I'm not sure what you mean by "performance" with respect to these cap's,
but the
reason we used to justify the somewhat more costly tantalum cap's back in the
'80's was that their frequency response was much better than electrolytics,
going pretty far into the range normally associated with ceramic cap's, yet they
would perform as well as the comparable values of electrolytics at lower
frequencies. That meant one didn't have to put a small ceramic in parallel with
a larger electrolytic in dense structures like DRAM arrays, which, with 3
supplies, and lots of noise, along with pretty significant currents, switching
at relatively high speeds, required both high-frequency response compatible with
the sequences of control strobes at least the three major ones, (nRAS, nCAS,
nWE) not considering the addresses and their associated muxes, all switching
within a tyipically 40-60 ns window. The electrolytic's function was to hold
the supply up so the power-to-ground surge didn't cause problems as devices were
suddenly selected, while the ceramic's function was to pick up the slack while
the electrolytic figured out what to do. I remember supply fluctuations over a
volt in magnitude in some memory circuits. That didn't make for high
reliability.
As for the S-100 supply spec, the 8-volt spec is a minimum because the forward
voltage of the regulators typical in the '80's was over 2 volts. Likewise, the
+/- 16 volt spec called out in the standard is a minimum for the same reason,
and, in fact, because some circuits used Zener regulators or the like, with
fairly low dropout, to supply ananlog circuits requiring bipolar 15 volt
supplies. It's not unusual to see the no-load voltages on these supplies to
look like 10 volts for the +8 and 21-22 volts for the bipolar 16 volt supplies.
I always preferred to stay close to the lower limits, but since then-current
design practice was to design transformers such that they produced a 10% voltage
drop from no-load to full load, and since many vendors pushed that to 15 or even
20%, the >20 volt values were not uncommon. Even my own designs tended to look
closer to 10 volts on the +8 in order to save on cost and size. Of course my
stuff was not with custom designed transformers, and COTS products tended to
have 18VCT and 36VCT secondaries.
BTW, the cardcage protects you from stupidly plugging a board in backward, which
I recently did on motherboard that was on the benchtop with no cage. That can
lead to real problems (smoke!), so be careful. I did that once before, back in
the '70's and a piece of an IC hit me right between the eyes when the thing
burst. That was with a regulated supply to the bus, (not S-100) so there was no
regulator to protect the circuits.
It is, by the way, a good idea to solder a hefty diode conducting from regulator
output to regulator input on S-100 boards, to protect the regulators against the
occasional situation during power-down, when stored charge in the load holds the
voltage up while loads elsewhere in the system pull the regulator's input down.
When that happens, next time you power the system up, you may no longer have a
regulator. These diodes would not have cost much to include on the PCB, but the
combination of cost and space made folks leave them out.
Dick
----- Original Message -----
From: "Clint Wolff (VAX collector)" <vaxman(a)qwest.net>
To: <classiccmp(a)classiccmp.org>
Sent: Sunday, June 03, 2001 8:41 AM
Subject: Re: S-100 Bus Voltages
Actually, modern ceramic caps are very close to tantalums
in terms of performance, and are approaching the size of
tantalums. Those of you who are engineers (or am I the
only one?) should investigate ceramics before plopping
a half dozen tantalums onto a board...
Clint
On Sun, 3 Jun 2001, John Foust wrote:
At 10:59 PM 6/2/01 -0500, Eric Dittman wrote:
The boards had several tantalum capacitors.
I thought I'd recently heard a news story that
claimed that 80% of the world's supply comes from
a single mine in the Australian outback, but that
may only be a slight exaggeration:
http://www.cellular.co.za/technologies/phones/tantalum.htm
which leads to tantalum ore poaching:
http://www.ebnews.com/digest/story/OEG20010427S0086
Perhaps a richer source of ore is old computers.
- John