Analog Devices an273 Application Notes

AN-273

a

APPLICATION NOTE

One Technology Way • P.O. Box 9106 • Norwood, MA 02062-9106 • 781/329-4700 • World Wide Web Site: http://www.analog.com

Use of the AD590 Temperature Transducer

in a Remote Sensing Application

by Paul Klonowski

INTRODUCTION

The AD590 is a two-terminal integrated circuit tempera­ture transducer that produces an output proportional to
absolute temperature. For supply voltages between +4 V
and +30 V the device acts as a high impedance, constant

current source supplying 1 µA/K. Laser trimming of the

chip’s thin-film resistors is used to calibrate the device

to an output of 298.2 µA at 298.2K (+25°C).

A typical application for the AD590 is a remote tempera­ture-to-current transducer. Figure 1 shows a thermom-

eter circuit that measures temperature from –55°C to
+100°C and whose output voltage is 100 mV/°C. Since

the AD590 measures absolute temperature (its nominal

output is 1 µA/K), the output must be offset by 273.2 µA

in order to read out in degrees Celsius. The output cur-

rent of the AD590 flows through a 1 kΩ resistance, devel-

oping a voltage of 1 mV/K. The output of the AD580 2.5 V
reference is divided down by resistors to provide a

273.2 mV offset, which is subtracted from the voltage

across the 1 kΩ resistor by an AD524 instrumentation

amplifier. The amplifier provides a gain of 100, so that

the output range corresponding to –55°C to +100°C is
–5.5 V to +10 V (100 mV/°C). An operational amplifier can

substitute for the instrumentation amplifier, although
care must be taken when designing with the
op amp since the gain at the two input terminals will be
different.

THE PROBLEM

A question often asked of Analog Devices Applications
Engineers by customers using the AD590 in a remote
temperature-to-current application is, “How long can I
make the cable and how can I eliminate any noise that
the cable picks up?” Experiments were performed in an
effort to provide some guidelines for answering this
question using the circuit of Figure 1 with a 1000'
shielded, though initially ungrounded, twisted pair cable
(Belden 9461, style 2092). In order to duplicate actual
conditions the experiments were performed in an indus­trial environment.

TYPES OF NOISE

There are three basic types of noise inherent in a
data-acquisition system. The first type is

noise

: noise received with the original signal and indis-

tinguishable from it. The second type is

transmitted

intrinsic noise:

noise generated within the devices used in a circuit, e.g.,
resistors, op amps, etc. Included in this category are
Johnson, shot and popcorn noise. The third type is

induced noise

: noise picked up from the outside world

and coupled into the circuit. This application note dis­cusses methods of reducing induced noise, which is the
only form of noise that can be influenced by choices of
wiring and shielding.

AD590

I

T

I

T

I

T

+15V

+

7V

–

0.1% LOW

TCR RESISTOR

1mV/K

<

Figure 1. Thermometer Circuit

1kV

AD580

909kV

200V

1kV

G = 100

RG

AD524

2

INSTRUMENTATION AMP

GAIN = 100, 100mV/8C

2kV

E

O

AN-273

NOISE FACTORS

Three elements are involved in any noise problem. The
first is the source of the noise. Possible noise sources
include AM radio signals, logic signals, magnetic fields
and power line transients. The second element is the
coupling medium. That is, how is the noise source en­tering the circuit? Possible coupling mediums include a
common circuit impedance (Figure 2), stray capacitance
(Figure 3) and mutual inductance (Figure 4). A brief de­scription of each follows.

Common impedance noise is developed by an imped­ance common to several circuits. This might occur, as
shown in Figure 2, when a pulse output source and an
op amp’s reference terminal are both connected to a
“ground” point having tangible impedance to the
power supply terminal. The noisy return current of Cir­cuit 1 develops a voltage, V

, across the common im-

NOISE

pedance Z which will appear as a noise signal to Circuit

2. Possible solutions to this problem include proper cir­cuits for distributing power and the use of isolation
transformers and optical isolators.

Capacitively-coupled noise is produced by stray capaci­tance which couples the voltage changing noise source
into high impedance circuits, as shown in Figure 3. The
nature of the impedance Z determines the shape of the
response. Methods of reducing capacitively-coupled
noise include reducing the noise source, properly imple­menting shields and reducing the stray capacitance.

Magnetically coupled noise is produced by mutual in­ductance and can occur, for example, in an incorrectly
shielded cable as shown in Figure 7. Figure 4 is a simple,
model of this incorrectly shielded cable, where L
sents the inductance of the shield, L

represents the in-

C

repre-

S

ductance of one of the center conductors, and L
represents the mutual inductance between the two. The
noise current l(t) flows through L

and establishes a

S

magnetic flux; this time-varying flux also surrounds L
and produces a voltage V
rate of change of the current I(t) flowing through L

proportional to the time

NOISE(t)

. This

S

voltage can be expressed as

dI(t)

V

NOISE(t)

=

LM

dt

The third element involved in any noise problem is the
receiver, or the circuit susceptible to the noise. It is im­portant to understand the role of each of the three ele­ments (the noise source, the coupling medium, and the
receiver) in order to solve the noise problem*. In this
experiment it was determined that the noise sources
were 60 Hz pickup and AM radio signals, the coupling
medium was stray capacitance and the receiver was the
AD524.

L

M

L

I

(t)

L

S

V

C

NOISE(t)

M

C

CIRCUIT 2

= I

V

NOISE

NOISE

POWER SUPPLY COMMON

V

S1

CIRCUIT 1

+

I

Z

COMMON

IMPEDANCE

NOISE

Figure 2. Common Impedance Noise

C

S

V

NOISE

SOURCE

Z

Figure 3. Stray Capacitance Noise

Z

V

NOISE

Figure 4. Mutual Inductance Noise

INITIAL NOISE EFFECT

The photograph in Figure 5 shows the output of the cir­cuit in Figure 1 with the shield ungrounded and with the

+

V

S2

remote AD590 at 30°C. Ideally, the output should be a
3 V (100 mV/°C) dc signal. However, a 60 Hz signal result-

ing from an electric field has been capacitively coupled
into the circuit via the stray capacitance of the cable and
then amplified by a gain of 100. Note, however, that the
60 Hz signal is offset by a dc signal; when the output
voltage of the AD524 was measured with a dc voltmeter
the value read was 3.0 V. This is because the voltmeter
read the value of the dc signal due to the AD590 along
with the average value of the 60 Hz sine wave noise sig­nal. The average value of a sine wave is zero. In essence,
notwithstanding the interfering signal the average value
was correct. The accuracy of all the measurements were
verified through the use of an RTD measurement sys­tem; the AD590 under test was physically attached to
the RTD.

*An excellent article dealing with this subject is “Understanding Inter-

ference-Type Noise” by Alan Rich, found in Section 20 of the Analog
Devices Databook.

–2–

AN-273

Figure 5. EO (Figure 1) with Shield Ungrounded

SHIELDING

One effective way to eliminate electrostatic noise is to
use shielding. A charge resulting from an external po­tential cannot exist on the interior of a closed conduct­ing surface. A shield is, in effect, a closed conducting
surface that surrounds the twisted pair of wires within
the cable.

In order for a shield to be effective it should be con­nected to the reference potential of any circuitry con­tained within the shield. If the signal is connected to
earth ground, then the shield should be connected to
earth ground (see Figure 6). No voltage should exist be­tween the reference potential and the shield conductor.

V

SIG

REFERENCE POTENTIAL

LOAD

Figure 6. Correctly Shielded Cable

Figure 8. EO with Shielded Grounded

Although the 60 Hz signal noise has been eliminated,
voltage spikes are still visible in Figure 8. Figure 9, which
is another view of the AD524 output with the scope ac
coupled and the scale magnitude increased, shows the
high frequency noise still being picked up by the cable.
A look at Figure 10, which is a view of the signal being
fed into the noninverting terminal of the AD524, shows
that the noise being picked up is actually an AM radio
signal.

RF NOISE

RF noise is a combination of an electric field and a mag­netic field, or an electromagnetic field. This electromag­netic field extends into and beyond the space between
conductors. That is, the electromagnetic field and hence
the RF energy is “steered” by the conductors.

V

SIG

"GROUND 1"

V

COMMON

I

"GROUND 2"

LOAD

Figure 7. Incorrectly Shielded Cable

Only one end of the shield should be tied to “ground.”
Tying both ends of the shield to “ground” will produce a
shield current equal to the difference in the potential of
the two “grounds” divided by the series resistance of
the shield (see Figure 7). As previously discussed, the
mutual inductance between the shield and conductors
will couple this noise current into the conductors in the
form of a series voltage, V

NOISE

.

–3–

Figure 9. RF Noise at E

O

AN-273

I

NOISE

1kV

I

DC

0.33mF

In this instance the radio carrier frequency was 550 kHz.
It is important not to ignore RF noise, even if the band­width of the external circuitry is smaller than the band­width of the RF noise. A large RF component will
overload the inputs of the external circuitry and be de­tected, causing an apparent dc shift in the component’s
output signal.

Figure 10. RF Noise at AD524 +Input

RF energy both enters and is reflected in a system wher­ever there is an impedance mismatch or discontinuity in
the system. This includes discontinuities at the end of
the signal run; at the AD590 end and the AD524 end of
the cable. In order to eliminate these discontinuities it
would be necessary to RF shield the entire circuit sys­tem, perhaps using conduit and metal boxes. In most
cases however, this is rather impractical.

What is usually sufficient for most systems is to provide
RF filtering using passive components at the critical
point of interest. It is important to note that this filter
may not eliminate the RF energy but it may just reflect
the energy and redistribute the problem. A circuit topol­ogy which ensures that none of the external circuitry is
affected by the RF noise is discussed later.

BYPASS CAPACITORS

A bypass capacitor can be used to divert some of the
high frequency noise current to ground. Figure 11 is a
simple schematic that shows the effect of the bypass
capacitor. Recalling that the reactance of a capacitor
X

= 1/2 π fC and assuming that we have the minimum

C

radio carrier frequency of 550 kHz, using a 0.33 µF ca-

pacitor the impedance seen by the noise current is:

Figure 11. Bypass Capacitors Reduce RF Noise Effect

Figure 12. AD524 +Input with Bypass Capacitor

X

=1/2 π (550 × 103)(0.33 × 10–6) = 0.88 Ω

C

Of course the direct current supplied by the AD590
will be unaffected by the bypass cap and will continue to

flow into the 1 kΩ resistor. Only the high frequency

noise current will be affected. Figure 12, which is a view
of the signal at the AD524 noninverting terminal, shows

the result of tying a 0.33 µF capacitor across the 1 kΩ

"load” resistor with the shield still grounded. Figure 13
is a look at the output signal of the AD524 with the

0.33 µF capacitor. Compare Figures 10 and 12, and Fig-

ures 9 and 13. The bypass capacitor reduces the noise
magnitude by a factor greater than 5.

Figure 13. EO with Bypass Capacitor

–4–

SERIES RESISTORS

Although an improvement, Figure 13 shows that noise is
still entering the circuit. Another method of reducing the
amount of noise seen at the noninverting terminal of the
op amp is to limit the noise currents through the cable

by adding 1 kΩ resistors in series with the AD590.

This in effect forms a noise-voltage divider between the

load impedance (the 1 kΩ resistor and the 0.33 µF ca-

pacitor) and the series resistors. Figure 14 shows the
output of the AD524 with this final circuit configuration.
The 10 mV p-p amplitude of the output noise is actually
100

× the amplitude of the noise seen at the input of the

AD524, which is 100 µV p-p. Note also that the high fre-

quency spikes in Figure 13 have been eliminated in Fig­ure 14. Figure 15 is the schematic of the final circuit
which has reduced the noise by a factor of 2000 over the
circuit in Figure 1.

AN-273

Figure 14. EO with Bypass Capacitor and Series Resistors

I

T

AD590

<

1kV

I

T

I

T

1kV

0.33mF

+15V

AD580

909kV

+

7V

–

200V

1kV

1kV

G = 100

RG

2kV

E

O

AD524

2

Figure 15.

–5–

AN-273

CONCLUSION

In order to eliminate the effects of RF noise, the circuit of
Figure 16 is recommended.

In Figure 16, the resistors and capacitors are placed at
both ends of the cable. This ensures that the RF noise
remains within the cable and does not affect the external
circuitry. Further note that these resistors can be arbi­trarily large as long as the voltage potential is large
enough to supply the current (V = IR). The use of a by-

R

C

R

AD590

<

R

R

0.33mF

pass capacitor across the AD590 with the series resistors
will filter RF signals. Theoretically the RF signal could
be rectified by the AD590 and offset the accuracy of the
device.

Using the techniques outlined above, noise and interfer­ence can be effectively eliminated through the use of
shielded twisted pair cable and resistors and capacitors.
Thus it is possible to drive the AD590 over 1000 feet of
cable without a loss of accuracy.

E920a–0–7/98

+15V

C

AD580

909kV

+

7V

–

200V

1kV

1kV

G = 100

RG

2kV

E

O

AD524

2

Figure 16.

PRINTED IN U.S.A.

–6–