What a godawful mess that must have been to debug. I've never used wirewrap, it looks awful to me.
I am trying to imagine what it would have been like to design such a system using only pencil and paper. Going from block diagram to the lowest level, just on big sheets of paper... the pencil sharpeners must have been emptied twice a day.
I made wirewrap boards back in the early 90s and it was extremely tedious, both following the netlist instructions to do the initial work, and then the inevitable debugging when it didn't work. Of course you can have two kinds of bugs - either you made a mistake doing the wirewrap, or your design has a problem! For the complexity of the boards I did, I think it took 2-3 days to do the initial wirewrap, and then days or weeks to debug and get it working.
It was, however, both cheaper and easier than doing a prototype PCB. For that, I'd have to use the institution's darkroom with their flatbed photoplotter connected to a PDP-something that you had to boot from reel-to-reel tape. The plot happened overnight, and then had to be developed next day in the darkroom, and then if you were lucky you'd have transparencies of each layer of the PCB that you could send off to a local company who would etch you a single PCB for a lot of money in a few weeks. Even that wasn't trouble-free, since PCBs can have manufacturing faults, or you could screw up when soldering the components to the board, or your design could be wrong.
It was very rare that I as the most junior person was allowed to go the PCB route. I think for my boards it happened only once on an ECL design that simply wouldn't have been possible with wirewrap. Although I was tasked with doing the transparencies for other team members. Since I was being paid only £40/week through a government benefits scheme, it was much cheaper to pay for my time than to pay an external company.
Also as the other reply says, I used CPLDs a lot which were much faster to iterate. With practice you could pull out the QFP package, put it in the programmer, recompile and upload the new logic, and put it back into the board in an hour. Luxury!
We never used pencil and paper (except for notes). The software for drawing schematics, laying out PCBs, making netlists, and compiling CPLDs was pretty advanced even then. Although all of it was horribly proprietary. No KiCAD for you.
Taught you to check everything in your design early and often.
Forbodingly, the article signs off with "Amiga, please don't join the sorrowful ranks that have wasted technological superiority through marketing muck-ups."
At VCFMW last month, my table was adjacent to Lorraine and her friends.
Ben Heck walked by during setup, and asked me what it was. I was clueless, so we started making educated guesses. The Amiga poster was a start.
I do wire-wrap. This thing is a marvel to behold. It is quite orderly, but could have used colors more effectively.
The three units implement the VLSI chips and the main board of the Amiga that was first shown at CES (I believe.)
Each VLSI is a stack of PCB such as you might get from Vector, with columns of pads for ICs in wire-wrap sockets, buss bars, and edge areas having mounting holes for connectors. The layers are connected by ribbon cables.
(they are not called breadboard!)
Wire wrap is a superior technology. There are no cold solder joints. They are gas-tight.
It is not hard to debug. If you follow some rules, and don't make a spaghetti bird's nest.
Such workmanship can be seen on minicomputers of the early 1970s.
Whole computers were made by wire-wrap around MSI chips. My wire-wrapped PDP-11/10 functioned perfectly thru the 1990s.
Recently, I implemented a microcomputer design in wire-wrap. That was enjoyable!
My design was captured in KiCad, laid out as a PCB, which I translated to perf-board and wire-wrap sockets
This approach is perfect for prototyping, as you can simply add new blocks.
I probably still have my wire-wrap gun in storage somewhere but I would much rather use an FPGA. If you don't mind, answering, why do you still use this tech instead of FPGAs?
As for cold solder joints: no, they don't do that. But when you start making modifications you have to be extremely careful not to cause any damage because tracing a loose joint on a wire wrap board is the stuff of nightmares.
Because I want to play with original chips. Many of them are in DIP sockets and are 5V. Not to mention linear and analog functions! No matter what, there's always wires or traces between packages.
When I want to do FPGA, I start with an IceStick or BlackIce board (Lattice ICE40 FPGA). But then it comes down to interfacing it to some other chip. So if I make a little PCB that needs more than 16 I/O pins at the edge, I've got myself into extra ribbon cable to jumper to more I/Os..
I even made one wire-wrap board with 4 socketed 74LVC245 level shifters that the BlackIce board would plug into. From there, I could do what I wanted.
I agree that damaging your wires is a serious risk. I guess I'm lucky that this never derailed me. I follow the color rule: blue on the bottom layer, yellow on the second. No daisy chains! Rather than unwrapping to get to a hidden blue wire, I can just cut it.
Also I should mention that I'm not doing anything above 25 MHz.
For a clumsy person like me wire-wrap seems even harder than the usual stuff around soldering, that said I'll look into it more because it seems incredibly interesting.
Also it's nice hearing about Ben Heck in the wild, when I first started fiddling with Arduino I watched his YouTube videos on the element14 channel. Though from what I remember he's long since left. Another channel that I remember fondly is EEVblog even though both were about electronics the content itself didn't have much overlap if I recall correctly.
Never seen wire wrapped boards besides photos of this and maybe some other early micro. So of course I had to do a little search and one of the first results has Bil Herd from Commodore (Plus/4, C128...) explaining it.
Thanks for sharing that! Never got to see any pro wrapping.
By the time this stuff started, I'd started forgetting all the hobbyist hardware electronics I'd learned (thinking it would last) and had moved to software ... at right about the time that manu's stopped documenting their internals ... but while disassemblers still existed.
If you really take seriously what you could have done for a home computer if you had started with fully integrated chips, its actually insane.
Imagine if you had an Amiga Chipset and you had combined it with a RISC like chip. If you did that in late 70s with 3.5μm HMOS (like 68k). The resulting system would be insane, in terms of performance to cost. You could outperform minicomputers that cost 10-100x more.
The ARM2 like chip and the complete Amiga chipset seem to have less transistors then a single 68k, so the price off such a system would be very low. And we can see that with the Amiga, what really blows my mind is how cheap Amiga ended up being, an unbelievable achievment.
Its seem the issue really was the the companies that had the resources to do that amount of chip design knowlage and finances were not interest in making a home computers/workstation. Workstation ended up being made by startups who didn't have the resources to do so much costume work. Appollo was a split-off group DEC because DEC was not interested in workstations. IBM was just to slow and couldn't really do prodcut design, and we all know how the eventually got around that problem with the PC. Apple for the Mac did try to do one ambitious chip with VSLI but didn't end up using it.
The split between computer companies and chip design company was just to big to get the needed amount of integration, and there was clearly a lacking vision for what home computer could be. Jobs vision for the Macintosh went in the right direction, but really Jay Mine had the right vision, and he had it because he build a computer for himself. He wanted a home comptuer that was fast, had a proper operating system and enough media capability to run a flight simulator software. Sadly manamgent most of the time wanted him to develop a console and later when they allowed a home computer they didn't share his full vision.
But then also actually plulling this vision off, multi-chip costume design with very few resources is just an amazing achievment. And many of the people didn't even have that much knowlage in chip design, there was a lot of competition for chip design people. Getting into Commmodore where they had the actual semiconductor teams to get these designs over the line was lucky, many other companies who could have bought them might have messed this up.
In a perfect world you add ARM2-like RISC chip, a Sun-like costume MMU to something like Amiga Chipset and you move computing forward by 10+ years. In reality the exact opposite won, a 16-bit PC that had basically no costume design in it what so ever.
When I was a kid I had a Dragon32 and my little brother had an Amiga 500. I thought it was so cool with the demos and the sound but he was always getting worms that spread via floppy disc.
The failure of the Amiga and the near-failure and resurrection of Apple is what makes me believe in parallel universes/alternate timelines more than anything :)
26 comments
[ 3.0 ms ] story [ 58.7 ms ] threadI am trying to imagine what it would have been like to design such a system using only pencil and paper. Going from block diagram to the lowest level, just on big sheets of paper... the pencil sharpeners must have been emptied twice a day.
It was, however, both cheaper and easier than doing a prototype PCB. For that, I'd have to use the institution's darkroom with their flatbed photoplotter connected to a PDP-something that you had to boot from reel-to-reel tape. The plot happened overnight, and then had to be developed next day in the darkroom, and then if you were lucky you'd have transparencies of each layer of the PCB that you could send off to a local company who would etch you a single PCB for a lot of money in a few weeks. Even that wasn't trouble-free, since PCBs can have manufacturing faults, or you could screw up when soldering the components to the board, or your design could be wrong.
It was very rare that I as the most junior person was allowed to go the PCB route. I think for my boards it happened only once on an ECL design that simply wouldn't have been possible with wirewrap. Although I was tasked with doing the transparencies for other team members. Since I was being paid only £40/week through a government benefits scheme, it was much cheaper to pay for my time than to pay an external company.
Also as the other reply says, I used CPLDs a lot which were much faster to iterate. With practice you could pull out the QFP package, put it in the programmer, recompile and upload the new logic, and put it back into the board in an hour. Luxury!
We never used pencil and paper (except for notes). The software for drawing schematics, laying out PCBs, making netlists, and compiling CPLDs was pretty advanced even then. Although all of it was horribly proprietary. No KiCAD for you.
Taught you to check everything in your design early and often.
https://www.atarimagazines.com/creative/v10n4/150_Amiga_Lorr...
Forbodingly, the article signs off with "Amiga, please don't join the sorrowful ranks that have wasted technological superiority through marketing muck-ups."
Ben Heck walked by during setup, and asked me what it was. I was clueless, so we started making educated guesses. The Amiga poster was a start.
I do wire-wrap. This thing is a marvel to behold. It is quite orderly, but could have used colors more effectively.
The three units implement the VLSI chips and the main board of the Amiga that was first shown at CES (I believe.)
Each VLSI is a stack of PCB such as you might get from Vector, with columns of pads for ICs in wire-wrap sockets, buss bars, and edge areas having mounting holes for connectors. The layers are connected by ribbon cables.
(they are not called breadboard!)
Wire wrap is a superior technology. There are no cold solder joints. They are gas-tight.
It is not hard to debug. If you follow some rules, and don't make a spaghetti bird's nest.
Such workmanship can be seen on minicomputers of the early 1970s.
Whole computers were made by wire-wrap around MSI chips. My wire-wrapped PDP-11/10 functioned perfectly thru the 1990s.
Recently, I implemented a microcomputer design in wire-wrap. That was enjoyable!
My design was captured in KiCad, laid out as a PCB, which I translated to perf-board and wire-wrap sockets
This approach is perfect for prototyping, as you can simply add new blocks.
As for cold solder joints: no, they don't do that. But when you start making modifications you have to be extremely careful not to cause any damage because tracing a loose joint on a wire wrap board is the stuff of nightmares.
When I want to do FPGA, I start with an IceStick or BlackIce board (Lattice ICE40 FPGA). But then it comes down to interfacing it to some other chip. So if I make a little PCB that needs more than 16 I/O pins at the edge, I've got myself into extra ribbon cable to jumper to more I/Os..
I even made one wire-wrap board with 4 socketed 74LVC245 level shifters that the BlackIce board would plug into. From there, I could do what I wanted.
I agree that damaging your wires is a serious risk. I guess I'm lucky that this never derailed me. I follow the color rule: blue on the bottom layer, yellow on the second. No daisy chains! Rather than unwrapping to get to a hidden blue wire, I can just cut it.
Also I should mention that I'm not doing anything above 25 MHz.
Also it's nice hearing about Ben Heck in the wild, when I first started fiddling with Arduino I watched his YouTube videos on the element14 channel. Though from what I remember he's long since left. Another channel that I remember fondly is EEVblog even though both were about electronics the content itself didn't have much overlap if I recall correctly.
https://www.youtube.com/watch?v=IXvEDM-m9CE
By the time this stuff started, I'd started forgetting all the hobbyist hardware electronics I'd learned (thinking it would last) and had moved to software ... at right about the time that manu's stopped documenting their internals ... but while disassemblers still existed.
Imagine if you had an Amiga Chipset and you had combined it with a RISC like chip. If you did that in late 70s with 3.5μm HMOS (like 68k). The resulting system would be insane, in terms of performance to cost. You could outperform minicomputers that cost 10-100x more.
The ARM2 like chip and the complete Amiga chipset seem to have less transistors then a single 68k, so the price off such a system would be very low. And we can see that with the Amiga, what really blows my mind is how cheap Amiga ended up being, an unbelievable achievment.
Its seem the issue really was the the companies that had the resources to do that amount of chip design knowlage and finances were not interest in making a home computers/workstation. Workstation ended up being made by startups who didn't have the resources to do so much costume work. Appollo was a split-off group DEC because DEC was not interested in workstations. IBM was just to slow and couldn't really do prodcut design, and we all know how the eventually got around that problem with the PC. Apple for the Mac did try to do one ambitious chip with VSLI but didn't end up using it.
The split between computer companies and chip design company was just to big to get the needed amount of integration, and there was clearly a lacking vision for what home computer could be. Jobs vision for the Macintosh went in the right direction, but really Jay Mine had the right vision, and he had it because he build a computer for himself. He wanted a home comptuer that was fast, had a proper operating system and enough media capability to run a flight simulator software. Sadly manamgent most of the time wanted him to develop a console and later when they allowed a home computer they didn't share his full vision.
But then also actually plulling this vision off, multi-chip costume design with very few resources is just an amazing achievment. And many of the people didn't even have that much knowlage in chip design, there was a lot of competition for chip design people. Getting into Commmodore where they had the actual semiconductor teams to get these designs over the line was lucky, many other companies who could have bought them might have messed this up.
In a perfect world you add ARM2-like RISC chip, a Sun-like costume MMU to something like Amiga Chipset and you move computing forward by 10+ years. In reality the exact opposite won, a 16-bit PC that had basically no costume design in it what so ever.
https://www.amiga-news.de/pics/L/lorraine/FB_IMG_17610203985...