I do note that you admit to regular failures. My rate of need for
debug is under 1%... usually due to the customer's failure to spec
what he really wanted.
I have no design "failures" since I started designing professionally
at the age of 18. I often change parts values on a new first-article,
and sometimes add a kluge, usually a cut/jumper and sometimes an added
resistor, cap, or even a diode, like if we find a latchup mode in a
chip or something like that. Sometimes we think of an improvement or a
feature, and squeeze it in if possible. It's very rare around here
that we can't sell a presentable rev "A" board. I do design a *lot* of
stuff.
How careful you need to be depends on the consequences of defects and
how hard they are to fix. An IC is expensive to turn, so a lot of
simulation is justified. Software is the other extreme, easy to hack
and easy to edit, so programmers are the sloppiest of designers. The
point is for everyone to optimize his particular design process, and I
believe my optimum involves careful design and checking and very
little breadboarding or simulation.
But the flat statement that complex analog design can't be done
without simulation may be mostly true for IC design, but it's sure not
universal. I wonder how much the uA709 was simulated.
The current educational paradigm, design by simulation, seems very bad
news to me. For Pete's sake, guys are simulating battery-resistor-LED
circuits and the most basic transistor switches.
John