Expansion
----------

Sept 30 2005
There are many applications that require large amounts of
memory, yet computing speed and power requirements are low. Text
editing, data logging, and games come immediately to mind.

Virtual memory systems are good when the problem is large and you need
to randomly access many different parts of it quickly. I agree that we
rarely need to solve problems of this sort with an 1802. We could -- but
there are better choices.

What we're really after for the 1802 is mass storage memory. For
instance, the Infocom Adventure games, or a micropower data logger. But
the traditional route for mass storage (disk drives) are ill-suited to
interfacing with very small computers. The disk interface would be far
larger, far more expensive, and use far more power than the computer
itself!

So, I think large 1802 memories are perfect for implementing silicon
disk drives, where the data is stored in memory chips instead of disks.
That's why it is wonderful to see an operating system for the 1802 which
can use these large memories as if they were a disk drive.

Oct 2 2005
What exactly is the purpose for having more memory? If it is to simulate a
disk drive, maybe it would be better to use serial memory chips. For
instance, a 24C512 is an 8-pin DIP, holds 64k bytes, and costs $2.29 each
from Jameco. They only need 2-3 wires to interface (can use Q, EF4, and TPB
directly). Extremely low power. Nonvolatile (EEPROM based, keeps data
without power).

Once you have the serial read/write software, there is a whole world of
other serial chips that become easy to interface; RAMs, UARTs, A/D
converters, real-time clocks, etc.

Oct 3 2005
-- Winfrey wrote:
> Lee, It sounds like fun to experiment with the serial [memory] devices at
> such a low cost and I would like to put a few on my next order.
> Have you actually tried any of these serial devices with an 1802?

Not on an 1802 yet; but I have used them many times in commercial
products with a variety of other CPUs; PICs, 68HCxx, and most recently,
a Z80.

BASIC Stamp microcomputers use these serial EEPROMs to hold their BASIC
programs and data. The CPU itself has the BASIC interpreter in its
internal ROM, and fetches each tokenized BASIC statement from the
EEPROM, and executes it. It's a bit slower than storing it in bytewide
memory, but *lots* cheaper, smaller, lower power, simpler, and
nonvolatile. This scheme would work very well for Pittman's Tiny BASIC.

These chips use a bunch of different "standard" serial protocols; SPI,
microwire, etc. As usual, each manufacturer has his own favorite, which
he tries to coerce everyone else to use. We can probably devise a
no-hardware approach for any of them using Q, TPB, the N lines for the
clock and data out, and one of the flag lines (EF1-EF4) for data in. But
a more general approach is to have two output bits that you can
independently control; then your software can talk to any of these
chips.
--