SBC interest?

[cosam]

If you can start a column/thread about "How to Troubleshoot a Dead Computer" (just an example), you would be a star!

Or consider expanding on the (currently rather stubby) troubleshooting articles on the wiki.

Things like the fundamentals of O-scope and how to use it, how to use a voltmeter, where to start, and some step by step guidelines; there would be more people who would join in, and it would be a blast. These are just a few suggestions and you can apply your own interests.

Also great candidates for wiki articles, IMHO. Some work has already been done in the hardware category.

 

[barythrin]

I'm one of those types that would have the interest but can't figure out how to get started on learning what is required to do math or build a basic computer. I've brushed across documents but likely due to my own laziness I haven't read through them or sometimes they're a nice layout of the design but never go into "why" which is what I need to understand to build something.

Anyway, naturally I say build it and think it's great. My problem is they tend to be way over my head in understanding the concepts. I can solder something together from a kit, monkey see monkey do, it's just frustrating when I'm not necessarily learning from the experience.

That's the huge gap I've seen at least. Writing up a layman's version of building a computer and explaining what is going on and why I'm soldering this component and how it works would be awesome. I don't think there's any sort of point to "it's already been done, why would you do that?". That would be like saying "the computer has been built, so everyone immediately stopped and just went with IBM right?"

- John

 

[Chuck(G)]

That's the huge gap I've seen at least. Writing up a layman's version of building a computer and explaining what is going on and why I'm soldering this component and how it works would be awesome. I don't think there's any sort of point to "it's already been done, why would you do that?". That would be like saying "the computer has been built, so everyone immediately stopped and just went with IBM right?"

Isn't that pretty much what's happened? Who uses non-Intel-architecture platforms anymore (aside from embedded apps, that is--where ARM seems to be the equivalent of x86).

But why use a microprocessor at all? isn't that like claiming you made dinner, when all you did was heat up a tray of frozen lasagna?

I'm not quite old enough to have professionally used an all-vacuum tube computer, but I've used computers made up of individual transistors and other discrete components (still have some in my hellbox)--no ICs.

Debating whether to build up from a microprocessor or to use a pre-assembled evaluation kit is like deciding whether or not to grate your own cheese or buy a bag of pre-shredded cheese to put on your frozen pizza. It's a matter of some small degree--and is certainly not "making a pizza" nor is using a microprocessor "building a computer".

Why not build a CPU--even a small one? That would certainly be a learning experience.

 

[leaknoil]

Why not build a CPU--even a small one? That would certainly be a learning experience.

That is a cool idea

 

[vbriel]

Ok, now that I've read all the posts, I'll chime in. As a kit designer I look to design kits of computers that are rare and hard to come by. Most people here have never seen a Apple 1 running, yet they get my kit and see what one was like. That being said, a good kit is one that has a unique appeal that makes it stand out. The N8VEM stands out because of the RAM & ROM drives. It makes for a really cool SBC that doesn't need physical drives.

I think there is plenty of SBC interest, and I plan to design one or two more in my day if time permits. The VC forum is actually perfect because most SBC's are 8 bit format which falls into the the same class of computer as vintage machines.

I personally find it difficult to do large group opinion projects because nobody will agree. Get ideas, choose what you want it to do, build it, see if anybody else wants one. This might sound selfish but I design my boards for me first, if I like it, maybe somebody else will too.

I do think the CPU idea is really a good thought! A kit would be really fun!

Good luck,

Vince

 

[MikeS]

Actually that's why I specifically wrote that I understood the desire to build something like this. I just don't know that the world really needs yet another z80 based SBC. By all means go for it though. It just seems maybe the effort could be better spent on something there aren't already so many of out there.

Well, if the criteria are that it should be better, cheaper and not already out there, then I'd say that almost all the folks on here are wasting their time and could spend their effort better than writing web browsers for XTs, souping up their Windows 3.1 installations, wasting money on 387 coprocs, etc. etc.

If your criteria are ease of construction, simplicity, flexibility, expandability, likelihood of finding the parts in your junkbox, etc. then boards like the N8VEM are indeed better than many of the other boards even though those are cheaper and more powerful, not to mention the fun and cameraderie of being a member of a relatively large group sharing ideas and experience.

 

[ziloo]

....troubleshooting articles.... on the wiki.

........Some work has already been done .....

Our wiki is surely expanding!!!
Aside from the wiki, we must have a Question/Answer thread about
the material that is being added to the wiki. That way, the author/s
know what is missing, what needs more expansion, or what is not
well explained.

Hello scorch.....hellooooooooo....any one.............

ziloo

 

[barythrin]

Right, a CPU kit would be completely awesome, and you're right I often joked with folks in my bachelor days about being able to cook and point out that I cooked some cheerios for dinner the night before and some popcorn and beer the night before that.

I just (and I know scorch is gonna run with this idea anyway lol) didn't want the majority of responses to be "it's been done. why bother?", everything's been done.. why do it? uh.. because I haven't is the correct answer. All of us are here playing around because it's fun, educational, and often because we haven't. That's certainly been my answer when folks are wondering the stupid projects my friends and I have been up to.

VInce's and Lynch's products are all great I love that people are recreating things and working on improving original designs still. It's even better that is IS educational. I mean I actually have the opportunity to solder together an Altair kit and Apple I and Kim-1! How could that ever be possible without that work.

In any event this is becoming a preaching to the choir post so I'll stop. Scorch, I wish I was smart enough to help or produce something but I'm not lol. Eventually when I build everyones kits and get off my keister and read all the books I've obtained I'll start whipping out some things too. I just think it's great for the young folks here to be able to learn this stuff and also someday contribute their own knowledge and efforts to the scene and these projects make that more possible.

 

[Chuck(G)]

So, on this CPU thing, what would be the ground rules? I mean, college kids cook up CPUs in FPGA regularly without so much as warming up the soldering iron.

Does it have to be TTL or is CMOS okay? Can one use programmable logic, such as CPLDs to reduce parts count? Is Turing-completeness a requirement?

 

[Dr_Acula]

I agree Chuck, ground rules would be most helpful.

If it is TTL chips then check out http://www.mycpu.eu/

Or are we talking a CPU based on transistors? Or relays? http://web.cecs.pdx.edu/~harry/Relay/

Or marbles? http://www.allthebestbits.net/a-mechanical-marble-computer/

I hope scorch is still reading these threads because I think there are a number of projects out there that could be built in one classroom lesson by anyone who can solder. And they would be very educational.

 

[cosam]

Interesting! I could quite get into a homebrew CPU project.

So, on this CPU thing, what would be the ground rules?
...
Does it have to be TTL or is CMOS okay? Can one use programmable logic, such as CPLDs to reduce parts count?

Personally I'd say keep it to TTL and CMOS parts and draw the line somewhere around the 74181 mark. If you're going to allow CPLDs, where do you draw the line regarding what you do with them? I think you'd start venturing into black box territory pretty quickly and, assuming the idea would be to learn about how CPUs work, that'd be missing the point. It'd also be nice if folk could build the thing without too much specialist equipment.

If it is TTL chips then check out http://www.mycpu.eu/

The Magic-1 is another nice example.

 

[ziloo]

Personally I'd say keep it to TTL and CMOS parts ...
venturing into black box territory pretty quickly ......
that'd be missing the point.

Right on Cosam, great point :thumbsup:! Having a sophisticated
Pentium running a sophisticated windose program to solve
logic equations, and then pouring it unto an FPGA belongs to
future projects.

ziloo

 

[Chuck(G)]

I think you could go considerably smaller scale than the two TTL designs, particularly if speed doesn't matter. How about a bit-serial or 4-bit word architecture? A move machine architecture is pretty simple to implement and easily extended.

Just tossing some ideas out here.

I tend to think of CPLDs as substitutes for blocks of TTL logic, rather than a way to implement "black box" stuff. If I need an 18-bit shift-left, shift-right register with dual-ported parallel load and sign-extension and tristated outputs, it's a pain in TTL, but easy in small-scale CPLDs. Would it be reasonable to impose a limitation of 44 pin CPLDs with no more than 64 macrocells?

As far as non-IC architecture goes, the old Univac SS80 and SS90 machines (magnetic core logic) or even the NEC Parametron machines are very interesting. ISTR that Seymour Cray also did some military computer work using core logic. But I don't know where you'd find "hard" magnetic toroids anymore.

 

[cosam]

I think you could go considerably smaller scale than the two TTL designs, particularly if speed doesn't matter. How about a bit-serial or 4-bit word architecture? A move machine architecture is pretty simple to implement and easily extended.

Oh yes, I'm not suggesting building anything quite as large as either of those examples. I like the idea of a 4-bit word design; not too complicated and yet still enough similarities with "normal" processors to be able to draw parallels.

I tend to think of CPLDs as substitutes for blocks of TTL logic, rather than a way to implement "black box" stuff. If I need an 18-bit shift-left, shift-right register with dual-ported parallel load and sign-extension and tristated outputs, it's a pain in TTL, but easy in small-scale CPLDs. Would it be reasonable to impose a limitation of 44 pin CPLDs with no more than 64 macrocells?

I totally understand the attraction although I have a niggling feeling that packing logic into boxes might put potential builders off. Maybe some people would relish the challenge of building complex stuff in TTL. I must admit I've no practical experience with CPLDs so I really don't know the ins and outs of using them. I am a little worried that requiring a CPLD programmer, or having to rely on others to provide pre-programmed chips, might discourage people from participating.

 

[Chuck(G)]

I totally understand the attraction although I have a niggling feeling that packing logic into boxes might put potential builders off. Maybe some people would relish the challenge of building complex stuff in TTL. I must admit I've no practical experience with CPLDs so I really don't know the ins and outs of using them. I am a little worried that requiring a CPLD programmer, or having to rely on others to provide pre-programmed chips, might discourage people from participating.

Many (if not most) CPLDs can be programmed with JTAG, requiring little more than a PC parallel port--and best of all--can be done in-circuit, leaving room for lots of playing around without having to remove the component.

I do understand that FPGA is quite another story in some people's minds--mostly involving writing lots of VHDL or Verilog and little else. (OTOH, there have been some breathtakingly fast PDP8 and Z80 implementations done in FPGA). But CPLD is little more than a tarted-up PAL.

 

[cosam]

Many (if not most) CPLDs can be programmed with JTAG, requiring little more than a PC parallel port--and best of all--can be done in-circuit, leaving room for lots of playing around without having to remove the component.

I have to say I'm warming to the idea but I still have my reservations that it's going to make the project less accessible and the design less transparent. I really like the philosophy behind the N8VEM project: all DIP parts and basically nothing but a soldering iron required to get the job done (the exception of course being that you need something to program your ROM but there's no practical alternative to get around that). Maybe limiting things to DIP components would automatically place a reasonable limit on the complexity of CPLDs are the like?

I guess we should wait and see what kind of direction the project takes before making too many major decisions about what are probably pretty minor issues. How about banging some more ideas out regarding the architecture? What kind of instruction set do people want to achieve? Are we going for a microcoded design?

 

[Chuck(G)]

Gotcha.

Personally, I think it'd be a kick going back to some of my earliest programming experiences. Decimal, variable word-length machines...but that would probably be too weird for most people.

 

[cosam]

Decimal, variable word-length machines...but that would probably be too weird for most people.

Heh, yes, probably getting a little obscure there for some ;-) I guess it depends on what kind of crowd you want to attract. I'm sure there are plenty of folk who would love to throw the conventional aside and work on something more unusual. I suspect however that the group of people interested in building something that resembles a "mainstream" CPU, in which they can recognise concepts they're familiar with, is quite a bit larger!

But hey, if it turns out it's just you and me interested in this thing, we might as well do the variable word length one ;-)

 

[Chuck(G)]

But hey, if it turns out it's just you and me interested in this thing, we might as well do the variable word length one

If it's just you and me, I'll do it in an FPGA. At my age, I'd rather crank out Verilog than push solder around.

 

[barythrin]

So my $.01 (I already gave someone else some cents). If I knew how to do a project like this, I would make it a simple design that's somewhat practical for learning/understanding semi-current architecture, or at least one that could be obtained on an easier effort. 4-bit is great, it's simple and practical and translates to other higher architectures without much difficulty.

To me being the uneducated builder of one of these designs, the more I can see the better I like it. I'd feel quite underwhelmed if I built a processor by uploading someones FPGA code into a chip on a board and then "there. you just created a processor!" "..um.. not really.." but as you all point out the BIOS code is a necessary step. I think it'd be cool to document that though and have some educational grounds be taught in that as well (ie explaining what you're typing in and/or implementing, why, etc).

Of course it's a lot of work for these projects but I think the educational benefit is so extremely useful. Essentially I'd love to just sit down with a kit and book that's understandable and go through it step by step similar to a college (low level school) tutorial. That's what I'd love.

- J