Actually Add
noise to circuits? What? That's the basic idea I had when I
started out too. But over time I have found that noise is
useful for testing, linearizing, actually measuring other
noise and securing.
There are two basic types of noise sources. Pseudo generated
noise - usually produced by a digital shift register scheme
and analog random noise - usually produced by some sort
of semiconductor junction.
Both types of noise generation circuits have their
place.
Pseudo random noise is useful because the statistics of the
noise can be very well known. Also the exact sequence of the
noise can be repeated. This is very useful in testing - so
that the results of one test can be compared to another.
There is a practical limitation on the upper frequency that
this noise can be generated however - due to the speed of
the digital logic used.
Random noise from a semiconductor junction has the advantage
of being wider bandwidth and the disadvantage of being
harder to control the statistics of. These types of noise
sources are useful in GHz noise figure measurement and
secure communications applications.
Many factors come into play when specifying a random noise
source. Things like noise "color" (it's power versus
frequency), environmental susceptibility and control of
the output amplitude are just few of the
problems. Testing is the key - this must involve both
time and frequency domain measurements to truly have
confidence that the noise source is truly useful for the
application at hand.
As an example consider the two plots below. Both noise
sources has similar noise spectra in the desired frequency
range, but upon closer inspection the first diode was
corrupted with "Popcorn" noise. In a general noise source
this probably wouldn't be a problem. But Popcorn noise is
highly dependent upon environmental factors and is
oftentimes not repeatable. Another problem is injection
locking of the noise source or the circuitry connected to
the noise source. The last problem depends on the noise
bandwidth and the frequency of the popcorn noise. Since the
pop's usually happen at a low frequency they can really mess
up the histogram of the noise, again adding "color" to the
noise and the color many not be consistent. Even worse it
may come and go during a test as the diode operating
conditions change.
Figure 1 - A noise source that exhibits "Popcorn" noise.
This particular diode performed well on a spectrum display,
but has out of band problems. The popcorn noise was also
very dependent on temperature and diode current.
Figure 2 - The time domain plot of a properly behaving noise
source.
Home...
