Page 30 - The Art and Science of Analog Circuit Design

Barry Harvey

groups, or require the process development engineers to create models. I
have found these groups' data inaccurate in the previous companies where
I've worked. We recently checked for accuracy between some device
samples and the models created by a modeling group at a well-known
simulator vendor, and the data was pure garbage. We modeled the devices
correctly ourselves.
This being a general design need, I would have the young engineer
create model parameters from process samples, guided by a senior engi-
neer with a knack for the subject. This would also be an opportunity to
steep the engineer in the simulation procedures of the department, since
the models are verified and adjusted by using them in the circuit simulator
to play back the initial measurements. It's a pretty tedious task, involving
lots of careful measurements and extrapolations, and would probably take
three months, part-time, to re-characterize a process. Modeling does give
the engineer truly fundamental knowledge about device limitations in
circuits and geometries appropriate to different circuit applications, some
really arcane and useful laboratory techniques, and the appreciation for
accuracy and detail needed in design.
Because of the tedium of modeling, few companies have accurate
ongoing process data.
3. A couple of layouts would then be appropriate. Most of our de-
signers at Elantec have done the mask design for some of their circuits,
but this is rare in the industry. The usual approach is to give inadequate
design packages to professional mask designers and waste much of their
time badgering them through the layout. The designer often does an inad-
equate check of the finished layout, occasionally insisting on changes in
areas that should have been edited earlier. When the project runs late, the
engineer can blame the mask designer. You see it all the time.
I would have the young engineer take the job of mask designer for
one easy layout in the second three months of half-time. He would lay
out another designer's circuit and observe all the inefficiencies heaped
upon him, hopefully with an eye to preventing them in the future. Actu-
ally, we designers have found it very enlightening to draw our own cir-
cuits here; you get a feel for what kind of circuitry packs well on a die
and what is good packing, and you confront issues of component match-
ing and current/power densities. The designer also gains the ability to
predict the die size of circuits before layout. The ultimate gain is in im-
proving engineers* ability to manage a project involving other people.
4. The first real design can be started at the beginning of the second
year. This should be a design with success guaranteed, such as splicing
the existing circuit A with the existing circuit B; no creativity desired but
economy required. This is a trend in modern analog 1C design: elaborating
functions around proven working circuitry. The engineer will be overseen
by a senior engineer, possibly the designer of the existing circuitry to be
retrofitted. The senior engineer should be given management power over

13