[Coco] Any news on the so called CoCo4 or NextCoCo projectthatBjork was heading?

Mark McDougall msmcdoug at iinet.net.au
Wed Nov 17 15:52:56 EST 2010


On 18/11/2010 2:47 AM, jdaggett at gate.net wrote:

> One suggestion that I would throw out that has already been suggested.
> Instead of one board do all have several boards that plug into a main
> board.

I'm going to have to disagree with this one. In fact, I would argue that it 
is *almost* a step backwards to the days of discrete chips.

There are three problems I can see with this approach.

1. Connectivity. Any time you take signals off-chip you're constrained by 
the physical circuit and I/O count that have been designed into the system. 
You actually *lose* flexibility because you can't arbitrarily connect any 
two parts of the system.

2. Speed. Moving data on/off chip slows things down, not only in terms of 
raw clock rate, but also the fact that you need to shuffle data across in 
organised bus protocols. Even more-so when off-board connectors are involved.

3. Cost. Several boards with several FPGAs will always be more expensive 
than a single board with a single FPGA.

I can see what you're trying to do - future-proof the design in terms of 
expandability. But that's very difficult to do - I'd even argue impossible - 
the way technology moves so quickly.

It's more difficult than you would think to simply connect two (different) 
FPGAs together these days. With both core and I/O voltages changing and 
coming down all the time, it becomes a nightmare of multiple power rails and 
level translation. As well as the obvious cost implications, this also 
further reduces connectivity. Trust me, been there.

I would argue that it's not difficult to predict the medium-term 
requirements of a Coco4 project. You can get FPGAs now large enough to run 
several complete Coco3FPGA instances at once in parallel. Once you've 
implemented the CPU, there really isn't much that consumes a lot of 
resources. Add the obvious I/O - VGA, PS/2, USB, SD/MMC, some sort of 
joystick interface, some sort of network, sound. Then add 1 or 2 expansion 
connectors for cartridge port (extremely expensive to add) and maybe legacy 
floppy (again, expensive to add for dubious gain).

That'll do anyone for years to come. Beyond that, re-spin the board when the 
time comes. It's likely that technology has moved along far enough that 
it'll be cheaper than trying to expand the original design - smaller, faster 
FPGAs with different power requirements, new media/bus standards, new video 
standards, etc.

And it's not as if the Coco4 project evolves particularly quickly! ;)

Just my $0.02 worth...

Regards,

-- 
|              Mark McDougall                | "Electrical Engineers do it
|  <http://members.iinet.net.au/~msmcdoug>   |   with less resistance!"



More information about the Coco mailing list