An introduction to acoustic thermometry
Use an ultrasonic transducer to measure air temperature in an olive jar.
Jim Williams and Omar Sanchez-Felipe, Linear Technology -- EDN, April 21, 2011
At A Glance
|
Acoustic thermometry is an arcane, elegant
technique that measures temperature using
the temperature-dependent transit time of
sound in a medium (references 1 through
4). The medium can be a solid, a liquid, or a
gas. Acoustic thermometers function in environments,
including extreme temperatures,
destructive physical abuse, and nuclear reactors,
that conventional sensors cannot tolerate. Sonic speed in
air varies predictably as the square root of temperature. The sonic
transit time in a gas-path thermometer is almost entirely insensitive
to pressure and humidity. Gas-path acoustic thermometers
respond quickly to temperature changes. They have essentially
no thermal mass or lag.
You take a measurement across the entire body of an acoustic thermometer. The measurement represents the total path-transit time. Conventional sensors, in contrast, measure at a single point. An acoustic thermometer is thus blind to temperature variations within the measurement path. It infers temperature from the isothermal or the nonisothermal measurement path’s delay.
Practical considerations
A practical acoustic-thermometer
demonstration begins with selecting
a sonic transducer and a dimensionally
stable measurement path. A wideband
ultrasonic transducer promotes
fast, low-jitter, high-fidelity response
free of resonance and other parasitic
losses. An electrostatic ultrasonic sensor
(Figure 1) meets these requirements
(Reference 5). An ultrasonic transducer
serves as both transmitter and
receiver. The device is rigidly mounted
on the metal cap of a glass enclosure.
You should stiffen the cap to ensure
dimensional stability of the measurement
path. This design uses the bottle
and cap from Reese Cannonball olives
(Figure 2). Barometric pressure is not
a major variable in transit time because
it does not bow the stiffened cap, averting
errors.
You should remove the olives and
their residue, and bake out the bottle and
cap at 100°C. Pass the transducer leads
through the cap using a coaxial header.
The glass enclosure has a relatively small
thermal-expansion coefficient. This
arrangement makes the path distance
stable with temperature, pressure, and
mechanical changes. The round-trip
path length is approximately 12 in. The
speed of sound in air is 1.1 feet/msec.
Thus, the round-trip time is 900 μsec.
The path’s temperature-dependent variation
is approximately 1 μsec/°F at 75°F.
To achieve a 0.1°F resolution requires
mechanical and electronic variations of
less than 100 nsec, which in turn requires
a 0.001-in. dimensional stability referred
to the 12-in. path length. When you
examine the likely error sources, you will
see this stability as a realistic goal.You bias the transducer, which acts as a capacitor, at 150V dc (Figure 3). The start-pulse clock drives the transducer with a short impulse, launching an ultrasonic event into the measurement path. The start-pulse clock simultaneously sets the width-decoding flip-flop high. The sonic impulse bounces off the enclosure’s bottom, travels back, and impinges on the transducer, resulting in a minuscule mechanical displacement, which in turn changes the capacitance of the transducer. Based on the equation relating charge, Q, to capacitance and voltage, Q=C×V, the capacitance change creates a voltage change at the receiver amplifier’s input.
The trigger comparator converts
the amplifier’s output excursion into
a logic-compatible level, resetting the
width-decoding flip-flop. The flip-flop’s
output width represents the measurement
path’s temperature-dependent sonic transit time. You program the
microprocessor with the measurement
path’s temperature and delay calibration
constants (see sidebar “Measurement-path
calibration”). The microprocessor
can then calculate the temperature
using the pulse width and supply this
information to the display.
You derive a second output from
the start-pulse generator, which gates
off the trigger’s output during most of
the measurement cycle, enabling the
trigger output only during the time
when you expect a return pulse. This
method eliminates false triggers by discriminating
against unwanted sonic
events that originate outside the measurement
path. The transducer’s return
pulse amplitude is less than 2 mV. The
high-gain, wideband receiver amplifier
is vulnerable to parasitic inputs, so you
must shut down the 150V bias supply
during the measurement to prevent its
switching harmonics from corrupting
the amplifier. You derive a second gate
from the width-decoding flip-flop. It
shuts down the 150V bias supply during
the measuring interval.A measurement cycle begins with a start pulse driving the transducer (Trace A in Figure 4), setting the flip-flop (Trace B) high. After the sonic impulse’s transit time, the amplifier responds (Trace C), tripping the trigger, which resets the flip-flop (Trace D). The gate signals protect the trigger from unwanted sonic events, start-pulse artifacts, and shut off the high-voltage regulator during measurement (traces E and F).
 |
|
Click here to view a larger version of Figure 5 |
Detailed circuitry
A silicon oscillator furnishes the 100-Hz
clock (Figure 5). Monostable pulse generator
ICA provides a 10-μsec pulse to a
driver you make with Q1 and Q2. You
capacitively couple the driver output’s
start pulse into the ultrasonic transducer
(Trace A in Figure 6). The monostable pulse generator simultaneously sets the
flip-flop high (Trace E). The high output
from the flip-flop shuts down the
high-voltage switching regulator during
the measurement. You set up the
monostable pulse generator, ICB, to produce
a second pulse. This pulse gates off
comparator IC1’s output for a time just
shorter than the expected sonic return
pulse (Trace B).
The sonic pulse travels down the
measurement path, bounces, and returns
to impinge on the transducer. The
switching regulator biases the transducer
at 150V dc. It operates as a cascode of the
internal switching transistor in the IC
and high-voltage transistor Q3. This high
voltage allows the small capacitance
change that the diaphragm’s motion creates
to generate an appreciable voltage
change. You send this voltage change
to the receiver amplifier. Capacitive
coupling isolates the high-voltage dc-transducer
bias, and diode clamps prevent
destructive overloads. The cascaded
receiver amplifier has an overall gain of
17,600. You can monitor the amplifier at
the low-impedance output of A2 (Trace
C). A3, the last stage in the cascade,
further amplifies the signal. You send
the amplifier output to comparator IC1,
which creates an output trigger (Trace
D) that occurs at the first event that
exceeds its negative-input threshold.
The trigger resets the width-decoding
flip-flop. The flip-flop pulse width then
represents the temperature-dependent
acoustic transit time. You send this pulse
to the microprocessor, which determines
and displays the temperature (Reference
6). See sidebar “Software code” for the complete processor-software code.
You can expand your oscilloscope’s
signals at the return impulse trip point
(Figure 7). You monitor the amplifier’s
response at the output of A2 (Trace A).
Amplifier A3 adds gain, which softly saturates
the signal’s leading-edge response (Trace B). Comparator IC1’s trigger output
creates multiple triggers (Trace C).
However, the flip-flop output remains
high after the initial trigger, providing an
accurate transit time (Trace D).Gate the high-voltage supply to
prevent switching harmonics from producing
spurious amplifier outputs. The
start pulse sets the flip-flop output high
(Trace A in Figure 8), shutting down
high-voltage switching at the onset of
the measurement period (Trace B). This
state persists during the entire transit
time and prevents erroneous amplifier-trigger
outputs. A return-pulse trigger
resets the flip-flop (Trace A in Figure
9)
You send the flip-flop output to a circuit that gates off the switching regulator by modulating its compensation pin, VC, delaying the high-voltage turn-on until after the measurement period (Trace B) and ensuring a clean, noise-free trigger signal.
You send the flip-flop output to a circuit that gates off the switching regulator by modulating its compensation pin, VC, delaying the high-voltage turn-on until after the measurement period (Trace B) and ensuring a clean, noise-free trigger signal.
Gating the trigger output prevents interference from outside sources. Gating the 150V converter prevents its harmonics from corrupting the receiver’s amplifier. The 150V supply value is a gain term. The higher it is, the more signal it returns. Gating off its regulation during the measurement is a potential concern. Practically, the 1-μF output capacitor decays only 30 mV, or approximately 0.02%, during this time. This small variation is constant and insignificant, and you can ignore it. You derive the trigger trip point and the start pulse from the same 15V supply, enhancing stability because it makes the trigger voltage vary ratiometrically with the received signal’s amplitude.
The wideband, highly
sensitive, and resonance-free transducer
descends from 1970s-era Polaroid SX-70 automatic-focus
cameras, promoting repeatable, jitter-free
operation. All of these attributes directly contribute to the 100-nsec, 0.1°F
resolution of the circuit for a 1-msec
travel time, representing less than 100-ppm uncertainty. Once you calibrate the
circuit, the absolute accuracy at 60 to
90°F is within 1°F.Further investigations might have you attempt to trigger the receiver after multiple bounces (Figure 10). This approach offers the benefit of easing timing tolerances. The return signals decay into noise that acoustic dispersion in the glass enclosure creates. Triggering on a later bounce would relax your timing margins but also gives you an unacceptable SNR (signal-to-noise ratio). Signal-processing techniques could overcome this problem, but the effort would have to justify the increased resolution.
|
References |
|
Measurement-path calibration
Theoretically, you can calculate temperature-calibration constants from the measurement-path length. In practice, it is difficult to determine the path length to the required accuracy because of glass-jar, transducer, and mounting dimensional uncertainties. Instead, you must calibrate the device over known temperatures (Figure A). Place the enclosure in a controllable thermal chamber with an accurate thermometer and ensure that the thermometer is isothermal with the enclosure. Next, vary the chamber temperature over 10 evenly spaced points at 60 to 90°F. The enclosure has a 30-minute time constant to settle within 0.25°F. You must allow adequate time for each step to stabilize before taking readings. Note the pulse width at each temperature and record the data. You then load this information into the microprocessor memory.
Theoretically, you can calculate temperature-calibration constants from the measurement-path length. In practice, it is difficult to determine the path length to the required accuracy because of glass-jar, transducer, and mounting dimensional uncertainties. Instead, you must calibrate the device over known temperatures (Figure A). Place the enclosure in a controllable thermal chamber with an accurate thermometer and ensure that the thermometer is isothermal with the enclosure. Next, vary the chamber temperature over 10 evenly spaced points at 60 to 90°F. The enclosure has a 30-minute time constant to settle within 0.25°F. You must allow adequate time for each step to stabilize before taking readings. Note the pulse width at each temperature and record the data. You then load this information into the microprocessor memory.
Talkback
-
Wow! Jim did a project with a Micro???? Is there no stability in the world???? (he, he, he...) Welcome to the "Dark Side" Jim.... :-)
BTW - Cool project
Steve Hageman - 2011-23-5 08:23:23 PDT -
Built PCB is on my blog, anablog
edn.com/blog/Anablog/40300-Jim_Williams_acoustic_thermometer_redux.php
If that link won't hunt, search for Jim_Williams_acoustic_thermometer_redux
in EDN' search box. Google doesn't love me.
Paul Rako - 2011-9-5 09:42:00 PDT -
I would like to see pictures of the built pcb.
Wanderer - 2011-9-5 08:23:28 PDT -
The method and circuit design is interesting. But the temperature measurement of gas is based upon propagation time which can be not uniform over the length or volume of gas.
I think a more interesting circuit or design could be used to observe temperature distribution in a small or larger volume.
Jiri Polivka - 2011-26-4 10:38:37 PDT -
Wonderful! Our firm Silacon Corporation would have another application of your circuit. Please contact us.
Charles Nutter - 2011-21-4 12:58:49 PDT
