Talk on the Apollo 11 computer

[Geoff S]

I came across this long, detailed and entertaining talk on YouTube about the computer used on the Apollo programme last night. If you're interested in the early days of computing it's well worth a look.

 

Geoff

[Caveman]

Just watched this, absolutely brilliant. Thanks for posting it Geoff.

[Geoff S]

I thought one of the most amazing features (though I suppose relatively trivial) was the way the 'ROM' was programmed with the way toroids were wired - or more accurately laced!  Not easy either to debug or correct.  Even a few years later we were using ROM on 2k x 8bit chips we could program electrically - though sometimes it was a once only shot - and in  any case we were able to run the s/w before blowing the ROM.

 

The RAM in the first computer I had anything to do with in 1961 had core (toroidal ferrites) store (though it was called IAS - Immediate Access Store) and that was volatile because switching off caused random currents to flow and changed the content.  Permanent memory was a magnetic drum with a head for every track and so didn't need stepper motors to position the head to read, which made it faster - clock speed was a staggering 1Mhz ?

[Trevor]

An excellent talk. Thanks Geoff for pointing it out.

[Phil McCavity]

This is an excellent space race channel Homemade Documentaries and has a chapter on the software configuration made to the Lunar module on the fly because of the conflict it was causing during the lunar take-off stage.

[Gary Manuel]

I started watching it but my missus had other ideas. I'll finish watching later.

 

What I did see took me back a few years. I was railway signalling engineer and a few years ago came across a working Elliot computer at Healey Mills SIgnal Box. It was in dire need of replacement as it kept breaking down and causing loss of train describer information. The local lineman had a nack for keeping it working but he took ill and it needed a bit of support. That's where I stepped in as Technical Support Engineer.

 

It was a steep learning curve; not so much because of how technical it was but because of how antiquated it was.

It had a similar Iron Core Memory to the one in the Apollo video. I recognised it immediately as I has been taught about these things many years ago at college.

The computer itself was 4 bit (1 Nibble = half a Byte) architecture, which limited it to 16 machine code instructions. Each instruction i.e. ROLC (ROtate Left with Carry if I remember correctly) was a dedicated circuit card measuring about 10 inches by 12 inches. All discreet components, hand soldered onto hand made circuit boards. The position of the instruction card in the mainframe was critical because if positioned wrong you could implement a ADD when you really wanted a NOR instruction. Fortunately the cards had a system of slots cut into the rear of them to only allow them to be inserted into the correct position. 

The computer was programmed using punched paper tape, which was about 1 inch wide paper tape with holes representing 1s and 0s. 

 

There was no keyboard or screen, just a bank of switches with an implement button and a row of lights as output indications.

 

I absolutely loved that period in time and the technology - it almost made what was going on under the bonnet of the computer visible. Not like the box of black art we live with today.

 

Ah the good old days.

[GrumpyGnome]

Early days seems just as much black magic as now....... just how do people think of these things?  

GG

[GrumpyGnome]

Not just science and technology.... like who thought it'd be a good idea to take some beans, roast them, grind them up, pour hot water over them and drink them ?  I can see alcohol may be a (happy) accident, but coffee is a deliberate and quite convoluted process.......

 

Sorry for wandering off topic!

 

GG

[Christopher Wolfe]

Your mobile phone has probably 10,000's times processing power of these early Apollo computers.

 

I used to work in the early 70's on weapon tracking radar systems, magnetic drum memory systems and TTL processing were the order of the day.

 

What a sweet way to learn about digital electronics when I was young enough to actually understand it. Clock speed maybe 1MHz.

 

So congratulations to the pioneers who actually had one chance to get the moonshot right 42 years ago with such 'primitive' resources.

 

[Cuban8]

Not forgetting that apart from the hardware that had to be made small and reliable enough for man rated spaceflight, software and the operating systems to use it had to be invented. See the documentary series 'Moon Machines' and a brilliant early pioneer of software engineering Hal Laning.

BTW, when I went to college in the early 80s studying electronics after work (evening classes, whatever happened to them?) magnetic core memory was still on the syllabus as there were plenty of examples of that technology still around at the time. The most ancient bit of electronics that was still in daily use and that I had the 'pleasure' of maintaining up to the millennium, when it was finally replaced, was an automated mailing system that used dozens of foot square PCBs each stuffed with hundreds of TTL logic chips - You'd probably get away with a couple of Arduinos to do the same thing now.

Edited March 18, 2021 by Cuban8

[Nigel R]

4 hours ago, Christopher Wolfe said:

Your mobile phone has probably 10,000's times processing power of these early Apollo computers

 

2mhz clock, 12 steps per instruction...

I make that 2/12 = 0.167 MIPs

 

current top mobiles have 8 core CPUs, each core running around 4.5MIPS per MHz, at around 2.5GHz. I make that

2.5 x 1000 x 4.5 x 8 =90,000 MIPS

 

in other words, around 500,000 times faster ?

[Geoff S]

However, as the speaker explained trying to compare the Apollo computer with modern devices is comparing apples with oranges. 

 

The Apollo computer was dealing with far more I/O and attempting multitasking real-time calculations.  It had no other purpose and was dedicated to its tasks.  Real-time multi-tasking is actually impossible the two things are incompatible as I found when I was trying to measure probe timing to calculate fan blade untwist with fairly simple trig.  I was using a Motorola 68000 processor and an operating system called OS9. In order to make the measurement I had to stop the operating system when I did it or I got erratic readings.  The only similar systems to the Apollo computer I've been involved with are the engine control computers for gas turbines which also need to process a lot of I/O in real time.  The first one I worked with wasn't even all-electronic - it used magnetic amplifiers and was used on the Phantom Spey engines back in 1965.

 

btw There was another YouTube video flagged that I intend to watch about the development of the original BASIC language at Dartmouth which must have been a bit earlier than the Apollo programme.

 

[Nigel R]

Somewhat apples/oranges. But I think the raw processing capability stands as a comparison of how much quicker the current technology is. Which is to say nothing of the application it is used for, nor the peripherals, IO or comms capabilities attached to it.

 

Your particular problem with OS9 is familiar. The use of multitasking OS is not de rigeur in aerospace - they are too difficult to test. Software with fixed schedule timeslice type approaches of varying implementation is much easier to debug and test. Interrupts have to be accounted for of course, but if you have only the two things to contend with, life is easier.

 

By the way, real time true multitasking is (now) perfectly possible with multi core chips.

[Geoff S]

1 hour ago, Nigel R said:

Somewhat apples/oranges. But I think the raw processing capability stands as a comparison of how much quicker the current technology is. Which is to say nothing of the application it is used for, nor the peripherals, IO or comms capabilities attached to it.

 

Your particular problem with OS9 is familiar. The use of multitasking OS is not de rigeur in aerospace - they are too difficult to test. Software with fixed schedule timeslice type approaches of varying implementation is much easier to debug and test. Interrupts have to be accounted for of course, but if you have only the two things to contend with, life is easier.

 

By the way, real time true multitasking is (now) perfectly possible with multi core chips.

 

 

Of course, I'm talking about things I did over 20 years ago.  We were all self-taught programmers learning as we went along and writing mostly assembler embedded code for  interrupt driven measurement systems you can probably buy off the shelf now,  but a lot of it was fun.  We started long before either home computers  or any sort of PC were available.  I even once wrote a floating point maths routine for a Motorola  6809 - I think I've still got the listing somewhere.  Good job we weren't too closely managed!

[Christopher Wolfe]

17 hours ago, Nigel R said:

Current top mobiles have 8 core CPUs, each core running around 4.5MIPS per MHz, at around 2.5GHz. I make that

2.5 x 1000 x 4.5 x 8 =90,000 MIPS

 

in other words, around 500,000 times faster ?

Lovely stuff.

 

Our mortar tracking radar had a drum memory and also used magnetic amplifiers.

 

The primitive computer took it's inputs from rotary encoders for bearing and elevation and from the radar for slant (actual) target range.

 

From that, given that the system was accurately surveyed in and the antenna was precisely calibrated to a known bearing on a good day it could be fine tuned for surprising accuracy and could locate a mortar baseplate to within ± 30m at a range of 10km, sometimes before the mortar shell actually landed.

 

With a quick counter battery response the area for 1 sq km in and around the location could be saturated with nasty air-burst projectiles. (Splintex)

 

The computer was hard wired to calculate true range from the slant range and elevation data, X-Y map coordinates from the bearing and then integrate the results with gravity and air drag to obtain the parabola of the projectile. Then it was a matter of extending the parabola backwards to the launch area and observing the map.

 

Model aircraft input: I discovered that a Phil Kraft Ugly Stick with a Taipan .61 was ideal to fine tune the optical telescope alignment to precisely the centre of the radar tracking beam at about 3 km. The radar locked on, the scope was aligned and so the tracking results matched the map and survey data.

 

All in all, sad news for the bad guys.

 

Yeah, my Lenovo tab is an Octo Core and it;s already 'obsolete' after just 4 years.

 

Also, the actual moon landing was 52 years ago, showing my age, but it makes the achievement even more impressive for 1969!

[Nigel R]

There were a few really good things on the TV last year about the moon landings.

 

Back to the now, we had Mars rover landing to watch a couple of weeks ago. That was impressive!

 

 

14 hours ago, Geoff S said:

writing mostly assembler embedded code for  interrupt driven measurement systems you can probably buy off the shelf now,

 

Can't say too much but assembler is only just being phased out in some of my projects.

 

1 hour ago, Christopher Wolfe said:

Yeah, my Lenovo tab is an Octo Core and it;s already 'obsolete' after just 4 years.

 

?

[Dale Bradly]

Another one on the Apollo computer.

https://www.youtube.com/watch?v=dI-JW2UIAG0

 

One of my favourite YT channels, Destin is awesome, if you have any interest in science, this dude is well worth many hours.

Edited March 19, 2021 by Dale Bradly

[Alex Ferguson 2]

When bringing up the link, first in this thread, there were in the list on the right, a lot about Turing, Enigma and Bletchley Park. A step back further than a number of the thread responders here, around about when we were born. OK, I won't mention Charles Babbage and his programmer Ada Lovelace.

 

What I'm saying is progress is so blindingly fast. OK, cynical jest. It was only about a 150 years ago, 1876 when we could first send an email to Britain, a digitally coded electronic message, originally called a telegram.

[Geoff S]

1 hour ago, Alex Ferguson 2 said:

When bringing up the link, first in this thread, there were in the list on the right, a lot about Turing, Enigma and Bletchley Park. A step back further than a number of the thread responders here, around about when we were born. OK, I won't mention Charles Babbage and his programmer Ada Lovelace.

 

What I'm saying is progress is so blindingly fast. OK, cynical jest. It was only about a 150 years ago, 1876 when we could first send an email to Britain, a digitally coded electronic message, originally called a telegram.

 

... and, of course, ultimate telegram delivery was by a lad on a red BSA Bantam motor bike.  When I worked  for Post Office Telephones  in 1957 I quite envied them ?