• Please review our updated Terms and Rules here

Honeywell 200 resurrection

Your answers are so thorough that they inspire more questions!

1000 logic boards? Wow. How many gates per board? Transistor logic? How many types of logic board?

Is the backplane already wired with an H-200 design, or do you have to replicate that?

Have you posted pictures of the hardware?
 
The logic boards hold from six to thirty-six gates each. They aren't from an H-200 but from seven Keytape machines, Honeywell tape drives with keyboards attached for manually encoding data. These were direct replacements for the old card punching machines with little more facilities. Their design dates from around 1969, some years after the original H-200, and they use early DTL ICs for the logic along with transistors for special functions. Despite packing in more gates per board than the original second generation transistorised boards of the H-200 they were limited by still having only 36 pins on the backplane connectors, so even with only eight gates not all the IC pins could be brought out onto the connector for external access. There are two distinct styles to their design. Some boards are general purpose ones with no internal connections between gates for use where all the logic design is in the backplane wiring and others perform very specific functions and have the logic design in the internal connections between the gates. They illustrate very well the way that the architecture of a computer gradually moved from the wiring of the backplanes to the tracks on the PCBs until eventually the backplanes just became a general purpose bus.

I have 84 types of board, some of which will only serve as sources of parts. The two backplanes that I have were from one Keytape machine and I spent hours removing all the existing wiring to get back to bare pins. I have just obtained an electric wire-wrap gun to make wiring up the new design a little easier. What I intend to replicate is the physical appearance, external and internal, of the machine as far as possible (I even have a sample of the original blue paintwork on the cabinets) and the operational and programming aspects, again as far as possible, but the actual logic design will have to be a compromise between authenticity, available components and available space on the backplanes. I will keep to using original Honeywell components as far as possible and only use other components which are appropriate to the era. I really don't want to use an original control memory though, even though one may be available, as they were extremely sensitive to variations in temperature and prone to burning out. This computer must operate reliably in a domestic environment in England (too damp a climate for reliable punched cards as well).

Vintage computer projects vary from dedicated careful conservation of original machines to building replicas using modern technology or software emulators. This project will be fun because it has one foot firmly in the 1960s and the other in the 21st century. If I do manage to complete the machine I will be able to claim that it is the newest 1960s computer in the world.

So far I haven't posted anything but this thread as I intend to put everything on the project website honeypi.org.uk which isn't up and running yet. Designing the website is just another little task that I have to master in my spare time as I was hoping to leave that particular skill to the next generation, so never took an interest in it. I do have an old friend who runs a web-mastering company though, so ...
 
Yes, that was one of the series. Our company kept upgrading so I never kept up with what the current machine was called, but I think that after the H-200 we had a 1250 and a 2200. At some stage we moved over to the series 60 and 66 running the GCOS operating system, originally the GECOS system from General Electric which Honeywell took over. If you hunted around in the operating system you could find places where Honeywell hadn't even changed "GECOS" to "GCOS" in the text when they adopted it. The GCOS family were fundamentally word-oriented machines and didn't have the simple style of the series 200 character-oriented machines although they included the EIS feature, the "Extended Instruction Set" which emulated the series 200 at hardware level. I remember an instructor on a training course enthusing over EIS and we couldn't understand why until we found out that he had only just come to Honeywell from working on IBM machines and thought that EIS was a great new feature, not realising that it was actually included for compatibility with the way that older Honeywell machines had been working for years. That in itself said something about the 200 series architecture.
 
You could be right. By the time we had the level 66 I had moved from programming to analysis and didn't need to understand such things so well, so that was hearsay. Our machine had both capabilities, so I don't know whether they were separate hardware or two aspects of the same hardware. What you got in the box and what you'd paid for weren't necessarily the same anyway. Our engineer once mentioned that one of our machines had a go-faster option which he could activate virtually instantly at the flip of a switch -- if we'd paid for it of course. History doesn't record whether he ever flipped the switch for a while but our work usually seemed to get done on time ... and PC gamers thought they were the first to use overclocking.

As another example the smallest memory configuration marketed for the original H-200 was 2k but I can't find any mention in the system description of memory modules being anything but 4k. The documentation states that the first 4k module was logically divided into the basic 2k and an optional second 2k, but there is no suggestion that the modules themselves were cut in half and all the supporting driver boards were designed to support 4k anyway, so the machine must have had the 4k capability as standard. The smallest assembler for EASYCODER needed 4k, so a 2k machine couldn't even assemble its own programmes. Honeywell envisaged 2k machines as front end processors for larger computers, so maybe they weren't bothered about giving away 2k of unused precious core memory to a good customer with other large computers and the reduced selling price was just a marketing ploy. A strong feature of the H-200 was its time-sharing peripheral handling which meant that it was more suited to being a sophisticated peripheral controller than a number-cruncher, but Honeywell managed to play it all ways. Interestingly the later Level 6 minicomputers were also employed as front ends to mainframes, handling communications with many dumb terminals. The Level 6 series followed the same principle of modularity established much earlier during the design of the H-200. I know because I have a "DIY" DPS6 kit made up of the components of seven such computers all using Honeywell's "Megabus" architecture.
 
There were lots of examples of that type of thing on the L-66. There were the 100/100Mb disks that only used the first 100 tracks and were upgraded by moving a jumper. My favourite was the "gear shift" on the L66/17 (and I think 27 and 37) models.

So the 17 performed like a L66/10 in batch and like a L66/20 for time sahring and on-line TP. The secret was a few lines of code in the dispatcher which disabled the cache. A one line path would make the machine run like a 20 for batch as well. A wonderfull marketing trick. Of course to upgrade you simply changed a link that set a bit in a control word....

The same pracrices continue today. Look at the BROCADE 24 port swith:-

http://www.brocade.com/products/all/switches/product-details/6510-switch/index.page

note the "port on demand" feature. Basically this means you always get a 24 port switch, but they charge you extra to use all 24 ports. IBM have a similar "storage on demand" feature in its SAN. It comes full of disks, you pay when you need them...
 
Now here's a funny thing. For one brief moment there was a posting from marcelvanherk on this thread. Marcel is the man with the H-200 memory modules and therefore a key person in my project. He subscribed to my thread at my invitation but I suspect that his posting may have been zapped by a zealous moderator, which is quite understandable and laudible, but I hope that Marcel sees the funny side and returns as I would be at a distinct disadvantage without him. Are you still there Marcel?:)
 
We're getting too far off topic for me now. G4UGM and all things analog? Being unable to tear myself away from the early 1960s I still have a small collection of valves somewhere. I may even still have a thyratron in the garage. It was taken from the "Mark 1 Televisor" that we used to watch the Queen's coronation in 1953. Any use for a 1962 printing of the 1961 third edition of The Amateur Radio Handbook, or is that even further off topic?:)
 
An update on my project, now permanently dubbed HoneyPi.

Marcel and I met up and pooled our resources, hardware and skills, so the project is definitely on. We now have enough memory and logic boards for the CPU and also enough switches and button tops to construct the control panel, although some buttons will be slim ex Univac buttons beefed up to the size of Honeywell buttons with a back moulding of clear acrylic. I've made one prototype just to prove that it works. I intend to use the remaining acrylic from the batch to make decorations for the Christmas tree; well, one must keep a sense of proportion.

Currently most of the work is putting together the infrastructure rather than logic. My checks on the PSU capacitors indicate that they are still in good condition after forty years out of use, which is a surprise. Having discovered how many supporting logic boards are needed just to get the core memory working, I am concerned about our shortage of backplane sockets. We only have 200, less than half the number used in the original model 201 CPU, so we may well have to pack the ICs that we are using onto custom-made boards to fit everything in. The memory support logic is the original second generation H200 stuff and we don't want to mess that around more than is necessary. At best HoneyPi is only going to be the illegitimate offspring of the H200, but should be identical performance-wise as we are working to original system timings and specifications as far as we know them.

Timing trials on the logic ICs gave a gate delay in the range 20-25nS, which will make designing logic around the 2 microsecond memory read/write cycle comfortable. We still don't know exactly what the ICs are, although we know their functions, as the Honeywell identifiers, AAAL4, AAAL5, AAAL6, AAFL1, AAVL1, etc are unhelpful but I have noticed secondary identifiers on some made by TI. These are 4295P, 4296P, 4297P and 4298P. I don't know of a 4200 series of ICs coming from TI in the 1960s but maybe somebody else does. Anyway, we know enough about them to use them.

Unfortunately I'm too busy (and probably incompetent) to set up the HoneyPi website at present but will get it under way when we've made demonstrable progress. Meanwhile it's all happening.
 
I look forward to your progress.

Are the ICs PMOS? Do they have negative supply voltage? In the early 70s TI made PMOS ICs for their first calculators. Maybe that's the P in your part numbers?
 
Last edited:
RobS,

Sounds like great progress. It's truly amazing what can be done with enough determination & resources!

I, too, look forward to progress. Maybe a few photos too.

Clay
 
The ICs use +5 volts and have feedback capacitors back to an intermediate connection point within each gate. We assume that they are DTL. The current levels on inputs point to that. I have schematics dated 1969 which use them and some ICs with 1969 batch numbers on them but that's the only clue to when they were introduced and they probably date from earlier.
 
Well, this is what's meant by starting from scratch, the main frame of the mainframe so to speak. I was about to give it to a scrap metal dealer when I relented and put all the parts back to start this project.
Frame 2.jpg
... and here's my stockpile of original Honeywell components. Also Marcel has contributed the memory components but they aren't shown here.
Components 2.jpg
We are taking photos but they'll have to wait until I get that website up.
 
Now we're really starting from scratch. One backplane is too restrictive on the design, so I've removed all the components again and arranged to take the frame to a foundry to get a copy made. That will enable us to put the CPU and memory on separate backplanes as in the original H200 design. The H200 hardly qualified as big iron, but it's certainly small aluminium. I would have loved to follow David Gingery's example and done the work myself, there even being a few crucibles in my garage just in case, but there's just too much for me to do already.

We don't have enough sockets to fill two backplanes, but I am considering several measures which will enable the logic boards to be spread across the two. One measure may be to get additional socket cases made by a 3D printing service. Pins can be transplanted from other unsuitable edge connectors to complete the sockets. I already have a stockpile of over 700 wirewrap pins which will do. Also Marcel believes that sockets from Univac machines may be directly suitable, if there are any going spare anywhere.

It would be poetic to use sandcasting, one of the oldest known ways of making shapes, in conjunction with 3D printing, one of the newest, on this project.
 
Here's the top view. The connectors are in blocks of five. It might look like ten but each board uses two across. The pins are spaced at 0.125 inch pitch. The break-apart provision in the centre suggests that they evolved from earlier eighteen pin connectors. I think earlier boards were narrower than the H200 ones. Marcel said that the eighteen pin connectors are compatible with Univac but the gap between the pairs is a different width. For the H200 it's exactly four pins, half an inch. If the gap is the only problem then one could put Univac connectors side by side with a half inch gap and let each H200 board straddle two.
Socket block2.jpg
 
Here's a bit, several bits actually, of a Honeywell 200 memory.Bits of memory.jpg The picture is rotated 45 degrees to frame up the central core, so the red and green diagonal wires are the x and y axes, the vertical green wire is the inhibit line and the thin red wire parallel to the green diagonal wire is the sense line. That's how I understand it anyway. The quality of the picture isn't too bad considering that the microscope I used is even older than the Honeywell 200, having been bought at a police station sale of lost property by my late father-in-law in the distant past, so I believe. I have some other Honeywell memory planes which have five wires through each core, but they work differently from these and weren't used in the H200.
 
Cool micrograph! - from a fellow glider pilot (Schweizer 2-33, 1-26, Grob 103, Blanik L-13).
 
Appearances can be deceptive. I am no more a glider pilot than Honey Pi will be a real Honeywell 200; both are a matter of creating a convincing illusion. My avatar photo was taken while on a day out with our pensioners' group several years ago. I did get a flight with an instructor though and even took the controls for a while. However, my only other experience of flying was as an RAF cadet in my school days and flying a glider is different from flying aerobatics in a Chipmunk with an engine. The winch launch was definitely memorable though, getting things happening very quickly indeed, just like finding like-minded people on the Internet is.

The backplane is now at the foundry waiting to be copied but it won't be done until after Christmas probably.
 
Back
Top