Lake Shore 372, Руководство пользователя, Страница: 52 / 214

40 c
HAPTER
2: Cooling System Design and Temperature Control
Model 372 AC Resistance Bridge and Temperature Controller
2. Step the current down a range, approximately a decade in power, and measure
resistance again. A change in measured value indicates self-heating was present
on the higher range.
3. Continue reducing current until the change is no longer significant to the mea-
surement.
4. If the first step down did not indicate self-heating, step current up to improve res-
olution.
5. Thermal resistance changes with temperature so this process must be repeated
at several temperatures.
2.8.2 Excitation vs.
Signal-to-Noise Since excitation power has such a dramatic effect on self-heating, why doesn't every-
one use the lowest excitation range? The answer is noise, or more appropriately sig-
nal-to-noise. Noise from a variety of sources affects the small signals used to make
bridge measurements. Lowering excitation makes the signals even smaller, but
unfortunately doesn't decrease the noise. The noise sources described in section 2.6
are a fact of life. Even with careful installation, their effects cannot be reduced below
a few microvolts. That is why the Model 372 does not have voltage excitation below
2 µV. The goal of most experiments is to choose an excitation that is the best compro-
mise between self-heating and signal to noise. The monitor output (section 3.7) can
be used to view the amplified signal coming into the input of the Model 372. Seeing
the amplitude and frequency of noise at this point can help identify its sources as an
aid in troubleshooting.
2.8.3 Signal-to-Noise
vs. Measurement
Bandwidth
The noise sources described in section 2.6 combine to reduce the resolution of small
signal measurements. There are some common themes in the way noise interacts
with measurements: noise effects worsen when signal size decreases, when
resistance increases and when bandwidth increases. Excitation is limited by self-
heating and cannot always be increased to create larger signals. Resistance can
seldom be changed because it is the quantity being measured.
That leaves bandwidth to overcome signal-to-noise and improve resolution. If noise
cannot be reduced at its source, it has to be averaged out. Reducing measurement
bandwidth can help with nearly every continuous noise source including induced
noise at discrete frequencies, 1/f circuit noise and broadband thermal noise.
Unfortunately there is a trade-off, and it is time.
All small signal measurement instruments include some type of bandwidth limiting
to improve resolution. The band limiting capabilities are so closely tied to instrument
performance that most resolution specifications are given at a specified time
constant or bandwidth. The Model 372 uses a combination of analog band pass with
digital low-pass filters and averaging. The digital phase sensitive detector (PSD) in the
Model 372 is a very narrow band pass filter. The output of the PSD is run through a
digital low-pass filter with 200 ms time constant to give an equivalent, sampled DC
voltage. The user cannot change these two filters, making 200 ms the widest
bandwidth possible on the Model 372. Once the input signal has been digitally
rectified and filtered using the 200 ms low-pass filter, the user can choose additional
digital averaging to reduce bandwidth further and increase resolution.
Unfortunately, reducing bandwidth increases time constant and settling time slows
the instrument response to real changes in resistance; because those changes are
averaged along with the noise. Probably the most frustrating aspect of using a resis-
tance bridge is realizing that when noise can be as large as the signal being measured,
it takes a lot of time averaging to separate the two. Returning to the thermal noise