Starting this as a new thread as recommended, and expanding upon my design plans.

I'm in the process of designing my third SBC. Unlike my previous builds this one will go straight from schematic to PCB; no breadboarding. For this reason, and to avoid death by feature creep like my last SBC, I plan to start with a relatively simple base system. As a secondary goal I am also trying to avoid buying any new parts as much as possible, and instead reusing what I've already got in stock. Since I'm still in the early design phases on this project I want to get some input and advice before I get too far long.

My initial specs look something like this:

- 65816 CPU, initially 4 MHz, with plans to go faster down the road
- 1 MB of RAM (2 x AS6C4008-55)
- Either a flash ROM, or loading bootloader from a microcontroller. Not yet decided.
- Two 6522 VIAs
- NXP 28L92 UART
- PS/2 keyboard port (probably through a separate microcontroller)
- Two NES-compatible game controller ports
- Compact Flash, and/or SD card via SPI
- Four layer board with internal power and ground planes.
- Glue logic in GALs

Once the system is up and running over serial console I want to implement VGA video. This is something I really want to do myself (it has fascinated me since childhood), so slapping on a ready-made solution such as a VGA terminal emulator is not an option. I also do not want the performance hit of trying to run the video (which will include graphics) through a VIA or SPI. I want something that least maps registers (and ideally all of VRAM) to the CPU address space. Thus, I need an expansion connector that provides access to the CPU busses.

When it comes to the topic of running CPU busses off board I've read through Garth Wilson's entire website, and scoured the first 40 or so pages of the hardware forum. From what I have read on this topic it seems like the answer is always just "don't do it". I am hoping there is some middle ground here besides that answer and "become a full EE".

So let's say I want to add just a single expansion connector (I'd prefer two or three but I'll settle for one). The type of connector does not matter to me; I like card edge for the retro feels, but I'll go with whatever works best. I understand having this is going to have a trade-off in the form of a lower maximum speed, but roughly how badly? Would I cap out at 4 Mhz? 8 MHz? Are there some mitigation strategies I can use, like buffering and/or terminating the expansion bus?

Note that I am willing to make some trade-offs. If I have to cap out at a few MHz in the interest of having a system on which to experiment, then I can live with that. At least then I could settle on a final design I like, and build a future integrated system around it.

I'm very interested in this topic too as it almost mirrors my goals in many ways too. I started the opposite way around so I have already made a working VGA board (not the best on the planet, but it's 100% my own creation) and now I'm working at designing the host.

And like you, my takeaway from Garth's (very awesome work) was "don't do it" also, which was a bit dissuading at the time of reading.

But when you are designing two mammoth projects like this (VGA & Computer), rebuilding your entire project really isn't an option. You might do it once or twice, but when things aren't working you'll end up walking away, which no one wants.

When you look at the size of the Amiga 2000 mainboard, it was massive! And had expansion ports galore.



Our projects are miniscule in comparison and would be lucky to be a quarter this size when completed (and probably run at a quarter of the speed too).

I guess when it comes down to it, expansion ports are part of our design decision, they will be in there. I guess the question is more of a 'what to watch out for' type thing.

From my projects perspective, I am looking at a system board with CPU/RAM etc. One or two input devices like Keyboard/Joystick. Then maybe two expansion slots that expose the address, data, signals, and other 'future use' pins. Expansion slots being for VGA and Sound at a minimum.

I actually have one already working VGA controller that I built for my last project. The funny bit here is that originally that project was going to be like this one, with expansion slots, but I got discouraged about adding them and tried to cram everything into one design. Since it was all on a breadboard it became unreliable and difficult to debug, and I finally just tore it all down. For a while though it was running decently at 6 MHz and had a video display (but no keyboard...)

That controller I developed is called the TIVI (TIny Video Interface). It's implemented in Verilog on a TinyFPGA-BX. However it is limited by the resources available on that board; the ice40 chip it uses only has 18K of dual-port RAM, and board does not have enough exposed I/O pins to allow the CPU direct VRAM access. So, it implements VRAM access via registers ala the TMS9918. It's not a bad design otherwise. It will do 80x25 text with selectable FG/BG colors, plus hardware blink, a hardware cursor, and hardware vertical scroll. It also supports a 320x200 monochrome mode in which each 8x8 pixel block can have its own fg/bg colors, kind of how the VIC-II works in a C64. With the hardware text scrolling the text display was pretty snappy, even with the CPU writing 2 bytes per char through a single byte data register.

There was supposed to be a larger TinyFPGA-EX that may have overcome those limitations, but it seems like it's stuck in a "coming soon" state. So, I ended up buying a ULX3S, which has much beefier ECP5 FPGA. That chip has enough DP-RAM for a huge (by '816 standards) frame buffer, and enough I/O pins to easily allow it to be memory-mapped. The dev board also has an on-board HDMI port so I don't need to build resistor ladder DACs anymore. At one point I had the TIVI code running on it and outputting a test image over HDMI, but without an SBC around which to develop it I've temporarily put it on the back burner.

_________________
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?

From my projects perspective, the max I would need is 7.5 MB/s if that's any indication and running a W65C02 at 12.5Mhz. If I can't pull that off, I'm happy to run at half of both those figures (and so on, until it did work - with 1.5MHz being my minimum target).

Well, I need to eat my words, at least a little...I didn't dig quite as far back into the hardware forum's post history as I thought! I found a nine year old thread on bus amplifiers and termination that answered some of my questions. The post is here: http://forum.6502.org/viewtopic.php?f=4&t=2260

Some takeaways i got from that thread:

1. I should be using 74AC or 74ABT logic. I was already planning to do that part (well 74AHCT for my buffers, because I have them already, but I can always swap them out down the road as speeds ramp up)

2. Keep line lengths short. This one is kinda a no brainer. My target size for my board is going to be somewhere in the Mini-ITX (170x170 mm) to Mini-ATX (200x200 mm) range. Assuming the slots are all grouped together and behind their own buffer, the line lengths should be relatively short. Connector choice still TBD but I'm pretty sure I have some PCI connectors around here I got as part of a bag of random electronics from ebay.

3. Possibly add some passive termination to the expansion bus. 220 ohm to Vcc and 330 ohm to ground was the number I saw suggested in the thread above.

With all this in mind my main remaining question for the experts here is: what lines should get buffered? A0-A23 and D0-D7 seem obvious, but I'm not as sure about other bus support signals like my /RD and /WR signals and the Phi2 clock. The clock I'm particularly worried about as I feel like delaying that through a buffer may not be a great idea. For reference my clock will be driven by a 74AC74 flip-flop, and won't be feeding anything directly other than the CPU, the VIAs, and my glue logic.

_________________
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?

Absolutely. And many thanks for pointing all of these caveats out. So much there to consider. And admittedly bus impedances etc, is well over my head right now. With your huge experience, you are able to foresee thing that I couldn't even begin to imagine yet.

The tricky part here of course is that hobby level board manufacturer generally won't have that level of control, since your project is sharing panels with other folks. So unfortunately I think it's going to be some trial and error. With that in mind I think my best best is to try to minimize those trace lengths, in particular keeping the trace delay under the rise time of mysignals. Quoting you from 2012:



Assuming I used ABT buffers here, and that I go with the most liberal termination guidelines, it sounds like if I keep my expansion bus lengths under 15cm or so I will have decent luck, even without termination. I should be able to fit 2-3 slots in well under that, leaving a bit of wiggle room for trace lengths on actual cards, which I will keep to an absolute minimum as well. If things are flakey I can swap down to parts with slower rise times and just limit my board's maximum speed. As i said in my opening post, I'm willing to make that sacrifice if the result is a working board against which I can develop my video controller



I understand, and I most appreciate the help! I currently feel like I'm stuck between a rock and a hard place. I can't do complete development on my video solution without a CPU to try talking to it, so I need something to use as a development platform. My previous attempt did this by building the whole thing out on a breadboard, but it's so unreliable I'd rather not go down that road again...plus the ULX3S is not at all breadboard friendly due to it's 2x40 expansion headers.

In the last 3 years I've built a number retro single board computers without address/data/control buffers that either plug into a small (5 slots) RC2014 backplane and interface to various I/O boards over the backplane or have 3-4 expansion slots on the baseboard. The SBC are based on 68K, Z80, Z280, 6502 processors and every pc board design are 2-layer pcb. The key is making the SBC small, from 50mmX100mm SBC that plugs into a backplane to 100mmX100mm baseboard that has 3-4 expansion slots. When a design is physically small and driving few components, it can be fast, even overclocked 50% (or more) over rated frequency. Termination, transmission line effects are not an issue when the physical board is small.

Make it small (100mmX100mm or less) and most signal integrity problems will go away; in addition, the pc boards are dirt cheap.
Bill

Yeah 100x100 seems to the cutoff for the super cheap boards. My first SBC was that size, so I use that as a mental reference. I'm not sure I can really fit it all of this in that size unless I do go to a backplane approach, which is an option but I am hoping to avoid it. The sizes I mentioned (170x170 or 200x200) seem more realistic IMHO, and if I keep the expansion slots on their own bus in one corner of the board it should have roughly the same effect.

Although not a strict requirement, sticking to the mini-ITX or mini-ATX form factors would be nice as it would let me use off-the-shelf cases. But again I'm not glued to that, and I can always design and 3-D print something. That would actually be cooler if I was gonna have a built-in keyboard, but speaking PS/2 seems like the less painful route to a keyboard since they are ubiquitous, and talking to them isn't that hard. My design will actually offload that to an AVR; I already have code written for that from the previous project.

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

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

As well as Ben Eaters "Worst VGA" thing, I've been following George Foot on YouTube - has some really promising stuff including hardware sprites on a 6502 system (on breadboards). Here is a link to his latest:

https://www.youtube.com/watch?v=A9b6BdymlyE

Might give you some food for thought...

Cheers,

-Gordon

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

Many thanks for digging so deep and unearthing something useful! Things haven't changed so much - advice and wisdom from the earliest days of this forum may still be relevant, and useful.