Hi again!

I am transferring my VGA board design onto a 100x100 mm 4-layer PCB. To access memory on it, it needs to be fed with up to 14 address lines, 8 data lines, some control lines, interleaved with ground and VCC. I am planning to design a second adapter board that among other things takes a 24-bit serial input signal and converts it to address/data using shift registers (74HC595s). The goal is to have a graphics card I can use my current 6502 SBC (which has no graphics port, only a 65C51 and a 65C22) and for a future SBC maybe create the "graphics port" directly on the motherboard. Fwiw I am currently using the serial interfacing algorithm with my breadboard design and it works very well (shifting out using 65C22 under the control of PHI2), but it is a bit slow (around 60-70 cycles per byte written to graphics memory), so in the future I am aiming for a faster interface.

My current though itsto use 40-pin pinsockets and 40-pin pinheaders for interfacing but I am a bit hesitant whether they will work well or not. For example, an angled pin header on one board and an angled pin socket on another, will they fit perfectly or will there be offsets in some dimension? I have considered using a stacked design (one board on top of the other) but the downside of this I don't get to see all components.

So my question is: How do you go about designing PCB interfaces, where two PCBs are to be talking to each other, and what are your thoughts? Any better connectors/design options available?

As a side note, perhaps very relevant: I am considering integrating the shift registers directly onto the VGA board, thus making it the only way of interfacing. I would then have to make a new PCB or integrate the graphics directly on a future SBC. It might actually be a better option (way less pins in the interface) but I would like to explore the option above first.

I attach a b&w PDF of the current PCB (not many traces added as of yet as the layout is not final) to give an idea of how it looks.

Any of the 'pin header' connectors - which are available male and female - will mate correctly and accurately (though you have to think about pin numbers as you change boards sometimes, and obviously use the same size on all: 2.54mm, 2mm, 1.27mm, 1mm).

Din41612 connectors in their many variants offer one, two, and three rows of pins, and in some cases also high power pins and coaxial connectors. Nice, but expensive. Beware that right angle connectors don't sit at right angles without some external support; the boards will want to lean a bit otherwise.

You might also consider 'feed through' pin header types as are used most famously on Arduino and Nucleo development boards.

Neil

90 degree pin socket and pin header will mate in a flat plane if you connect them together first, lay them flat then solder the pc boards. Mounting the combined boards on a baseplate will help keeping them flat. You can also have another row next to the 90 degree header and socket so additional boards can be plugged in vertically.
Bill

Edit, Few years ago I was exploring a generic 8-bit processor concept with a main board that has CPLD, memory, IO, and mass storage that can accommodate different processors. The processor board plugs in one edge, other edge has prototyping area, and there is a row next to the 90-degree connector that can be populated with a header to accommodate another pc board vertically. It worked fairly well. The generic main board supported 6502, Z80, 6809, 8085, and 68008.

Thanks for the mounting tips! Any particular reason you chose a single row connector over a double row connector in that build?

Mm I was a bit concerned about the laying-flatness of the whole setup.

Stacking a la Arduino using sockets with extra long pins is certainly an option. Having the socket on only one side might having them working against gravity in that case as well though.

I have been thinking about the option to put the card in vertically as well although stability is a concern plus it feels a bit unpractical if there are to be cables connected to it. I am not great at mechanical design so I wouldn't know how to go about creating support ala PCIe-cards (the metallic end which you hold when you push in the card).

Maybe integrating the serial-thing on the board is easiest after all. I will work my way from left to right on the PCB with traces before making the final decision i think

Single row connector is a lot easier to disconnect than double rows. It also takes less pc board space and can be plugged into a solderless breadboard.
Bill

My local supplier don't have enough angled connectors it seems. Trying to find them on Mouser is a pain. Any good keywords to search for?

If you can cope with lcsc.com in China, consider HCTL PZ254-2-20-S and HCTL PM254-2-20-W-8.5 as a matching pair; both right angle so you can mount at the edge of boards for a flat build.

Neil

(just the first I came across; there are plenty of others from the usual suspects: I happened to have an lcsc tab open.)

On Mouser, options might be TE Connectivity 2-826634-0 for the male and Preci-dip 803-87-040-20-001101 for the female.

Female right angle headers seem a bit thin on the ground at mouser, though I restricted myself to 2x20 0.1" pitch. Preci-dip looks as if they might offer single row options, too.#

Neil

Thank you for the suggestions!

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

I'd personally stay away from a 40-long row of header pins, having used them on early prototypes of my project.

Every time they're removed there's a risk of bending pins unless you are very strict about how you pull on the boards. Even when I was careful I had to spend considerable time straightening pins with pliers.

That's why I switched to mezzanine connectors like the TX24-40R-6ST-N1E and its mating counterpart TX25-40P-8ST-H1E. They take up less board space and are easy to connect or disconnect. The connection is mostly pretty stable, though if you plan to tuck the boards inside an enclosure or otherwise leave it connected long term you'll want to use standoffs to mechanically secure the connection as they can get unseated if the whole thing is jostled around enough.

Again thank you for the suggestions! I have been contemplating about this and I think it might be a better design choice to move the serial-to-address/data-pin functionality onboard, thereby reducing the need for external connectivity to only a few serial communication pins + vblank-status pin which are connected directly to the 65C22 on my SBC. One reason is that I think my next SBC will have graphics on board, encouraging me to keep the design simple to keep board size down. Doing so would eliminate the need for a graphics port completely. This would also simplify logic for the CPU accessing the video memory as it can be a part of the normal address decode. (EDIT: using the ridiculously expensive beast that is Renesas 7007 32KB dual port ram I could even have a single RAM-chip doing both code/data, and video RAM)

You could also just use some standard pin headers and connect the boards together with IDC cables, like Daryl does in some of his designs: https://sbc.rictor.org/pics4.html

IDC crimpers and kits are pretty cheap.

_________________
"The key is not to let the hardware sense any fear." - Radical Brad