I get the impression back planes are a dirty word around here. That said....


Assuming a 4 layer PCB on both the plane and all cards, and assuming, say, 4 MHz max clock with 74HC logic (slower edges), why not?

Howard Johnson's book (High speed design) seems to imply I'm under the speed where even putting ground traces between signal traces on the connectors is needed.

Also, everything I've done so far has worked well on breadboards with long jumper wires everywhere.

Also, weren't all the backplane style boards done back in the day 2 layer boards? Or at the least, wasn't 2 layers common for that?

I guess my thing is this: A back plane design makes so much sense for a hobby computer that is in a constant state of evolution. It seems ith modern PCB construction, and the low speeds involved, I am actually in better shape than the backplanes that were common before. That I have a stable build on breadboards with tons of janky leads says to me a stable backplane build would be no problem.

Finally, anyone here have experience with this? I was thinking of using standard 8 bit PC ISA connectors. Probably more contacts than I need, but I'd just scatter ground connections throughout. I for sure want a connector that is easy to get.

I've never looked at a typical back plane. My idea was to have the processor, ROM, RAM, and address decoding on one card, text video on one, graphics video on one, IO (VIAS and serial chips) on another. Id probably put +12, -12 , 9, and 5 volt rails on the backplane. That way cards could either use them directly, or have their own regulators as an option. Just pondering really.

Ideas?

Adding the ground plane is of course very helpful, but you'll still need to avoid getting the lines super long. Make sure the ground plane is truly a plane, not just a pour with breaks between sections, like islands or peninsulas. It should run between all the pins in your connectors too, not just go around the ends of the connectors and cutting through the middle in a couple of places. You can get away with a lot with 74HC(T), if you'll be using a slow-enough clock speed for it. The faster 74AC(T), not so much.

The 1MHz home-computer backplanes of the 80's were probably all 2-layer; but the industrial-bus backplanes where they needed (and could afford) the added performance were multilayer. I have an STD-bus buyers' guide from 1990 here, and most of the backplanes in it are multilayer, even though STD bus was an economy-oriented industrial bus compared to something like VME. STD bus was usually limited to a few MHz max.

_________________
http://WilsonMinesCo.com/ lots of 6502 resources
The "second front page" is http://wilsonminesco.com/links.html .
What's an additional VIA among friends, anyhow?

I just spun a 4 layer board for a portion of my video board. Never doing 2 layer again, so no ground pours for me. Also, I never put traces on the ground/vcc layers, so they are as unblemished as can be.

Its strange to me that backplanes are so frowned on here, when the means to make them feasible seem so trivial. 4 layer boards cost more, but for hobby one offs, its not that big a deal. Guard traces are also simple to add if wanted. Its well documented here
that avoiding fast edged logic families is good. That said, I did run out of HC counters awhile back, and used a bunch of AC on breadboards with long leads. Not saying I'd recommend it, but I had zero problems. The PCB has HC on it, so I don't go LOOKING for trouble

Also, to me, a backplane should encourage short traces automatically.

Just seems like an obvious route for a hobby computer where you want to make major changes all the time.

I don't see the point of avoiding the VCC layer for signal traces. If there's a low inductance path between cap and power pin, then the VCC layer doesn't carry much high frequency energy, and becomes much less critical.

As far as backplanes are concerned, biggest issue becomes length of traces. If they get too long, it increases chance of crosstalk and interfering reflections. Termination can help with the latter.

I wouldn't say that myself! There is no group consensus, on this or any forum. Instead, there are a large number of opinions, and a very skewed distribution as to who speaks up, how often, and how emphatically. It's quite possible that some of the wisest and most accomplished members never post at all!

It may be that not many backplane projects are publicly started here. One possible reason is expense: you need to make at least three PCBs and need at least two sturdy connectors. When considering that, and comparing it to an SBC, an SBC will always be cheaper now, although less expandable later.

Back in the day, one might need two memory cards full of RAMs to make a good-sized system. Now, all that RAM and more fits in a single DIP. So that certainly shifts the tradeoffs.

Take a look at the RC2014 design: it's been such a successful kit that the designer, Spencer Owen, has given up his day job. While it was conceived as a Z80 system, there are now numerous CPU options including 6502. There's a 5-slot and 8-slot backplane.

I'm pretty sure that RC2014 is a two-layer design, and also that Commodore, Atari, Apple, and Acorn's designs were two-layer only.

Another practical reason not to use backplane systems is that they can be hard to access with probes, when trying to find a fault.

I like backplanes and slots.

Considering most of my SBC designs are inspired from the Apple II.

Attached is my SBC (Potpourri6502) with a CRUDE (but working!) ACIA 65C51 serial card.

I run at 4 MHz and haven't seen any noticeable issues (yet). This is a 2 layer board.

My next design will be 4 layer and run between 1-8 MHz (haven't decided yet).

_________________
Cat; the other white meat.

I think the biggest question with a backplane is what connectors to use. Many of the available options are hard to work with in one way or another, or resemble a modern interface (such as PCIe) too closely to discourage muggles from attempting to fit a modern expansion card in. The ISA bus essentially began life as a backplane, so it's a logical choice, but a simple 0.1" pin header is smaller and might be more suitable for small cards that don't need unplugging often, or where a ribbon cable might want to be fitted.

The other big question might be precisely which signals go over the backplane. The choice might influence whether it's really considered a backplane or just an I/O bus, eg. if some of the address lines are replaced by device select signals. In general I'd assume that in a true backplane, the address decoding logic goes on each slave card, not on the master.

As Arlet says, a rigid backplane isn't "flexible" enough when looking for errors, so Chromatix' suggestion of using a ribbon_cable_backplane is a solution if you are not using a typical 19" rack. Back in those days we use "extender cards" to gain access to the board in question.

The problem whether signals degrade or not depends (beside overloading weak outputs) only to the edge speed. The higher the speed the more ringing and overshoot will appear if not correctly terminated. Using a rigid backplane the first and the last place should be populated first - open stubs are poison for fast edges. Just check if slow drivers (LS or HC) would be sufficient. If so use them. Don't use faster drivers if not necessary.

BTW consider todays cheap and small switchmode voltage regulators to generate the required voltages on board. Only one +15V...+24V supply rail would then be required. See this Recom Application Note on page A-21 how to use them to generate negative voltages as well. I would recommend using a +1V regulator and a linear low-drop series regulator when using this voltage for ADC/DAC and video stuff.


Regards,
Arne

A bit of ringing isn't so bad when it happens outside the setup/hold window. For example, you can generate bus signals with fast logic, and let them ring, while protecting the clock signal with ground shielding on both sides, and good termination. If a nice clean clock edge arrives after the ringing, it should be okay.

The Northstar Horizon, an S100 bus computer only had every other socket populated - You could solder in more sockets, but those empty sockets made it easier to get probes in - so then it boils down to working out what you're testing - the motherboard/bus or a new plug-in card...



Modern SMPS / DC-DC converter units are great now - and I'll mention the S100 bus again - mid 1970's - and it carried +/- 8v and +/- 16v. Great idea with lots of flexibility, but waiting for the old linear regulators to cool before swapping a RAM board was a thing...

Cheers,

-Gordon

_________________
--
Gordon Henderson.
See my Ruby 6502 and 65816 SBC projects here: https://projects.drogon.net/ruby/