AD844
Rev. F | Page 14 of 20
NONINVERTING GAIN OF 100
It is important to understand that the low input impedance at
the inverting input is locally generated and does not depend on
feedback. This is very different from the virtual ground of a
conventional operational amplifier used in the current summing
mode, which is essentially an open circuit until the loop settles.
In the AD844, transient current at the input does not cause
voltage spikes at the summing node while the amplifier is
settling. Furthermore, all of the transient current is delivered
to the slewing (TZ) node (Pin 5) via a short signal path (the
grounded base stages and the wideband current mirrors).
The AD844 provides very clean pulse response at high
noninverting gains. Figure 32 shows a typical configuration
providing a gain of 100 with high input resistance. The feedback
resistor is kept as low as practicable to maximize bandwidth,
and a peaking capacitor (C
PK
) can optionally be added to
further extend the bandwidth. Figure 33 shows the small signal
response with C
PK
= 3 nF, R
L
= 500 Ω, and supply voltages of
either ±5 V or ±15 V. Gain bandwidth products of up to
900 MHz can be achieved in this way.
The current available to charge the capacitance (about 4.5 pF) at
the TZ node is always proportional to the input error current,
and the slew rate limitations associated with the large signal
response of the op amps do not occur. For this reason, the rise
and fall times are almost independent of signal level. In practice,
the input current eventually causes the mirrors to saturate.
When using ±15 V supplies, this occurs at about 10 mA (or
±2200 V/μs). Because signal currents are rarely this large,
classical slew rate limitations are absent.
The offset voltage of the AD844 is laser trimmed to the 50 μV
level and exhibits very low drift. In practice, there is an
additional offset term due to the bias current at the inverting
input (I
BN
), which flows in the feedback resistor (R1). This can
optionally be nulled by the trimming potentiometer shown in
Figure 32.
OFFSET
TRIM
C
PK
3nF
20
4.7
0.22µF
0.22µF
R
L
V
IN
+
V
S
–V
S
AD844
R1
499
R2
4.99
4.7
1
2
3
8
7
4
6
00897-032
This inherent advantage is lost if the voltage follower used
to buffer the output has slew rate limitations. The AD844 is
designed to avoid this problem, and as a result, the output
buffer exhibits a clean large signal transient response, free
from anomalous effects arising from internal saturation.
RESPONSE AS A NONINVERTING AMPLIFIER
Because current feedback amplifiers are asymmetrical with
regard to their two inputs, performance differs markedly in
noninverting and inverting modes. In noninverting modes, the
large signal high speed behavior of the AD844 deteriorates at
low gains because the biasing circuitry for the input system (not
shown in Figure 31) is not designed to provide high input
voltage slew rates.
Figure 32. Noninverting Amplifier Gain = 100, Optional Offset Trim Is Shown
FREQUENCY (Hz)
GAIN (dB)
46
16
100k 1M 20M10M
40
34
28
22
00897-040
V
S
= ±5V
V
S
= ±15V
However, good results can be obtained with some care. The
noninverting input does not tolerate a large transient input; it
must be kept below ±1 V for best results. Consequently, this
mode is better suited to high gain applications (greater than
×10). Figure 23 shows a noninverting amplifier with a gain of 10
and a bandwidth of 30 MHz. The transient response is shown in
Figure 26 and Figure 27. To increase the bandwidth at higher
gains, a capacitor can be added across R2 whose value is
approximately (R1/R2) × C
t
.
Figure 33. AC Response for Gain = 100, Configuration Shown in Figure 32
AD844
Rev. F | Page 15 of 20
USING THE AD844
BOARD LAYOUT
As with all high frequency circuits considerable care must be
used in the layout of the components surrounding the AD844.
A ground plane, to which the power supply decoupling capaci-
tors are connected by the shortest possible leads, is essential to
achieving clean pulse response. Even a continuous ground plane
exhibits finite voltage drops between points on the plane, and
this must be kept in mind when selecting the grounding points.
In general, decoupling capacitors should be taken to a point
close to the load (or output connector) because the load
currents flow in these capacitors at high frequencies. The +IN
and −IN circuits (for example, a termination resistor and Pin 3)
must be taken to a common point on the ground plane close to
the amplifier package.
Use low impedance 0.22 μF capacitors (AVX SR305C224KAA
or equivalent) wherever ac coupling is required. Include either
ferrite beads and/or a small series resistance (approximately
4.7 Ω) in each supply line.
INPUT IMPEDANCE
At low frequencies, negative feedback keeps the resistance at the
inverting input close to zero. As the frequency increases, the
impedance looking into this input increases from near zero to
the open-loop input resistance, due to bandwidth limitations,
making the input seem inductive. If it is desired to keep the
input impedance flatter, a series RC network can be inserted
across the input. The resistor is chosen so that the parallel sum
of it and R2 equals the desired termination resistance. The capacit-
ance is set so that the pole determined by this RC network is
about half the bandwidth of the op amp. This network is not
important if the input resistor is much larger than the termination
used, or if frequencies are relatively low. In some cases, the
small peaking that occurs without the network can be of use in
extending the −3 dB bandwidth.
DRIVING LARGE CAPACITIVE LOADS
Capacitive drive capability is 100 pF without an external net-
work. With the addition of the network shown in Figure 34,
the capacitive drive can be extended to over 10,000 pF, limited
by internal power dissipation. With capacitive loads, the output
speed becomes a function of the overdriven output current limit.
Because this is roughly ±100 mA, under these conditions, the
maximum slew rate into a 1000 pF load is ±100 V/μs. Figure 35
shows the transient response of an inverting amplifier (R1 =
R2 = 1 kΩ) using the feedforward network shown in Figure 34,
driving a load of 1000 pF.
AD844
V
OUT
C
L
750
22pF
6
5
00897-034
Figure 34. Feedforward Network for Large Capacitive Loads
00897-035
5V
10
500ns
100
0
90
Figure 35. Driving 1000 pF C
L
with Feedforward Network of Figure 34
SETTLING TIME
Settling time is measured with the circuit of Figure 36. This
circuit employs a false summing node, clamped by the two
Schottky diodes, to create the error signal and limit the input
signal to the oscilloscope. For measuring settling time, the ratio
of R6/R5 is equal to R1/R2. For unity gain, R6 = R5 = 1 kΩ, and
R
L
= 500 Ω. For the gain of −10, R5 = 50 Ω, R6 = 500 Ω, and R
L
was not used because the summing network loads the output
with approximately 275 Ω. Using this network in a unity-gain
configuration, settling time is 100 ns to 0.1% for a –5 V to +5 V
step with C
L
= 10 pF.
R5
D1 D2
R1
R2
R3
R
L
R6
V
IN
C
L
V
OUT
TO SCOPE
(TEK 7A11 FET PROBE)
NOTES
1. D1, D2 IN6263 OR EQUIVALENT SCHOTTKY DIODE.
AD844
00897-036
Figure 36. Settling Time Test Fixture
AD844
Rev. F | Page 16 of 20
DC ERROR CALCULATION
Figure 37 shows a model of the dc error and noise sources for
the AD844. The inverting input bias current, I
BN
, flows in the
feedback resistor. I
BP
, the noninverting input bias current, flows
in the resistance at Pin 3 (R
P
), and the resulting voltage (plus
any offset voltage) appears at the inverting input. The total
error, V
O
, at the output is:
()
R1I
R2
R1
RIVRIV
BNINBN
OS
PBP
O
+
+++= 1
Because I
BN
and I
BP
are unrelated both in sign and magnitude,
inserting a resistor in series with the noninverting input does
not necessarily reduce dc error and may actually increase it.
R2
V
N
I
NN
I
NP
R
P
R1
I
BP
I
BN
V
OS
AD844
R
IN
00897-037
Figure 37. Offset Voltage and Noise Model for the AD844
NOISE
Noise sources can be modeled in a manner similar to the dc bias
currents, but the noise sources are I
NN
, I
NP
, V
N
, and the amplifier
induced noise at the output, V
ON
, is:
()()()
2
2
2
2
1 R1I
R2
R1
VRIV
NNNPNP
ON
+
++=
Overall noise can be reduced by keeping all resistor values to a
minimum. With typical numbers, R1 = R2 = 1 kΩ, R
P
= 0 Ω,
V
N
= 2 nV/√Hz, I
NP
= 10 pA/√Hz, I
NN
= 12 pA/√Hz, and V
ON
calculates to 12 nV/√Hz. The current noise is dominant in this
case, because it is in most low gain applications.
VIDEO CABLE DRIVER USING ±5 V SUPPLIES
The AD844 can be used to drive low impedance cables. Using
±5 V supplies, a 100 Ω load can be driven to ±2.5 V with low
distortion. Figure 38 shows an illustrative application that
provides a noninverting gain of +2, allowing the cable to be
reverse-terminated while delivering an overall gain of +1 to the
load. The −3 dB bandwidth of this circuit is typically 30 MHz.
Figure 39 shows a differential gain and phase test setup. In video
applications, differential-phase and differential-gain characteris-
tics are often important. Figure 40 shows the variation in phase as
the load voltage varies. Figure 41 shows the gain variation.
V
IN
50
50
R
L
50
300
300
3
2
+5
V
–5V
7
6
4
2.2µF
2.2µF
Z
O
= 50
V
OUT
0
0897-038
Figure 38. The AD844 as a Cable Driver
HP8753A
NETWORK
ANALYZER
HP11850C
SPLITTER
CIRCUIT
UNDER
TEST
HP3314A
STAIRCASE
GENERATOR
V
OUT
V
IN
V
IN
OUT
OUT
OUT
INRF OUT
RF IN
EXT
TRIG
SYNC OUT
50
(TERMINATOR)
OUT
470
0
0897-039
Figure 39. Differential Gain/Phase Test Setup
V
OUT
(IRE)
DIFFERENTIAL PHASE (Degrees)
0.3
0.2
–0.3
90180
36 54 72
0.1
0
–0.1
–0.2
IRE = 7.14mV
00897-040
Figure 40. Differential Phase for the Circuit of Figure 38
V
OUT
(IRE)
DIFFERENTIAL GAIN (%)
0.06
0.04
–0.06
018 936 54 72
0.02
0
–0.02
–0.04
0
IRE = 7.14mV
00897-041
Figure 41. Differential Gain for the Circuit of Figure 38

AD844JRZ-16-REEL

Mfr. #:
Manufacturer:
Analog Devices Inc.
Description:
Operational Amplifiers - Op Amps 60MHz 2000V/uS Monolithic
Lifecycle:
New from this manufacturer.
Delivery:
DHL FedEx Ups TNT EMS
Payment:
T/T Paypal Visa MoneyGram Western Union