durgadas311
Veteran Member
I did some work on punch card data processing equipment, although not the 129 Keypunch machine. You can see that here: http://sims.durgadas.com/punchcards/ibm029.html
- VCF South West - June 14 - 16, Davidson-Gundy Alumni Center at University of Texas at Dallas
- VCF West - Aug 2 - 3, Computer History Museum, Mountain View, CA
- VCF Midwest - Sept 7 - 8 2024, Schaumburg, IL
- VCF SoCal - Mid February 2025, Location TBD, Southern CA
- VCF East - April 2025, Infoage Museum, Wall NJ
-
Please review our updated Terms and Rules here
Honeywell 200 resurrection
- Thread starter RobS
- Start date
Chuck(G)
25k Member
Very similar to the IBM 1620 and 1401, but in particular, the 1620.
Fields and records, addressed respectively, from the low-order position and from the high-order position.
On the 1620, there's no way to directly (or indirectly) read the P-counter, though, I suppose, it's possible to guess it by modifying instructions. The calling sequence generally looked like.
Exit: Branch (filled in by caller)
Entry:
...
Branch to Exit
The 1620 did have a "hidden" register that would save the location of the next instruction before a "branch and transmit" instruction was executed, but said register was operated on only by the "branch back" instruction, which made the facility useful only for a single nesting level.
The subroutine calling method was shared by quite a number of machines, including the CDC 6000-7000 supercomputers. ("Return jump" instruction). PDP-8 certainly uses that as well.
Fields and records, addressed respectively, from the low-order position and from the high-order position.
On the 1620, there's no way to directly (or indirectly) read the P-counter, though, I suppose, it's possible to guess it by modifying instructions. The calling sequence generally looked like.
Exit: Branch (filled in by caller)
Entry:
...
Branch to Exit
The 1620 did have a "hidden" register that would save the location of the next instruction before a "branch and transmit" instruction was executed, but said register was operated on only by the "branch back" instruction, which made the facility useful only for a single nesting level.
The subroutine calling method was shared by quite a number of machines, including the CDC 6000-7000 supercomputers. ("Return jump" instruction). PDP-8 certainly uses that as well.
Cool! I'll be trying your emulators.
Looks like the FORTRAN manual you posted (http://sebhc.durgadas.com/hw2000/docs/H200-027_Fortran_Compiler_D.pdf) is the next edition of the one I mentioned upthread. I have not scanned mine yet.
durgadas311
Veteran Member
I should note, that the FORTRAN manual is just a redistribution from bitsavers - for convenience sake. Not a personal collection or otherwise unavailable document. Same goes for nearly all documents I post.
Also, a disclaimer about my FORTRAN and EasyCoder emulations. Neither is a complete implementation of the respective language. I implemented enough to (hopefully) make them useful. Plus, some features require more information, that I don't yet have.
Also, a disclaimer about my FORTRAN and EasyCoder emulations. Neither is a complete implementation of the respective language. I implemented enough to (hopefully) make them useful. Plus, some features require more information, that I don't yet have.
durgadas311
Veteran Member
I'm pretty sure the problem is Java version. I compiled on Java 8, but used backward-compatibility options for Java 7. Java 6 is very old. The error message is odd, and not very helpful, but I would not expect that JAR to run on Java 6. Can you get Java 7 or 8?
RobS
Experienced Member
I have just received a scanned copy of the H200 Main Memory Setup Manual from a contact in Switzerland. This is a very useful document for me as it isn't just a field engineer's guide but the full 49 page instruction manual for setting up the memory when it is first built. Hence it's exactly what I need to ensure that I adjust the unit for optimum reliability ... if it works at all of course. That's yet to be discovered. Some of the driver boards definitely aren't working correctly at present, so it may be a while before I know.
RobS
Experienced Member
I have now installed an Arduino Mega 2560 board as a temporary USB interface between the memory unit buses and a PC, so yesterday I was able to run dynamic tests on every memory location in the two magnetic core stacks to verify that they are fully functional, which they are. These stacks were salvaged from a scrap heap decades ago and were just static mementos until now. Knowing that they are in full working order is an incentive for me to get on with the project work this year. There is much more to do before I can actually store data in the memory unit but this is a positive step forward and good news.
Before I ran the read-write tests I dumped the existing contents of the memories. Magnetic core memories retain data persistently, so there was still data present from the last time that the computer that originally contained them was used decades ago. Although the data was still readable it wasn't possible to extract any meaningful information from it. In fact it was so ingrained into the magnetic cores that it took several read-write cycles to erase it entirely. For a while I thought that the stacks were faulty, but they just needed some exercise after being inactive for so long. At my age I know how they feel sometimes, especially just after the Christmas holiday.
A Happy New Year to all at VCF.
Rob
Before I ran the read-write tests I dumped the existing contents of the memories. Magnetic core memories retain data persistently, so there was still data present from the last time that the computer that originally contained them was used decades ago. Although the data was still readable it wasn't possible to extract any meaningful information from it. In fact it was so ingrained into the magnetic cores that it took several read-write cycles to erase it entirely. For a while I thought that the stacks were faulty, but they just needed some exercise after being inactive for so long. At my age I know how they feel sometimes, especially just after the Christmas holiday.
A Happy New Year to all at VCF.
Rob
RobS
Experienced Member
I have now completed construction of the main memory unit. During testing one of the two core memory stacks produced some erratic results, but my colleague Marcel has others in his collection and will give them to me to try out the next time that we meet up. Apart from any minor problems that might need fixing I consider this phase of the project to be finished. However, the unit will need fine tuning for optimum reliability, but that is best done when the final operating environment is in place, e.g. a proper cooling system and power supplies set at the standard voltages.
Even though the memory unit has a digital interface it is fundamentally analogue technology. It has fifty-two Helitrim potentiometers which set the important voltages, currents and timings and also some components can be replaced to adjust other key factors such as pulse rise times. I now have a copy of the official fifty page set-up instruction manual for the unit, kindly sent to me by a contact in Switzerland, so I know what has to be done when the time comes. Hence the unit is hardly plug and play and no doubt I will spend some time making final adjustments to it.
I am now moving on to design of the control memory and other registers. For this I will use the 1960's Honeywell ICs in my collection instead of the original high speed magnetic core memory stack. I do have one of those but don't have the circuit boards needed to operate it or their schematics and don't relish the idea of designing my own. That device was a differential magnetic memory running at four times the normal speed for a core memory, so the circuitry was critical and it was always the most likely component to fail as well as being the only relatively advanced technology in the Honeywell 200. As the machine was aimed at the bottom of the market most of its technology was relatively simple but inclusion of the high speed control memory was what gave it its outstanding operating speed. By using a stack of flip-flop ICs instead I will have a far more reliable device that can easily be designed to run at the same speed without breaking the constraint of using only 1960's Honeywell 200 technology in my design.
I have dismantled the main memory unit now in order to add the sockets to the other half of the same backplane where the control memory will be located, so it is currently just a pile of circuit boards and a lot of wiring on the backplane.
The project is now getting more interesting as, rather than just getting Honeywell's original circuit boards to work together as they intended, I am now designing my own solutions to the problems that they faced, albeit using only the small scale integrated circuits that they had back then with just two logic gates per chip. The functional modules of the machine will follow their original design as much as possible, but the internal design of each module will be my own. In particular, the timings of all the internal operations will be the same as in the original machine so that it will perform in exactly the same way as that did. I have to suppress the urge to build a better machine than the H200, but just building the equivalent is likely to be enough of a challenge.
Even though the memory unit has a digital interface it is fundamentally analogue technology. It has fifty-two Helitrim potentiometers which set the important voltages, currents and timings and also some components can be replaced to adjust other key factors such as pulse rise times. I now have a copy of the official fifty page set-up instruction manual for the unit, kindly sent to me by a contact in Switzerland, so I know what has to be done when the time comes. Hence the unit is hardly plug and play and no doubt I will spend some time making final adjustments to it.
I am now moving on to design of the control memory and other registers. For this I will use the 1960's Honeywell ICs in my collection instead of the original high speed magnetic core memory stack. I do have one of those but don't have the circuit boards needed to operate it or their schematics and don't relish the idea of designing my own. That device was a differential magnetic memory running at four times the normal speed for a core memory, so the circuitry was critical and it was always the most likely component to fail as well as being the only relatively advanced technology in the Honeywell 200. As the machine was aimed at the bottom of the market most of its technology was relatively simple but inclusion of the high speed control memory was what gave it its outstanding operating speed. By using a stack of flip-flop ICs instead I will have a far more reliable device that can easily be designed to run at the same speed without breaking the constraint of using only 1960's Honeywell 200 technology in my design.
I have dismantled the main memory unit now in order to add the sockets to the other half of the same backplane where the control memory will be located, so it is currently just a pile of circuit boards and a lot of wiring on the backplane.
The project is now getting more interesting as, rather than just getting Honeywell's original circuit boards to work together as they intended, I am now designing my own solutions to the problems that they faced, albeit using only the small scale integrated circuits that they had back then with just two logic gates per chip. The functional modules of the machine will follow their original design as much as possible, but the internal design of each module will be my own. In particular, the timings of all the internal operations will be the same as in the original machine so that it will perform in exactly the same way as that did. I have to suppress the urge to build a better machine than the H200, but just building the equivalent is likely to be enough of a challenge.
Last edited:
RobS
Experienced Member
The links and contact page on my website was entirely out of date, so I have now updated it including the links to this site and thread. That page also now contains my email address for use by anyone who prefers to contact me directly about the project. I have also added a link to the article that I wrote about the project for the Computer Conservation Society here in the UK back in 2014. You may have to refresh the page in your browser to get the up to date version.
At present this thread still remains the sole place where I am reporting progress on the project as I haven't organised my own website to do this yet. Once I have done that I will start posting photographs there.
At present this thread still remains the sole place where I am reporting progress on the project as I haven't organised my own website to do this yet. Once I have done that I will start posting photographs there.
durgadas311
Veteran Member
Great to see this progress! I'll bet you can almost hear the system running now!
Do you think it will be ready for a cameo spot in the remake of "Billion Dollar Brain"? ;-)
Do you think it will be ready for a cameo spot in the remake of "Billion Dollar Brain"? ;-)
RobS
Experienced Member
I can hear it running already actually, but that's just the fans cooling the vintage PSU's.
The problem with "Billion Dollar Brain" is that the key scene portrays cards being read into a model 223 card reader. Okay, so maybe building a 223 from scratch may be within someone's capabilities, but finding a supply of punched cards could prove more difficult. Anyway, there seems to be a misconception that the H200 was the billion dollar brain in the title, but it clearly wasn't. Apart from being a free-standing small mainframe in its own right the H200 was also used as a satellite machine for far larger ones, which is why it was designed with such versatile peripheral capabilities. The scenes in that film evidently portray such a set-up.
At least in that film the H200 shown was a real complete machine actually being operated normally. H200 control panels were very photogenic and therefore appeared in other films where there was no evidence of there actually being an H200 computer as well. In the title sequences of that film they did put a lot of bogus stickers on the control panel and even when the real machine was shown it had a prominent HONEYWELL sticker on the control panel where there shouldn't have been one to provide a little extra point of action product placement.
I'm currently putting together a test circuit to determine the precise operating characteristics of the Honeywell flip-flop ICs that I will be using in the control memory so that I can design the PCBs for it. Sixteen identical PCBs with gold-plated edge connectors will cost a fair bit for my supplier to manufacture, so I need to get the design right. I know the basic logic functions of the ICs but also need to assess the fan-out and line driving capabilities as well as the effect of the feedback capacitors that Honeywell included in their designs. I noticed that in older designs they were included but in later versions of the same boards some were omitted, so there seems to have been some lack of clarity about when they were really needed. Modern digital ICs don't have analogue feedback pins to permit wave-shaping adjustments, so design with them is much simpler, but these ICs were used in circuits on wire-wrapped backplanes with long interconnecting wires and plenty of scope for induced noise.
So far I have discovered that with an 18MHz clock pulse into a flip-flop switching at 9MHz a 33pf feedback capacitor changes the (sort of) square output pulse into a triangular one because the rise and fall times are slowed down so much. Interesting. No doubt at slower switching speeds that might be useful. I can see why the standard logic boards were described as containing "slow gated buffer amplifiers". Oh well, I'd better get back to the workbench to add the next frequency divider stage. The basic cycle clock rate of the H200 was 4MHz, so I should be using a 16MHz crystal anyway but don't have one handy, not that it matters for now.
The problem with "Billion Dollar Brain" is that the key scene portrays cards being read into a model 223 card reader. Okay, so maybe building a 223 from scratch may be within someone's capabilities, but finding a supply of punched cards could prove more difficult. Anyway, there seems to be a misconception that the H200 was the billion dollar brain in the title, but it clearly wasn't. Apart from being a free-standing small mainframe in its own right the H200 was also used as a satellite machine for far larger ones, which is why it was designed with such versatile peripheral capabilities. The scenes in that film evidently portray such a set-up.
At least in that film the H200 shown was a real complete machine actually being operated normally. H200 control panels were very photogenic and therefore appeared in other films where there was no evidence of there actually being an H200 computer as well. In the title sequences of that film they did put a lot of bogus stickers on the control panel and even when the real machine was shown it had a prominent HONEYWELL sticker on the control panel where there shouldn't have been one to provide a little extra point of action product placement.
I'm currently putting together a test circuit to determine the precise operating characteristics of the Honeywell flip-flop ICs that I will be using in the control memory so that I can design the PCBs for it. Sixteen identical PCBs with gold-plated edge connectors will cost a fair bit for my supplier to manufacture, so I need to get the design right. I know the basic logic functions of the ICs but also need to assess the fan-out and line driving capabilities as well as the effect of the feedback capacitors that Honeywell included in their designs. I noticed that in older designs they were included but in later versions of the same boards some were omitted, so there seems to have been some lack of clarity about when they were really needed. Modern digital ICs don't have analogue feedback pins to permit wave-shaping adjustments, so design with them is much simpler, but these ICs were used in circuits on wire-wrapped backplanes with long interconnecting wires and plenty of scope for induced noise.
So far I have discovered that with an 18MHz clock pulse into a flip-flop switching at 9MHz a 33pf feedback capacitor changes the (sort of) square output pulse into a triangular one because the rise and fall times are slowed down so much. Interesting. No doubt at slower switching speeds that might be useful. I can see why the standard logic boards were described as containing "slow gated buffer amplifiers". Oh well, I'd better get back to the workbench to add the next frequency divider stage. The basic cycle clock rate of the H200 was 4MHz, so I should be using a 16MHz crystal anyway but don't have one handy, not that it matters for now.
durgadas311
Veteran Member
Yeah, I got that impression watching the scene with Karl Malden. From what I could follow, he was doing little more than printing out a message from a mag tape, possibly after decrypting it - which was probably all local to the H200... not that there was actually a "billion dollar brain" attached to the H200...
RobS
Experienced Member
No, I doubt that there was. The film was made with the cooperation of Honeywell and that set was quite possibly a Honeywell data centre with many computers in it.
I've now added the next flip-flop to halve the frequency again, so I'm now getting nine pulses in two microseconds. It should be eight if I use the correct crystal, but who's counting? Oh yes, I am. I must add a 16MHz crystal to my shopping list. Two microseconds was the cycle time of the main memory and there were eight distinct time slots within each such cycle. Control memory cycled four times faster, so executed a read-write cycle every half microsecond, which is my design target. The logic gates appear to have around a 25 nanosecond propagation delay, so with just 250 nanoseconds per time slot I have to bear that in mind. Those feedback capacitors really slow response times down though. Perhaps I don't need to provide for them in my PCB design at all.
durgadas311
Veteran Member
I'm guessing the parts are RTL/DTL ICs then? I never worked with that technology, but saw some in the Wang 600/700 programmable calculators. I didn't realize you could still get those parts, or is this recycled chips? Anyway, kudos to having the courage to work in that space! I saw some wire-wrap backplanes (newer tech) that used twisted-pair wiring... did Honeywell use any of that to reduce noise?
RobS
Experienced Member
I am using original ICs manufactured around 1971 althought the designs for many of the boards that they are on date from 1969 or earlier, so they were certainly in use in the 1960s. The original transistorised H200 used discrete diode and transistor logic, as is evident from the schematics. The earlier circuits had single ended outputs with pull-up resistors but the later circuits used push-pull transistor output stages. From my tests on the ICs they also appear to have push-pull output stages. I think that they are more advanced than RTL (I have some old RTL ICs which are much more primitive.) but I have no reason to believe that they are TTL, so DTL is the most likely. I suspect that they are basically integrated circuit equivalents of the technology used in the previous discrete diode transistor boards. Those had feedback capacitors and Honeywell appear to have included them in their DTL design as well as external components, all the rest being within the ICs. The flip-flops are relatively sophisticated, being master-slave devices with three pairs of anded input gates and a "recirculate" pin which causes the current state to stay latched during clock pulses although the inputs can still set the flip-flop. This style of flip-flop logic may be the reason why on the H200 control panel you could turn individual bits on but could only turn them off by clearing an entire display line. Hitting the "Clear" button on the left hand side presumably killed the recirculation of the data in the flip-flops. One might assume that the buttons worked in press-on-press-off fashion, but the majority of them didn't. The H200 was designed to be as cheap and easy to build as possible, for which I am grateful.
The boards in my collection from the 1970's contain in total around 5000 ICs, which should be enough to build the machine. There are only seven main types of IC on them, so I have about a thousand of each of the common types. They each contain two logic gated buffers or one flip-flop, so I actually have around 9000 to 10,000 gated buffers to play with. If I really get short I may be able to add extra diodes to extend the fan-in of some of the gates. I can also add transistors as output boosters to increase fan-out where necessary. I am likely to spend as much time experimenting with the ICs as actually building the machine, but it's all part of the fun for me.
My test circuit is still sitting here clocking up cycles while I'm typing, so I ought to get back to it. My oscilloscope must be getting bored with tracing it by now.
RobS
Experienced Member
Gosh, it's now a year since I last posted to this thread. I'm so busy working on this project that I haven't even found the time to update my website for it either, but rest assured that I am making significant progress daily at present and the enterprise is in good health. I'll try to give more details but don't have time today now, having just written a lengthy post about wire-wrapping in another thread.
I'll be back.
I'll be back.
daver2
10k Member
On a project I am currently working on - we have some wire-wrap configuration posts on one of the cards we are remanufacturing. The PCB assembly firm had to train some operators to perform the wire wrapping...
NASA have a very good article on their wire wrapping Standards and Expectations. They describe the 'brick wall method' that Rob describes as a requirement such that future work/modifications can be accommodated. They also recommend leaving one spare set of turns available on each wire wrap post for future modifications. it's a good read...
Dave
NASA have a very good article on their wire wrapping Standards and Expectations. They describe the 'brick wall method' that Rob describes as a requirement such that future work/modifications can be accommodated. They also recommend leaving one spare set of turns available on each wire wrap post for future modifications. it's a good read...
Dave
durgadas311
Veteran Member
Great news! I've been wondering if the project was advancing.
Probably not of use to you, but I've been working on a "C" cross-compiler for the H200/2000, and have GUN/Unix style assembler/linker working. It is a bit of a odd fit, since the H200 architecture is so foreign to common Unix architectures, but it's already got me making progress on reconstructing some sort of OS for the H200/2000. Of course, if you've only got 2K RAM then there's not much use for an OS. I think you've already got a cross-assembler anyway.
billdeg
Technician
thanks for the update.