Date: Fri, 15 Feb 2008 13:37:46 -0800
From: Brent Hilpert
Reading the Osborne book some time ago I got the
impression the SMS300
might be considered the first (micro-)DSP. Extracts:
"The SMS300 is described by its manufacturer as a "microcontroller"
rather than a "microprocessor". This distinction draws attention to the
very unique capabilities of the SMS300 which make it the most
remarkable device described in this book."
"The SMS300 is designed to serve as a signal processor, operating
at very high speed. The SMS300 can handle applications of this type at
more than 10 times the speed of any other device described in this
book."
"If yours is a high speed signal processing application, then give
the SMS300 serious consideration; otherwise, the SMS300 is probably not
for you."
It's true that the SMS300/8X300 was pretty fast for the time
(instruction time was 250 nsec., consisting of 4 cycles of 62.5
nsec), the instruction set was very limited, as was the data
addressing (256 bytes each for right and left I/O register banks).
As to the limits of the implementation, consider the benchmark
implementation in the Osborne book.
If I had to guess, I'd say that where the 8x300 saw most of its
application was in hard disk controllers. About the only other
alternative in the day was to employ bit-slice logic or discrete TTL.
I suppose I'd characterize the 8x300 as being halfway between bit-
slice such as 2901 and a full-blown microprocessor like an 8080.
Since this was a Harvard-architecture design, one way to extend the
instruction set was to add additional bits onto the normally 16-bit
wide instruction store and use them to implement additional
functions. The limit of addressability of the instruction store was
8192 words.
In fact, the 8x305 couldn't support its own development system. The
SDK was initially offered as a board that could be plugged into the
GI GIMINI CP1600 development system. Later, Signetics bundled the
8X305 as a signle-board SDK using an Intel 8035 to handle user
interface functions.
Cheers,
Chuck