Anybody can make an oscillator these days at
almost any imaginable frequency. But how do
you make a good one at low power? A couple
of people recently asked me why I add so much
complexity by using an op-amp oscillator in my Cold Toe
Detector. (That general-purpose slow oscillator is in the LH
upper corner of www.national.com/rap/coldtoes.html.) “What
kind of analog freak are you?” they asked. “Why don’t you just
use a CD4060, which has a built-in oscillator?”
I responded that everybody knows that CMOS is a lowpower
way to do things—except when it does not provide lowpower
operation. If a CMOS analog circuit is designed to run
at low power, it can be very efficient. But if it is not engineered
for low power, it can waste a lotta microwatts.
CMOS amplifiers and oscillators are not inherently lowpower,
but run at low power only when designed to do so.
Operated in a linear region, they can be very power-hungry.
Conversely, bipolar transistor circuits can easily be designed to
be quite efficient as they have more gm per microampere.
So I thought some more. If I want to design a 1-Hz clock,
is the CD4060 really that bad and the op-amp oscillator that
much better? And if a CMOS clock oscillator based on a
CD4007 was optimized, would it have any advantages? Maybe
it was time for me to build and measure some oscillators.
Time for Homework
We had a guy at National who was really helpful and knowledgeable
about oscillators. If any of us NSC guys or any customers
had a question about oscillators, Tom Mills always
made himself available to help.
Tom had the bad taste to die in his sleep about five years ago,
leaving us bereft of a great engineer and a great friend—and
an oscillator expert. But somewhere up in Heaven, I am sure
Tom is saying, right now, “That’s right, Bob. Do your darned
homework. Build and measure things.”
The oscillator I built for the Cold Toe Detector to oscillate
at 1.6 Hz was not optimized per se. I just grabbed the first
low-power op amp that I could get, an LMC6041 (not the
lowest-power one). I slapped on the R’s and C’s and it ran, just
fine, and I thought little more of it until today.
- As I observed in “What’s All This CD4007 Stuff, Anyhow?”
back in April of 1999 (see www.electronicdesign.com, ED
Online 6073), the CD4007 is a very versatile and powerful linear
circuit. But to get the best results, you may have to do some
real engineering.
- The original cold-toes oscillator (1.6 Hz) using an LMC6041
drew about 18 µA. I took all the data on my low-frequency
“1-Hz” test oscillators running at 70 Hz to make it easier to
average the current drain. I didn’t think that would change the
power requirement appreciably from the current at 1 Hz—
only a few percent. So we are in the right ballpark.
- The CD4060 self-oscillating counter/timer used about 180
µA. I used the cookbook circuit from the Fairchild CD4060
datasheet. I just used 10M, 10M, and 0.001 µF—not terrible,
not wonderful.
- A basic MM74C14 Schmitt trigger used about 90 µA, using
just 10M and 0.002 µF.
- The basic CD4007 circuit, per the figure, also used about 180
µA. That’s funny. I was hoping I could tweak it to do better
than that. I fooled around with it. Those 2.2M resistors were
not a great idea. Finally I made a lucky guess, and if I shorted
out one of the 2.2M resistors, or the other, the drain would
fall to 22 µA. But if I shorted both, it would go back up to 180
µA. Ha! The joke’s on me!
- I threw in a real low-power op amp, with a rated Is of 2 µA,
and the drain fell below 3 µA.
- A comparator is really the right way to make a low-power
oscillator. I got one of our lowest-drain comparators, the
LPV7215 at 0.58 µA, and put it into the basic oscillator
shown in that Cold Toes circuit. It did the best job at 1.4 µA.
Go ahead and beat that!
See associated figure