1 Apr
1999
1 Apr
'99
6:19 p.m.
YES! Finally, a point of agreement!
One of the S-100 boards I'm looking to place has a component suspended in
space above the board, and, in fact floating aboug an inch above one of the
regulators. This was undoubtedly built by one of the self-styled experts I
had working on automation software back in the early '80's. These guys had
me buy all kinds of stuff we ended up not using and they excused the
occasional waste by saying they bought a kit in order to save money. Of
course they didn't consider that the cost of building it was a cost to me as
well.
I rather suspect that, in spite of the strange insertion of this floating
component, the board worked.
On the other hand, I had an expert prototype a low-current highly
noise-immune PLL, for use in data/clock separation in hard disk drive
interfaces, which was constructed as you describe, with a piece of
copper-clad, double-sided, with the upper layer at Vcc and the lower at Vss.
It had wires tacked to other wires and suspended inches above the board,
while the components were soldered to pads cut from the solid plane, with
machined pins soldered through the board and the IC's plugged in belly-up
and wires soldered to their pins. It looked pretty tentative, but worked
very well. It extracted clock with lock in about 5-6 microseconds all the
time! My target was 10 microseconds. What's more, it required only a
single supply. Tentative though it was, it had an order to it that you
don't easily overlook.
About the simulators . . . I've found that fewer and fewer of my clients
blindly, and it is blind, believe me, accept a "rock solid" prototype, until
after a simulation reveals that it not only does work, but, rationally,
should work. The simulator rules out easily overlooked synergies between
miscalculations. These are immediately revealed in a simulator when you run
sensitivity analyses and worst-case analyses. If the circuit is
misdesigned, it often works well at one of the extremes, but doesn't work at
nominal spec.
Dick
-----Original Message-----
From: Tony Duell <ard(a)p850ug1.demon.co.uk>
To: Discussion re-collecting of classic computers
<classiccmp(a)u.washington.edu>
Date: Thursday, April 01, 1999 2:32 PM
Subject: Re: Kits vs ready-made (was RE: Rebirth of IMSAI)
of
necessarily
> >> senior engineers. The excuse was that
"it's not a deliverable," but
> often
> >> the shoddy technique (air-wires, etc) made for problems which couldn't
> >
> >If that's another name for dead-bugging, there's nothing wrong with it
if
> >used correctly. In fact IMHO it's the
_only_ way to prototype
> >high-frequency circuits with any sort of reliability
>
>
> I have used dead-bug patches quite a few times myself. More
specifically,
> dead-bugging is typically gluing or taping an IC
onto another's back and
> running wires between it and the appropriate points in the circuit. I
don't
> mean that, so much, but using multiple feet of #40
magnet wire with the
> shellac sanded or scraped off and having the scabbed-in IC floating on a
web
of wires
3" above the board . . . ???
Oh well, now that I would object to, even on a prototype. When I
dead-bug, all the large compoents (chips, etc) are fixed to the
groundplane (often a piece of copper-clad board). And the ground
connections are made with short pieces of 22swg-ish wire, soldered firmly
(they'd normally support the components even if they weren't stuck down).
Of course, one other thing is that my circuits have generally been
intended to be used by myself, or perhaps somebody else who can read a
schematic only. I don't give them to unsuspecting 'customers' in that
state. Even so, my prototypes are solidly constructed.
>
> >And if you trust simulations to correclty predict the behaviour of even
> >simple circuits, well, have I got some storys to tell you...
> >
> Yes, I have a few, too, but . . . Careful now . . . I've spend thousands
> hours in front of a big tube waiting for a
simulation. I am a big
believer,
and believe
further, that anyone who claims that simulators don't have a
place, as some old-timers do, just hasn't investigated sufficiently.
This list is based on the principle that a new idea/method isn't better than the old method. It might be, of course. And
that sums up my
views
on simulators.
Oh, simulations have a place, that much is certain. They are very useful
tools if used correctly.
What I object to in particular is :
a) Designers who couldn't prototype the circuit if they tried
b) Designers who trust the simulator implicitly (even if the simulator
has no bugs, which is by no means certain, they might not have given it
all the right information)
c) Circuits that are 'delivered' after only having been tested on a
simulator. IMHO the real test of a circuit is does it work when
constructed.
d) Simulators that take longer to provide less
information that actually
building the circuit
e) Simulators that can't handle some common occurances (one classic FPGA
simulator can't handle external memory linked to the pins on the FPGA,
for example).
f) Designers who fiddle with the simulation 'until it works' rather than
using good solid design principles. Yes, fiddling with real hardware
'until it works' is equally bad, but I've found that because it's easier
to make changes on the simulator than on real hardware, the use of
simulators encourages that behaviour.
g) Simulators that plain get it wrong. Don't get me started here, suffice
it to say that I've spent too long tracking down glitches in other
peoples designs that the simulator claimed didn't exist.
In short, in the hands of a good designer, a simulator is another useful
tool to be used alongside all the other tools. In the hands of a bad
designer, it generally leads to disaster. But alas Management often
believe the ads that say that %simulation-program allows anybody to do
design.
-tony