Chapter V: Applying In-Amps Effectively

Chapter V
APPLYING IN-AMPS EFFECTIVELY
Dual-Supply Operation
The conventional way to power an in-amp has been
from a split or dual polarity power supply. This has
the obvious advantage of allowing both a positive and
a negative input and output swing.
Single-Supply Operation
Single-supply operation has become an increasingly
desirable characteristic of a modern in-amp. Many
present day data acquisition systems are powered from
a single low voltage supply. For these systems, there
are two vitally important characteristics. First, the
in-amp’s input range should extend between the positive
supply and the negative supply (or ground). Second, the
amplifier’s output should be rail-to-rail as well.
The Need for True R-R Devices in Low Voltage,
Single-Supply IA Circuits
Sometimes problems arise when designers forget
about amplifier headroom and use standard (non
rail-to-rail) products in low voltage, single-supply
in-amp applications. Many dual-supply in-amps will
only have an output swing within about 2 V of either
rail. However, even the very best cannot swing as
close to the rails as a single-supply device.
In the example shown in Figure 5-1, the p-p output swing
would only be about 1 V p-p for a standard product vs.
4 V p-p or more for a rail-to-rail in-amp.
Power Supply Bypassing, Decoupling, and
Stability Issues
Power supply decoupling is an important detail that
is often overlooked by designers. Normally, bypass
capacitors (values of 0.1 F are typical) are connected
between the power supply pins of each IC and ground.
Although usually adequate, this practice can be ineffective
or even create worse transients than no bypassing at all.
It is important to consider where the circuit’s currents
originate, where they will return, and by what path. Once
that has been established, bypass these currents around
ground and other signal paths.
In general, like op amps, most monolithic in-amps have
their integrators referenced to one or both power supply
lines and should be decoupled with respect to the output
reference terminal.This means that for each chip, a bypass
capacitor should be connected between each power supply
pin and the point on the board where the in-amp’s reference
terminal is connected, as shown in Figure 5-2.
A good quality rail-to-rail in-amp, such as the AD623, can
swing its output to within 0.5 V of the supply and within
0.1 V above ground. Its input voltage range is similar.
Note that these numbers are conservative: with very light
output loading the performance is even better. So, when
using a 5 V single supply, the amplifier has at least a 4 V
output swing, in many cases, more than that.
+VS
).!-0
M&
M&
M&
M&
63
+VS
6/54
–IN
R-R
IN-AMP
VOUT
+VS
2
+IN
n63
GND
Figure 5-1. A modern rail-to-rail in-amp can
swing more than 4 V p-p with a 5 V supply, but
this is NOT possible using a standard dualsupply device. Note that the in-amp’s VREF pin
would normally be set at VS/2 for maximum
output swing.
Figure 5-2. Recommended method for
power supply bypassing.
For a much more comprehensive discussion of these
issues, refer to Application Note AN-202, An IC Amplifier User’s Guide to Decoupling, Grounding, and Making
Things Go Right for a Change, by Paul Brokaw, on the
ADI website at www.analog.com.
5-1
THE IMPORTANCE OF AN INPUT GROUND
RETURN
+VS
AC coupling is an easy way to block dc voltages that
are present at the in-amp’s inputs. But ac coupling into
a high impedance in-amp input without providing a dc
return renders the circuit nonfunctional! This is one of
the most common applications problems that arises when
using in-amp circuits. Figure 5-3 shows two common
(incorrect) arrangements.
R1
IB2
VOUT
IN-AMP
REFERENCE
INPUT
R2
–VS
VOS = (IB1 R1) – (IB1 R2)
Figure 5-4. A high value resistor between
each input and ground provides an effective
dc return path (see Table 5-1).
C1
–IN
IN-AMP
C2
+IN
+VS
C2
IB1
–IN
THESE CIRCUITS WILL NOT WORK!
A
C1
This is a simple and practical solution for dual-supply
in-amp circuits. The resistors allow a discharge path for
input bias currents. Now both inputs are referenced to
ground. There will be a small offset voltage error due to
the mismatch between the input bias currents flowing
through the two nonidentical resistors. To avoid errors
due to an R1/R2 mismatch, a third resistor, about onetenth their value, can be connected between the two
in-amp inputs.
VOUT
+IN
–VS
TRANSFORMER COUPLED
WITH NO DC RETURN
Figure 5-5 shows the recommended dc return for a
transformer-coupled input.
B
+VS
– INPUT
Figure 5-3. Nonfunctional, ac-coupled
in-amp circuits.
RG
In Figure 5-3A, the input bias currents will charge up
the ac coupling capacitors until the input common-mode
voltage is exceeded. In other words, the caps will charge
up to the supply line or down to ground depending
on the direction of the input bias currents. Now, with
a FET input device, and very large capacitors, it could
take several minutes before the in-amp is rendered inoperative. As a result, a casual lab test might not detect
this problem, so it’s very important to avoid it altogether.
Figure 5-3B shows a transformer-coupled input with no
center tap or other means for a dc return; so, the same
problem occurs.
IN-AMP
VOUT
LOAD
+ INPUT
–VS
REFERENCE
TO POWER
SUPPLY
GROUND
Figure 5-5. Recommended dc return path
for a transformer-coupled input.
For transformers without a center tap, two resistors, one
from each input pin to ground, can be used to provide
a dc return path.
A simple solution for the circuit in Figure 5-3A is to add
a high value resistance (R1, R2) between each input and
ground, as shown in Figure 5-4. The input bias currents
can now flow freely to ground and do not build up a large
input offset as before.
5-2
Providing Adequate Input and Output Swing
(“Headroom”) When AC Coupling a Single-Supply
In-Amp
AC coupling using an in-amp powered by a single
supply is more complicated than dual-supply operation
and normally requires applying a dc common-mode
voltage, VCM, to both inputs, as shown in Figure 5-6.
This is necessary because the internal buffer amplifiers
of a typical 3-op amp in-amp cannot swing more than
a few millivolts below the negative supply (in this case
ground) without clipping the signal. In addition, the
output can never swing below the negative supply.
63
6##
).
2
).!-0
Output centering is similar: estimate the amount and the
direction of the in-amp’s output swing (in most cases this
will be VIN 3 gain + VREF) and then apply a reference
voltage at VREF that is in the center of that range.
Selecting and Matching RC Coupling Components
In ac-coupled applications, selecting values for the
capacitors and dc return resistors is a trade-off between
–3 dB bandwidth, noise, input bias current, and the
physical size of the capacitors. RC components should be
reasonably matched so that the time constant of R1/C1
is close to that of R2, C2. Otherwise, a common-mode
voltage can be converted into a differential error.
#
n).
+2 V (or a bit higher), which will allow 2 V of headroom
in the minus direction. The trade-off is that there is now
2 V less swing in the positive direction. If the in-amp is
operating with gain, VCM needs to be tailored to allow
the buffer outputs to swing fully without clipping (that
is, without exceeding their maximum voltage swing in
either direction).
6/54
2
2%&%2%.#%
).054
Higher value input capacitors provide greater low
frequency bandwidths and allow smaller value input
resistors. But these caps are physically larger, requiring more board space. As a rule, capacitors larger
than 0.1 microFarad need to be polarized types such as
tantalums to keep their size manageable. But polarized
caps require a known, constant polarity dc voltage to
keep them properly biased.
62%&
490)#!,,963
Figure 5-6. An ac-coupled, single-supply,
in-amp circuit. For maximum dynamic range,
set VCM to the center of the maximum input
range. VREF is typically set to VS/2.
Choosing appropriate voltages for VCM and VREF is an
important design consideration, especially in low supply
voltage applications. In general terms, VCM should be
set to the middle of the expected input dynamic range,
while VREF should be centered on the expected output
dynamic rage. As an example, say the input signal
(the difference between +IN and –IN) is expected to
be between +1 V and –2 V. Under these conditions,
the in-amp’s input buffers will need to swing both
positive and negative with respect to ground. Assume
the in-amp is operating at unity gain, VCM can be set to
Smaller cap values require higher value input resistors
which have higher noise. And with larger input resistors,
dc offset voltage errors also become larger. So, there is
always a trade-off that needs to be made here.
Since (IB1R1) – (IB2R2) = DVOS, any mismatch between
R1 and R2 will cause an input offset imbalance
(I B1 – IB2). The input bias currents of Analog Devices
in-amps vary widely, depending on input architecture.
However, the vast majority have maximum input bias
currents between 1.5 nA and 10 nA. A good guideline
is to keep IBR < 10 mV.
5-3
Table 5-1. Recommended Component Values
for AC Coupling In-Amp Inputs
Input Bias VOS
RC Coupling Components Current
at Each
–3 dB BW C1, C2
R1, R2
(IB)
Input
2 Hz
2 Hz
30 Hz
30 Hz
100 Hz
100 Hz
500 Hz
500 Hz
1 M
1 M
115 k
53.6 k
162 k
162 k
162 k
162 k
0.1 F
0.1 F
0.047 F
0.1 F
0.01 F
0.01 F
0.002 F
0.002 F
2 nA
10 nA
2 nA
10 nA
2 nA
10 nA
2 nA
10 nA
Table 5-1 gives typical R and C cookbook values for ac
coupling using 1% metal film resistors and two values
of input bias current.
Properly Driving an In-Amp’s Reference Input
Another common applications problem occurs when
a high impedance source is used to drive an in-amp’s
reference terminal. In the example shown in Figure 5-7,
R2’s added resistance unbalances the otherwise closely
matched resistors in subtractor amplifier A3. A resistor
divider is shown here, but the same problem is created
with any input source that is even a small percentage of
RREF. This results in both a CMR error and a reference
voltage error.
REFERENCE
INPUT
R1
R2
–
IN-AMP
EXTERNAL
REFERENCE
VOLTAGE
OP AMP
BUFFER
RREF
EXTERNAL
VOLTAGE
DIVIDER
+
SUBTRACTOR
A1
Figure 5-8. A simple op amp buffer provides a
low impedance for driving an in-amp’s input.
2
A3
VOUT
6
3
VIN+
40 V
200 V
5 V
11 V
7 V
32 V
7 V
32 V
A CMR error is introduced because RREF2 no longer
equals RREF1 (R2 is in series). Note that RREF1 and RREF2
are not always equal but that introducing any significant
series resistance between the VREF terminal and ground
will unbalance A3 and cause a CMR error. A reference
voltage error is also produced because R2 becomes
loaded by the in-amp’s finite reference impedance.
IN-AMP
VIN–
2 mV
10 mV
230 V
536 V
324 V
1.6 mV
324 V
1.6 mV
VOS Error
for 2%
R1, R2 Mismatch
A2
RREF1 RREF2
REFERENCE
INPUT
PROBLEM: R2’s RESISTANCE
CAUSES CMR ERROR AND RREF1
AND RREF2 ARE NOW UNBALANCED.
AN ADDITIONAL VOLTAGE
REFERENCE ERROR IS INTRODUCED
BY THE SHUNTING OF R2 BY RREF1
AND RREF2.
EXTERNAL
VOLTAGE
DIVIDER
R1
R2
EXTERNAL
REFERENCE
VOLTAGE
Figure 5-8 shows a simple solution to this problem.
An op amp buffer is added between the voltage divider
(or other high-Z source) and the in-amp’s reference
terminal. Now, the in-amp only sees the very low
output impedance of the op amp, which typically is
much less than one ohm.
Many other solutions are possible, as long as the impedance driving the reference terminal is kept very low.
Figure 5-7. When using this simple voltage
divider, RREF1 and RREF2 are now unbalanced,
which introduces a large CMR error in A3.
5-4
Cable Termination
When in-amps are used at frequencies above a few hundred kilohertz, properly terminated 50  or 75  coaxial
cable should be used for input and output connections.
Normally, cable termination is simply a 50  or 75 
resistor connected between the cable center conductor,
and its shield is at the end of the coaxial cable. Note that
a buffer amplifier may be required to drive these loads
to useful levels.
Figure 5-9 provides details on the input architecture
of the AD8221 in-amp. As shown, they have internal
400  resistors that are in series with each input
transistor junction.
+VS
–IN
Input Protection Basics
For ADI In-Amps
Input Protection from ESD and DC Overload
As interface amplifiers for data acquisition systems,
instrumentation amplifiers are often subjected to input
overloads, i.e., voltage levels in excess of their full scale
for the selected gain range or even in excess of the supply
voltage. These overloads fall into two general classes:
steady state and transient (ESD, etc.), both of which
occur for only a fraction of a second. With 3-op amp
in-amp designs, when operating at low gains (10 or
less), the gain resistor acts as a current-limiting element
in series with their resistor inputs. At high gains, the
lower value of RG may not adequately protect the
inputs from excessive currents.
Standard practice is to place current-limiting resistors
in each input, but adding this protection also increases
the circuit’s noise level. A reasonable balance needs to be
found between the protection provided and the increased
resistor (Johnson) noise introduced. Circuits using
in-amps with a relatively high noise level can tolerate
more series protection without seriously increasing
their total circuit noise.
+VS
400
Q1
400
Q2
RG
–VS
+IN
–VS
6mA MAX INPUT CURRENT
Figure 5-9. AD8221 in-amp input circuit.
The AD8221 was designed to handle maximum
input currents of 6 mA steady state (or dc), at room
temperature. Its internal resistors and diodes will
protect the device from input voltages 0.7 V above
the positive supply, or 2.4 V more negative than
the minus supply (6 mA  0.4 k). Therefore, for
15 V supplies, the maximum safe input level is
+15.7 V, –17.4 V. Additional external series resistors
can be added to increase this level considerably, at the
expense of a higher circuit noise level.
The AD8221 in-amp is a very low noise device, with a
maximum (eNI ) 8 nV/√Hz. A single 1 k resistor will
add approximately 4 nV/√Hz of Johnson noise. The
addition of this resistor would raise the maximum
dc level to approximately 22.5 V above each supply or
37.5 V with 15 V supplies.
Of course, the less added noise the better, but a good
guideline is that circuits needing this extra protection
can easily tolerate resistor values that generate 30% of
the total circuit noise. For example, a circuit using an
in-amp with a rated noise level of 20 nV/√Hz can tolerate
an additional 6 nV/√Hz of Johnson noise.
Use the following cookbook method to translate this
number into a practical resistance value. The Johnson
noise of a 1 k resistor is approximately 4 nV/√Hz. This
value varies as the square root of the resistance. So,
a 20 k resistor would have √20 times as much noise
as the 1 k resistor, which is 17.88 nV/√Hz (4.4721 3
4 nV/√Hz). Because both inputs need to be protected,
two resistors are needed, and their combined noise will
add as the square root of the number of resistors (the
root sum of squares value). In this case, the total added
noise from the two 20 k resistors will be 25.3 nV/√Hz
(17.88 3 1.414).
+VS
+VS
+VS
+VS
+VS
–IN
4006
–VS
+VS
Q1
4006
Q2
RG
+IN
–VS
6mA MAX INPUT CURRENT
Figure 5-10. AD8222 and AD8225 in-amp
input circuit.
The AD8222 and AD8225 have a very similar input
section to that of the AD8221. Except that now, all six
diodes are located on the Q1/Q2 side of the 400  input
resistor (see Figure 5-10).
5-5
Figure 5-11 shows the input section of the AD620 series
(AD620, AD621, and AD622) in-amps. This is very
similar to that of the AD8221: Both use a 400  resistor
in series with each input, and both use diode protection.
The chief differences are the four additional AD8221
diodes. One set is tied between each input and the
positive supply, and the other set is connected between
the base of each input transistor and the negative supply.
The AD620 uses its 400  internal resistor and a single
set of diodes to protect against negative input voltages.
For positive voltage overloads, it relies on its own baseemitter input junction to act as the clamping diode.
+VS
case, adding two 5 k resistors will raise the circuit’s
noise approximately 13 nV/√Hz (30 percent) but would
provide an additional 100 V of transient overload
protection.
Figure 5-13 shows the input architecture of the AD623
in-amp. In this design, the internal (ESD) diodes are
located before the input resistors and, as a consequence,
provide less protection than the designs previously
discussed. The AD623 can tolerate 10 mA maximum
input current, but in many cases, some external series
resistance will be needed to keep input current below
this level.
+VS
+VS
+VS
400�
–IN
Q1
400�
Q2
1k
–IN
+IN
RG
–VS
6mA MAX INPUT CURRENT
+VS
–IN
2k
Q2
–VS
–VS
–VS
–VS
Figure 5-14 shows the simplified input circuitry for the
AD8230 zero-drift in-amp. As shown, the AD8230 only
has a single ESD diode connected between each input
and the negative supply line. This diode offers ESD
protection for negative pulses larger than 0.7 below the
negative supply. However, it was not designed to protect
the in-amp against positive voltage transients or long
duration voltage overloads in either direction.To protect
against these, external low leakage diodes and resistors
are needed, as shown in Figure 5-20.
+IN
200k
–VS
+IN
Figure 5-13. AD623 in-amp input circuit.
+VS
Q1
1k
Since the AD623’s device noise is approximately
35 nV/√Hz, up to 5 k of external resistance can be
added here to provide 50 V of dc overload protection,
while only increasing input noise to 38 nV/√Hz total.
+VS
2k
–VS
Q2
+VS
10mA MAX INPUT CURRENT
Figure 5-11. AD620 series (AD620, AD621,
and AD622) in-amp input circuit.
+VS
Q1
+VS
–VS
6mA MAX INPUT CURRENT
Figure 5-12. AD627 in-amp input circuit.
The AD627 can tolerate 20 mA transient input currents
(Figure 5-12). In addition, it has built-in 2 k resistors
and can handle input voltages 40 volts higher than its
supply lines (20 mA 3 2 k). This level of protection
is quite beneficial. Because of its low power, many of
the AD627’s applications will use a low voltage single
power supply. If even more protection is needed, quite
large external resistors can be added without seriously
degrading the AD627’s 38 nV/√Hz noise level. In this
5-6
–IN
1606
G
+IN
1606
D
D
Q1
Q2
S
G
S
–VS
–VS
6mA MAX INPUT CURRENT
Figure 5-14. AD8230 in-amp input circuit.
Figure 5-15 is a simplified diagram showing the input
structure for the AD8220 JFET in-amp.The input circuit
has a very high impedance: typically 1000 GV and 6 pF.
There are two ESD protection diodes at each input but
no internal series resistance between the input terminals
and the JFET input stage. Therefore, external resistors
need to be added to limit the current in applications
subject to an input overvoltage condition.
+VS
VCC
VCC
–IN
VCC
1906
M1
1906
M2
+IN
+VS
+VS
+VS
5mA MAX INPUT CURRENT
Figure 5-17. AD8553 in-amp simplified input circuit.
+IN
–IN
–VS
Figure 5-17 shows the input section of the AD8553
zero drift in-amp. This has a 190 V internal protection
resistor between each input pin and the diode clamping
circuitry.
–VS
The maximum input current of the AD8553 is approximately 5 mA, so additional external resistance may be
needed in some applications (see table 5-2).
5mA MAX INPUT CURRENT
Figure 5-15. AD8220 in-amp input circuit.
Figure 5-16 shows a simplified version of the AD8250
input circuitry. Here, two internal resistors and two diodes
protect each input. The AD8250 can tolerate maximum
sustained input currents of 20 mA. Note that there are
two pairs of input protection resistors—one between each
input and the two ESD diodes, and the other between
these diodes and the transistor base. This offers more
protection to the transistor bases than to the ESD diodes.
Therefore, designers should take precautions to protect
the ESD diodes from being destroyed.
Also note that when the input voltage approaches
VCC – 0.2 V (outside the input common-mode range),
current will begin to f low into the inputs of the
AD8553. If the enable pin is also held low during
this period of time, the output of the AD8553 will
no longer be high impedance. In some applications,
multiple AD8553 instrumentation amplifier outputs
can be connected together to mux many inputs to
one output. In these applications, only one AD8553
enable pin will be high, while all of the other AD8553
enable pins will be low.
+VS
+VS
+VS
–IN
2006
+VS
1006
–VS
1006
20mA MAX INPUT CURRENT
2006
–VS
Figure 5-16. AD8250 in-amp input circuit.
5-7
+IN
If an input overload voltage of VCC – 0.2 V or greater
occurs on any of the AD8553 instrumentation amplifier
inputs while its enable pin is low, the output of that
amplifier can overload the AD8553 that is driving the
output (enable pin is high).
The AD8555 and AD8556 products are instrumentation
amplifiers designed to be used in sensor applications.
Figure 5-18 is a simplified input circuit for the AD8555.
Here, protection diodes are connected between each
input terminal and the supplies. The input signal then
travels through a series resistor before arriving at the
amplification and switching circuitry.
+VS
+VS
+VS
3886
–IN
–VS
3886
5mA MAX INPUT CURRENT
+IN
–VS
Table 5-2. Recommended Series
Protection Resistor Values
Max
In-Amp
Input
Noise
Overload
Current
Device (eNI)
Recommended
External Protection
Resistors Adding
Additional Noise*
of 10%
of 40%
AD8220
AD8221
AD8222
AD8225
AD8230
AD8250
AD8251
AD8553
AD8555
AD8556
AD620
AD621
AD622
AD623
AD627
1.74 k
500 
500 
500 
4.99 k
1.30 k
1.30 k
6.98 k
8.06 k
8.06 k
634 
634 
634 
9.53 k
11.3 k
15 nV/√Hz
8 nV/√Hz
8 nV/√Hz
8 nV/√Hz
160 nV/√Hz
13 nV/√Hz
13 nV/√Hz
30 nV/√Hz
32 nV/√Hz
32 nV/√Hz
9 nV/√Hz
9 nV/√Hz
9 nV/√Hz
35 nV/√Hz
38 nV/√Hz
5 mA
6 mA
6 mA
6 mA
6 mA
20 mA
20 mA
5 mA
5 mA
5 mA
6 mA
6 mA
6 mA
10 mA
20 mA
*This noise level is for two resistors, one in series with
each input.
Figure 5-18. AD8555 in-amp simplified input circuit.
The AD8556 input circuitry (Figure 5-19) is very similar
to that of the AD8555, except that there is a larger value
resistor (3.6 kV) and other components between the
diode and the amplifying and switching circuitry. These
provide an internal RFI/EMI filtering capability.
Because the input circuitry in both products lacks an
internal resistor, before the first set of clamping diodes,
some external resistance is usually necessary to ensure
adequate overvoltage protection.
Table 5-2 provides recommended series protection resistor values for a 10% or 40% increase in circuit noise.
Adding External Protection Diodes
Device input protection may be increased with the
addition of external clamping diodes as shown in
Figure 5-20. As high current diodes are used, input
protection is increased, which allows the use of much
lower resistance input protection resistors that, in turn,
reduces the circuit’s noise.
Unfortunately, most ordinary diodes (Schottky, silicon,
etc.) have high leakage currents that will cause large offset
errors at the in-amp’s output; this leakage increases
exponentially with temperature. This tends to rule out
the use of external diodes in applications where the
in-amp is used with high impedance sources.
+VS
+VS
+VS
3.6k6
–IN
3.6k6
–VS
+IN
–VS
13pF
–VS
6.98 k
2.0 k
2.0 k
2.0 k
4.99 k
5.23 k
5.23 k
28.0 k
32.4 k
32.4 k
2.55 k
2.55 k
2.55 k
38.3 k
45.3 k
5pF
13pF
–VS
5mA MAX INPUT CURRENT
Figure 5-19. AD8556 simplified input circuit.
5-8
Specialty diodes with much lower leakage are available,
but these are often difficult to find and are expensive.
For the vast majority of applications, limiting resistors
alone provide adequate protection for ESD and longer
duration input transients.
+VS
+VS
–IN
D2
–VS
+IN
0.33MF
D1
RLIM
D3
RLIM
RG
0.01MF
IN-AMP
Although it is a simple matter to use a higher power protection resistor, this is a dangerous practice, as the power
dissipation will also increase in the in-amp’s input stage.
This can easily lead to device failure (see the preceding
section on input protection basics for input current
limitations of ADI in-amps). Except for ESD events, it
is always best to adopt a conservative approach and treat
all transient input signals as full duration inputs.
VOUT
D4
+VS
input resistors or the in-amps input stage. A 1 k resistor,
in series with an in-amp input terminal drawing
20 mA, will dissipate 0.4 W, which can easily be handled
by a standard 0.5 W or greater surface-mount resistor.
If the input current is doubled, power consumption
quadruples as it increases as the square of the input
current (or as the square of the applied voltage).
0.33MF –V 0.01MF
S
D1–D4 ARE INTERNATIONAL RECTIFIER SD101 SERIES
FAST SCHOTTKY BARRIER RECTIFIERS
Figure 5-20. Using external components
to increase input protection.
Despite their limitations, external diodes are often
required in some special applications, such as electric
shock defibrillators, which utilize short duration, high
voltage pulses. The combination of external diodes and
very large input resistors (as high as 100 k) may be
needed to adequately protect the in-amp.
It is a good idea to check the diodes’ specifications to
ensure that their conduction begins well before the
in-amp’s internal protection diodes start drawing current. Although they provide excellent input protection,
standard Schottky diodes can have leakage up to several
mA. However, as in the example of Figure 5-20, fast
Schottky barrier rectifiers, such as the international
rectifier type SD101 series, can be used; these devices
have 200 nA max leakage currents and 400 mW typical
power dissipation.
ESD and Transient Overload Protection
Protecting in-amp inputs from high voltage transients and
ESD events is very important for a circuit’s long-term
reliability. Power dissipation is often a critical factor,
as input resistors, whether internal or external, must
be able to handle most of the power of the input pulse
without failing.
ESD events, while they may be very high voltage, are
usually of very short duration and are normally one-time
events. Since the circuit has plenty of time to cool down
before the next event occurs, modest input protection is
sufficient to protect the device from damage.
On the other hand, regularly occurring short duration
input transients can easily overheat and burn out the
Designs that are expected to survive such events over long
periods of time must use resistors with enough resistance
to protect the in-amp’s input circuitry from failure and
enough power to prevent resistor burnout.
Design Issues Affecting DC Accuracy
The modern in-amp is continually being improved,
providing the user with ever increasing accuracy and
versatility at a lower price. Despite these improvements
in product performance, there remain some fundamental
applications issues that seriously affect device accuracy.
Now that low cost, high resolution ADCs are commonly
used, system designers need to ensure that if an in-amp is
used as a preamplifier ahead of the converter, the in-amp’s
accuracy matches that of the ADC.
Designing for the Lowest Possible Offset
Voltage Drift
Offset drift errors include not just those associated with
the active device being used (IC in-amp or discrete in-amp
design using op amps), but also thermocouple effects in
the circuit’s components or wiring. The in-amp’s input
bias and input offset currents flowing through unbalanced
source impedances also create additional offset errors. In
discrete op amp in-amp designs, these errors can increase
with temperature unless precision op amps are used.
Designing for the Lowest Possible Gain Drift
When planning for gain errors, the effects of board layout,
the circuit’s thermal gradients, and the characteristics of
any external gain setting resistors are often overlooked. A
gain resistor’s absolute tolerance, its thermal temperature
coefficient, its physical position relative to other resistors in the same gain network, and even its physical
orientation (vertical or horizontal) are all-important
design considerations if high dc accuracy is needed.
5-9
In many ADC preamp circuits, an external user-selected
resistor sets the gain of the in-amp, so the absolute
tolerance of this resistor and its variation over temperature, compared to that of the IC’s on-chip resistors, will
affect the circuit’s gain accuracy. Resistors commonly
used include through-hole 1% 1/4 W metal film types
and 1% 1/8 W chip resistors. Both types typically have
a 100 ppm/°C temperature coefficient. However, some
chip resistors can have TCs of 200 ppm/C or even
250 ppm/C.
Even when a 1% 100 ppm/C resistor is used, the gain
accuracy of the in-amp will be degraded. The resistor’s
initial room temperature accuracy is only 1%, and
the resistor will drift another 0.01% (100 ppm/C) for
every C change in temperature. The initial gain error
can easily be subtracted out in software, but to correct
for the error vs. temperature, frequent recalibrations (and
a temperature sensor) would be required.
If the circuit is calibrated initially, the overall gain accuracy
is reduced to approximately 10 bits (0.1%) accuracy for
a 10C change. An in-amp with a standard 1% metal film
gain resistor should never be used ahead of even a 12-bit
converter: It would destroy the accuracy of a 14-bit or
16-bit converter.
Additional error sources associated with external resistors
also affect gain accuracy. The first are variations in resistor
heating caused by input signal level. Figure 5-21, a simple
op amp voltage amplifier, provides a practical example.
R1 = 100
G = 1 + ——
R2
R1 = 9.9k�, 1/4W
R2 = 1k�, 1/4W
Figure 5-21. An example of how differences in
input signal level can introduce gain errors.
Under zero signal conditions, there is no output signal
and no resistor heating. When an input signal is applied,
however, an amplified voltage appears at the op amp
output. When the amplifier is operating with gain,
Resistor R1 will be greater than R2. This means that
there will be more voltage across R1 than across R2.
The power dissipated in each resistor equals the square
of the voltage across it divided by its resistance in ohms.
The power dissipated and, therefore, the internal heating
of the resistor will increase in proportion to the value of
the resistor.
In the example, R1 is 9.9 k and R2 is 1 k. Consequently,
R1 will dissipate 9.9 times more power than R2. This
leads to a gain error that will vary with input level. The
use of resistors with different temperature coefficients
can also introduce gain errors.
A1
R3
10k�
R4
10k�
SENSE
R1
25k�
RG
A3
OUTPUT
R2
25k�
A2
R5
10k�
R6
10k�
REF
Figure 5-22. A typical discrete 3-op amp in-amp
using large value, low TC feedback resistors.
Even when resistors with matched temperature coefficients (TC) are used, gain errors that vary with input
signal level can still occur. The use of larger (i.e., higher
power) resistors will reduce these effects, but accurate,
low TC power resistors are expensive and hard to find.
When a discrete 3-op amp in-amp is used, as shown
in Figure 5-22, these errors will be reduced. In a
3-op amp in-amp, there are two feedback resistors, R1
and R2, and one gain resistor, RG. Since the in-amp uses
two feedback resistors while the op amp uses only one,
each of the in-amp’s resistors only needs to dissipate half
the power (for the same gain). Monolithic in-amps, such
as the AD620, offer a further advantage by using relatively
large value (25 k) feedback resistors. For a given gain
and output voltage, large feedback resistors will dissipate
less power (i.e., P = V2/RF). Of course, a discrete in-amp
can be designed to use large value, low TC resistors as
well, but with added cost and complexity.
Another less serious but still significant error source
is the so-called thermocouple effect, sometimes
referred to as thermal EMF. This occurs when two
different conductors, such as copper and metal film,
are tied together. When this bimetallic junction is
heated, a simple thermocouple is created. When using
similar metals, such as a copper-to-copper junction,
a thermoelectric error voltage of up to 0.2 mV/C may
be produced. An example of these effects is shown in
Figure 5-23.
5-10
A final error source occurs when there is a thermal
gradient across the external gain resistor. Something as
simple as mounting a resistor on end to conserve board
space will invariably produce a temperature gradient
across the resistor. Placing the resistor flat down against
the PC board will cure this problem unless there is air
flowing along the axis of the resistor (where the air flow
cools one side of the resistor more than the other side).
Orienting the resistor so that its axis is perpendicular to
the airflow will minimize this effect.
RESISTOR
MATERIAL
T1
Option 2: Use a Fixed-Gain In-Amp
By far, the best overall dc performance is provided
by using a monolithic in-amp, such as the AD621 or
AD8225, in which all the resistors are contained within
the IC. Now all resistors have identical TCs, all are at
virtually the same temperature. Any thermal gradients
across the chip are very small, and gain error drift is
guaranteed and specified to very high standards.
At a gain of 10, the AD621 has a guaranteed maximum
dc offset shift of less than 2.5 µV/C and a maximum gain
drift of 5 ppm/°C, which is only 0.0005%/C.
The AD8225 is an in-amp with a fixed gain of 5. It has
a maximum offset shift of 2 V/C and a maximum drift
of 0.3 V/C.
T2
RTI AND RTO ERRORS
Another important design consideration is how circuit
gain affects many in-amp error sources such as dc offset
and noise. An in-amp should be regarded as a two stage
amplifier with both an input and an output section. Each
section has its own error sources.
RESISTOR LEADS
TYPICAL RESISTOR THERMOCOUPLE EMFs
• CARBON COMPOSITION
400�V/�C
• METAL FILM
20�V/�C
• EVENOHM OR MANGANIN
WIRE-WOUND
2�V/�C
• RCD COMPONENTS HP-SERIES 0.05�V/�C
Figure 5-23. Thermocouple effects inside
discrete resistors.
Practical Solutions
As outlined, a number of dc offset and gain errors are
introduced when external resistors are used with a
monolithic in-amp. Discrete designs tend to have even
larger errors. There are three practical solutions to this
problem: use higher quality resistors, use software correction, or, better still, use an in-amp that has all of its
gain resistors on-chip, such as the AD621.
Option 1: Use a Better Quality Gain Resistor
As a general rule, only 12-bit or 13-bit gain performance
is possible using commonly available 1% resistors,
which assumes that some type of initial calibration is
performed.
A practical solution to this problem is to simply use a
better quality resistor. A significant improvement can be
made using a 0.1%, 1/10 W, surface-mount resistor. Aside
from having a 10 better initial accuracy, it typically has
a TC of only 25 ppm/C, which will provide better than
13-bit accuracy over a 10C temperature range.
If even better gain accuracy is needed, there are specialty
houses that sell resistors with lower TCs, but these are
usually expensive military varieties.
Because the errors of the output section are multiplied
by a fixed gain (usually 2), this section is often the
principal error source at low circuit gains. When the
in-amp is operating at higher gains, the gain of the
input stage is increased. As the gain is raised, errors
contributed by the input section are multiplied, while
output errors are not. So, at high gains, the input stage
errors dominate.
Since device specifications on different data sheets often
refer to different types of errors, it is very easy for the
unwary designer to make an inaccurate comparison
between products. Any (or several) of four basic error
categories may be listed: input errors, output errors, total
error RTI, and total error RTO. Here follows an attempt
to list, and hopefully simplify, an otherwise complicated
set of definitions.
Input errors are those contributed by the amplifier’s input
stage alone; output errors are those due to the output
section. Input related specifications are often combined
and classified together as a referred to input (RTI) error,
while all output related specifications are considered
referred to output (RTO) errors.
For a given gain, an in-amp’s input and output errors
can be calculated using the following formulas:
Total Error, RTI = Input Error + (Output Error/Gain)
Total Error, RTO = (Gain  Input Error) + Output Error
Sometimes the specifications page will list an error term
as RTI or RTO for a specified gain. In other cases, it is up
to the user to calculate the error for the desired gain.
5-11
Offset Error
Using the AD620A as an example, the total voltage
offset error of this in-amp when operating at a gain of
10 can be calculated using the individual errors listed on
its specifications page. The (typical) input offset of the
AD620 (VOSI) is listed as 30 V. Its output offset (VOSO)
is listed as 400 V. Thus, the total voltage offset referred
to input (RTI) is equal to
Total RTI Error = VOSI + (VOSO/G) = 30 V + (400 V/10)
= 30 V + 40 V = 70 V
The total voltage offset referred to the output (RTO)
is equal to
Total Offset Error RTO = (G (VOSI )) +VOSO = (10 (30 V))
+ 400 V = 700 V
Note that the two error numbers (RTI vs. RTO) are
different: the RTO numbers are 103 larger, and
logically they should be, as at a gain of 10, the error at
the output of the in-amp should be 10 times the error
at its input.
Noise Errors
In-amp noise errors also need to be considered in a
similar way. Since the output section of a typical 3-op
amp in-amp operates at unity gain, the noise contribution
from the output stage is usually very small. But there
are 3-op amp in-amps that operate the output stage at
higher gains, and 2-op amp in-amps regularly operate the
second amplifier at gain. When either section is operated
at gain, its noise is amplified along with the input signal.
Both RTI and RTO noise errors are calculated the same
way as offset errors, except that the noise of two sections
adds as the root mean square. That is,
Input Noise = eni ,Output Noise = eno
Total Noise RTI =
Total Noise RTO =
(eni ) + (eno Gain )
2
2
Reducing RFI Rectification Errors in
In-Amp Circuits
Real-world applications must deal with an ever
increasing amount of radio frequency interference
(RFI). Of particular concern are situations in which
signal transmission lines are long and signal strength is
low. This is the classic application for an in-amp, since
its inherent common-mode rejection allows the device
to extract weak differential signals riding on strong
common-mode noise and interference.
One potential problem that is frequently overlooked,
however, is that of radio frequency rectification inside
the in-amp. When strong RF interference is present, it
may become rectified by the IC and then appear as a dc
output offset error. Common-mode signals present at
an in-amp’s input are normally greatly reduced by the
amplifier’s common-mode rejection.
Unfortunately, RF rectification occurs because even the
best in-amps have virtually no common-mode rejection at frequencies above 20 kHz. A strong RF signal
may become rectified by the amplifier’s input stage
and then appear as a dc offset error. Once rectified,
no amount of low-pass filtering at the in-amp output
will remove the error. If the RF interference is of an
intermittent nature, this can lead to measurement
errors that go undetected.
Designing Practical RFI Filters
The best practical solution is to provide RF attenuation
ahead of the in-amp by using a differential low-pass
filter.The filter needs to do three things: remove as much
RF energy from the input lines as possible, preserve
the ac signal balance between each line and ground
(common), and maintain a high enough input impedance over the measurement bandwidth to avoid loading
the signal source.
(Gain (eni )) + (eno )
2
2
For example, the (typical) noise of the AD620A
is specified as 9 nV/√Hz eni and 72 nV/√Hz eno.
Therefore, the total RTI noise of the AD620A operating
at a gain of 10 is equal to
Total Noise RTI =
(eni)2 + (eno Gain)2 =
(9)2 + (72 10)2 = 11.5 nV Hz
5-12
Figure 5-24 provides a basic building block for a wide
number of differential RFI filters. Component values
shown were selected for the AD8221, which has a typical
–3 dB bandwidth of 1 MHz and a typical voltage noise
level of 7 nV/√Hz. This same filter is recommended for
the AD8222 dual in-amp and for the AD8220 JFET
input in-amp. In addition to RFI suppression, the filter
provides additional input overload protection, as resistors
R1a and R1b help isolate the in-amp’s input circuitry
from the external signal source.
Figure 5-25 is a simplified version of the RFI circuit. It
reveals that the filter forms a bridge circuit whose output
appears across the in-amp’s input pins. Because of this,
any mismatch between the time constants of C1a/R1a
and C1b/R1b will unbalance the bridge and reduce high
frequency common-mode rejection. Therefore, resistors R1a and R1b and capacitors C1a and C1b should
always be equal.
As shown, C2 is connected across the bridge output
so that C2 is effectively in parallel with the series
combination of C1a and C1b. Thus connected, C2
very effectively reduces any ac CMR errors due to
mismatching. For example, if C2 is made 10 times
larger than C1, this provides a 20 reduction in CMR
errors due to C1a/C1b mismatch. Note that the filter
does not affect dc CMR.
The RFI filter has two different bandwidths: differential
and common mode. The differential bandwidth defines
the frequency response of the filter with a differential input
signal applied between the circuit’s two inputs, +IN and
–IN. This RC time constant is established by the sum of
the two equal-value input resistors (R1a, R1b), together
with the differential capacitance, which is C2 in parallel
with the series combination of C1a and C1b.
RFI FILTER
–IN
R1a
4.02k
0.01F
C1a
1000pF
1
–
2
C2
0.01F
+IN
3
4
C1b
1000pF
0.33F
8
G = 1+
AD8220
AD8221
AD8222
RG
R1b
4.02k
+VS
+
5
0.01F
49.4k
RG
7 15*
VOUT
6
REF
0.33F
*AD8222 ONLY
–VS
Figure 5-24. LP filter circuit used to prevent RFI rectification errors in AD8220, AD8221, and AD8222 in-amps.
R1a
C1a
+IN
IN-AMP
C2
VOUT
–IN
R1b
C1b
Figure 5-25. Capacitor C2 shunts C1a/C1b and very effectively reduces ac CMR errors due to
component mismatching.
5-13
The –3 dB differential bandwidth of this filter is
equal to
BWDIFF =
1
2πR(2C 2 + C 1)
The common-mode bandwidth defines what a
common-mode RF signal sees between the two inputs
tied together and ground. It’s important to realize that
C2 does not affect the bandwidth of the common-mode
RF signal, as this capacitor is connected between the
two inputs (helping to keep them at the same RF signal
level).Therefore, common-mode bandwidth is set by the
parallel impedance of the two RC networks (R1a/C1a
and R1b/C1b) to ground.
The –3 dB common-mode bandwidth is equal to
BWCM =
1
2πR1C 1
Using the circuit of Figure 5-24, with a C2 value of
0.01 F as shown, the –3 dB differential signal bandwidth
is approximately 1900 Hz. When operating at a gain of
5, the circuit’s measured dc offset shift over a frequency
range of 10 Hz to 20 MHz was less than 6 V RTI. At
unity gain, there was no measurable dc offset shift.
The RFI filter should be built using a PC board with
ground planes on both sides. All component leads should
be made as short as possible. The input filter common
should be connected to the amplifier common using the
most direct path. Avoid building the filter and the in-amp
circuits on separate boards or in separate enclosures, as
this extra lead length can create a loop antenna. Instead,
physically locate the filter right at the in-amp’s input
terminals. A further precaution is to use good quality
resistors that are both noninductive and nonthermal
(low TC). Resistors R1 and R2 can be common 1%
metal film units. However, all three capacitors need to
be reasonably high Q, low loss components. Capacitors
C1a and C1b need to be 5% tolerance devices to avoid
degrading the circuit’s common-mode rejection. The
traditional 5% silver micas, miniature size micas, or
the new Panasonic 2% PPS film capacitors (Digi-Key
part # PS1H102G-ND) are recommended.
Selecting RFI Input Filter Component Values Using
a Cookbook Approach
The following general rules will greatly ease the design
of an RC input filter.
1. First, decide on the value of the two series resistors while ensuring that the previous circuitry can
adequately drive this impedance. With typical values
between 2 k and 10 k, these resistors should not
contribute more noise than that of the in-amp itself.
Using a pair of 2 k resistors will add a Johnson noise
of 8 nV/√Hz; this increases to 11 nV/√Hz with 4 k
resistors and to 18 nV/√Hz with 10 k resistors.
2.Next, select an appropriate value for capacitor C2,
which sets the filter’s differential (signal) bandwidth.
It’s always best to set this as low as possible without
attenuating the input signal. A differential bandwidth
of 10 times the highest signal frequency is usually
adequate.
3. Then select values for capacitors C1a and C1b, which
set the common-mode bandwidth. For decent ac
CMR, these should be 10% the value of C2 or less.
The common-mode bandwidth should always be less
than 10% of the in-amp’s bandwidth at unity gain.
5-14
Specific Design Examples
An RFI Circuit for AD620 Series In-Amps
Figure 5-26 is a circuit for general-purpose in-amps such
as the AD620 series, which have higher noise levels
(12 nV/√Hz) and lower bandwidths than the AD8221.
Accordingly, the same input resistors were used, but
capacitor C2 was increased approximately five times
to 0.047 F to provide adequate RF attenuation. With
the values shown, the circuit’s –3 dB bandwidth is approximately 400 Hz; the bandwidth may be increased
to 760 Hz by reducing the resistance of R1 and R2
to 2.2 k. Note that this increased bandwidth does not
come free. It requires the circuitry preceding the in-amp
to drive a lower impedance load and results in somewhat
less input overload protection.
An RFI Circuit for Micropower In-Amps
Some in-amps are more prone to RF rectification than
others and may need a more robust filter. A micropower
in-amp, such as the AD627, with its low input stage
operating current, is a good example. The simple
expedient of increasing the value of the two input
resistors, R1a/R1b, and/or that of capacitor C2, will
provide further RF attenuation, at the expense of a
reduced signal bandwidth.
Since the AD627 in-amp has higher noise (38 nV/√Hz)
than general-purpose ICs, such as the AD620 series
devices, higher value input resistors can be used without
seriously degrading the circuit’s noise performance.
The basic RC RFI circuit of Figure 5-24 was modified
to include higher value input resistors, as shown in
Figure 5-27.
The filter bandwidth is approximately 200 Hz. At a gain
of 100, the maximum dc offset shift with a 1 V p-p input
applied is approximately 400 V RTI over an input range
of 1 Hz to 20 MHz. At the same gain, the circuit’s RF
signal rejection (RF level at output/RF applied to the
input) will be better than 61 dB.
RFI FILTER
R1a
4.02k
+IN
0.01F
C1a
1000pF
3
+VS
0.33F
7
+
1
C2
0.047F
R1b
4.02k
–IN
6
AD620
RG
8
2
C1b
1000pF
VOUT
5
–
4
REF
0.01F
0.33F
–VS
Figure 5-26. RFI circuit for AD620 series in-amp.
RFI FILTER
+IN
20k
0.01F
C1a
1000pF
3
+
+VS
0.33F
7
1
C2
0.022F
–IN
8
20k
2
C1b
1000pF
6
AD627
RG
5
–
4
REF
0.33F
0.01F
–VS
Figure 5-27. RFI suppression circuit for the AD627.
5-15
VOUT
An RFI Filter for the AD623 In-Amp
AD8225 RFI Filter Circuit
Figure 5-28 shows the recommended RFI circuit for
use with the AD623 in-amp. Because this device is less
prone to RFI than the AD627, the input resistors can
be reduced in value from 20 k to 10 k; this increases
the circuit’s signal bandwidth and lowers the resistors’
noise contribution. Moreover, the 10 k resistors still
provide very effective input protection. With the values
shown, the bandwidth of this filter is approximately
400 Hz. Operating at a gain of 100, the maximum dc
offset shift with a 1 V p-p input is less than 1 V RTI.
At the same gain, the circuit’s RF signal rejection is
better than 74 dB.
Figure 5-29 shows the recommended RFI filter for this
in-amp. The AD8225 in-amp has a fixed gain of 5 and a
bit more susceptibility to RFI than the AD8221.Without
the RFI filter, with a 2 V p-p, 10 Hz to 19 MHz sine wave
applied, this in-amp measures about 16 mV RTI of dc
offset. The filter used provides a heavier RF attenuation
than that of the AD8221 circuit by using larger resistor
values: 10 k instead of 4 k.This is permissible because
of the AD8225’s higher noise level. Using the filter, there
was no measurable dc offset error.
RFI FILTER
+IN
10k
0.01F
C1a
1000pF
3
+
+VS
0.33F
7
1
C2
0.022F
–IN
6
AD623
RG
8
10k
2
C1b
1000pF
VOUT
5
–
4
0.01F
REF
0.33F
–VS
Figure 5-28. AD623 RFI suppression circuit.
RFI FILTER
+IN
10k
0.01F
C1a
1000pF
2
+
C2
0.01F
–IN
+VS
0.33F
7
6
AD8225
5
10k
3
C1b
1000pF
–
4
0.01F
REF
0.33F
–VS
Figure 5-29. AD8225 RFI filter circuit.
5-16
VOUT
An RFI Filter For The AD8555 Sensor
Amplifier
In-Amps with On-Chip EMI/RFI Filtering
The AD8556 is very similar to the AD8555. The
AD8556 features internal EMI filters on the –IN, +IN,
FILT, and VCLAMP pins. These built-in filters on the
pins limit the interference bandwidth and provide good
RFI suppression without reducing performance within
the pass band of the in-amp. A functional diagram of
AD8556 along with its EMI/RFI filters is shown in
Figure 5-31.
The circuit in Figure 5-30 provides good RFI suppression
without reducing performance within the AD8555 pass
band. Using the component values shown, this filter has
a common-mode bandwidth of approximately 40 kHz.
To preserve common-mode rejection in the AD8555’s
pass band, capacitors need to be 5% (silver mica) or
better and should be placed as close to its inputs as possible. Resistors should be 1% metal film. The circuit’s
differential bandwidth is approximately 4 kHz when a
C3 value of 0.047 mF is used.
VDD
–IN
+IN
VSS
R2
4.02k6
C3
0.047MF
AD8556 has on-chip filters on its inputs, VCLAMP,
and filter pins. The first-order low-pass filters inside
the AD8556 are useful to reject high frequency EMI
signals picked up by wires and PCB traces outside
the AD8556. The most sensitive pin of any amplifier
to RFI/EMI signal is the noninverting pin. Signals
present at this pin appear as common-mode signals
and create problems.
1
VDD
2
FILT/DIGOUT
3
DIGIN
4
–IN
C2
1nF
VSS 8
VOUT 7
VCLAMP 6
VDD
The filters at the input of the AD8556 have two different
bandwidths: common and differential mode. The EMI
filters placed on the input pins of the AD8556 reject
EMI/RFI suppressions that appear as common-mode
signals.
+IN 5
C1
1nF
AD8555
R1
4.02k6
Figure 5-30. AD8555 RFI filter circuit.
DIGIN
VCLAMP
VDD
EMI
FILTER
DAC
LOGIC
1
2
+IN
A5 OUT 3
–IN
VSS
VDD
VPOS
1
EMI
FILTER
2
+IN
3
A1 OUT
–IN
R5
P4
R7
R2
VSS
P2
VDD
EMI
FILTER
R3
1
+IN
3
A3 OUT
2
–IN
P1
VDD
VDD
RF
EMI
FILTER
1
2
+IN
3
A4 OUT
–IN
VSS
1
VNEG
+IN
3
A2 OUT
2
–IN
EMI
FILTER
VSS
R1
R4
VSS
R5
P3
AD8556
VSS
FILT/DIGOUT
Figure 5-31. AD8556 block diagram showing on-chip EMI/RFI filter.
5-17
VOUT
Figure 5-32 simulates the presence of a noisy commonmode signal, and Figure 5-33 shows the response dc
values at VOUT.
+2.5V
+2.5V
–2.5V
U2
1
–2.5V
2
U3
1
2
3
VCC
VEE
FILTER
VOUT
8
7
AD8556
+IN
GAIN = 70
DC OFFSET = 2.5V
VCLAMP
AD8556
7
6
GAIN = 70
DC OFFSET = 2.5V
VOUT
+2.5V
+IN 5
+
–
V2
Figure 5-34. Test circuit to show AD8556
performance exposed to differential-mode
RFI/EMI signals.
The response of AD8556 to EMI/RFI differential signals
is shown in Figure 5-35.
Figure 5-32. Test circuit to show AD8556
performance exposed to common-mode
RFI/EMI signals.
!$
$#/&&3%4M6
$%6)!4)/.&2/-$#/54054M6
VOUT
8
200mV p-p
NOISE
INPUT
+
–
DATA
+2.5V
5
VEE
FILTER
4 –IN
VOUT
(0V)
VCLAMP 6
DATA
4 –IN
3
VCC
./.%-)02/4%#4%$3/,54)/.
n
./.%-)02/4%#4%$3/,54)/.
n
n
n
n
n
n
!$%.(!.#%$0!24&/2%-)
&2%15%.#9-(Z
Figure 5-35. DC offset shift of AD8556 due to
EMI/RFI differential signals vs. frequency.
n
&2%15%.#9-(Z
Figure 5-33. DC input offset values at VOUT
caused by common-mode RFI vs. frequency.
The differential bandwidth defines the frequency response
of the filters with a differential signal applied between
the two inputs, VPOS (that is, +IN ) and VNEG (that
is, –IN). Figure 5-34 shows the test circuit for AD8556
EMI/RFI susceptibility.
For the most effective EMI rejection, the printed circuit
board leads at VPOS and VNEG should be as similar as
possible. In this way, any EMI received by the VPOS
and VNEG pins will be similar (that is, a common-mode
input), and rejected by the AD8556. Furthermore, additional filtering at the VPOS and VNEG pins should
provide better reduction of unwanted behavior compared
with filtering at the other pins.
5-18
Common-Mode Filters Using X2Y® Capacitors*
Figure 5-36 shows the connection diagram for an X2Y
capacitor. These are very small, three terminal devices
with four external connections—A, B, G1, and G2.
The G1 and G2 terminals connect internally within the
device. The internal plate structure of the X2Y capacitor forms an integrated circuit with very interesting
properties. Electrostatically, the three electrical nodes
form two capacitors that share the G1 and G2 terminals.
The manufacturing process automatically matches both
capacitors very closely. In addition, the X2Y structure
includes an effective autotransformer/common-mode
choke. As a result, when these devices are used for
common-mode filters, they provide greater attenuation
of common-mode signals above the filter’s corner
frequency than a comparable RC filter.This usually allows
the omission of capacitor C2, with subsequent savings in
cost and board space.
C3 +V
C4
330nF S 10nF
–IN
+IN
Figure 5-36. X2Y electrostatic model.
+IN
R1B
4.02k� 1%
RG
IA
6
VOUT
5
C5
330nF
C6
10nF
D"
n
#/.6%.4)/.!,
#-&),4%2
n
n
n
n
K
K
K
&2%15%.#9(Z
-
'
Figure 5-38. RF attenuation, X2Y vs.
conventional RC common-mode filter.
8
U1
7
IA
6
AD8221
3
4 +
7
AD8221
3
4 +
n
C3 +VS C4
330nF
10nF
1
–
2
RG
8
U1
Figure 5-37a illustrates a conventional RC commonmode filter, while Figure 5-37b shows a common-mode
filter circuit using an X2Y device. Figure 5-38 is a
graph contrasting the RF attenuation provided by
these two filters.
C1A
220pF C2
C1B 10nF
220pF
1
–
2
Figure 5-37b. Common-mode filter
using X2Y capacitor.
2&!44%.5!4)/.D"
"
–IN
R1B
4.02k6 1%
89#-&),4%2
'
R1A
4.02k� 1%
C1*
10nF
–VS
!
'
R1A
4.02k6 1%
VOUT
5
C5
330nF
C6
10nF
–VS
Figure 5-37a. Conventional RC
common-mode filter.
*C1 is part number 500X14W103KV4. X2Y components may be purchased from Johanson Dielectrics, Sylmar, CA 91750, (818) 364-9800. For a full
listing of X2Y manufacturers, visit www.x2y.com/manufacturers.
5-19
Using Common-Mode RF Chokes for In-Amp
RFI Filters
As an alternative to using an RC input filter, a commercial
common-mode RF choke may be connected in front of
an in-amp, as shown in Figure 5-39. A common-mode
choke is a two-winding RF choke using a common
core. Any RF signals that are common to both inputs
will be attenuated by the choke. The common-mode
choke provides a simple means for reducing RFI with
a minimum of components and provides a greater
signal pass band, but the effectiveness of this method
depends on the quality of the particular common-mode
choke being used. A choke with good internal matching
is preferred. Another potential problem with using the
choke is that there is no increase in input protection as
is provided by the RC RFI filters.
Using an AD620 in-amp with the RF choke specified, at
a gain of 1000, and a 1 V p-p common-mode sine wave
applied to the input, the circuit of Figure 5-39 reduces
the dc offset shift to less than 4.5 V RTI. The high
frequency common-mode rejection ratio was also greatly
improved, as shown in Table 5-3.
Table 5-3. AC CMR vs. Frequency
Using the Circuit of Figure 5-39
Frequency
CMRR (dB)
100 kHz 333 kHz 350 kHz 500 kHz 1 MHz 100
83
79
88
96
Because some in-amps are more susceptible to RFI than
others, the use of a common-mode choke may sometimes
prove inadequate. In these cases, an RC input filter or
an X2Y-based filter is a better choice.
0.01F
+VS
0.33F
PULSE
ENGINEERING
#B4001 COMMON-MODE
RF CHOKE
+
+IN
RG
VOUT
IN-AMP
–
–IN
REF
0.01F
0.33F
–VS
Figure 5-39. Using a commercial common-mode RF choke for RFI suppression.
5-20
RFI TESTING
 VIN 
 2 


CMR = 20 log
 VOUT 
 Gain 


Figure 5-40 shows a typical setup for measuring RFI
rejection. To test these circuits for RFI suppression,
connect the two input terminals together using very
short leads. Connect a good quality sine wave generator
to this input via a 50  terminated cable.
Using an oscilloscope, adjust the generator for a 1 V
peak-to-peak output at the generator end of the cable.
Set the in-amp to operate at high gain (such as a gain
of 100). DC offset shift is simply read directly at the
in-amp’s output using a DVM. For measuring high
frequency CMR, use an oscilloscope connected to the
in-amp output by a compensated scope probe and measure
the peak-to-peak output voltage (i.e., feedthrough) vs.
input frequency.When calculating CMRR vs. frequency,
remember to take into account the input termination
(VIN/2) and the gain of the in-amp.
USING LOW-PASS FILTERING TO IMPROVE
SIGNAL-TO-NOISE RATIO
To extract data from a noisy measurement, low-pass filtering can be used to greatly improve the signal-to-noise
ratio of the measurement by removing all signals that are
not within the signal bandwidth. In some cases, band-pass
filtering (reducing response both below and above the
signal frequency) can be employed for an even greater
improvement in measurement resolution.
+VS
RF
SIGNAL
GENERATOR CABLE
+
RFI
INPUT
FILTER
VOUT TO
SCOPE OR DVM
IN-AMP
RG
–
REF
–VS
Figure 5-40. Typical test setup for measuring an in-amp’s RFI rejection.
5-21
The 1 Hz, 4-pole, active filter of Figure 5-41 is an example
of a very effective low-pass filter that normally would be
added after the signal has been amplified by the in-amp.
This filter provides high dc precision at low cost while
requiring a minimum number of components.
Furthermore, since the input bias current of these op
amps is as low as their input offset currents over most
of the MIL temperature range, there is rarely a need to
use the normal balancing resistor (along with its noisereducing bypass capacitor). Note, however, that adding
the optional balancing resistor will enhance performance
at temperatures above 100C.
Note that component values can simply be scaled
to provide corner frequencies other than 1 Hz (see
Table 5-4). If a 2-pole filter is preferred, simply take
the output from the first op amp.
Specified values are for a –3 dB point of 1.0 Hz.
For other frequencies, simply scale capacitors C1
through C4 directly; i.e., for 3 Hz Bessel response,
C1 = 0.0387 F, C2 = 0.0357 F, C3 = 0.0533 F,
and C4 = 0.0205 F.
The low levels of current noise, input offset, and input
bias currents in the quad op amp (either an AD704 or
OP497) allow the use of 1 M resistors without sacrificing
the 1 µV/C drift of the op amp. Thus, lower capacitor
values may be used, reducing cost and space.
C1
4C2
Q1 =
W=
1
W=
R6 C1C2
R6 = R7
C1
INPUT
R6
1M
R7
1M
C3
4C4
Q2 =
1
R8 C3C4
R8 = R9
1/2 AD706*
1/2 OP297*
C3
R8
C2
1M
R10
1M
* REFER TO THE ANALOG DEVICES WEBSITE AT
WWW.ANALOG.COM FOR THE LATEST OP AMP
PRODUCTS AND SPECIFICATIONS.
OUTPUT
C4
2M
R10
C5
A1, A2 ARE AD706 OR OP297
1/2 AD706
1/2 OP297
R9
2M
C5
OPTIONAL BALANCE
RESISTOR NETWORKS
CAN BE REPLACED
WITH A SHORT
0.01 F
CAPACITORS C 2 –C4 ARE
SOUTHERN ELECTRONICS
MPCC, POLYCARBONATE,
5%, 50V
Figure 5-41. A 4-pole low-pass filter for data acquisition.
Table 5-4. Recommended Component Values for a 1 Hz, 4-Pole, Low-Pass Filter
Desired Low-
Pass Response
Section 1
Frequency
Frequency
(Hz)
Q
(Hz)
(Q)
Section 2
C1
C2
(F)
(F)
C3
(F)
C4
(F)
Bessel
Butterworth
0.1 dB Chebychev
0.2 dB Chebychev
0.5 dB Chebychev
1.0 dB Chebychev
1.43
1.00
0.648
0.603
0.540
0.492
0.116
0.172
0.304
0.341
0.416
0.508
0.160
0.416
0.733
0.823
1.00
1.23
0.0616
0.0609
0.0385
0.0347
0.0290
0.0242
0.522
0.541
0.619
0.646
0.705
0.785
1.60
1.00
0.948
0.941
0.932
0.925
0.806
1.31
2.18
2.44
2.94
3.56
5-22
0.107
0.147
0.198
0.204
0.209
0.206
EXTERNAL CMR AND SETTLING TIME
ADJUSTMENTS
AC CMR trimming is accomplished in a similar
manner, except that an ac input signal is applied. The
input frequency used should be somewhat lower than the
–3 dB bandwidth of the circuit.
When a very high speed, wide bandwidth in-amp is
needed, one common approach is to use several op amps
or a combination of op amps and a high bandwidth
subtractor amplifier. These discrete designs may be
readily tuned up for best CMR performance by external
trimming. A typical circuit is shown in Figure 5-42. The
dc CMR should always be trimmed first, since it affects
CMRR at all frequencies.
The input amplitude should be set at 20 V p-p with
the inputs tied together. The ac CMR trimmer is then
nulled-set to provide the lowest output possible. If the
best possible settling time is needed, the ac CMR
trimmer may be used, while observing the output wave
form on an oscilloscope. Note that, in some cases, there
will be a compromise between the best CMR and the
fastest settling time.
The +VIN and –VIN terminals should be tied together
and a dc input voltage applied between the two inputs
and ground. The voltage should be adjusted to provide
a 10 V dc input. A dc CMR trimming potentiometer
would then be adjusted so that the outputs are equal and
as low as possible, with both a positive and a negative
dc voltage applied.
VIN#1
INVERTING
INPUT
AC CMR TRIM
R1
A1
CT
R2
SENSE
R5
RG
COMMON-MODE
INPUT SIGNAL
(AC OR DC)
VIN#2
NONINVERTING
INPUT
A3
OUTPUT
DC CMR TRIM
R6
A2
R3
1/2 CT
R4
Figure 5-42. External dc and ac CMR trim circuit for a discrete 3-op amp in-amp.
5-23
REF