The prototype first ran on 28th May 2021 whist perched on our dining room table, chairs and surrounding floor space.
The Valve.Computer is now firmly nailed to the study wall,
and is almost safe to touch.
Thermionic valves, (aka vacuum tubes), can switch several hundred million times a second, and in the 1950s were the basis for all computer designs. To work efficiently they require high voltages and are not for the faint hearted.
The Valve.Computer is an 8 bit computer, with the usual 12 bit address and data buses plus the rather unusual current demand of over 200 Amps. It can play a decent game of PONG using its valve and relay RAM, or run a 32 bit Fibonacci sequence using modern NVRAM.
After switch on you have to wait a while for the last thermionic valve to warm up. If you look from the side you see a few start to show a red glow. After about a minute the Valve.Computer has the pleasant homely aroma of 560 double valves, quietly burning off their dust.
When all the valves are glowing, I check the fire extinguisher is full, and run the code.
It has been a ridiculous amount of soldering
and a fantastic amount of fun.
After visiting Bletchley Park, it occurred to me that several thermionic valve computers had been rebuilt, and now run in museums, but that no new design of a valve computer had been constructed in over 50 years. The thought of building one seemed ridiculous, but I wondered if a modern design could overcome the issues of size, power and the very real danger of high voltages.
When I retired I looked at the problem again and realised it could be an interesting endeavour. I spent an enjoyable 18 months building both the prototypess and the final Valve.Computer.
The Valve.Computer is mainly built from my two prototype machines, ena.computer and fred.computer. Modifying the prototype PCBs required changing over 1,000 components.
The rebuilt main boards are now configured exactly as shown in the technical page schematics. The new system console incorporates the GUI from a prototype. There are also new auxiliary valve boards incorporating additional memory registers and modified oscillators.
The Valve.Computer has valve memory, reed relay memory, (lethal) memory switches and NVRAM.
I found it best not to use a separate room to build the computer, but to spread it all over the house, one bit in each room if I could, just so I knew where everything was. To my great surprise my lovely wife Judy was completely overjoyed when after only a year or so, I transferred all the components around the house onto the study wall.
Please note that high voltages are very dangerous and shouting bang when a friend has their back to the computer is very childish, but great fun. I must stop doing it!
The Valve.Computer is designed using 1,120 thermionic triodes. Conveniently each 6N3P valve contains 2 triodes around a single heater, halving the physical size and power requirements.
All the double valves are configured as identical NOR gates. Memory registers are built from groups of 5 NOR gates, and combined into D type flip flops. The NOR gates also form all the other functional components, including the 8 bit ALU, the two oscillators, and the relay drivers.
The amount of heat is ridiculous, but the warm cosy aroma is divine.
The Valve.Computer integrates eight large printed circuit boards and four auxiliary PCBs, which combine the thermionic valves into a functional general purpose computer.
It has now grown beyond its design, and for me, has become an art installation on the study wall.
Every one agrees that the Turner Prize is much more than just a display of virtue signalling by the cultural elite, and I have decided to enter the Valve.Computer for the prize.
I shall write about the blank canvas of machine code, the brush with danger, the sculpture of the valve, and a palette load of arty stuff like that. I am sure Tate Modern will be impressed!
Thermionic valves (or just valves) in the UK, are also named vacuum tubes (or just tubes) in the USA, and электронная лампа (or just electron tubes) in Eastern Europe. The 6N3P thermionic valve was produced in many communist factories in the 1950s and 1960s.
The finest 6N3P thermionic valves were selected at manufacture for military use, with up to 5000 hours life expectancy and printed with an additional code (Cyrillic 6Н3П-ЕВ). The rest were for domestic use, and with the ability to operate at over 210Mhz, the 6N3P was used in many 1960s East European VHF radios, televisions and in several beautiful radiograms.
The Valve.Computer has a new system console, which lifts open from the front.
Five AZTEC MP6 PSUs are now behind the Console. These second hand, 60 Amp professional power supplies replace the prototype's cheap and cheerful PSUs. It turns out that cheap PSUs do not like a metal inspection torch across their unprotected outputs.
The AZTECs enable the heaters to have a soft start, which hopefully will extend the life of the thermionic valves and I now use a plastic inspection torch.
Valve.Computer programs are written using just 16 instructions. Machine code is real code, a totally different world to all the posh high level languages. It's gloves off with no safety net. You simply talk directly to the computer. It is incredibly fast compared to other programming languages, and gives you control over data storage, memory, and computer hardware. It can be a very useful tool in fast decryption and modern graphics. It's use may give a deeper understanding of high level coding, but really it's just great fun.
On 15th August 2021, three months after the first table top run with manual ALU computation and storage, a prototype for the new Valve.Computer demonstrated a clocked, GUI displayed, 8 bit Fibonacci sequence.
The Fibonacci value is displayed vertically, in binary. It uses both the GUI relay memory and NVRAM, and is demonstrated by Henry the cat.
On 11th January 2022, Judy starred as the Mysterious black gloved lady, in the epic video production of The PONG Game. Using the prototype for the new Valve.Computer, she demonstrated a simple version of the game.
Less than 100 machine code instructions enable the ball to bounce around the court, and subroutines for gameplay and the GUI display complete the program.
It's is a fast, action packed, no holds barred game [NOT].
Spoiler alert, the final score is ONE - NIL, but guess who wins!
The Valve.Computer is a once in lifetime project for me, so I decided not to do a “build it now, fix it later” version.
Which reminds me, the ducks on the pond needed a lighthouse, as it is sometimes difficult to see the edge of the pond. I thought I would help, and after discussing it with no one I just built a lighthouse. Just doing it is far more fun than planning and project managing, the only problem is the quality of the result.
It was not even watertight, and a bit of a family joke, and so it went into the shed.
I decided to build a second one.
This time I had a plan.
Project management is not just a tool to mislead the client into thinking you're actually doing something.
It can also be used to create a high quality product, on time, and slightly over budget.
For the second lighthouse I project managed, designed and then 3D printed all the parts, It looks great, and fits the brief completely.
But guess which one I now prefer.
So two lighthouses later, I realised that the act of planning sometimes loses the spirit of the idea.
Even so, I decided to try to plan and manage the Valve.Computer project, and fortunately, as the build grew, it gained more spirit than I can shake a stick at.
My stepdaughter is a senior Product Manager and my stepson is a Chartered Engineer and he frequently works as a Project Manager. They would know how to manage this project. Luckily they are too successful and busy to ever read this, thank goodness!
I tried doing the project management on a PC but gave up, and instead used 3 coloured pens, A4 paper and self adhesive address labels to cover up my mistakes. I simply found it easier. The only problem I discovered is that the paper gets a bit thick after several layers of amendments, but the big advantage is that can always find your last version, it's simply the thickest.
The main thing I discovered over the year was that the first time I tried almost anything I failed, but the annoying experience often enabled future success.
I started by stabbing at various bits of the idea. Eventually I was able to define what I wanted to do. I now realise that this definition, was the most important part of the whole project management.
The program seemed to consist of many complex components, all of which I needed to understand, to be able to build the Valve.Computer.
How do thermionic valves work, how do you design a PCB, and how on earth do you get from a NOR gate, to a memory register, via microcodes from a clock and a ROM, to actually run software. And what exactly is a microcode anyway. At times, in fact most of the time, it all just seemed impossible.
The main solutions I found were, firstly to keep the number of unique sub systems to an absolute minimum, secondly to use double triode valves to halve the physical size, which also reduced the construction time, and thirdly, to design multi layer printed circuit boards to reduce the inevitable build errors and produce a robust construction.
So far it has only gone BANG twice. The first time I pulled out plugs like my life depended on it, and Judy cried, “Everything all right!”. The second time I only slightly panicked, and Judy continued to decorate the Christmas tree, but later bought me a surpise present, a fire extinguisher!
But most important of all, is to have a lovely wife, who knows you're daft as a brush, and that life together is brilliant.