The best way to make midrange,
low- to medium-accuracy temperature
measurements (considering size, cost,
performance, and ease of use) is to use
an IC temperature sensor. But most IC
temperature sensors are designed for
applications where the circuits to which
they connect are nearby. Therefore, the
inclusion of sensing, digitizing, and signal-
processing functions in one IC greatly
simplifies the design of such sensors
and the data-acquisition interface.
Some applications, however, require
the acquisition of temperature data from
locations quite remote from the power
supply and data-processing electronics.
These systems demand extra care,
because any deterioration in the signal
from the remote sensor can degrade the
measurement quality.
Choosing to digitize and process the
signal at the point of measurement
(near the sensor) greatly reduces the
problem of signal integrity. However, that
approach also complicates the interconnection
and raises the problem of supplying
power to the circuit. Either design
option - signal processing remotely or at
the sensor - requires cabling that's complex
or expensive or both.
Figure 1 shows a simple and economical
interface for remote IC thermal sensors.
IC1, a MAX6576, is an absolute
temperature-to-period converter that
integrates a sensor with the necessary
signal electronics. It connects to the
receiver (a simple comparator) using a
twisted-pair cable that simultaneously
carries power to the sensor and signals
from the sensor.
At the receiving end, you can recover
the temperature data from the comparator-
output pulses with a simple timer/
counter routine executed by a microcontroller.
Or, in analog form, you can use a
constant-slope, linear saw-tooth generator
synchronized with the received pulses,
followed by a peak sample/hold
(S/H) converter.
With its TS0 and TS1 terminals connected,
the MAX6576 exhibits a digitizing conversion
constant of 10 s/K (Fig. 1). So at
room temperature (300K), the output
pulse period should be 3000 s (3 ms),
which corresponds to a repetition rate of
about 333 Hz. Figure 2 shows this to be
the case, even with a 1000-m (approximately
3300-ft) cable. Similar results were
obtained using a 60-cm (2-ft) cable. Figure
3 shows the receiver's input pulses using
the 1000-m twisted-pair cable. The timing
results were similar to those measured
using the 60-cm cable.
Measurements of the receiver's pulseto-
pulse output jitter (where total measurement
error is the ratio of jitter to the
signal period) indicate that the jitter influence
is negligible, even with the long
cable. This transmission scheme can also
be used with temperature-to-frequency
converters and other sensors.
ALFREDO SAAB, applications engineering manager,
studied in Buenos Aires, Argentina. Previously,
he worked as an American Scientific Associate
at CERN, Geneva, Switzerland.
TINA ALIKAHI, applications engineer, holds a
bachelor's degree in electrical engineering from
Azad (Private) University of Garmsar, Iran.