"Reg Edwards" wrote in message
...


SNIP

I sometimes think that the relatively few engineers between 1790 and 1890
performed greater engineering feats than the many who followed them into
the
present age of electronic and genetic engineering. They devoted the whole
of
their lives to their work.

As for us poor souls, the best we can manage is haggling about imaginary
SWR
and conjugate matches which were all sorted out 120 years back. But it's
all
good fun.

Cheers, Reg.




That was back in the days when fantastic claims were settled with a working
model. If you wanted to argue about the efficiency of a venturi, or the
strength of a gear tooth profile, you built it and then actually used it. If
your drill bit stayed sharp longer, or you pumped more water with less coal,
you won your argument.

We spend a lot of time now arguing about how well the computer model
replicates reality, and whether the math has enough variables accounted for.
Working models seem so old fashioned.


Ed
wb6wsn

Ed said -
That was back in the days when fantastic claims were settled with a
working
model. If you wanted to argue about the efficiency of a venturi, or the
strength of a gear tooth profile, you built it and then actually used it.
If
your drill bit stayed sharp longer, or you pumped more water with less
coal,
you won your argument.

We spend a lot of time now arguing about how well the computer model
replicates reality, and whether the math has enough variables accounted
for.
Working models seem so old fashioned.

=======================================

It is a fatal mistake to treat a modelling program, even if you think it has
no bugs (errors), as a bible which always tells the gospel truth. ALL
programs have limitations.

Limitations result from the computer itself, those deliberately introduced
by the programmer, those accidentally introduced by the programmer because
he didn't understand how the thing being modelled really works, those
introduced by the user because he doesn't understand how the program is
supposed to work or what the programmer was thinking about when he wrote it.

The result is UNRELIABILITY.

Ideally, the originator of the thing being modelled and the programmer
should be one and the same person. Committies produce drumadaries with 3 or
more humps. Or elephants with trunks at both ends.

The definition of Reliability is Quality versus Time, and therefore
confidence (or lack of it) can be gained only with both use and time.

Given time, and use, with large programs, such statistics as
mean-time-between-failures can be produced. But when the next error might
arise and its magnitude is anybody's guess. One is always caught unawares.
More insidiously, one may not be aware that an error HAS occurred. Or most
insidiously, one may imagine an error has occurred when it hasn't.

Problems will surely persist - if a failure is suspected, is it the program
which has failed, is it the computer, is it the modelling, or is it the
actual thing being modelled (it may not exist) which is defective?

The proof of the pudding lies in the eating. Get off your ass, wrench
yourself away from the keyboard, do what you should have done in the first
place, erect the thing and use an instrument which purports to measure SWR,
hope for the best, don't swear by it, and take care to record the instrument
manufacturer's name and its serial number. ;o)

To summarise, the reliabilty of a modelling program is always worse than the
quality of the blamed programmer. Initially, don't believe anything it
produces.

And whatever you do, don't become depressed. Even if the program doesn't
work the radio will. Most happy-band radio amateurs don't realise how
fortunate they are - almost anything works thank goodness.

At present I'm on Spanish Red, Berberna, Reserva 2000. I know it's Spanish
because, unusually, the entire blurb on the bottle is in that language. But
I feel somewhat guilty because at the back of my mind there's the continuing
unbelievable horror of the enormous disaster in the countries surrounding
the Eastern Indian Ocean. The worst effects may still be to come.
----
Reg, G4FGQ

"Ed Price" wrote in message
news:gAdBd.6143$yW5.2@fed1read02...
We spend a lot of time now arguing about how well the computer model
replicates reality, and whether the math has enough variables accounted
for. Working models seem so old fashioned.

That's because they're so expensive to build. You'd probably never finish
designing something like a modern RF IC if all you could do was design it on
paper, build it, probe around a little to figure out what it 'really' does,
and repeat.

Likewise, few companies can afford to design the autopilot for a jet without
a great deal of simulation first. :-)

On Wed, 5 Jan 2005 12:03:37 -0800, "Joel Kolstad"
wrote:

"Ed Price" wrote in message
news:gAdBd.6143$yW5.2@fed1read02...
We spend a lot of time now arguing about how well the computer model
replicates reality, and whether the math has enough variables accounted
for. Working models seem so old fashioned.

That's because they're so expensive to build. You'd probably never finish
designing something like a modern RF IC if all you could do was design it on
paper, build it, probe around a little to figure out what it 'really' does,
and repeat.

Likewise, few companies can afford to design the autopilot for a jet without
a great deal of simulation first. :-)


Hi Guys,

Back when I designed the UFDR for the 757/767, they were the first
airplanes designed entirely in software (conventional drafting went
the way of the Dodo). Today's 7E7 was entirely modeled in software if
I'm not mistaken.

73's
Richard Clark, KB7QHC

"Richard Clark" wrote in message
...
On Wed, 5 Jan 2005 12:03:37 -0800, "Joel Kolstad"
wrote:

"Ed Price" wrote in message
news:gAdBd.6143$yW5.2@fed1read02...
We spend a lot of time now arguing about how well the computer model
replicates reality, and whether the math has enough variables accounted
for. Working models seem so old fashioned.

That's because they're so expensive to build. You'd probably never finish
designing something like a modern RF IC if all you could do was design it
on
paper, build it, probe around a little to figure out what it 'really'
does,
and repeat.

Likewise, few companies can afford to design the autopilot for a jet
without
a great deal of simulation first. :-)


Hi Guys,

Back when I designed the UFDR for the 757/767, they were the first
airplanes designed entirely in software (conventional drafting went
the way of the Dodo). Today's 7E7 was entirely modeled in software if
I'm not mistaken.

73's
Richard Clark, KB7QHC


There's nothing wrong with doing a lot of modeling. OTOH, Boeing didn't go
from the computer model directly to production. Computer models don't do
very well in predicting the unexpected; things like digital designers
discovering the concept of parasitics ("There's nothing in my design that
generates 832 MHz!") or mechanical designers exploring the wonders of RF
stray coupling paths ("The RF gets from this compartment to that compartment
through a BOLT?!"

My initial point was that wondrous and amazing things can be proven through
software being pushed to the outer edges of its parameters. Before you
invest in several years of modeling, someone needs to take a whack at a
brassboard model to calibrate the sanity of the software.

Ed
wb6wsn

On Thu, 6 Jan 2005 04:21:19 -0800, "Ed Price" wrote:

OTOH, Boeing didn't go from the computer model directly to production.

Hi Ed,

That is arguably the goal if not the actual reality:

"As computers have become faster and more powerful in recent years,
we have been able to do a better job in modeling the entire airplane and
predicting the three-dimensional effects of the airflow around it,"
Cogan said. "The codes we have developed allow us to look at more
potential design options faster than ever before."

Indeed, Cogan said the process for developing airplanes today begins
with the computer model. The coding is so accurate that designers can
evaluate miniscule changes in a design to determine impacts on
aerodynamic efficiency, he added.

In fact, the accuracy of the coding has also focused the application of
another aerodynamics tool: wind tunnel testing.

In the '80s, the Boeing 767 team took more than 50 wing designs into
the wind tunnel to verify their designs, Cogan said. In the '90s, the
Boeing 777 team took 18 designs into the tunnel. "We were really not
verifying the designs as much as we were verifying that our computation
tools were accurate and looking at performance at the extreme operating
conditions, which the coding couldn't do," Cogan said.

"With the 7E7, we will take fewer than 12 wings into the tunnel," Cogan said.
"We are still proving our coding and testing the extremes.
The tunnel is a great tool but it's not very cost-effective.
So, being able to really focus on a few designs to get the data we need
is helping us be more cost-effective."

73's from Jet City,
Richard Clark, KB7QHC