[time-nuts] Form Factor and such, Big Picture

[Chris Albertson]

Bob,

That is what I'm thinking also but when you say the main board has
nothing on it but a few connectors then why have a main board?

All the functions of the main board can be placed in a plug-in that is
easy to change as technology moves.  Rather then a main board you have
just a cable, maybe a ribbon cable for control and status and coax for
any analog signals.

The problem with a backplan is that they are physically very hard to
build.  You need a card cadge of some kind.  I think we want this to
be build-able by amateurs at home.  If you use a "flexible backplane"
(aka, "cable") then the mechanical design is very easy, you could
screw the cards down to a bread board or mount them ina 1U rack
chasis, put them in a diacast Hammond box or whatever.  But a ridgid
backplan would dictate only one mechanical design

The "soft backplane can evolve.  Right now people want a "counter" and
that application does not need any high speed data on the bus.  Even
100Kbps is overkill.  When you start taking about spectrum analyzers
then you need high speed data interconnects and we can add that when
the time comes.  You can't do that if the bus in a PCB with
connectors, you are stuck with what you have so you end up running
card to card cable jumpers.  May as well just start with the jumpers
from the beginning.

There are some really great embedded processors out.  The problem is
software.  If you pick some "exotic" processor then you will be the
only one to write software for it.  To avoid that I'd pick one that
"most" people already know. The bottleneck in these projects is always
software and you should design hardware to minimize that problem even
if it adds $20 to the cost.

Moving the processor into a "plug in" module seems pointless to some
people who already own a PC but if you are working with high speed
data then you find interfacing that data to a desktop PC is not easy
so if simply place the PC inside your project then "Poof" the
interface problem is gone.  The HPSDR project is a good example.  They
had to use a card with a FPGA on it to pre-process data to the point
where it would fit down the bandwidth of a USB cable.  Cost for that
is a couple hundred dollars and a year of engineering but, had they
simply placed the computer inside the box there would be no need for
high speed box-to-PC link.  Years ago this would have been expensive
but today powerful CPU that can run Linux or even Windows costs very
little

That said the basic counter project should NOT need such a powerful
CPU.  A little 8-bit uP with 14 pins should be enough.

On Thu, Dec 23, 2010 at 7:11 AM, Bob Bownes <bownes at gmail.com> wrote:

> Just thinking out loud here, I would imagine a main board that such a
> processor plugs into with locations to plug in input modules, A/D
> converters, a serial level converter (if there isn't one on the board), and
> a front 'control' panel. The main board holds not much more than connectors
> and traces to get signals from those modules to the processor board and/or
> the A/D boards.
>
> Hmmm.
>
> Bob

-- 
=====
Chris Albertson
Redondo Beach, California