John Larkin said:
You don't rely on device parameters? I suppose it is convenient to
only stock a single resistor value.
I rely on device parameters that I can rely on. This circuit, maybe not so
much. But Rds(on) is specified about as well as beta is, wouldn't you say?
I don't stock a single resistor value, but I do find I use an awful lot of
1k, 4.7k and 10k. There's quite a lot you can do with just those three.
Hard to do much without other values though.
All the stuff that I design works, usually the first time, and it's
manufactured and shipped in volume.
The last major project I designed worked the first time, requiring only
minor adjustments in the control loops. And it spans three (smallish)
boards.
None of these products are remarkable. They are all well within the known
laws of physics. Possible ways to design them are immediately obvious.
Actual implementation would take longer to work out, and the complete
design, a few months more.
The only thing that might be impressive about them are the tedious
refinements (the kind of thing Tektronix was expert at, making things like
your handsome square wave), and repetition in scale, which people find
amusing (like those entertaining domino videos), but which is mindless work
putting together. 256 input ADC? Sure, that's easy. It takes a lot of
op-amps and big board space to hold everything, but each channel is just as
boring as the next, just filters and switches and whatnot.
The data processing that brings everything together should be impressive, if
not for already being pedestrian today. (It's a sad age when billions of
transistors, and their programming, can be considered boring.) Just slap in
an FPGA, wave the magic VHDL wand, and there's your bus interface.
Microcontrollers, FPGAs -- computers in general, are tedious to make (>1G
transistors) and tedious to program (>10k LOC). They are neither hard to
make nor difficult to program.
It's too bad all that tedium has to be done by human hands, so much of it
could be so easily written automatically. Then, there would actually be
time to do impressive things. Too bad the "software design software"
required is somewhat more complicated than anything anyone can understand
today. If it's true that "software design software" can never design
itself, then since humans are software writers, they will certainly never be
able to write one.
Tim
