15 Nov
2010
15 Nov
'10
6:41 p.m.
Tony writes:
Tim writes:
> Having worked with 1103's I would argue that by modern standards
> Even the "cream of the crop" were little above floor sweepings.
Was it that a lot of new 1103s were DOA, oe what there
a problem with
them fialing in use, or what?
The reason I ask is that I have several old HP
calculators full of 1103s,
and AFAIK all are original (and gettign on for 40 years old). So far,
otuch wood, I've not had any problems with the RAM in these machines.
How many hundreds of 1103's are in each HP calculator? :-)
In the early minicomputer 1103-based memories (say 16Kbytes, so
128 1103 chips) it was, um, optimistic to expect a lot of stability.
In the bigger "supermini" systems (say 128 Kbytes, so over a
thousand 1103 chips) ECC was a necessity and continual scrubbing
helped a lot with stability - but that was a lot of work and power
consumption just to brag that you had semiconductor memory,
and the advantages over core were not all that obvious.
I'd expect that smaller systems probably didn't have ECC or scrubbing
but also didn't have high expected uptime. Were 1103's in the HP 9830?
I had a couple but they usually worked fine enough that I never had
to go rummaging through the insides.
My feeling is that in the 1103 era (which preceded my direct
involvement but I inherited several 1103-based systems), DRAM chips
were not quite ready for prime time. By the time the 4116 came
along things had settled down and a lot had been learned from
the 1103 and DRAM really was ready for prime time.
Tim.
16 Nov
16 Nov
4:33 p.m.
Tony writes:
It depends on the model. I think my 9830 has 0.96 hundreds of 1103s, for
example. And that';s not fully populated.
Tim writes:
> Having worked with 1103's I would argue that by modern standards
> Even the "cream of the crop" were little above floor sweepings.
Was it that a lot of new 1103s were DOA, oe what
there a problem with=20
them fialing in use, or what?
The reason I ask is that I have several old HP
calculators full of 1103s,=
=20
and AFAIK all are original (and gettign on for 40
years old). So far,=20
otuch wood, I've not had any problems with the RAM in these machines.
How many hundreds of 1103's are in each HP calculator? :-)
In the early minicomputer 1103-based memories (say 16Kbytes, so
128 1103 chips) it was, um, optimistic to expect a lot of stability.
that is the sort of size of the HP9830 memory. Problems in this area are
not at all common.
I'd expect that smaller systems probably
didn't have ECC or scrubbing
but also didn't have high expected uptime. Were 1103's in the HP 9830?
I had a couple but they usually worked fine enough that I never had
to go rummaging through the insides.
Yes, the HP9800 series (9810, 9820, 9830) are the machines I am thinking
of. I beleive the 9810 was one of the first commerical applications of
the 1103. Interestingly, the memory unit in the 9210 and 9820 is its own
module, a metal box cotnaining the RAM boards (nothing but 1103s) the ROM
board(s) (HP custom 512 byte ROMs) and the memory control boards. One
rumour was this was done so that a different memory technology could be
easily fitted if the 1103s had proved unreliable.
The 9830 hasa more conventional layout with all the boards going into one
main backplane at the bottom
The 9810 is probably the only machine where memory expansion involves
multiples of 3 1103s. User programs are stored as sequences of 6 bit
keycodes and there si hardware in the memory controller to store a word
in 2 consequtive memory locations (thus the ssytem 16-bit RAM is 8 chips,
user progam RAM is 3 chips). The 9820 and 9830 spoil the fun hby having
conventioanl 16 vti wide memory in 16 1103s.
My feeling is that in the 1103 era (which preceded my
direct
involvement but I inherited several 1103-based systems), DRAM chips
were not quite ready for prime time. By the time the 4116 came
along things had settled down and a lot had been learned from
the 1103 and DRAM really was ready for prime time.
Possibly. The 1103 was the first 1K bit DRAM IIRC after all. However, I
wonder how much was due to problems with the DRAMs themselves and how
much was down to problems with supplies (They are unconventional for the
PMOS 1103s, it's really a main -16V supply and a +3V bias supply, often
states as 16V and 19V), decoupling (they are touchy about this!), signal
timing and the like.
-tony
5:41 p.m.
On 16 Nov 2010 at 21:33, Tony Duell wrote:
Possibly. The 1103 was the first 1K bit DRAM IIRC
after all. However,
I wonder how much was due to problems with the DRAMs themselves and
how much was down to problems with supplies (They are unconventional
for the PMOS 1103s, it's really a main -16V supply and a +3V bias
supply, often states as 16V and 19V), decoupling (they are touchy
about this!), signal timing and the like.
It's worthwhile noting that nobody selected the 1103 for any speed
advantage over core. Less power or complexity, perhaps but it was
quite some time before MOS memories caught up to core in speed.
My own recollections of MOS memory weren't with the 1103s, but the
2107 4K variety. A 22-pin 3-supply chip with TTL compatible I/O,
with the exception of CE, which was kind of strange.
At any rate, I believe this chip finally had the mainframe makers
convinced about the benefits of MOS memory.
--Chuck
17 Nov
17 Nov
3:20 p.m.
On 16 Nov 2010 at 21:33, Tony Duell wrote:
Indeed...
Incidnetally, a trivia question. Does anyone (else) remember the main
difference between am 1103and an 1103A?
Possibly. The 1103 was the first 1K bit DRAM IIRC
after all. However,
I wonder how much was due to problems with the DRAMs themselves and
how much was down to problems with supplies (They are unconventional
for the PMOS 1103s, it's really a main -16V supply and a +3V bias
supply, often states as 16V and 19V), decoupling (they are touchy
about this!), signal timing and the like.
It's worthwhile noting that nobody selected the 1103 for any speed
advantage over core. Less power or complexity, perhaps but it was
quite some time before MOS memories caught up to core in speed.
My own recollections of MOS memory weren't with the 1103s, but the
2107 4K variety. A 22-pin 3-supply chip with TTL compatible I/O,
with the exception of CE, which was kind of strange.
Yes, I've come across those too. I think on oe my I2S image displays is
full of them (over 3000 of the darn things...) And in some HP devices
(1350 'graphics translator' and 9831 comptuer (my 9825 uses 4116-like 16K
DRAMs))
They had non-multiplexed addresses I think. Later DRAMs had multiplexed
address inputs. of course. Was there a good reason for this other than to
save package pins?
-tony
10:02 p.m.
Tony Duell wrote:
Later DRAMs had multiplexed
address inputs. of course. Was there a good reason for this other
than to
save package pins?
I don't remember any other justification being given back in the day.
Lower pin count allowed for more chips per unit board area. Prior to
Mostek's invention of the multiplexed-address DRAM, 4K DRAMs were in a
22-pin DIP, which occupied approximately 0.55 square inches, while the
multiplexed-address DRAM in a 16-pin DIP occupied only 0.32 square
inches. (Not counting decoupling capacitors or any additional spacing
needed for power distribution.)
In hindsight, it seems surprising that the "ZIP" package didn't appear
until many years later, since that allowed for twice the packaging
density without moving to surface mount.
Eric
5006
days inactive
5009
days old
5 comments
5 participants
participants (5)
-
ard＠p850ug1.demon.co.uk -
cclist＠sydex.com -
eric＠brouhaha.com -
pu1bzz.listas＠gmail.com -
tshoppa＠wmata.com