Apollo Guidance Computer - facts from the horses mouths

On a sunny day (Wed, 24 Jul 2019 19:13:05 -0700 (PDT)) it happened snipped-for-privacy@gmail.com wrote in :

Very nice.. All Ye Need is transistors. And do everything with NOR gates.

:-)

I like the power supply too:-) That sort of circuit was very common in those days.

yes or one Microchip PIC. Not sure how that would behave in radiation.

a AGC emulator for an 8 bit RISC seems plausible, usually Flash memory at least is read word-wise and the Apollo AGC has a 16 bit word length.

A e.g. PDP-8 emulator is more problematic there isn't any both time and space-efficient way I know of to pack a 12 bit word into 16 bits and read/write it from 16 bit/8 bit memory blocks.

There are also pictures of RTL logic used in AGC at

Similar RTL chips where also used by some Titan missiles. After the cold war, when some Titan missiles were converted to satellite launchers, they needed some RTL chips for replacement. They had some problems finding such chips any longer :-)

On a sunny day (Thu, 25 Jul 2019 16:53:47 +0300) it happened snipped-for-privacy@downunder.com wrote in :

Yes, funny I was thinking of that, because even in the sixties I designed and build a triggered scope with RTL logic, so was wondering maybe those were not rad-hard enough... Makes sense to use those, reduces part count considerably.

Next was DTL... have not done much with that. TTL yes, and CMOS :-), ECL no..

The multipart series about restoring an original AGC to working condition on YouTube (Curious Marc) mentions a gate-exact FPGA implementation multiple times, so no doubt others have succeeded in that.

I have a Honeywell Alert computer that was developed for the X-15 project. Design was started about a year before the AGC, but finished a year earlier, as it was a much simpler project, overall. It is a DTL 24-bit computer built from, I think, about 5 basic chip designs. There was an edge-triggered universal FF. It could do D, JK, T and RS styles. it fits in a half-ATR box and draws 125 W (25 A @ 5 V). It still works, as best as I can see. I got a memory box with it, but it is a wreck and I have no docs on it. I have some docs on the Alert, as well an an AMAZING 2-volume set about the development of the project.

Jon

Sure, you could write an emulator on just about any single-chip CPU - it has been done many times. I run simh (on a RPi) in parallel with my old iron PDP8E and PDP11/34, and my hybrid 11/34 fpga+simulator.

My thrust was to design/build a hardware gate-level engine (using VHDL to specify a 3-input nor gate which would most likely be in a LUT), and not just emulate (in software) the thing. j

As i vaguely remember,from speed testing of RTL, the rise time was about 6nSec and the delay was about 7nSec.

On a sunny day (Thu, 25 Jul 2019 21:52:55 -0800) it happened Robert Baer wrote in :

hehe, I was thinking "AGC" normally means 'Automatic Gain Control' and did a look up for what else it could stand for.

The wikipedia link about that AGC (the Apollo Guidance Computer one) I found is very informative for an overview of what hey were doing, the hardware and programming, with lots of pictures:

and indeed it was RTL (4100 chips), and the second model even less it had 2800 RTL chips more integration.

What I do not see is the point (with all respect) of rebuilding it in FPGA or something, apart from even the fact that you cannot do RTL in FPGA AFAIK :-) But to run the same software in a PIC makes sense (or on a PC). I remember playing moon landing game on a Commodore PET2000 I think it was in the late seventies. Very interesting in that wikipedia link is the 'PGNCS 1668' bug they encountered and that takes you right to the current state of affairs with embedded programming.

Whatever way you look at Apollo, it was a master piece.

Yes, I think that is right. The first AGC used ICs of a single NOR gate (3 transistors) in a round metal can like TO-99. The second version used ICs with dual NOR gates (6 transistors) in ceramic flat packs. So fewer chips but more gates in total :)

piglet

So you can get to the Moon in 30 seconds.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC / Hobbs ElectroOptics 
Optics, Electro-optics, Photonics, Analog Electronics 
Briarcliff Manor NY 10510 

http://electrooptical.net 
http://hobbs-eo.com

OOps, design was started a year LATER than the AGC.

Jon

That is one weird power supply. And the bridge seems to be rotated 90 degrees.

Yes, the bridge is not rectifying the output to supply DC, the output to the EL panels is AC - instead what I think the bridge is doing is rectifying a feedback current control into the saturable core. Cute?

piglet

On a sunny day (Fri, 26 Jul 2019 16:01:25 -0400) it happened "Tom Del Rosso" wrote in :

At the risk of being totally wrong, I think it works this way:

It is a feedback circuit. L2 is a transductor. The bridge senses current and saturates the transductor via the winding marked 'control'.

It is an old design, was it named 'westat'? Cannot find a circuit with google or, bing finds this:

Not exactly like this, but using transductors for stabilizing power was very common in the sixties. Anybody has some circuits? Should have payed more attention to it back then, nut was using transistors and thyristors.. In that company where I worked there was a guy who was very good at that sort of thing.

It may be part of the display HV generation (note the text near pin 8)?

Correct, 250V AC at 800Hz for the DSKY El panel display. The segments were switched by lots of tiny TO-99 sized Teledyne relays.

The mag-amp current limiting may have distorted the waveform a bit - not sure if C1 was to help resonate to more sine-like - I think waveform probably didn't matter as long as there was no DC.

piglet