On Wed, 15 Oct 2003 17:03:00 -0400, "Tarmo Tammaru"
wrote:

Richard,

Just a few words about modeling. Yes, they built a model of a 4004 before
they made the chip. They stopped doing that somewhere around the 386.

As for my analog simulation, the simulator I am using is distributed by
Linear Technology Inc so that people can model circuits using their chips. A
lot of real world stuff is sort of simulated, where an analog signal is
digitized, and all filtering and other audio stuff is actually done in a
DSP.

If you want to see something that will really blow your mind, check out the
Harris digital AM transmitter; and no, I don't mean a transmitter for
digital audio.

Tam/WB2TT


Hi Tam,

I've used many simulators, including Spice and Electronic Workbench.
I've also read and adhere to Robert Pease's comments, a columnist and
engineer extraordinaire who can be just as dismissive about these
tools, especially when they cloud common sense with the impression of
legitimacy through presenting a number on the basis of faulty
presumptions.

For example, there is no way you can simulate a perfect resistor
simply and confirm it at the bench. So which is more important, the
pipe dream of a virtual circuit, or the real circuit that fails to
perform as forecast? I've seen your recitation of equations. I've
seen many like them applied to precision work that falls flat on its
face when they hit the wall of reality. Basically those so-called
simulations frequently fail to attend to many matters that are
unnoticed to 1% accuracy when the equation supplies 6 places of
resolution in perfection. No such thing could ever be duplicated
without further elaboration of the model, and as evidenced by your
example (and the successive iterations), you are in jeopardy of just
such failures in this "proof" offered.

I've measured the Ohm to seven places of resolution and 6 places of
accuracy. Ohm's Law was not enough, but it did guide at every turn in
the road that found error galore. To accomplish it, it took the
understanding of the Galvanic series, temperature, expansion,
humidity, actual power dissipated, noise, drift, as a host of many
sources of error that go unnoticed and un modeled in ad-hoc
simulations. There is of course no demand for such precision to
reveal the characteristic Z of the source, that can be accomplished
far simpler and is documented by the same authorities who took such
care in measuring the Ohm. The point of the matter, and this has been
shown in Chipman, that the presumption of a matching source in the
discussion of SWR is often taken for granted, and then dismissed as a
necessity.

The proof of a simulator is found in its suite of tests that confirm
what is demonstrable. If your simulator cannot predict the loss of
real coaxial line when faced with mismatches at both ends, then it is
not simulating actual performing conditions. You did not express any
condition of source Z within your equations, when if fails from that
reason, it necessarily invalidates the simulation. You claim, and
Cecil says otherwise, that there was no line specified (making the
specification of SWR rather obscure as a model), and as the inclusion
of a line was a necessary correlative to the exhibition of source Z,
it follows that your model is two steps removed from that argument by
your own admission. So, what is this a model of? Just what do you
mean by SWR? Which model is the most perfect if they are all better
than bench measurements? What impelled the process of changing the
model when even the first one is so much better than the bench? The
80386 would have never beat an abacus with those metaphorical
questions hanging over its model.

I've seen digital transmitters. Circuit Cellar Ink has covered this
form of modulation at least half a dozen years ago or more where Don
Lancaster offered that a digital string can present (through a low
pass filter) a power curve with distortion 60dB down (simply a matter
of string length and clocking).

73's
Richard Clark, KB7QHC