I know card edge connectors died out with the Dodo's.

But I still have a fascination with them and have always wanted to put 8 (yes, 8 ) of them in my computer. That is, keeping the same spirit as the original Apple II.

Of course, I understand that brings challenges. Especially when you increase the clock speed. I don't want to say it was "easy" for Woz (or, actually Allen Baum who helped him with the slots) to add in the slots. But at the 1 MHz the Apple II runs, a lot can be forgiven. Not to mention it was the 70's.

Anyway, I've always wanted to keep my SBC (version 1.5) in the 1 - 4 MHz range with 8 MHz being top-end. Now, my Windows machine has many slots and runs at 3.5 GHz but there was a team of engineers behind that.

So I started looking on Mouser to create a price list and project plan. So far I'm < $80 for version 2. That is, until I get to the edge connectors! Mouser wants $19 EACH!

Now, maybe I got the wrong kind. But I was going for the same 50 pins connector the Apple used and, IIRC, it had flanges. Flanges are the mounting holes. Correct?

One day I am going to build my "big board" computer running with 8 slots. But I will be buying them on eBay. LOL

So, was Steve Jobs and Jef Raskin correct? Slots are a design flaw? Where's Woz to back me up on this?

_________________
Cat; the other white meat.

Most of the modern interfaces are actually serial in nature, not parallel. That makes things much simpler.

I was pretty amazed long ago when I learned that the DRAM in my Macintosh is accessed serially.

You can look at the modern Thunderbolt interfaces, which are effectively PCI interfaces, but on a few wires vs a parallel broadcasting noise fest.

25 pos (50 pin) 2.54mm Standard Edge II Through-Hole mount edge connector socket, £2.39 from Mouser.

I think you may have been mixing up quantities - they're normally listed in terms of the number of positions, each position being two pins, one on each side of the socket. The Apple II had a 50 pin connector, i.e. 25 position. A 50 position / 100 pin connector is indeed in the £12.xx range and would be enourmous.

I believe flanges/latches refers to plastic dohickies (technical term) at either end of the socket which can be used to hold the PCB card in place. Think the plastic retaining hooks on the sides of RAM DIMMs in modern PCs. You probably do not want these.

I don't think wanting to add a parallel bus to your SBC is "wrong", afterall no-one can tell you what to do on your SBC design, it just makes things more complex due to timing, skew, capacitance and a whole host of other issues. However, as you say, at a couple MHz a the laws of physics tend to not be too punishing. Nor do I think Slots are a design flaw - they certainly weren't in the 80s and the Apple II, indeed the expandability they offered was arguably one of the things that made the platform so successful.

Garth always recommends using modern serial buses buffered from an I/O chip instead of a parallel bus, e.g. SPI. This gives you access to a whole host of modern chips that don't come in Parallel versions, protects your main internal busses from outside schenigans and eases routing.

_________________
Want to design a PCB for your project? I strongly recommend KiCad. Its free, its multiplatform, and its easy to learn!
Also, I maintain KiCad libraries of Retro Computing and Arduino components you might find useful.

Ah! Yeah, I think you're right. I must have just put some wrong info into my query. Thanks!



I totally agree with that idea. But, there is just something about slots that screams retro. Not even just the Apple II, but the IBM and clones as well. I've always loved the utilitarian nature of those designs.

But, yes, SPI is the way to go for modern designs and easier routing. Plus, you're not going to overrun a fast SPI device with an 8 MHz computer.

More than likely, I'm just going to continue adding a few headers for tinkering. But that big board, 8 slot SBC is still calling for me.

Especially since I have a pretty good credit with PCB Way. I could make a large PCB for very little money. Hmmm.

_________________
Cat; the other white meat.

PC and server motherboards continue to be shipped with PCI slots, which are a parallel interface. It's not obsolete technology by any means.

As far as the practicality of slots, you may need to run them at a submultiple of the Ø2 frequency, and use one or more bus drivers and/or transceivers between the slots' data and address lines and the MPU buses. It's usually not very practical to directly connect slots to the MPU buses, unless the Ø2 rate is going to be quite slow.

Also, consideration has to be made about chip selects, IRQs, etc. Designing your expansion bus will be an interesting exercise for you.

_________________
x86?  We ain't got no x86.  We don't NEED no stinking x86!

Slots are great! definitely a good idea especially if you make some matching protoboards with card-edges so you can easily experiment with things.

What strategies do people use for the data buffer output enable?
I'm thinking there are a couple of options like having a specific memory range that always enables the buffer, or have the card(s) drive a signal to the glue logic notifying it that a peripheral address has been selected?
The reason I think of the latter is to have things like a ROM card, some other peripheral etc inhibit the on-board ram/rom so it can use that space instead.

I do dig the vibe with card-edge connectors and they're easy to insert and remove and still get a stable connection with. That said, if I ever roll my own homebrew standard, I think I'd go with 2-row header connectors simply because it's a lot less trouble to make. But an expansion bus of some kind is absolutely a must-have.

I have been dreaming of using some kind of PCI bus to put together several fpga-based boards. If I make one good board that can be used for several tasks, it would be possible to get a multi-cpu computer together in this way. It may end up more like a 68xxx type than a 65xxx type machine.... but with some fpga's, its kinder to the end user since its possible to choose (for oneself). I do like the Amiga-era architecture... but the processor kind of grew into a fat beast at some point (68060).

Well... its a dream. No idea if I manage to realize it sometime. Routing fpga's with hundreds of "pins" is more like a prolonging headache than a nice project.

These days it's more likely to be PCI Express, which is serial. You can have up to 16 lanes, and data will sliced and sent on all of the available lanes in parallel, but each lane is an independent serial bus. They all transmit at the same rate, but the clock is reconstructed independently at the receiving end of each lane, so you don't have to worry about skew. That allows much faster data rates. But it's not something that I would fancy implementing myself.

But anything with a 6502 in it is retro, and if you're going retro, slots are cool.

PCI Express has yet to replace PCI so much as just providing a faster expansion bus for components where very high bandwidth is useful (i.e. GPUs,) in my experience. Most systems, even micro-ATX boards, still come with one or more PCI slots, and plenty of manufacturers still produce PCI cards for various purposes.

I too dig the retro vibe of a motherboard or backplane PCB with cards that plug in, but like commodorejohn I wouldn't create such a design for myself. I would use ribbon cable as the backplane because its flexibility provides more mounting options and better access to any one particular board you might happen to be troubleshooting.

The actual connector is a separate decision. IDC ribbon connectors can mate with a card-edge or with a 2-row header on the daughterboard.

For example, here's photo of Daryl Rictor's SBC-4 computer.

-- Jeff

_________________
In 1988 my 65C02 got six new registers and 44 new full-speed instructions!
https://laughtonelectronics.com/Arcana/ ... mmary.html

That is a really nice computer. I like the solution with ribbon connectors to ease troubleshooting. Makes it much easier than stiff connectors.

I recall reading a white paper years ago (1980s) in which the author asserted that card-edge connectors were better for high frequency work than pin headers. At the time, a number of Motorola 68K-powered minicomputers were using SIP memory modules, which used pins (which often broke off at inopportune moments). Eventually SIPs died out and today we only have memory modules with edge connectors.

A ribbon cable interface is best implemented with alternating grounds. That is, first lead is ground, second lead is signal, third lead is ground, fourth lead is signal, etc. Most ribbon cable has a characteristic impedance of about 110 ohms, so it would make sense to include impedance matching at both ends, such as with a Thévenin circuit.

_________________
x86?  We ain't got no x86.  We don't NEED no stinking x86!

The Micro-ATX PC I built last year (and upgraded this year) has no PCI nor ISA slots, and I would have had to look *very* hard to find one; it has only PCIe, SATA and USB for primary expansion. The board I chose has one concession to legacy I/O: a PS/2 keyboard/mouse port - which is also serial.

You'll likely find it has a Serial port header, too. Most motherboards still have one, though I'm honestly not sure why, because 99% of buyers will never use it in this day-and-age. Maybe for debugging at the factory?

_________________
Want to design a PCB for your project? I strongly recommend KiCad. Its free, its multiplatform, and its easy to learn!
Also, I maintain KiCad libraries of Retro Computing and Arduino components you might find useful.