ETC AB-193

APPLICATION BULLETIN
®
Mailing Address: PO Box 11400 • Tucson, AZ 85734 • Street Address: 6730 S. Tucson Blvd. • Tucson, AZ 85706
Tel: (602) 746-1111 • Twx: 910-952-111 • Telex: 066-6491 • FAX (602) 889-1510 • Immediate Product Info: (800) 548-6132
THE CURRENT-FEEDBACK OP AMP
A HIGH-SPEED BUILDING BLOCK
1
By Anthony D. Wang, Burr-Brown Corp.
Although current-feedback amplifiers (CFAs) have been in
use for quite some time, there is a reluctance to view them
in the same light as voltage-feedback amplifiers (VFAs). For
instance, the gain-bandwidth curve of VFAs has a parallel
in a transimpedance-bandwidth curve for CFAs. This parameter can be used to determine the closed-loop behavior
of the CFA in the same way that GBW can for the VFA. Not
all the fault is with the users—the amplifier manufacturers
have not standardized the CFA characterization as they
have done with VFAs. This paper describes the CFA and its
behavior in an intuitive manner.
2
The three transistor amplifier of Figure 1 is arranged in a
series-shunt configuration. However, in order to analyze the
amplifier, the circuit is rearranged as shown in Figure 2.
3
The feedback network shows up in two places—a series
network at the output and a parallel network at the emitter of
the input transistor. This allows for open-loop analysis while
keeping the effects of loading intact.
4
The loading of the output by the feedback network is
generally not a problem. However, the gain of the first
transistor stage is dependent on the values of the resistors in
the feedback network. Thus the open-loop response will
change with closed-loop gain (as the feedback network
changes), which could make frequency compensation an
iterative chore.
HISTORICAL PERSPECTIVE
The term “operational amplifier,” or “op amp” in typical
engineering shorthand, has generally been associated with
the transistorized voltage-feedback amplifier. It is becoming
more acceptable now to include the current-feedback amplifier in the same category.
The discrete transistor circuit of Figure 3 circumvents this
difficulty. Adding another transistor, Q4, to buffer the input
stage transistor, Q1, from the feedback network illustrates
this modification. This is the first step to a voltage feedback
amplifier topology.
Interestingly enough, the basic architecture for the CFA
might have predated the VFA although it was not until the
1980s that the CFA was itself repopularized. To appreciate
R1
the evolution of the beast, it helps to look back to some early
discrete transistor circuits.
R2
R1
R2
Q2
Q2
CT
VIN
CT
Q1
VIN
Q3
Q1
Q3
RF
VOUT
RG
VOUT
R3
RE =
FIGURE 1. Three Transistor Amplifier.
RF • RG
RF + R G
R3
RL = R F + R G
FIGURE 2. Amplifier Redrawn for Analysis.
AB-193
5
6
R1
R2
Q2
CT
Q3
RF
VIN
Q1
+In
–In
VOUT
Q4
RG
VOUT
R3
FIGURE 3. Adding a Buffer Transistor.
The added transistor presents a high impedance input to the
feedback network. It also features the benefits of a balanced
input, such as low offset voltage and equal input bias
currents.
FIGURE 4. Basic CFA Topology.
The NPN current mirror of Figure 5a, which provides
double-ended to single-ended conversion, still maintains
balance in the circuit because the second stage output voltage is determined by the current that flows into the high
impedance presented by the collectors.
Of greater significance is the fact that the dynamic emitter
resistance(1) of the added transistor is substituted for the
parallel resistance of the feedback network in Figure 2. The
first stage gain, and consequently the open-loop gain, no
longer depends on the feedback network. The process of
frequency compensation has one less degree of variation to
be concerned with.
Inspection of Figure 4 shows that the axis of symmetry for
the CFA is centered horizontally. Therefore, the half-circuit
used for analysis is the same circuit as presented in Figure
1, ignoring the input emitter follower. However, as pointed
out previously, the feedback network is closely intertwined
with the analysis. Therefore, the circuit of Figure 2 can be
used for the analysis. The compensation capacitor, CT, can
be the intrinsic base-collector capacitor of Q2 or an extrinsic
capacitor deliberately added for compensation.
These two circuits illustrate the basic distinctions between
current-feedback and voltage-feedback amplifiers. In both
cases, the feedback network is connected to an (inverting)
input node. In Figure 1, the emitter presents a low impedance input, while in Figure 3 the base presents a high
impedance input.
The only real difference between Figure 5b and Figure 2 is
the presence of the parallel combination of the feedback
network resistors in the emitter of the CFA’s input transistor.
Needless to say, the three transistor amplifier of Figure 1 can
be considered the forebear for the CFA as it is known today,
while Figure 3 is the template for the VFA. Figure 4 shows
the same amplifier connected to a mirror-image of itself,
whose transistors have been converted to the opposite polarity type. The input transistors are buffered by emitter followers for level shifting to ensure low offset voltage. This is the
basis of the modern current-feedback architecture.
The CFA analysis is straightforward and the DC gain can be
determined by inspection of Figure 2.
A VDC =
R1 R 3
•
RE R2
The open-loop pole can be approximated quite accurately as
the interaction of the resistor, R1, with the Miller multiplied
capacitor, CT.
ANALYZING THE CFA
The study of the differential input, voltage-feedback amplifier is simplified with a technique known as “half-circuit
analysis.” This technique, illustrated in Figure 5, recognizes
that the symmetry of the circuit presents an opportunity for
simplification whereby only half the signal path needs to be
considered.
ωP ≅
1

R
R1  3 • C T 

 R2
This analysis presumes that re1, the dynamic emitter resistance of Q1, can be neglected (RE>>re1) and that R2 includes
re2.
NOTE: (1) The dynamic emitter resistance is tangent to the
slope of the I-V curve for the base-emitter diode
2
–In
+In
VOUT
FIGURE 5a. Basic VFA Topology.
Removing RE, the feedback network term, from the equation
for open-loop voltage gain yields a more general expression
that describes the amplifier’s open-loop performance in
terms of its intrinsic characteristics. This equation would
have units of ohms and would be better identified as a
complex impedance, or transimpedance, ZT:
+1
ZT =
VOUT
RT
1 + jωR T C T
This is the true measure of performance for CFAs. It is now
obvious why the amplifier is known as “current-feedback.”
The output voltage is responsive to a current at the low
impedance inverting input node (the emitter of Q1) that
interacts with the open-loop transimpedance, ZT.
+In
FIGURE 5b. VFA Half-Circuit.
Furthermore, the open-loop response of the amplifier is
completely described by the DC transresistance, RT, and the
compensation capacitor, C T , which is called the
transcapacitance. RT interacts with CT to form the open-loop
pole. This is graphically depicted in Figure 6.
It would be convenient at this point to define the
transresistance as:
RT =
R1 • R 3
R2
ZT
Note that the transresistance has the dimensions of ohms and
is determined solely by elements internal to the amplifier.
The previous equations can be rewritten more simply.
A VDC =
RT
RE
and ω P =
RT
1
R T • CT
Now the open-loop gain can be completely described by:
AV =
RT
R
1
1
•
= T•
RE 1 + j ω
R E 1 + jωR T C T
ωP
In order to arrive at this equation, it was assumed that the
feedback network was known. This is the crux of the issue—
the open-loop voltage gain of a CFA depends on the value
of the feedback network.
(RTCT)–1
FIGURE 6. Open-Loop Transimpedance.
3
log ω
The DC value of closed-loop gain is set by the feedback
network, while the closed-loop pole is determined by the
interaction of the transcapacitance with the feedback resistor. This latter term is what gives the CFA its much touted
characteristic of gain-independent bandwidth.
The ordinate axis has the dimension of ohms and is scaled
logarithmically.
Having described the CFA with just two components suggests a simplified version of the half-circuit used for analysis. Figure 7 shows a convenient model that has all the
essentials necessary for quick hand calculations. The inverting buffer preserves the sense of the signal as it is amplified
by the Q2 stage in Figure 2.
RT
Figure 8 illustrates graphically that the open-loop gain curve
slides vertically to keep the closed-loop intercept frequency
constant. Since RF is kept constant, the area of the curve
above the closed-loop gain stays constant.
CT
–1
+In
A closer look at the unsimplified equation for the closedloop gain helps to clarify this property. The DC portion of
open-loop gain in the numerator is modified by the parallel
combination of the feedback network, which changes with
desired closed-loop gain. As long as RF is kept constant, the
loop gain expression in the denominator does not vary, nor
do any of the frequency dependent terms.
VOUT
Q1
RL = RF + RG
–In
AV
RF • RG
RE =
RF + RG
RT
RE´
RT
RE
FIGURE 7. CFA Model for AC Analysis.
When determining which op amp to use for an application,
comparisons with voltage-feedback amplifiers will inevitably be made. Presumably the closed-loop gain is known,
which means that a feedback network can be established.
Therefore, the open-loop voltage gain can be calculated for
the CFA and a fair comparison with VFA can be established.
RF
RE´
RF
RE
Note that the analysis described here is based on a fairly
simple current-feedback topology. Although the design of
integrated circuit CFAs has become more sophisticated, the
open-loop transimpedance approach (ZT) is still valid.
(RTCT)–1
AV
1 + AV • β
where β =
The closed-loop gain expressions have been expressed as a
ratio of the feedback resistor to the equivalent feedback
network. This can be verified algebraically as:
R + RG
RF
RF
=
= F
RE  RF • RG 
RG


 RF + RG 
RG
RF + RG
Substituting for AV yields the following expression:
Thus, the open-loop gain varies directly with the closed-loop
gain for changes in RE as long as RF is kept constant.
RT
1
•
R E 1 + jωR T C T
Open–Loop Gain
A CL =
=
RF
Loop Gain
1+
+ jωR F C T
RT
RT
•
RF
1 + jωR T C T
The loop gain, of course, limits the accuracy of the closedloop gain. Note that RT>>RF (typically RT>100k and RF<5k),
therefore the equation can be easily simplified to:
A CL =
log ω
FIGURE 8. Variation of Open-Loop Gain.
CLOSED-LOOP PERFORMANCE
The closed-loop response of the CFA can be described by
using classical analysis:
A CL =
(RFCT)–1
RF + RG
1
•
RG
1 + jωR F C T
4
NONIDEAL CONSIDERATIONS
The assumption that the re of Q1 can be neglected has limits.
For ease of analysis, Figure 6 has been redrawn to include it
as a finite input resistance, RIN (Figure 9). Note that RIN is
internal to the CFA terminals.
The second generalization is that the closed-loop bandwidth
will become gain-bandwidth limited when
R IN
≥ R F ⇔ R IN ≥ R E
β
The latter expression makes use of the fact that the feedback
factor, β, is a function of the feedback network resistors.
Once this limit has been reached, the CFA can be associated
with a gain-bandwidth product, GBW.
RT
CT
GBW =
1
R IN C T
VOUT
–1
The graph in Figure 10 shows an asymptotic approach to
estimating a CFA’s closed-loop response.
+In
RL = RF + RG
RIN
–In
RE =
AV
RF • RG
RF + RG
RT
RIN
FIGURE 9. CFA Model Modified for Finite RIN.
The open-loop gain equation can be modified by inspection,
while a new closed-loop gain equation can again be derived
using the classical approach.
RF
RIN
RT
1
AV =
•
R E + R IN 1 + jωR T C T
A CL =
RF + RG
•
RG
1

R 
1 + jω  R F + IN  C T
β 

(RTCT)–1
RIN decreases the open-loop gain but not its corner frequency. On the other hand, RIN does not affect the DC
closed-loop gain but does modify the intercept frequency. In
practice, RIN includes more than just the dynamic emitter
resistance—it also includes bulk resistances that are in series
with the inverting input, as well as parasitic resistances
external to the amplifier. Obviously, RIN should be as low as
possible to get the maximum benefit from a CFA.
(RINCT)–1 log ω
FIGURE 10. CFA Closed-Loop Performance.
To be technically accurate, it should be pointed out that the
inverting input is characterized by an impedance, ZIN, which
does vary with frequency. Fortunately, the resistive portion,
RIN, dominates over most of the CFA’s useful bandwidth. At
high frequency, the inverting input impedance increases,
which only further degrades the closed-loop performance,
although the extent of the increase is generally well under an
order of magnitude.
The modified equations lead to some practical generalizations when using CFAs. The first is that the open-loop gain
has a theoretical maximum and this can be conveniently
estimated as:
A V(max) ≅
(RFCT)–1
FREQUENCY COMPENSATION
The analysis so far has centered on the gain versus frequency
performance without taking into account any phase shift
considerations. Excess phase plagues the CFA just as it does
the VFA. The open-loop transimpedance curve of Figure 6
depicts a single-pole response which would have only 90° of
phase shift. Parasitic poles introduce additional phase shift
to the open-loop phase response. Figure 11 displays the
more complete open-loop transfer curves—both magnitude
and phase.
RT
1
•
R IN 1 + jωR T C T
This is an ideal value that can never be realized since any
feedback network will automatically reduce the open-loop
gain. However, it is useful for estimating a CFA’s merits
against a particular VFA.
5
ZT
UNITY GAIN FREQUENCY RESPONSE vs
PHASE MARGIN
5
RT
45°
0
Gain (dB)
–5
60°
90°
–10
75°
–15
–20
ZT(f1)
–25
fi/100
fi
10fi
Relative Frequency
log ω
(RTCT)–1
fi/10
fi
FIGURE 12. Phase Margin’s Effect on Frequency Response.
0°
A more general way of looking at this is to make the
observation that the closed-loop response can be extended if
the open-loop phase has fallen 120° at fi, the frequency
where the asymptote for closed-loop gain intersects the
open-loop gain curve.
–45°
–90°
In VFAs, the phase margin is set by design and the user does
not change it. There are a few amplifiers which allow access
to the high impedance node to tailor compensation, but these
are in the minority. In general, VFAs break out into two
categories—compensated and decompensated.
–135°
ØM
–180°
Ø
The compensated amplifiers allow operation at unity gain
but at the expense of bandwidth in higher gains. Decompensated, or undercompensated, amplifiers must be operated in
gains greater than unity but have a higher gain-bandwidth
product. In either case, the phase margin is predetermined.
FIGURE 11. CFA Open-Loop Transfer Curves.
Since the feedback network sets the open-loop gain for the
CFA, it also sets the phase margin, ΦM. This is the crucial
factor that actually determines the selection of the feedback
network resistors.
For the CFA, phase margin is set by the user via the
feedback network. However, rather than use phase margin as
the design criterion, higher performance can be attained by
making use of the general observations regarding phase shift
and bandwidth. In other words, guarantee that the open-loop
phase has fallen 120° at fi.
The significance of phase margin would benefit from a brief
review of its properties. Phase margin for operational amplifiers is measured at that frequency, fU, where an amplifier’s
open-loop voltage gain has fallen to unity. It is the difference
between the open-loop phase shift and –180°, where the
amplifier would lose negative feedback and become unstable.
The mechanics are rather straightforward because, as illustrated in Figure 8, varying the feedback network causes a
simple vertical translation of the open-loop gain curve. The
open-loop pole does not move and so the attendant openloop phase shift is unaffected. The excess phase shift is also
insensitive to the feedback network change. Thus, selection
of a desired phase shift automatically sets the intercept
frequency.
Φ M = Φ(f U ) − (−180° )
The concept of phase margin is best illustrated by plotting
unity gain frequency response curves as phase margin is
varied (Figure 12).
Once the intercept frequency, fi, is determined, so is the
magnitude of transimpedance, ZT(fi). This is depicted graphically in Figure 11 by following the dashed lines up from the
open-loop phase curve to the intersection with the open-loop
transimpedance curve.
As the plot shows, the optimum value for phase margin is
60°. This gives the desirable combination of broad bandwidth with flat frequency response. Note that an amplifier
with 90° of phase margin, which implies a lack of excess
phase, has a –3dB bandwidth less than half of the optimum
response.
6
To realize the benefit of the –120° phase shift, the feedback
network has to be selected so that the open-loop gain equals
the closed-loop gain at fi. A convenient way to visualize this
problem is to concentrate on the essentials of the model in
Figure 9.
MODEL REPRESENTATION
The single transistor model of Figure 9 is a satisfactory
vehicle to provide intuitive insight. It is by no means an
accurate representation of the CFA but offers a good visual
aid for the user.
The CFA model can be simplified further by ignoring the
inverting buffer and focusing on that portion of the circuit
which provides gain. In Figure 13 the CFA model has been
reduced to an elementary transistor amplifier. The gain for
this circuit is
A more generally accepted model for the CFA is depicted in
Figure 14. This model is a very faithful rendition of the CFA
from a block diagram standpoint. It can accurately account
for the bipolar input and output swings that are possible with
the CFA’s complementary symmetry.
AV =
Comparing it to Figure 4, it is readily apparent that the unity
gain buffer at the input is an accurate portrayal of the input
stage between the input pins. The finite input resistance, RIN,
is included for completeness.
Z T (f i )
R E + R IN
ZT(fi)
VOUT
+In
Q1
+In
RIN
+1
–In
RE =
+1
RF • RG
RF + RG
RIN
IIN
RT
VOUT
CT
–In
IIN
FIGURE 13. Elementary Amplifier.
The goal, therefore, is to select the necessary feedback
network so that AV equals the desired closed-loop gain.
Since ZT has previously been defined as a complex impedance, direct substitution yields a closed form solution.
RF + RT
RG
FIGURE 14. Block Diagram CFA Model.
RT
1 + j2πf i R T C T
Z T (f i )
=
=
R E + R IN
R E + R IN
The current-controlled current source, IIN, translates the
current from the inverting input to the open-loop
transimpedance, again composed of RT and CT. The unitygain buffer provides a low impedance source to the external
load.
which can be reduced to a less bulky equation:
RF ≅
R
1
− IN
β
2πf i C T
Either of the models is sufficient to appreciate the CFA and
its performance features. Figure 9 bears a strong resemblance to the ancestral antecedent of the CFA while the latter
is more readily adaptable to generating a SPICE macromodel.
Not surprisingly, this expression conforms to the plot of
CFA closed-loop performance (Figure 10). For low gains,
the RIN term is negligible and RF is set by fi. As closed-loop
gain increases and RIN/β can no longer be neglected, RF
should be adjusted according to the equation to maintain
optimum performance. When RF approaches zero, the CFA
is becoming gain-bandwidth limited and the intercept frequency must be lowered.
Other properties of the CFA are apparent when studying
these models. The slew rate is limited by the current available to charge the transcapacitance. Decreasing (RIN + RE)
will certainly benefit slew performance. Minimizing CT will
increase slew rate as well as the small-signal performance.
7
Potential for trouble exists when parasitic capacitance is
present at the inverting input. This parasitic capacitance can
be the result of poor layout techniques, inappropriate use of
a socket or even the wrong package. If CP is the lumped
parasitic capacitor, the open-loop gain will become:
AV =
RT
•
R E + R IN
It is not very common practice to specify power supply
rejection for each supply separately but, for the CFA, it is
essential. The complementary devices, NPN and PNP, should
not be expected to match each other closely and usually the
PNPs are the weaker. PSR measured with tracking supplies
typically tend to partially cancel the errors. Real world
applications usually rely on independent positive and negative voltage regulators.
1 + jωR E C P


R •R
(1 + jωR T C T )1 + jω R E + RIN C P 


E
IN
The table below is for a medium performance CFA and
exemplifies the amount of detail that should be provided.
This expression has added a zero and a pole to the transfer
function. The zero will always occur before the pole and can
be the source of trouble in some cases. If instability arises
because of CP, move the closed-loop pole to a lower frequency by adjusting the feedback network.
PARAMETER
To model excess phase, the addition of a delay line can be
more expedient than trying to add multiple poles and zeroes
to the open-loop transimpedance. The modified transfer
function is still quite compact.
RT
ZT =
• e − jωTD
1 + jωR T C T
The exponential adds phase shift without affecting magnitude. A reasonable technique is to use the phase shift at the
highest intercept frequency the circuit is expected to encounter.
1
TD =
Φ(f i ) − 90°
2πf i •
360°
Here, subtracting 90° from the open-loop phase, Φ(fi), removes the phase shift due to the open-loop pole.
TYP
UNIT
INPUT OFFSET VOLTAGE
Initial
vs Temperature
vs Common-mode
vs Supply (Tracking)
vs Supply (Non-tracking)
5
8
60
85
60
mV
µV/°C
dB
dB
dB
+INPUT BIAS CURRENT
Initial
vs Temperature
vs Common-mode
vs Supply (Tracking)
vs Supply (Non-tracking)
5
30
200
50
150
µA
nA/°C
nA/V
nA/V
nA/V
–INPUT BIAS CURRENT
Initial
vs Temperature
vs Common-mode
vs Supply (Tracking)
vs Supply (Non-tracking)
25
300
200
300
1500
µA
nA/°C
nA/V
nA/V
nA/V
5M || 2
30 || 2
Ω || pF
Ω || pF
OPEN-LOOP TRANSIMPEDANCE
Transresistance
Transcapacitance
440
1.8
kΩ
pF
OUTPUT CHARACTERISTICS
Voltage
Current
Output resistance, Open-loop
12
150
70
V
mA
Ω
INPUT IMPEDANCE
+Input
–Input
DATA SHEET SPECIFICATIONS
The open-loop transimpedance terms, RT and CT, and the
input resistance, RIN, have already been identified as necessary features to describe a CFA. Additionally, the open-loop
transimpedance and phase versus frequency curves should
be provided as well.
TABLE I. Source: BB OPA603 Data Sheet.
The block diagram presentation of Figure 14 suggests the
other specifications that should not be overlooked. The
presence of a buffer between the noninverting and inverting
inputs of the CFA guarantees that the input characteristics
will not match. This is the main difference between the VFA
and the CFA data sheets.
SPICE SIMULATION
The combination of declining hardware costs with increasing computing horsepower has made circuit simulation a
required part of the design cycle. This has forced the op amp
vendors to supply the macromodels for their product offerings.
The VFA data sheet typically specifies the power supply and
common-mode rejection for the offset voltage only. The
input bias currents are also subject to disturbances from
these sources but good VFA design encourages matching
impedances at the inputs to mask the effects.
These simulation tools have been offered in varying degrees
of complexity, from the simple Boyle model to simplified
circuit models, which utilize full transistor models in the
signal path. There has been a growing consensus that this
latter approach is necessary for the high bandwidth amplifiers.
The CFA does not have the privilege of bias current match,
so the same effects that are specified for the offset voltage
need to be measured for the two input currents. In particular,
the inverting input, which is the true signal input is often the
biggest source of error. It is not uncommon to see a CFA
constrained to operate in an inverting gain configuration to
circumvent common-mode effects.
8
The circuit of Figure 15a was simulated with the following
listing:
There can be no doubt that having these models available
helps to fill in the gaps from incomplete data sheets. Although the models may not necessarily be configured for
worst case process extremes, there may be some performance peculiarities that can be discovered through their use.
The pitfall to be aware of is that even the simplified circuit
models generally idealize the biasing circuitry, which may
mask some second order PSR and CMR effects.
* CURRENT-FEEDBACK OPEN-LOOP SIMULATION *
* file: CFA-OL.CIR
***** Simulation Commands *****
.options noecho nomod numdgt=8
.op
.ac dec 20 10 200meg
.probe
***** Library Files *****
.lib burr_brn.lib
***** Circuit Listing *****
vp 7 0 15
vm 4 0 -15
*ginv 2 0 6 0 -1
inv 2 0 dc -38.3pa ac 1
x603 0 2 7 4 6 opa603
rl 6 0 100k
.end
Figure 15 shows two alternative simulation schemes. In
Figure 15a, the CFA is driven open-loop to measure the
open-loop transimpedance and input resistance. This requires two separate simulations. The first uses a voltagecontrolled current source to find the dc value of inverting
input current to servo the output to zero. The second pass is
the ac simulation to actually measure the transimpedance.
VDIF
Figure 16a is the plot of input resistance as measured by
dividing the ac voltage by the ac current. Note that for the
useful frequency range of the amplifier (roughly 100MHz),
RIN varies less than 10Ω. The open-loop transimpedance is
displayed in Figure 16b. Here the magnitude has fallen from
a DC value of 790kΩ to 1.5kΩ at 51.6MHz, which is where
the open-loop phase has fallen to –120°.
VOUT
RL
CL
IIN
FIGURE 15a. Open-Loop Simulation.
Figure 15b uses a zero volt battery to measure the inverting
input current while the op amp is in a closed-loop configuration. This measures an effective transimpedance that includes the common-mode effect.
VDIF
VOUT
RL
FIGURE 16a. Measuring the Inverting Input Impedance.
CL
VINV
RF
RG
FIGURE 15b. In Circuit Measurement.
FIGURE 16b. Measuring Open-Loop Transimpedance.
9
MEASUREMENT CIRCUITS
If companies could ship only simulation files to their customers, life would be so easy. Sooner or later, a reality check
has to be made. The following circuits have been proven to
be quite reliable for measuring the CFA performance parameters.
The circuit of Figure 15b was simulated with the following
listing:
* CURRENT-FEEDBACK CLOSED-LOOP SIMULATION *
* file: CFA-CL.CIR
***** Simulation Commands *****
.options noecho nomod
.ac dec 20 1000 200meg
.probe
***** Library Files *****
.lib burr_brn.lib
***** Circuit Listing *****
vp 7 0 15
vm 4 0 -15
vin 3 0 dc 0 ac 1
x603 3 inv 7 4 6 opa603
vinv inv 2 dc 0
rf 6 2 1450
rg 2 0 1450
.end
The low impedance of the inverting input node presents a
special problem for the test engineer. Conventional op amp
test circuits cannot easily separate the individual parameter
variations. The most logical solution is to test the CFA with
a current mode test circuit.
Figure 18 shows the basic current pump topology used in the
DC test circuit. It consists of a JFET input op amp, a Pchannel MOSFET and a unique current reference circuit
which includes two very accurate current sources and a high
precision current mirror.
The plot in Figure 17 shows the intersection of the open-loop
gain curve with the closed-loop gain asymptote which occurs at 45.7MHz. The open-loop phase has the value of
–120° at this frequency and the broadbanding of the closedloop gain is quite evident. Note the technique used to
generate the open-loop gain curve.
IS1
100µA
The equation relies on the calculation of open-loop
transimpedance (via the current in the battery) which is
divided by the sum of the equivalent feedback network plus
the input resistance.
IS2
100µA
RIN
VIN
RL
1/2
OPA2111
vm(output)
Z T (f i )
im(battery)
AV =
=
R E + R IN
R E + R IN
REF200
FIGURE 18. Current Pump Topology.
The high gain of the JFET input op amp (VFA) constrains
its inverting input to stay at null ground by controlling the
current flowing through the MOSFET. If VIN is positive, a
current equal to VIN/RIN is shunted to the current mirror
input. If VIN is negative, a matching VIN/RIN is provided by
the 100µA current source, IS1, and the input to the current
mirror decreases. This is an inverting current pump, a
positive voltage causes the output to sink current and a
negative input causes the output to source current.
FIGURE 17. Slope Intercept Curves for CFA Circuit.
10
The full test circuit is shown in Figure 20. The input offset
voltage of the DUT is measured directly by the instrumentation amplifier, A1. The RC filters minimize noise and
protect the inputs of A1 from overload transients.
integrator servos the output to zero by sensing the DUT
output and feeding a small current back to the input. A2
buffers the DUT output and drives the 50Ω input of the
network analyzer.
Amplifier A2 maintains the common-mode bias by forcing
the current pump (A3, M1, IC1) to keep the noninverting
input of the DUT equal to the input, VCM. The output of A2
driving the 100kΩ input resistor to the current pump is a
measure of +Ib.
The only caveat is to take into account the gain and phase
rolloff of A2. Automated network analyzers allow for compensation by storing an “offset” sweep which is subtracted
from the actual signal sweep.
Although the network analyzer will scale the output in dB,
the transimpedance can be determined by using the following equation:
Amplifier A4 constrains the DUT output to be the negative
of the input voltage, Vf, by forcing the current pump (A5, M2,
IC2) to drive the low impedance inverting input. The amount
of inverting current drive is reflected by the output of A4.
Z T = 500 • log −1 

All DC parameters, including RT and RIN, can be measured
independently and directly. When adapted to a measurement
card for the HP Semiconductor Analyzer, the test parameters
can be displayed as slopes to determine the limits of linearity.
dB magnitude 

20
The transcapacitance can be found by extrapolating the
open-loop pole.
CONCLUSION
Figure 19 details an open-loop transimpedance test circuit
which, when mated with a network analyzer, will provide
the open-loop frequency response curves.
CFAs are not difficult to comprehend and work with if the
basic relationships between RT, CT, RIN and open-loop phase
are kept in mind. The lack of balanced input nodes require
extra care be taken with applications requiring DC accuracy.
Simulation is a wonderful tool for the early design stages but
only actual measurements will grant peace of mind.
The input ladder network divides the input by 20,000 to
provide a low current level signal to the inverting input of
the DUT. The 500Ω value for the input resistor dominates
the small but finite input resistance of the CFA. The A1
0.2µF
500kΩ
0.4µF
A1
A1: OPA177
A2: OPA621
NOTE: (1) Short these points
to compensate test circuit.
500kΩ
100kΩ
100Ω
9750Ω
DUT
500Ω
50Ω
VIN
A2
CL
50Ω
500Ω
RL
(1)
250Ω
250Ω
(1)
FIGURE 19. Open-Loop Frequency Response Test Circuit.
11
VOUT
100kΩ
A2
A3
+IB
IC1
M1
1000pF
100kΩ
VCM
10kΩ
+VS
1000pF
100kΩ
VOS
A1
x200
VOUT
DUT
100kΩ
1000pF
–VS
A1: INA110
A1–A5: 1/2 OPA2111
IC1–IC2: REF200
M1–M2: VPO300
200kΩ
1000pF
200kΩ
Vf
A5
M2
IC2
A4
100kΩ
–IB
FIGURE 20. Current Mode CFA Test Circuit.
12