IEEE Solid-State Circuits Magazine - Winter 2014 - 16

had a product containing an onchip crystal clock oscillator which
was suffering serious failures in
the field. The oscillator circuit had
been designed from a cookbook
with little attempt at analysis, and
had a fault mode where some parts
would fail to start at low temperatures with some (high Q it turned
out) crystals. This was a very serious defect, and the bad units were

with the realization of the first tunable Si monolithic gigahertz-range
oscillator [22], made possible by our
earlier research in on-chip inductors. Other interesting pathologies
such as multiple-frequency oscillation in high-frequency oscillators
were explored and documented [23].
A rather interesting consulting
job involving an oscillator that I performed around this time involved

That was an exciting time for me, learning to
make real circuits work and performing difficult,
sensitive measurements on real devices.

very difficult to screen out in production. I took the opportunity to
expand my repertoire of oscillator
expertise to circuits using quartz
crystals as resonators by digging
deeply into quartz resonator characteristics. I used linear root-locus
analysis to examine the initial pole
positions of the oscillator in question, and found that the designers
had set up the circuit with an initial loop gain of about a hundred!
(The assumption was, I think, that
if some positive loop gain is good,
then more is always better.) Instead
of ensuring oscillations, this brute
force approach pushed the poles of
the circuit transfer function back
into the left half plane under some
conditions, so that the oscillation
would not start. The problem was
worst for low temperatures and high
Q crystals, both of which increased
the loop gain. The problem was easily solved by the counter-intuitive
move of lowering the loop gain. I saw
an opportunity for some interesting
research on this topic and took on
a graduate student to flesh out the
theory. The resulting publication
[20] has been widely referenced, as
was our contemporary research [21]
on systematic methods of MOS crystal oscillator design. My research on
oscillators continued over the years

w I n t E r 2 0 14

an oscillating amplifier. I was contacted by a company making burglar
alarms that were exhibiting a major
failure mode, and the company
wanted some consulting advice on
the problem. It turned out they had
a need in one of their alarms for an
op amp with 741 characteristics but
with more output current capacity
than the standard part. They had
simply added a power transistor
to the output of the 741, but they
added it inside the overall op amp
feedback loop. My calculations and
simulations showed that the addition of the power transistor reduced
the nominal phase margin of the
op amp feedback loop to about 5°.
Apparently, the circuit tested out
just fine in the lab using nominal
parts (the gain peaking indicating
the onset of oscillatory behavior
occurred at a frequency well above
the signal bandwidth) but in production they got a batch of "hot" 741s
with higher than usual unity-gain
frequency (but still within the data
sheet spec). At low temperatures
the loop gain of the composite loop
got high enough to cause oscillation
of the op amp and triggering of the
alarm. As a result, a large number of
the company's alarms were calling
the police in the dead of night during cold snaps in the Midwest. The

IEEE SOLID-STATE CIRCUITS MAGAZINE

company's solution to this disaster
was to sue the op amp manufacturer
for selling them high-gain parts!
This tactic failed in court so they
declared bankruptcy and neither I
nor anybody else got paid. This was
a learning experience for me, both
technically and financially.

To the Present
In the 1990s, as the cellphone boom
intensified, there was great demand
for increasing levels of integration
in cellphone electronics. I teamed
up with Bill Mack and others at Philips to design early BiCMOS RF ICs
at medium levels of integration in
the gigahertz frequency range for
high-volume commercial cellphone
applications [24]-[28]. As in previous collaborations, the challenge
of breaking new ground in circuit
function and operating frequency
played into and fed my academic
research activities. My research on
a monolithic low-power RF peak
detector [26] created a large amount
of commercial interest and led to
several patents (by other people)
on variations of the circuit. One of
my graduate students on a summer internship designed a commercial version for Maxim which sold
extremely large volumes into the
cellphone power detector market.
I retired from teaching in EECS at
UCB in 2004 after 36 years, but still
keep my ties to the department. I
indulge my passion for electronic
circuit design and keep my circuit
design skills honed by taking on challenging consulting jobs as they arise.
I would like to acknowledge the
contributions made to my professional life by a number of people.
First the late Daryl Hooper who was
my mentor and role model, and the
late Don Pederson who brought me
to Berkeley and gave me sage advice
on navigating the pathways of academia. Ed Cherry's work on amplifiers and feedback has been of great
use to me in my technical career and
has led me to many crucial insights
in electronic circuit design. My colleague and co-author Paul Gray has

Table of Contents for the Digital Edition of IEEE Solid-State Circuits Magazine - Winter 2014