other employment.
Illiac 3 hardware engineer reported that Illiac 3 designer said “The most original idea he ever had was to
go to the bathroom.” :(
Illiac 3 designer left UI.
Illiac 2 project leader introduced poster to his and his phd students computer arithmetic bit by bit elementary
function computations.
Bruce DeLugish UI phd thesis dissertation might lead you to discover how public key cryptography was broken in
~1991/2?
Public key vulnerable to encrypting messages close to the square root of the two prime number modulus?
Simultaneous equations decryption possible? Poster does not know, of course.
But NSA started removing public key from its products in ~1991/2, poster knows.