5 bit Friden Flexowriter punching paper tape for coding in machine language. Coding sheets. Weird hexadecimal digits 0-9fgjkqw (it took me a while to adapt to 0-9a-f when I first saw them!) Ordering instructions so they'd appear under the drum reader when it was time to execute them. Spending hours debugging code to the point where I was flagged in the weekly publication as the person most likely to be on the computer.
The most amazing thing I ever saw on the LGP-30 was that someone had managed to implement floating point on it -- we got a chance to try it out near the end of our summer course. The second most amazing thing was a program to compute e to several thousand decimal places.
FGJKQW makes sense because almost all the rest of the alphabet was used for the instruction set (one letter per instruction) so these were left over.
For the life of me, I can't figure out why ORDVAC use KSNJFL. If you Google around you'll find the claim that this was because in figure shift mode the input teletype produced 01010 to 01111 in binary on the tape for those letters. Except, I can't find a 5 level code that actually conforms to that. It's certainly not Baudot as some claim. And if you look at the manual the claim about figure shift isn't even true for the actual code used. https://www.ideals.illinois.edu/items/55048
On p4 the code has KSNJFL punch binary 10–15. It starts with a permutation of Baudot ITA2 (notice the QWERTYUIOP/1234567890) to make 0–9 punch their values.
Yes. But why that permutation? And although there is a relationship (like Q and 1 being on the same key) there isn't a relationship with Baudot or ITA2 (e.g. Q/1 isn't 00001 it's 11101).
In case you're not already aware (and/or for readers who aren't, because I wasn't yesterday), they made substantial modifications to Model 15 Teletypes to get their chosen coding and key layout (numeric-pad-like rather than QWERTY), including modifying and permuting levers, type bars, and keycaps. It could be that KSNJFL was somehow most convenient in that respect, though it's hard to imagine.
Yeah. It's pretty interesting because they clearly did this super intentionally. As you say they moved everything around. Since they are using the existing keycaps which have Q/1, W/2 etc. they lose QWERTYUIOP but they've still got ASDFGHJKLZXCVBNM from which to get the hex digits.
The one clue I haven't explored deeply is that there are references to ORDVAC and ILLIAC I using the same teletype equipment as prepared for the Institute of Advanced Study (IAS) Machine (i.e. von Neumann's machine). These machines all have the same principles and it's possible that these teletypes (prepared by the US Bureau of Standards) are the source of KSNJFL.
My late father used to be a service technician with Royal McBee who sold the LGP-30 in Europe. Before I got my ZX-81, he gave me the LGP-30 manual as an introduction to programming. He had some stories about programmers optimizing drum memory access and what headaches this caused when the timing of the motor was slightly off.
It’s such a weird architecture by modern standards - self-modifying code is the norm and you have to pay attention to the rotational speed of the drum. I can only imagine what programming on a real one must’ve been like :)
LGP-30 had 113 tubes (for logic switching? or memory?) of mostly type 5687, which elsewhere had industrial uses and were used in hifi audio (I'm unsure how), but are not appropriate for guitar amplifiers. Apparently they take a lot of plate voltage and get pretty hot. I found this brief writeup[1] on them (with pictures) mildly interesting. They can be readily found on eBay in a wide variety of pricing of various brands with NOS pretty expensive, as usual. And there's a neat page dedicated to LGP-30 here.[2]
I can fill in since I recently dug through the LGP-30 schematics[0] on my current (incomplete) journey of designing my own tube based computer.
Tubes on computers from this era were used almost exclusively as amplifiers, inverting and otherwise. [1]
Inverting amplifiers are super helpful to turn the and/or gates you can build out of diode logic into nand/nor gates which then give you access to all of the other logic functions. But if you only need and/or for a particular circuit diode logic is much cheaper with much higher MTBF. [2]
Non-inverting amplifiers are useful to buffer signals when the required current draw is too high for a single tube/diode to push.
So in the LGP-30, you primarily see them being used as the inverters for the nor gates on it's flip flops, the amplifiers needed for the read/write heads of the drum, and inverters/buffers needed in the general logic. [1]
And FWIW, I don't really see why it wouldn't do a good job within a guitar amp. It wouldn't be a good fit for the final output stage before the output transformer because of it's relatively low heft compared to a beam power tube or something, but it'd be a good fit for an input stage before that. They really don't take that much plate voltage (the LGP-30 runs them with a B+ of 150v). Maybe they're simply too clean of an amplifier and guitar amps like that distortion of tubes that are more prone to voltage sag?
[1] I want to add in two places using [1] that in addition to what I'm saying about amplifiers, diode tubes are also used as rectifiers in the power supply. There'll only be a handful of the in each design, but because they're used in the power supply of pretty much every design they're still ubiquitous even if they don't count for much versus all tube use in these computers.
[2] This use of germanium diodes in logic came pretty late in tube computers though, really in the late 50s at the cusp of transistorized machines. Royal Precision's next computer, the RCP-4000 was basically the same design as the LGP-30 but with tubes swapped out for transistors. Leads to some goofy schematic choices like the negative voltages being at the top of the schematic pages because it lines up better when compared with a tube based design.
The 5687 has a low mu and high transconductance. If you put it on a chart with the 12AX7 and the 12AU7, the 12AU7 would be in the middle. The 5687 has even higher transconductance and a little lower gain.
If it ran hot, part of that was probably the fact that the 6.3 V heaters ran at 900 mA, instead of the 300 mA. This is a full 3x as much power, just for the heater.
You could use it in a guitar amp, but I would probably just use a 12AU7 instead. The 5687 has a different pinout.
> It wouldn't be a good fit for the final output stage before the output transformer because of it's relatively low heft compared to a beam power tube or something, [...]
There are guitar amps out there with less powerful dual triodes used as output. Yes, they're low-power. There are amps out there using a 12AU7 for 500 mW of output power. It seems like such a small amount of power, but it can be loud--loud enough to damage your ears, loud enough to get evicted, not loud enough to keep up with a drummer.
> Leads to some goofy schematic choices like the negative voltages being at the top of the schematic pages because it lines up better when compared with a tube based design.
I'm guessing this is because it was using early germanium transistors which would most likely be PNP. It was a long time before NPN transistors became the "default", because they were easier to make in silicon which came along later.
Probably right. It's actually still normal/correct to draw a PNP circuit between negative rail and ground with the negative supply at the top. Schematics should convey how the engineer thinks about the circuit, and inputs come in the top and left and outputs exit at the bottom/right. Except no one really designs PNP-oriented circuits with negative supply voltages these days because eugh.
I don't know if 5687 are still made, but there are tube manufacturers, Shuagang and Linlai in China, JJ Electronic in Slovak Republic, and Reflektor Corporation in Russia.[1][2] But 5687 are not rare.[3]
I have a small ambition to write an LGP-30 emulator & to run Mel's poker code; sadly, in my poking around, while there's a tiny amount of Mel Kaye-authored source code archived[1] (in the form of scans of hand-written documents that appear to be the equivalent of pull requests, with one engineer writing them and another signing off on reviewing it) it's not very useful & is not from the poker program.
It'd be amazing for that program is on this computer and someone can pull it off, but I imagine the magnetic drum memory would've decayed into noise by now.
University of Stuttgart has copies of quite a few LGP-30 paper tapes, including Mel Kaye's blackjack. ftp://ftp.informatik.uni-stuttgart.de/pub/cm/lgp30/ [edited link]
ETA2: For anyone who's interested, I had better luck with the Python ftplib than a command line client; despite following their instructions to use a - in the password, my command line client still froze up.
Even the http mirror only allows to browse, not download. Please consider uploading the files to Internet Archive so other people can have easier access.
I wonder if that blackjack program still exists? Getting that to run again on this machine would probably get you the Nobel Prize for retrocomputing (if that was a thing).
I used to program one! Then a PDP 11-34. In school.
Some years after graduate school, in Los Alamos, visiting the Black Hole a very strange govt surplus store in a very strange town, I saw a bunch of old PDP 11-34 stuff along with printing terminals (DECWriters?) etc. Standing out in the rain. Because there wasn't room in that old warehouse for one more thing.
No, "new" would not be fine. Being forced to make an account to see certain content is not fine. Being shown one comment depth at a time is not fine. Burning vast amounts of RAM is not fine. It's a trash fire. Intentionally so, one suspects, to shepherd people towards the app. The whole reason they keep "old" around at all is because they know that "new" sucks.
IIRC I think a Reddit dev commented here on HN saying that that was the case. Can't find it again unfortunately, but it was the most popular top level comment on a thread about Reddit
> I mean if they got rid of the popups and nagging to install the mobile app, new would be fine
Sure, but they haven't, and they won't, so it's not. They constantly run A/B tests that include some pretty aggressive variants, too. Many of those completely lock out reading posts until you're signed up or using the app.
In old Reddit you open the link, see the content, and close the tab. New Reddit is heavier, slower, more aggressive towards requiring an account, specially on mobile.
It took me a moment - the LGP-30 is a computer that uses magnetic drum memory and is not a musical instrument from the 1980's.
I'm curious if the ferrous oxide is still attached to the drum. I would suspect the heads need cleaning - being clogged with magnetic particles that have flaked off the drum.
A friend of mine got one back in the 1980s when a steel mill threw it out... sadly it was destroyed in a flood. It brings back nostalgia any time I see them mentioned.
I've seen tons (quite literally!) of ancient computers, but the sight of the LGP-30 still surprised me. When I saw it, I thought more along the lines of jukebox or background music system, not computer. I think the typefaces used, and the condition of the surfaces, might play a role in that.
I'm definitely used to seeing "retro" computers from the 60s and 70s on the Web, but not from the 1950s.
It's really jarring to me, to see what looks to me like "WWII era" or "Korean War era" styling on a computer (that isn't the size of an entire room). The fact that its input and output device looks like a traditional typewriter and not a modern computer keyboard/printer combo thing is also a weird mix of older and newer styles.
It's a really interesting transitional era in design, technology, and culture all at once. And I feel like this computer is straddling both sides of all 3 transitions. It feels like a link to a past that is otherwise locked away in a handful of black-and-white photos. I love it.
The PDP is also cool, if only because the colors and styling are so clearly "of that era". I love how chunky and rectangular the design feels, without seeming bulky, paired with the bright orange and yellow. Those switches look like they'd be really satisfying to flip, like the switches on the electric organ in my family friend's parents' basement that they had when I was a kid (this kind: https://i.ebayimg.com/00/s/Nzk0WDEwMjQ=/z/Vo8AAOSwqMtdzDYV/$...).
The LGP-30 made ingenious, economic use of the technology available in the mid-1950s.
Even the CPU registers were on the drum memory. The control panel displayed the register contents on a small oscilloscope - it showed the bits as they rolled by the read head!
5 bit Friden Flexowriter punching paper tape for coding in machine language. Coding sheets. Weird hexadecimal digits 0-9fgjkqw (it took me a while to adapt to 0-9a-f when I first saw them!) Ordering instructions so they'd appear under the drum reader when it was time to execute them. Spending hours debugging code to the point where I was flagged in the weekly publication as the person most likely to be on the computer.
The most amazing thing I ever saw on the LGP-30 was that someone had managed to implement floating point on it -- we got a chance to try it out near the end of our summer course. The second most amazing thing was a program to compute e to several thousand decimal places.