PDA

View Full Version : Considering an S-100 "replica" of sorts



Grindar
September 13th, 2008, 07:38 PM
Well guys, new registeree, long time reader here at vintage-computer. Finally getting off my lazy keester and doing something, hopefully.

My dream: As I grew up too late to be part of the computer revolution, I've always had a fascination with these machines. I've been pricing them lately...and I came to the conclusion if I was going to own one, I'd go ahead and get the Altair kit....right, if there's ever another batch of them. So barring that, and loving the thrill of building something myself, I've decided to cobble together a "replica" kit of my own. I use replica loosely here because I'm not reproducing any specific computer, but parts will be duplicated.

For instance, if I can ever find where I put all my documentation, I'm thinking the Cromemco Z80 CPU board as a good start. Also, some random memory card, not terribly important which one. I'm thinking a real-time clock, and a front panel based on the Altair's. I say based, because I'll probably build in a 16-key pad and LCD display for when I don't feel like toggling and interpreting binary.

This will probably be split into 2 boards, the display board and the one with everything else. Space isn't a real concern, and I might change this if I find an enclosure I really like. I'm also planning on having a couple s-100 slots on the back of the board for future expansion, if I can find the parts.

Has anyone tried this before? Any advice on design? Any suggestions for choice of schematics on the major components, like which mem card to use?

I already know that cross-talk is going to be one of my worst enemies in designing this monstrosity, what else should I be looking out for?

MikeS
September 14th, 2008, 07:44 PM
A worthy project indeed!

I've got a fair bit of Cromemco stuff and documentation, but most of the docs are on line anyway.

Sounds like fun; keep us posted if/when it gets rolling.

mike

NobodyIsHere
September 16th, 2008, 06:42 AM
Hi!

This sounds interesting!

What sort of project are you considering? A "from scratch" redesign of an S-100 bus machine or cloning an existing design? What PCB construction techniques are you planning? Do you plan to make your own PCBs or have them manufactured?

Please post more details on your plan. Thanks!

Andrew Lynch

Grindar
September 16th, 2008, 02:09 PM
Well, I plan to reuse any modules that are practical. But if I don't like the design, I'm open to starting from scratch. I'll probably get them manufactured, so I can really take advantage of multi-layer boards. Of course, when I get them completed, I'd be willing to share :D

What I'm looking at doing from scratch already are the front panel/terminal (haven't decided if I want the terminal on the front panel, or attached by a ribbon cable to the main board.

I'm gonna sit down tonight and start going over some schematics, and hopefully start drafting stuff out in Eagle.

I'm gonna start with going through modules that seem like they'd work and make sure I can get all the components fairly easily today. First in line, I think I'll borrow the cromemco z-100 for the cpu...I don't remember any "obselete" or unique parts on it, but that was years ago last time I looked at it.

I also need to look at the power requirements, do you guys think it would be practical to modify an atx power supply for this? or do I need to build or buy one?

NobodyIsHere
September 16th, 2008, 04:00 PM
Hi!

Well, I think you are undertaking a massive and very difficult project. You might want to discuss this with Grant Stockly who did an Altair replica. It was a huge effort, cost a small fortune, and took months. He documents his project on his site:

http://www.altairkit.com/creation_of_a_kit_story.html

As for modifying S-100 power requirements, modifying an ATX power supply might work depending on which one you start with. I think Todd Fischer had some luck with that on the IMSAI replica. Most likely though you'll need separate supplies for the 8v and 18v rails or at least a fairly complex and powerful one to supply the range of voltages at the current ratings needed.

You might want to look into the costs of manufacturing PCBs. The S-100 form factor is about 5" x 10" or 50 square inches of area plus a 50 position edge connector with fingers. Those are going to be expensive units to manufacture even for the simplest designs. Going with multilayer PCBs will drive the costs up exponentially. I expect you are looking at several hundred dollars for even the smallest lot of PCBs. Given the costs involved with the PCBs, I recommend making working prototypes of your designs on prototype boards before you send anything to manufacturing.

Thanks and have a nice day!

Andrew Lynch

Grindar
September 16th, 2008, 05:26 PM
The biggest difference, I think, between the two projects is the level to which he duplicated it, which was a large part of his costs. His enclosure was a custom order...if I use one, odds are it'll be converted from something we have laying around at work.

I expect to spend weeks just building the schematics in Eagle, let alone arranging the pcbs, routing lines...

And as far as PCB costs, I'm gonna do my best to keep it down. One thing I'm gonna try is actually condensing the basic components all onto the same board, leaving hopefully just the front panel as it's own board. It'll be bigger, costing more, but I think minimizing boards will cut the cost more, albeit increasing the complexity....

I forgot about the 8 and 18...I might be able to regulate a 12 to the 8...anybody got any idea where I would begin to look for the transformer if I wanted to build a replica of an old s-100 ps? That's the only thing that worries me as far as that goes...and I really, really don't wanna wind it by hand.

One thought that has occurred to me, i could possibly reduce the level of complexity of the pcb, where the bus would have run, connect the "modules" with ribbon cables. I mean, I wouldn't run power through there, and it's not exactly the neatest solution, but it could save a lot of work.

What's the worst that could happen, I design it, realize there's no way I could afford to build it, and release the files to everyone here? Or I give up halfway done, still releasing if anybody wants it? I've got lots of spare time and nothing to do with it. If nothing else, I'll know far more about the workings of a computer than Microprocessors class ever taught me.

Anyone care to suggest a RAM module? The only one I seem to have printed out is a CDC 64k one, and it's done in banks of 2116s...of course I could always redo the addressing logic and use a larger chip, I suppose. I'm pretty sure I've got another notebook around here somewhere with more of this stuff in it.

NobodyIsHere
September 17th, 2008, 02:25 AM
The biggest difference, I think, between the two projects is the level to which he duplicated it, which was a large part of his costs. His enclosure was a custom order...if I use one, odds are it'll be converted from something we have laying around at work.

I expect to spend weeks just building the schematics in Eagle, let alone arranging the pcbs, routing lines...

And as far as PCB costs, I'm gonna do my best to keep it down. One thing I'm gonna try is actually condensing the basic components all onto the same board, leaving hopefully just the front panel as it's own board. It'll be bigger, costing more, but I think minimizing boards will cut the cost more, albeit increasing the complexity....

I forgot about the 8 and 18...I might be able to regulate a 12 to the 8...anybody got any idea where I would begin to look for the transformer if I wanted to build a replica of an old s-100 ps? That's the only thing that worries me as far as that goes...and I really, really don't wanna wind it by hand.



Hi!

One of the later S-100 systems (CompuPro? Cromemco? I forget but I am sure someone will remember...) used standard regulated PS rails of 5v and 12v on their boards and did away with local regulators. That is a good approach too and then you could use an ATX PS. Making a good S-100 PS is non-trivial and will drive cost due to the number of components and weight.






One thought that has occurred to me, i could possibly reduce the level of complexity of the pcb, where the bus would have run, connect the "modules" with ribbon cables. I mean, I wouldn't run power through there, and it's not exactly the neatest solution, but it could save a lot of work.



I would look into the DIN 41612 series connectors. They are cheap, reliable, strong and easy to work with. They come in a variety of sizes and are fairly standard items.

http://www.jameco.com/webapp/wcs/stores/servlet/ParametricSearchResultsView?langId=-1&storeId=10001&catalogId=10001&jameco_page=46&drill_parents=category_root%24%2430%24%243040&drill_children=30%24%243040%24%24304061&drill_displays=Interconnects%20%2F%20Rectangular%2 0Connectors%20%2F%20EURO%20Type%20(DIN41612)







What's the worst that could happen, I design it, realize there's no way I could afford to build it, and release the files to everyone here? Or I give up halfway done, still releasing if anybody wants it? I've got lots of spare time and nothing to do with it. If nothing else, I'll know far more about the workings of a computer than Microprocessors class ever taught me.

Anyone care to suggest a RAM module? The only one I seem to have printed out is a CDC 64k one, and it's done in banks of 2116s...of course I could always redo the addressing logic and use a larger chip, I suppose. I'm pretty sure I've got another notebook around here somewhere with more of this stuff in it.

As for RAM, I recommend to stick with SRAM for design simplicity and low cost. Designing reliable DRAM boards is quite an art and there is some material from Tim Olmstead (deceased) on the subject. DRAM is highly complex and requires special attention to grounding and other factors. You can get SRAM parts that are dense enough to support any practical S-100 like machine. A board full of 512Kx8 SRAMs should be more than sufficient for an 8 bit machine. There are denser parts available but prices and technologies go up accordingly. Personally, I like these:

http://www.jameco.com/webapp/wcs/stores/servlet/ProductDisplay?langId=-1&storeId=10001&catalogId=10001&productId=157358&

Regarding Eagle, do you have the full version already? The free and hobbyist versions will not support large enough PCBs for S-100 style projects. I think the restrictions are 80x100mm and 160x100mm respectively or something like that. You might want to investigate KiCAD and/or gEDA as they are free and support any size PCB.

Thanks!

Andrew Lynch

Grindar
September 17th, 2008, 06:30 AM
I would look into the DIN 41612 series connectors. They are cheap, reliable, strong and easy to work with. They come in a variety of sizes and are fairly standard items.

http://www.jameco.com/webapp/wcs/sto...pe%20(DIN41612)

I'll keep an eye on these, though as I said, I'm going to minimize board count. I was referring to not having to run 100 bus lines through the pcb from the area where I assemble the ram to the area where I put the cpu(I of course meant the Cromemco ZPU, not the z100) to the ...that will get messy fast, and require probably a couple extra layers. I was thinking of using standard IDE ribbon cables basically as giant jumpers. to be more specific. Think kind of like they could be cut into separate PCBs, if they didn't share power rails...


Regarding Eagle, do you have the full version already? The free and hobbyist versions will not support large enough PCBs for S-100 style projects. I think the restrictions are 80x100mm and 160x100mm respectively or something like that. You might want to investigate KiCAD and/or gEDA as they are free and support any size PCB.

Work proves to be handy...we have a license there, which they allowed for me to put on my laptop. But if I have difficulties with it, I'll definitely give those a shot.


As for RAM, I recommend to stick with SRAM for design simplicity and low cost. Designing reliable DRAM boards is quite an art and there is some material from Tim Olmstead (deceased) on the subject. DRAM is highly complex and requires special attention to grounding and other factors. You can get SRAM parts that are dense enough to support any practical S-100 like machine. A board full of 512Kx8 SRAMs should be more than sufficient for an 8 bit machine. There are denser parts available but prices and technologies go up accordingly. Personally, I like these:

http://www.jameco.com/webapp/wcs/sto...ductId=157358&

I already knew that SRAM would be much easier, if only just for sheer number of available schematics. I think I might have to go something cheaper than your suggested chip, given the anticipated expense of the boards(in the hundreds) and decent toggles already...though that is a pretty big hunk of memory in itself, with just one of them. What I'm hoping to find is schematics of a fairly easily reproducible RAM card for the s-100 bus. Dealing with addressing logic might be fun, but this project will be complicated beyond all belief, and that'd be a major hassle to eliminate.

MikeS
September 17th, 2008, 07:25 AM
A couple of 32Kx8 cache RAM chips from an old 486 MB will give ya the basic 64K.

Speaking of Cromemco, they did have a supplementary bus like you're talking about for the high-end CPU/RAM/MMU combos, with dual row 90 degree headers on the edge opposite the S100 bus, in the same location on all boards, and ribbon cable going across forming a 'bus'. If you're not concerned about bus compatibility that's probably the easiest and cheapest way to go.

m

Grindar
September 17th, 2008, 07:46 PM
Now 486s I should have access to!

Going through schematics, I have one major question. I'll try and pull up data sheets tomorrow, but do you guys think the 74LS376 would be an acceptable substitute for the 8t97 and 74367s I see all over the place? Given it's just TTL signals basically, I don't see the problem right off.

A good sign: The ZPU card only has +5V connections shown, except for the 8V that runs into the 7805, which then goes to the +5v....
Is the 16V always regulated down to 12? If so, couldn't I just leave those off and use the atx supply we talked about before?

NobodyIsHere
September 18th, 2008, 03:42 AM
Now 486s I should have access to!

Going through schematics, I have one major question. I'll try and pull up data sheets tomorrow, but do you guys think the 74LS376 would be an acceptable substitute for the 8t97 and 74367s I see all over the place? Given it's just TTL signals basically, I don't see the problem right off.

A good sign: The ZPU card only has +5V connections shown, except for the 8V that runs into the 7805, which then goes to the +5v....
Is the 16V always regulated down to 12? If so, couldn't I just leave those off and use the atx supply we talked about before?

Hi!

Check the datasheets to be sure but generally I have found the 74LS367 parts to be compatible with the 74367/8T97 parts. The latter can be a bit faster but use generally more power. I have substituted them often though without problem.

From a circuit perspective, you'll almost always see 5v TTL and some 12v rails with a rare -5v section or possibly other voltages. The board regulators are a function of what power is available on the bus and what the circuit needs.

If you have clean regulated 5v on the S-100 bus power rails (highly non compatible but I know some S-100 manufacturers did this towards the end -- CompuPro I think) you don't need the regulators at all. In fact, I have some S-100 boards that came without 7805's etc and I ended up adding them as needed. So yes, if you use an ATX power supply to power the rails on your S-100 bus project you won't need the regulators. However, you won't have a "true" S-100 bus either.

The commonly accepted S-100 bus has a variety of power rails such as 8v, 18v, -18v, etc. Most boards are expecting those to be present. You can build your own system using regulated switching PS but I would make it use a different connector so people aren't accidentally plugging things into non-compatible busses and "releasing the magic smoke".

Getting away from the strict S-100 form factor and power requirements will reduce the cost of your project dramatically. Everything will be simpler and less costly. S-100 boards are large, require special power, and the connectors are non-trivial.

Like Mike S suggested, even using a 5"x10" PCB with a dual row header pins for a bus would be MUCH less expensive. If you did the bus as two 25x2 headers you could use two commonly available 50 conductor SCSI cables for the bus interconnects with IDC connectors. Providing regulated 5v / 12v from an ATX PS would reduce costs and simplify the circuitry tremendously.

I think Thrashbarg did something akin to this on his home brew system. Do a search of the forum to see his works. He did an incredibly great job with this.

However, with the above simplifications, the project has drifted away from any S-100 physical compatibility, however, the underlying circuits could be adapted to the new format.

If the original S-100 bus designers were given a chance to do it over again today, I doubt very much they would have chosen the format, connector, or power supplies they did. I believe/suspect it was what they had on hand at the time which dictated the design more than anything.

Personally, I find the ECB "standard" much more hobbyist friendly and buildable than S-100 since it is 1) smaller format, standard 160x100 mm PCBs 2) standard PS voltages 5v & 12v 3) standard DIN 41612 connectors.

As a result, the making an ECB prototype board is very inexpensive and costs less than $10 ($5 generic prototype board + $4 DIN 41612 connector) whereas getting an S-100 prototype board is difficult to even find and will be much more expensive due to size and the edge connectors. They occasionally go on sale on eBay but usually are at least $20 and I have seen them go for much more -- their availability is rather spotty too.

Your project is to make an S-100 replica machine and I don't want to dissuade you from it. However, I am suggesting if you are going to build a project machine to take a look at some of these considerations before you spend big $$$ on a project or commit to one with an excessive cost. There are ways to make home brew computers economically and still keep the intent.

For what its worth...

Thanks and have a nice day!

Andrew Lynch

Grindar
September 18th, 2008, 09:35 AM
I guess maybe the best way of putting it is I want to make an s100 compatible then, given that the form factor is more or less going out the window. The idea is to integrate a couple real s100 slots onto the "back" of the machine for later, and the core components be electrically identical to an s100 machine, if not in the form factor. Like I said, I used "replica" loosely.

No need to fear of me wasting a bunch of money on it anytime soon, either. I won't spend a dime until I'm convinced the project is entirely feasible, I've compiled a BOM, and in general planned out construction and testing.

To be honest, it's the power supply that worries me the most, for all the reasons we've already covered. I might just watch on ebay and snatch one up. Or I'll go the switching power supply route like the altair replica. How often were those 8 and 18 v lines in the bus used without being regulated down to 5 and 12 is what I'm trying to find out. Tonight, the plan is to go over the zpu schematic again, with a printout of the bus lines handy, and go through each one and check for that kind of thing. Though for cpu and ram, at least, I would suspect it's all just regulated down to TTL voltage, with the exception of a mem card build to handle writing to (e)proms?

I would like to use those connectors for attaching my front-panel at least. Speaking of which, I don't see any reason why that board should be more than 2 layers, so I should be able to etch it myself. This might change, of course, depending on just how many bells and whistles I put on the thing.

And if there's a merciful deity around I can get the "motherboard" to 2 layers with less than a bazillion vias....I could conceivably make that one as well. Conceivably. Likelihood of only having that many vias or only 2 layers? Probably roughly equivalent to the chance of 4chan's /b/ becoming a place of polite conversation.

NobodyIsHere
September 18th, 2008, 10:59 AM
I guess maybe the best way of putting it is I want to make an s100 compatible then, given that the form factor is more or less going out the window. The idea is to integrate a couple real s100 slots onto the "back" of the machine for later, and the core components be electrically identical to an s100 machine, if not in the form factor. Like I said, I used "replica" loosely.

No need to fear of me wasting a bunch of money on it anytime soon, either. I won't spend a dime until I'm convinced the project is entirely feasible, I've compiled a BOM, and in general planned out construction and testing.



Hi!

Good idea. Investigation and research are cheap. Hardware/software is expensive.

You might want to get some price quotes for prototype PCBs. There are many places and I made some here which work fine. Prices are OK for small quantity.

https://www.barebonespcb.com/!BB1.asp

There may be cheaper ones out there but the prototypes will be low quantity and thus have relatively large unit cost.







To be honest, it's the power supply that worries me the most, for all the reasons we've already covered. I might just watch on ebay and snatch one up. Or I'll go the switching power supply route like the altair replica. How often were those 8 and 18 v lines in the bus used without being regulated down to 5 and 12 is what I'm trying to find out. Tonight, the plan is to go over the zpu schematic again, with a printout of the bus lines handy, and go through each one and check for that kind of thing. Though for cpu and ram, at least, I would suspect it's all just regulated down to TTL voltage, with the exception of a mem card build to handle writing to (e)proms?



Yes, the 8v and 18v rails were/are unregulated and almost entirely converted to regulated 5v and 12v rails. Its possible that some cards used them as is but since they represented more of a range of values than an actual level due to their being unregulated. The linear power supplies of the time actually produced voltages which decreased with load and the regulators made TTL circuits possible with predictable power sources.

Even EPROM programming cards like the SD-100 EPROM burner raised its own special voltage levels from the rails through regulators and conversion since the actual voltages on the raw supply rails were rather unpredictable.




I would like to use those connectors for attaching my front-panel at least. Speaking of which, I don't see any reason why that board should be more than 2 layers, so I should be able to etch it myself. This might change, of course, depending on just how many bells and whistles I put on the thing.



That should work fine. Be careful of current loads if there are lots of LEDs though.





And if there's a merciful deity around I can get the "motherboard" to 2 layers with less than a bazillion vias....I could conceivably make that one as well. Conceivably. Likelihood of only having that many vias or only 2 layers? Probably roughly equivalent to the chance of 4chan's /b/ becoming a place of polite conversation.

Yes, it can be done with 2 layer PCBs. I made an ECB backplane (6 DIN 41612 connectors) using 2 layer PCB techniques and it works fine. Two row connectors like S-100 are actually easier than three row like DIN 41612 but in either case 2 layer PCBs will work with a reasonable number of vias.

I use KiCAD and manual routing for the bus backplane. I use FreeRouting.net for the rest of the circuit boards.

Good luck!

Andrew Lynch

Grindar
September 18th, 2008, 11:15 AM
That's actually why I was saying I was considering I was considering running the bus as ribbon cable, as opposed to routing it through the PCB...Should free up some space on the board, lowering production cost.

What features would you guys say are indispensable for a computer like this? maybe on optional segment display that decodes the binary into octal or hex for readability? an octal/hex keypad? A multi-level vector-interrupt? These are the kind of things I'd like to plan for in the beginning, because there's nothing worse in development than creep...other than managers that allow it.

NobodyIsHere
September 18th, 2008, 01:20 PM
Hi!

If it were me, I would just reuse the CCS85 miniterminal design and put an interface socket on the Z80 CPU board. The CCS85 is an already proven design and not reaccomplishing its functions would save time and cost in development. Besides, Rolf is a great guy and has been endless help to me. He is quite the home brew computer enthusiast.

http://www.hd64180-ecb.de/html/boards.html

Personally, I would use a serial UART interface and skip the hex keypad/LED characters. Serial ports are common on PCs and are easy to implement.

If you are going Z80, you can use its IEI/IEO prioritized interrupt scheme for simplicity. Using Zilog peripherals may make life a little easier in the long run too. Those are decisions only you can make though. What do you like?

Thanks and have a nice day!

Andrew Lynch

Grindar
September 18th, 2008, 02:05 PM
I just glanced at it, but I liked what I saw, and would let me pretty easily go back to the having the terminal as a separate bit rather than integrated into the display panel...we'll see how the rest of the design goes.

I plan on integrating a UART, of course! but I'd rather have the keypad for low level tinkering, rather than having to haul the lappy over and deal with it's flaky serial port, or even worse, have to get it somewhere close to my desktop computer(shudder).

I had mostly gone with the z80 to keep costs down, it only runs like $3 last time I checked, as opposed to when I last checked an 8080...It's data sheet is in my download queue now (dialup is a drag, let me tell you).

So, let's see what's on my list so far:

1. CPU - ZPU still looks good at this point
2 . RAM - still an unknown, but I can kludge something together if I have to. Still Researching
3. Power - looking at ebay/switching supplies are best bet, or with careful selection, an atx might do, but unlikely.
4. Display board - simple enough, probably will largely be based on the altair's
5. Serial UART - need to do research on it but what I remember in Microprocessors was pretty easy
6. Terminal - Go over CCS85, make sure it fits my desires, but looking good
7. Case - Not exactly a priority, can probably scrounge up something at work, an old UPS case or something
8. Permanent storage? - what do you guys think on this? simulate a tape drive with the laptop? I've heard rumors that it was fairly easy to hook up an IDE drive to s100, though I wouldn't be aware of where to begin to look.

Oh, interesting note: BG Micro's website came up as having 74376s last night, for about .60, if anybody's interested. Just thought I'd mention it.

NobodyIsHere
September 18th, 2008, 02:37 PM
Hi!

I was thinking about your design plan while out running this afternoon and it occurred to me there may be a low cost solution to make an S-100 bus machine.

How about using a N8VEM SBC and ECB backplane as a foundation and designing a "bridge" board to S-100 bus slots? Basically this would be an ECB card which exports Z80 bus signals to a small S-100 backplane.

You could use the Z80 (TRS-80) to S-100 bus circuit from this book starting on page 115:

http://www.hartetechnologies.com/manuals/Unclassified/S-100_and_Other_Micro_Buses.PDF

You could also use the Dick Smith System-80 S-100 expansion unit (X4010)design guide:

http://classic-computers.org.nz/system-80/manuals_technical.htm

One idea I have been bouncing around for a while is an ECB to S-100 adapter board. The circuitry is pretty basic and could be incorporated into a single ECB card with a ribbon cable to an S-100 backplane. The S-100 could use either its own power supplies for "compliant" 8/18v rails or use a PC power supply for noncompliant 5/12v rails.

Alternatively, you could make a single PCB with an ECB backplane (like 6 DIN 41612 connectors), glue logic chips, and 4 S-100 bus connectors.

This approach has the added benefit of not having to redesign the CPU board in addition to designing the S-100 motherboard. You could build it in phases to spread the cost over a long enough period to keep it affordable. The hardware and software for the N8VEM SBC are available for free at the website.

The N8VEM SBC and ECB backplane PCBs are available now and relatively inexpensive. It would be a worthwhile project just to explore the concept and could be a good source of ideas.

Thanks and have a nice day!

Andrew Lynch

http://groups.google.com/group/n8vem

NobodyIsHere
September 18th, 2008, 04:25 PM
8. Permanent storage? - what do you guys think on this? simulate a tape drive with the laptop? I've heard rumors that it was fairly easy to hook up an IDE drive to s100, though I wouldn't be aware of where to begin to look.



I am making a Disk IO board for the N8VEM ECB right now. It has an IDE section that is already working and software written for it in the CBIOS. It works great and is based on the all TTL Hans Summers design. I use a 1.2GB IDE hard disk on my system for development all the time.

The Disk IO board also has a FDC section which I am working on right now. It uses the i8272 (NEC 765) FDC chip like the IBM PC and some CP/M computers. It is working on my bench right now but is still in development and test. It reads 360K DSDD and 1.2MB DSHD 5.25" floppy disks and 720K DSDD and 1.4MB DSHD 3.5" floppy disks.

Eventually I'll be making a PCB for the Disk IO board but not before shaking out the bugs and getting some semi-decent software for it. I may include a CBIOS depending on time available.

If you go with a S-100 design using the N8VEM SBC as a base you could share peripherals with the N8VEM project further saving costs and development including the Disk IO board and the ECB bus monitor. Then you could focus on what you'd like to work on rather than re-inventing everything from scratch.

Thanks!

Andrew Lynch

RichCini
September 18th, 2008, 04:57 PM
So, let's see what's on my list so far:

1. CPU - ZPU still looks good at this point
2 . RAM - still an unknown, but I can kludge something together if I have to. Still Researching
3. Power - looking at ebay/switching supplies are best bet, or with careful selection, an atx might do, but unlikely.
4. Display board - simple enough, probably will largely be based on the altair's
5. Serial UART - need to do research on it but what I remember in Microprocessors was pretty easy
6. Terminal - Go over CCS85, make sure it fits my desires, but looking good
7. Case - Not exactly a priority, can probably scrounge up something at work, an old UPS case or something
8. Permanent storage? - what do you guys think on this? simulate a tape drive with the laptop? I've heard rumors that it was fairly easy to hook up an IDE drive to s100, though I wouldn't be aware of where to begin to look.

Oh, interesting note: BG Micro's website came up as having 74376s last night, for about .60, if anybody's interested. Just thought I'd mention it.

I like the idea of changing the form factor. I know you want to start from scratch, but I wonder if it would be easier to add a bridge board to something like the P112 to get your S100 bus? Just an idea from someone who's designed a couple of 6502 things from scratch...

Regarding the drive, you can use the Altair Peripheral Emulator. It works over a 9600 baud serial link connected to a host PC (I use the same laptop for the console for my IMSAI). You can work with Frank Barberis on it to create a custom boot disk image for your system. You give him the hardware specs and he'll prepare a CP/M 2.2 image.

I've gotten a lot of my vintage chips (regular TTL and S versions) at BG Micro. Great place to get that kind of stuff. Unicorn Electronics is another to consider as well.

RAM is easy because you can use high-density SRAM. I have a schematic on my Web site for a memory board design using high-density RAM. Look under random projects.

Grindar
September 18th, 2008, 06:01 PM
I like the idea of changing the form factor. I know you want to start from scratch, but I wonder if it would be easier to add a bridge board to something like the P112 to get your S100 bus? Just an idea from someone who's designed a couple of 6502 things from scratch...


How about using a N8VEM SBC and ECB backplane as a foundation and designing a "bridge" board to S-100 bus slots? Basically this would be an ECB card which exports Z80 bus signals to a small S-100 backplane.

pretty similar ideas, and not bad. They sound intriguing, and alleys to be explored if what I want to do doesn't pan out. I'm going to end up with something like an SBC, anyway, with a display and expansion slots. It'll be a large, expensive board, but really, isn't it just as expensive to get 2 or 3 different boards for the same circuits?

Now that EBC to S100 circuit, how easily could it work the other way? Adding EBC slots to an s100 computer? I would think it's mostly just matching up i/o lines, right? I'm not familiar at all with the EBC, and my Google Fu is weak tonight, apparently.


Then you could focus on what you'd like to work on rather than re-inventing everything from scratch.
Well, the nice thing about "re-inventing" it is that I understand every bit of it from the ground up. The only reason I choose to use someone else's schematics at all is because I don't think I've had enough classes in electronics to design it completely from scratch, though I've had enough to know what the components are doi8ng where they're placed, to an extent. I'm using this as a way to teach myself. I want to understand what every "wire" is there for...even if I never physically build it, I learn. That's why I'm getting ready to print out these ZPU schematics on 11 x 18 and hunt down a seemingly reliable version of the s100 bus and spend probably an hour or two just making notes on which lines on the schematic do what...though this'll probably wait until I finish downloading all the datasheets for the components. Darn this dialup! I live like 3 miles away from where I'd be able to get DSL...and no cable, either.

NobodyIsHere
September 19th, 2008, 02:38 AM
pretty similar ideas, and not bad. They sound intriguing, and alleys to be explored if what I want to do doesn't pan out. I'm going to end up with something like an SBC, anyway, with a display and expansion slots. It'll be a large, expensive board, but really, isn't it just as expensive to get 2 or 3 different boards for the same circuits?



Hi!

Well with either approach you can build something and get it working (N8VEM or P112) which is worthwhile. I recommend an incremental build rather than doing everything all at once. I think building on an existing platform and leveraging it may be a lower risk, lower cost method than a pure "from scratch" ground up approach.

Making a large PCB is expensive and risky. Making several smaller ones is also expensive but in a different way. The difference is by using several smaller boards you can better customize and isolate problems easier. With a single large PCB, you have to design everything all at once *and* get it right the first time since there is no going back. Changing the circuit in a PCB after it is manufactured is difficult and messy. It can be done with "cuts and jumpers" but I try to keep those to a minimum.

The WaveMate Bullet and Ferguson Big Board are "all in one" SBCs and they are fine but represent some high end expertise to create such a complex devices.





Now that EBC to S100 circuit, how easily could it work the other way? Adding EBC slots to an s100 computer? I would think it's mostly just matching up i/o lines, right? I'm not familiar at all with the EBC, and my Google Fu is weak tonight, apparently.



Yes. S-100 to ECB would be similar but I don't have a circuit schematic. Since ECB is basically just the Z80 bus fully buffered it would be easy to just export the Z80 pins to an ECB connector. Due to the timing issues, it would probably be more difficult to go straight from S-100 to the ECB but I am sure it is possible. I don't have a design to do that though and have never heard of or seen one either.





Well, the nice thing about "re-inventing" it is that I understand every bit of it from the ground up. The only reason I choose to use someone else's schematics at all is because I don't think I've had enough classes in electronics to design it completely from scratch, though I've had enough to know what the components are doi8ng where they're placed, to an extent. I'm using this as a way to teach myself. I want to understand what every "wire" is there for...even if I never physically build it, I learn. That's why I'm getting ready to print out these ZPU schematics on 11 x 18 and hunt down a seemingly reliable version of the s100 bus and spend probably an hour or two just making notes on which lines on the schematic do what...though this'll probably wait until I finish downloading all the datasheets for the components. Darn this dialup! I live like 3 miles away from where I'd be able to get DSL...and no cable, either.

Yes, few things will teach like building your own home brew computer. I have learned much with TestPrototype and now the N8VEM project.

Good luck with your project and let me know if I can help out.

Thanks!

Andrew Lynch

PS, the P112 (as is the SB-180) is a fine SBC and they are very nice, however, it is my understanding they are not available. Possibly you could find a used one for sale or incomplete kit or broken unit? If so, please let me know as I would like to get one myself. However, I do not think they are available and have not ever seen one for sale in any form. Please let me know if this is incorrect. Thanks!

Grindar
September 19th, 2008, 10:51 AM
Well, small update, I do believe I've found how I'm going to make my enclosure:

http://www.geocities.com/homemadecpu/aidil/index3.htm

It's just so...beautiful...And by the looks of it, fairly cost-effective.

Of course, I probably won't have as many output LEDs as the Magic-1...

This weekend should work out well for research, I have nothing that I know of planned.

NobodyIsHere
September 21st, 2008, 05:55 AM
Hi!

The home brew CPU case is definitely nice but personally, I would concentrate on making the electronics work first. I do all my stuff with bare boards. Once the system is more fully functional I'll start doing something about case.

My current thinking for cases is to re-use an old mini tower case and drill some holes in the bottom for nylon stand-offs. Then mount the ECB backplane to the stand-offs. By using DIN 41612 and ECB format boards they are strong enough to stand freely without extra mounting. That should be a good temporary solution.

At some point in the future, I may start looking at a card cage design but that is so far off as to be not on my list yet. Then we'll need a metal bender or plastic forming to make the guides. Cases are nice but only if they contain working computers IMO.

Thanks and have a nice day!

Andrew Lynch

Terry Yager
September 21st, 2008, 09:53 AM
Cases are nice but only if they contain working computers IMO.

Thanks and have a nice day!

Andrew Lynch

Aww, ya mean I should just scrap my "collection" of stripped-out cases? I was kinda hoping to donate them to a Computer Case Museum someday :evil:.

--T

NobodyIsHere
September 21st, 2008, 01:50 PM
Hi Terry!

There is a computer case history museum? ;-)

I suppose there are a few exceptions... The original wooden NorthStar Horizon is a nice case and the original IBM PC/XT/AT. Maybe some of the S-100 crates like the VG, CompuPro, or Cromemco.

However, most cases are of the XT/AT clone variety and are barely worth anything except maybe as scrap metal. They take up a lot of space and are just in the way during repairs. In recent years, the quality of the cases has deteriorated and are now very thin metal and weak. I have sent more than a few to recycling after stripping out anything of value.

Thanks and have a nice day!

Andrew Lynch

Grindar
September 21st, 2008, 04:33 PM
Well, of course :cool: I mostly just posted that link here so I can refer back to it later in case of hard drive failure in the lappy.

Now an idea occurred to me yesterday. Those DIN 41612s, what if I were to use two of them per card, and do the multiple cards? I didn't see it, but I guess they're about 4" wide, assuming the pins are .1" apart, right? Anyway, the thought is, 2 of them side by side, and for running bus lines, just ignore the middle row. Eliminates a lot of work routing lines to them, simpler to solder, reducing my chance of screwing up, retains the mechanical strength of the wider connector. Would force my card to be at least 8 or so inches wide at the minimum, but maybe it'll be made up for by a lower profile.

Your thoughts on this, before I start playing around in eagle?
Also, anyone got a source on s100 slots for integrating onto the m/b? OR I could possibly rig up some kind of adapter that would run from the 41612s presumably maybe a couple couple inch long ribbons, which are on a pcb that is basically just an s100 stab, to be placed into a real s100 motherboard. Maybe a real s100 extender could be modified. Any input on this is welcome too!

NobodyIsHere
September 21st, 2008, 05:00 PM
Well, of course :cool: I mostly just posted that link here so I can refer back to it later in case of hard drive failure in the lappy.

Now an idea occurred to me yesterday. Those DIN 41612s, what if I were to use two of them per card, and do the multiple cards? I didn't see it, but I guess they're about 4" wide, assuming the pins are .1" apart, right? Anyway, the thought is, 2 of them side by side, and for running bus lines, just ignore the middle row. Eliminates a lot of work routing lines to them, simpler to solder, reducing my chance of screwing up, retains the mechanical strength of the wider connector. Would force my card to be at least 8 or so inches wide at the minimum, but maybe it'll be made up for by a lower profile.



Hi,

DIN 41612 connectors come in a variety of pin counts. Some ECB configuration use the 96 pin type and just use the outer rows. That gives 64 pins which is quite enough for a Z80 style bus.

You could use an entire 96 pin to route all S-100 pins (several pins are unassigned or duplicates). I have had no problems routing all 96 pins with KiCAD using a 2 layer board.

When I did the ECB backplane, I needed to manually route the traces due to complexity but still it was no problem with a two layer PCB.

I think VME is similar to what you describe as it uses two Eurocards side by side. There are extensions which use dual DIN 41612 connectors.

If you want to preserve the routing of the S-100 card maybe the dual 50 pin header approach may work better. I think the Heathkit "Benton Harbor" bus does something similar as it is a single row of pins the card connects to.





Your thoughts on this, before I start playing around in eagle?
Also, anyone got a source on s100 slots for integrating onto the m/b? OR I could possibly rig up some kind of adapter that would run from the 41612s presumably maybe a couple couple inch long ribbons, which are on a pcb that is basically just an s100 stab, to be placed into a real s100 motherboard. Maybe a real s100 extender could be modified. Any input on this is welcome too!


Well, try lots of configurations until you find one that suits your preferences.

There are S-100 compatible edge connectors available but you'll find they are not easy to find. Try Unicorn Electronics as I think I bought some replacements from them a while back.

Thanks!

Andrew Lynch

PS http://catalog.tycoelectronics.com/catalog/finf/en/c/11535/10054/290?RQS=C~1%5EM~BYPN%5EPID~213499%5EPN~1612163-4

if you do find a supplier, please post the URL here as I am sure there will be lots of interested hobbyists.

Grindar
September 23rd, 2008, 01:03 PM
There was a possible match at unicorn for a wire-wrap connector, wasn't much about it.

If anyone knows how to enhance the image enough to make it legible, the brand name on the side of the connector on the altairkit motherboard is just too blurry to be read...I can make out either electronics or electronica as the last word, and highly doubt that it's electronica. Maybe one of you can guess the manufacturer from what's visible? I'll email Mr. Stockly tonight, if I remember, and see if he'll share his source with us.

UPDATE: Found it in another picture. It's from www.sullinselectronics.com, part number EBA50DRTH. Digikey sells them, at 16 or so bucks a pop, in 10 packs. A bit more than I want to spend on connectors, but if I can split an order with somebody later on, that won't be bad. I just can't justify 160 dollars on connectors when I only need 2 or 3 tops.

I've spent the last couple days flipping through catalogs, looking at connectors. I think I'm gonna go with my double-wide 41612 idea. Lets me pretty easily preserve the bus, while not forcing me to solder that middle row, which is what really terrifies me, my hands tend to shake when doing that precise of work.

The backplane is a secondary concern for right now, so I have time to find the s-100 connectors, and if I don't, I can always leave them off, for a future rendition of the backplane.

So I can make sure I don't forget anything, what do you guys like to have on the front panel. Address and data value switches are obvious. LEDs for address and examine would be standard too, I suppose. Another row of LEDs to indicate value of switches? LEDs are a lot cheaper now, better to have a couple too many than kick myself later.

Tonight's a coin-toss, start digging through memory schematics til I find something I like or can modify pretty easily, or start dropping stuff into a schematic in eagle. Really, it's gonna hinge on how much longer I can look at a computer screen today. Been working with group policy all day.

Druid6900
September 23rd, 2008, 08:29 PM
I was thinking that you could put the connectors on the top of the boards and use old 50-pin SCSI cables (which must abound) to distribute the signals and a strip PCB running perpendicular to the channels holding the cards to distribute the power. The strip would also allow you to use smoothing caps.

Grindar
September 24th, 2008, 05:00 AM
I'm designing the m/b anyway, smoothing caps are an option...

I elected not to do it the original way I was thinking with the ribbon cable bus because I had forgotten to consider the distance the clock would have to travel. It could be minimized, but there were better alternatives.

With your idea, I'd have to still have a mechanical mount on the "bottom", which means I'd still have to set up a similar connector or at least a bracket on the bottom. Not to mention I'd still have to dream up an s-100 adapter...which I insist on doing, so I can pick up parts like a floppy drive/controller later and essentially plug it right in.

Just curious, anyone know where to find documentation of the Extensis ex3000? Google didn't turn up anything, and though I wouldn't want to implement it, I'm curious to see how they rigged the multiple processors on the bus "in action"

Another thought for displaying status of switches: lighted toggle switches. I know I saw some somewhere...

Chuck(G)
September 25th, 2008, 02:24 PM
If I was after recreating something with a standardized bus, I'd probably look at Multibus I before S-100, unless I had a lot of S-100 cards already available.

Multibus is older than S-100 and better designed. Instead of the 100 pin edge connector it uses an 86 pin connector with wider "fingers".

If I wanted a smaller profile, I'd probably pick STD Bus.

Grindar
September 25th, 2008, 02:37 PM
If I was after recreating something with a standardized bus, I'd probably look at Multibus I before S-100, unless I had a lot of S-100 cards already available.

Multibus is older than S-100 and better designed. Instead of the 100 pin edge connector it uses an 86 pin connector with wider "fingers".

If I wanted a smaller profile, I'd probably pick STD Bus.
Reply With Quote

Profile's not really an issue, for except considering the cost of the pcb in square inches :mrgreen:.
I'm fabricating my own enclosure, and most of my own cards. I'm sticking with the s100 because I fell in love with the altair, and when I finally get enough $$ I either want an original or a replica kit.

I'm not bothering with card edge connectors for anything I'm designing, going with the connectors that we talked about for like an entire page, so finger width isn't a concern. I'm only attaching a few s-100 slots for when I can scrounge up some vintage equipment (gotta be lurking in a closet at work somewhere).

NobodyIsHere
September 25th, 2008, 02:43 PM
Hi!

MultiBus I *is* a much better designed bus than S-100. It is well thought out, extremely reliable, and generally very tough. On my previous job they had original MultiBus I machines running in the labs under very heavy duty cycles and in a rough environment for many years. We bought spares but the cards almost never broke. They were/are awesome machines and are no doubt still chugging away today. I'll bet those machines are at least 20 years old and maybe older than 25. Nice!

MultiBus, S-100, and STD all use edge connectors so making prototypes is expensive. Also getting the connectors themselves is no easy task either. As a result, I recommend either ECB bus which I already mentioned or the STE bus since both use DIN 41612 connectors. As a result making your own prototype boards will be much less expensive.

http://en.wikipedia.org/wiki/STEbus

Grindar, have you considered just building an N8VEM SBC and ECB backplane system and then building some peripherals? I sure could use some help on the project and if you are interested in home brew computing, that would give you a big head start for low cost. It is also something you can do now rather than in months.

Your ideas for a display and input terminal are good. Please consider and you are welcome to join up on N8VEM regardless.

Thanks!

Andrew Lynch

http://groups.google.com/group/n8vem

Chuck(G)
September 25th, 2008, 10:21 PM
There was a low-cost hobbyist system in the 70's that I recall. I believe it was called "Jolt" made by a couple of guys in Los Altos. It used ribbon cable as a bus to hook together various boards that were simply stacked with threaded through-rods and spacers.

I would have gladly traded my MITS 8800 and my Integrand S100 box for a Multibus MDS-80 any day of the week in a flash. Andrew's right that those things were industrial-strength. I recall that a few friends and I had a discussion as to whether there were more S-100 or Multibus boxes made. After I pointed out that Sun, SGI and HP/Apollo used Multibus in addition to Intel, it was no contest. I think one can still buy Multibus boards.

A previous poster remarked that DRAM is not to be toyed with lightly. My Altair had 2 4K DRAM boards in it; used the TI 4Kx1 DRAM chips. It was miserable--dropped bits every time you hit the reset switch.

You know, designing the whole affair using SMT is a very viable option--why not build a replica half- or quarter-size?

Grindar
September 29th, 2008, 11:20 AM
Sorry guys, work's been keeping me busy. Darn these kids and their desire to wreck everything the District owns! OK, enough griping, on to computers!

You know, designing the whole affair using SMT is a very viable option--why not build a replica half- or quarter-size?
Because SMT is terrifying to me. My hands get jittery, ironically enough, the more control I need. I can handle DIPs fine in short bursts, but it gets bad after a little while. I am planning on picking up one of those SMT practice kits I've seen around, but I'd not like to build a project this complicated as my first real experience with it.

I got in a couple s100 edge connectors saturday in the mail, so that more or less finalizes it. A colleague had some stored with his electrical gear at home (across the state) so he had his parents mail 'em on over. I have 2, if I need more I'll have to buy them or scrounge them.

I'm actually satisfied with the bus I selected, even if isn't industrial-strength, and it's rather picky to build it for...all the more endearing to me.

When you first mentioned it, I checked out the N8VEM, something just didn't click with me, unlike this ZPU...I can't put my finger on it. I am considering a move to KiCad before I get started with any serious circuit drawing...how is it compared to Eagle? I rather like how eagle works, but I can put KiCad on my ubuntu box at work for those odd afternoons the kids don't destroy something, and we don't have a single bad port take a campus offline(seriously, 3 bad superstack ports in one week, on 3 different campuses?)

Backplane, given my current supply of of s-100 connectors, I'm thinking those 2, and 6 of the double-Din-41612s I've been planning on, 3 for currently planned cards, the other 3 for later expansion with custom-made cards. I shouldn't need more than that, right? Or would you recommend kicking it up to 12 total slots, being 10+2.

Grindar
October 10th, 2008, 07:48 AM
Not dead, just busy...mid-terms.

NobodyIsHere
October 30th, 2008, 09:27 AM
Hi! Long time, no hear! What's the latest on this project?

Thanks and have a nice day!

Andrew Lynch

Grindar
October 30th, 2008, 11:00 AM
Poorly, though not due to any element of the design. Just haven't had much time between work, class, and homework. This weekend's looking good and open though.

As of right now, have the z80 and some of the addressing logic done (really, really haven't had much time for anything...got a pile of pre-ordered video games waiting for their turn too). Really not too shabby given the amount of free time I've actually had and the fact that I'm learning how to use Eagle at the same time.