Page 74 - Photodetection and Measurement - Maximizing Performance in Optical Systems

Fundamental Noise Basics and Calculations

Fundamental Noise Basics and Calculations 67

3.12.2 Use of an oscilloscope
The first job is always to have a good look on the oscilloscope. It is amazing
how often one sees detectors plugged straight into instrumentation and PC
loggers, without any idea of what the signal is like. This is asking for problems.
With a scope and the sensitivity turned up so that you can see the variations,
you will be able to get some idea of the magnitude and frequency content of the
noise. Gaussian noise has a certain “look,” and it is often possible to detect vari-
ations from the correct one. It should for instance be bunched around a mean
value, be symmetrical, and have the right fuzziness. A problem is that you can
get almost any fuzziness you like by changing the scope time base. Wind it up
to high speed and the trace might be a straight line. Slow it down and the noise
will just look like a thick band. It is therefore important to have an idea of the
frequency band that the noise occupies. If white noise is band-limited at 1Hz
and displayed with 1s/cm on a scope or chart recorder, it should look the same
as noise band-limited at 10kHz, and displayed at 100ms/cm. The look is defined
by the ratio of time-base (or spreadsheet X-axis scale) to noise bandwidth.
The problem is compounded these days with the dwindling availability of
analog scopes. The analog scope usually draws many traces during the persist-
ence time of the phosphor screen. Faster time-bases give more overlaid traces,
which the screen and the eye tend to average. The digital scope, and especially
that with a digital LCD screen, erases the previous trace before drawing a new
one and typically makes only a few updates per second. The averaging of a per-
sistent phosphor is lost. We still have some averaging (about 100ms) due to the
eye’s response time, but the difference in look can be drastic. Neither is right
or wrong, but it’s best to have both types of scope available.
Even with these difficulties, the scope can be used to estimate noise ampli-
tudes. If it is Gaussian, then the displayed voltage should spend 99.7 percent
of its time within ±3s of the mean (s is the standard deviation). We can take
the visible peak-to-peak voltage as 6s. Of course, if it is Gaussian, you will
always see a bigger voltage if you wait long enough, so the peak-to-peak value
should not be taken as a precision estimate. It is said that two displayed noisy
waveforms, aligned and offset on the scope to fit together without visible join,
can be used to improve precision. In practice it is probably easier to just take
a guess at the peak-to-peak.
The scope is also useful to detect non-Gaussian character. Popcorn noise
from poorly passivated photodiodes and uncleaned PCBs can give a waveform
with a large number of steplike jumps. Asymmetry can suggest problems with
dynamic range, for example with a DC voltage approaching the supply rails.

3.12.3 Spectrum analyzer
After the scope the electrical spectrum analyzer is the most powerful
instrument for noise measurements. With its hundreds of frequency-resolved
voltage measurements per scan, it is usually easy to separate spot-frequency
interference from white noise signals, see the effects of bandwidth limitation,

Downloaded from Digital Engineering Library @ McGraw-Hill (www.digitalengineeringlibrary.com)
Copyright © 2004 The McGraw-Hill Companies. All rights reserved.
Any use is subject to the Terms of Use as given at the website.