Noty an07f Linear Technology

Application Note 7

February 1985

Some Techniques for Direct Digitization of Transducer Outputs

Jim Williams

Almost all transducers produce low level signals. Normally,
high accuracy signal conditioning amplifi ers are used to
boost these outputs to levels which can easily drive cables,
additional circuitry, or data converters. This practice raises
the signal processing range well above the error fl oor,
permitting high resolution over a wide dynamic range.

Some emerging trends in transducer-based systems are
causing the use of signal conditioning amplifi ers to be
reevaluated. While these amplifi ers will always be useful,
their utilization may not be as universal as it once was.
In particular, many industrial transducer-fed systems are
employing digital transmission of signals to eliminate
noise-induced inaccuracies in long cable runs. Addition­ally, the increasing digital content of systems, along with
pressures on board space and cost, make it desirable to
digitize transducer outputs as far forward in the signal chain
as possible. These trends point toward direct digitization
of transducer outputs—a diffi cult task.

Classical A/D conversion techniques emphasize high level
input ranges. This allows LSB step size to be as large
as possible, minimizing offset and noise-caused errors.
For this reason, A/D LSB size is almost always above a
millivolt, with 100μV to 200μV per LSB available in a few
10V full-scale devices. The requirements to directly A/D
convert the output of a typical strain gauge transducer are
illuminating. The transducer ’s full-scale output is 30mV,

meaning a 10-bit A/D converter must have an LSB incre­ment of only 30μV. Performing a 10-bit conversion on a
type K thermocouple monitoring a 0°C to 60°C environment
proves even more stringent. The type K thermocouple
puts out 41.4μV/°C over the 0°C to 60°C range. The LSB
increment is found by:

60 41 4

°°

CμVC

•. /

These examples furnish extraordinarily small step sizes,
far below commercially available A/D units and seemingly
impossible to digitize without DC preamplifi cation. In fact,
both transducers’ outputs may be directly digitized to stable
10-bit resolution using circuitry specifi cally designed for
the function.

This application note details circuit techniques which
directly digitize the low level outputs of a variety of trans­ducers. The approaches described are unique in that they
do not utilize any DC gain stage. The transducer outputs
receive no DC signal conditioning; A/D conversion is directly
performed at low level. The circuits produce a serial data
output which may be transmitted over a single wire with
the characteristic noise immunity of digital systems. By
eliminating the traditional DC gain stage, these circuits
furnish a direct, economical way to digitize low level
transducer outputs without sacrifi cing performance.

1024

242

=

μV LSB

./

an7f

AN7-1

Application Note 7

Figure 1 shows a simple way to convert the current output
of an LM334 temperature sensor to a corresponding output
frequency. The sensor pulls a temperature-dependent cur­rent (0.33%/°C) from A1’s positive input node. This point,
biased from the LM329-driven resistor string, responds
with a varying, temperature-dependent voltage. The voltage
varies the operating point of A1, confi gured as a self-reset­ting integrator. A1 integrates the LM329 referenced current
into its summing point, producing a negative-going ramp
at its output. When the ramp amplitude becomes large
enough, the transistors turn on, resetting the feedback
capacitor and forcing A1’s output to zero. When the capaci­tor’s reset current goes to zero, the transistors go off and
A1 begins to integrate negatively again. The frequency of
this oscillation action is dependent on A1’s DC operating
point, which varies with the LM334’s temperature. The
circuit’s DC biasing values are arranged so that a 0°C to
100°C sensor temperature excursion produces 0kHz to
1kHz at the output. Additionally, only 2V appear across
the LM334, minimizing sensor power dissipation related
errors. The differentiator-transistor network at A1’s output
provides a TTL compatible output. To calibrate this circuit,
place the LM334 in a 0°C environment and trim the “0°C
adjust” for 0Hz. Next, put the LM334 in a 100°C environ­ment and set the “100°C adjust” for 1kHz output.

Repeat this procedure until both points are fi xed. This circuit
has a stable 0.1°C resolution with ±1.0°C accuracy.

Figure 2 shows another temperature-to-frequency convert­er, but this circuit uses the popular type K thermocouple as
a sensor. The design includes cold junction compensation
for the thermocouple over a 0°C to 60°C range. Accuracy
is ±1°C and resolution is 0.1°C.

The thermocouple’s extremely low output (41.4μV/°C) and
the requirement for cold junction compensation make it
one of the most diffi cult transducers to directly digitize.
The approach used is based on the 50nV/°C input offset

®

drift performance of the LTC

1052 chopper-stabilized

amplifi er.

In this circuit, A1’s positive input is biased by the ther­mocouple. A1’s output drives a crude V→F converter,
comprised of the 74C04 inverters and associated com­ponents. Each V→F output pulse causes a fi xed quantity
of charge to be dispensed into a 1μF capacitor from the
100pF capacitor via the LTC1043 switch. The larger ca­pacitor integrates the packets of charge, producing a DC
voltage at A1’s negative input. A1’s output forces the V→F
converter to run at whatever frequency is required to bal­ance the amplifi er’s inputs. This feedback action eliminates
drift and nonlinearities in the V→F converter as an error

560 1k* 1k*

15V

6.2k* 6.2k* 6.2k*

LM329

500Ω

0°C ADJ

820Ω*

510Ω

Figure 1. Temperature-to-Frequency Converter

2k

100°C ADJ

0.01

POLYSTYRENE

–

A1

LT1056

+

*1% FILM RESISTOR
NPN = 2N2222
PNP = 2N2907

510pF

15V

10k

2V

2.7k

10k

TTL OUT
0kHz TO 1kHz
0°C TO 100°C

4.7k

LM334-3

137Ω*

AN07 F01

an7f

AN7-2

–
+

TYPE K

THERMOCOUPLE 41.4μV/°C

OPTIONAL INPUT

FILTER-OVERLOAD

CLAMP

50k

60°C TRIM

A

A1 LTC1052

STABILIZING AMP

1N914100k

1μF

150k**

†

COLD JUNCTION BIAS

0.1μF

33k**

1μF

301k*487Ω*

3

2

+

LTC1052

–

1

5V

–5V

7

4

16

LTC1043

8

2

6

0.1μF

3300pF

100pF

CHARGE
PUMP

65

R

T

COLD JUNCTION TEMPERATURE TRACKING

1.8k*

33k

1μF

74C04

0.68μF

74C04

4.75k*

1k*

1N4148

A

10k

5V

F

–5V

LT1004

1.2V

Application Note 7

470Ω

B

74C04C74C04D74C04E74C04

820pF

VmF

5V

3k

OUTPUT
0Hz TO 600Hz
0°C TO 60°C

187Ω*

0.01% FILM-TRW MAR-6

*

TRW/MTR/5/ + 120

**

= YELLOW SPRINGS INST. #44007

R

T

100pF = POLYSTYRENE

†

FOR GENERAL PURPOSE (1mV FULL-SCALE)
10-BIT A/D, REMOVE THERMOCOUPLE—
COLD JUNCTION NETWORK, GROUND POINT A
AND DRIVE LTC1052 POSITIVE INPUT

Figure 2. Thermocouple-to-Frequency Converter

term and the output frequency is solely a function of the
DC conditions at A1’s inputs. The 3300pF capacitor forms
a dominant response pole at A1, stabilizing the loop.

A1’s low drift eliminates offset errors in the circuit, despite
an LSB value of only 4.14μV (0.1°C)!

, a thermistor, and the 1.8k, 187Ω, 487Ω and 301k

R

T

values form a cold junction compensation network which

®

is biased from the LT

1004 1.2V reference. In addition
to cold junction compensation, the network provides
offsetting, permitting a 0°C sensor temperature to yield
0Hz at the output.

AN07 F02

Figure 3 details circuit operation. A1’s output drives the
33k-0.68μF combination, producing a ramp (Trace A,
Figure 3) across the capacitor. When the ramp crosses
inverter A’s threshold, the cascaded inverter chain switches,
producing a low output at E (Trace B). This causes the

0.68μF capacitor to discharge through the diode, resetting
the capacitor to 0V. The 820pF unit provides positive AC
feedback to inverter B’s input (Trace C), assuring a clean
reset. The frequency of this ramp-and-reset sequence varies
with A1’s output. Inverter F ’s output controls the LTC1043
switch. When the inverter output is high, Pins 2 and 6 are
connected, allowing the 100pF capacitor to charge to a

an7f

AN7-3

Application Note 7

A = 100mV/DIV

B = 10V/DIV

C = 10V/DIV

Figure 4 is another temperature measuring circuit, but
the transducer used is unusual. The circuit measures
temperature by utilizing the relationship between the
speed of sound and temperature in a medium. In dry air
the relationship is governed by:

D = 10μA/DIV

HORIZONTAL = 200μs/DIV

Figure 3. Thermocouple Digitizer Waveforms

AN07 F03

potential derived from the LT1004 1.2V reference. When
the inverter goes low, Pin 2 is connected to Pin 5. During
this interval, the 100pF capacitor completely discharges
(Trace D) into the 1μF unit. The amount of charge delivered
is constant over each cycle (Q = CV), so the voltage the
1μF capacitor charges to is a function of frequency and
discharge path resistance. This voltage is summed with
the LT1004-derived offsetting potential at A1’s negative
input, closing a loop around A1. The –120ppm/°C drift
of the 100pF charge-dispensing polystyrene capacitor is
compensated by the opposing tempco of the specifi ed
resistors used in the 1μF’s discharge path. Typical circuit
gain is 20ppm/°C, allowing less than 1LSB (0.1°C) output
drift over a 0°C to 70°C ambient operating range.

The thermocouple’s known characteristics, combined with
A1’s low offset and the cold junction/offsetting network
components specifi ed, eliminate zero trimming. Calibration
is accomplished by placing the thermocouple in a 60°C
environment and adjusting the 50kΩ potentiometer for
a 600Hz output. Beyond 60°C the cold junction network
departs from the thermocouple’s response and output
error increases rapidly. Although the digital output will
be a function of the thermocouple’s temperature over
hundreds of degrees, linearization by a monitoring pro­cessor is required.

It is worth noting that this circuit can directly convert any
low level, single-ended signal. If the offsetting/cold junction
network is removed and the 50kΩ potentiometer returned
directly to ground, inputs may be applied to A1’s positive
terminal. The circuit produces a 10-bit accurate output
with a full-scale range of only 1mV (1μV per LSB)! The
high impedance of A1’s input allows fi ltering or overload
clamping of the input signal without introducing error.

C = 331 5,

T

meters/second

273

where C = speed of sound.

Acoustic thermometry is used where extremes in operat­ing temperature are encountered, such as cryogenics and
nuclear reactors. Additionally, acoustic temperature stan­dards have been built by operating the acoustic transducer
inside a sealed, known medium.

The inherent time domain operation of acoustic ther­mometers allows a direct conversion into a digital output.
Figure 4 shows a circuit that does this. A1, the inductor,
and their associated components for a simple fl yback
type regulated 200V supply which biases the transducer.
The transducer is composed of the Polaroid ultrasonic
element noted, mounted at one end of a sealed, 6-inch
long Invar tube. The Invar material minimizes mechanical
tube deformation with temperature. The medium inside
the tube is dry air. The transducer may be thought of as a
capacitor, composed of an insulating disc with a conduc­tive coating on each side.

Each time the TTL clock (Trace A, Figure 5) goes high,
the transducer receives AC drive via the 0.22μF capacitor.
This drive causes mechanical movement of the disc and
ultrasonic energy is emitted. The clock input simultane­ously sets the 74C74 fl ip-fl op output (Trace E) low and
pulls the 0.01μF capacitor to ground. This cuts off drive
to C1’s 3k output pull-up resistor (Trace C), forcing C1’s
output (Trace D) to zero. During the clock pulse’s period,
A2’s output (Trace B) is saturated due to excessive signal
at its input. When the clock pulse ceases, A2 comes out
of bound and amplifi es in its linear region. The ultrasonic
transducer now acts like a capacitance microphone, with
the 200V supply providing bias. Residual disc ringing is
picked up and appears at A2’s output. This signal cannot
trigger C1, however, because the 0.01μF capacitor has not
charged high enough to allow the inverter to chain output
to bias C1’s output pull-up resistor.

AN7-4

an7f

Application Note 7

The ultrasonic energy emitted by the transducer travels
down the tube, bounces off the far end and heads toward
the transducer. Before it returns, the 0.01μF capacitor
crosses the inverter’s threshold and C1’s 3k resistor
(Trace C) receives bias. Upon returning, the sonic energy
causes a mechanical displacement of the transducer, forc­ing a shift in capacitance. This capacitance shift causes
charge to be displaced into C2’s summing point, and the

INVAR

TUBE

0.22

600V

100M**

ENCLOSURE

TRANSDUCER

10k

200V

1N645

0.01

2k 39k

1N645

1N914

0.02

–

LT1056

+

10M

A2

–15V

output responds with an amplifi ed version of this signal
(Trace B). C1’s output (Trace D) triggers, resetting the fl ip­fl op. The fl ip-fl op’s output pulse (Trace E) represents the
transit time down the tube and will vary with temperature
according to the equation given. A monitoring processor
can convert this pulse width into the desired temperature
information.

1N4148

15V

2N3440

10pF

14.7k

250Ω

L

–

+

15V

15V

C1

LT1011

–15V

7

1

3k

22k

INPUT

TTL CLOCK INPUT

100Hz, 10μs

–15V

1.2M*

10k

LT1004

2.5V

1/6 74C04

A = 20V/DIV

B = 20V/DIV

C = 20V/DIV

D = 20V/DIV

–

100Ω

15V

R Q

74C74

S

+

22

1N4148

LM307

+

A1

Figure 4. Acoustic Thermocouple

180k

L = AIE VERNITRON-24-104 1MHy
TRANSDUCER = POLAROID-604029
ENCLOSURE = INVAR 6" TUBE SEALED, DRY AIR

*

1% FILM RESISTOR

**

VICTOREEN #MOX-300

74C04

0.01

WIDTH
OUTPUT

E = 20V/DIV

HORIZONTAL = 200μs/DIV

Figure 5. Acoustic Thermocouple Waveforms

AN07 F05

an7f

AN7-5