ADL5391
Rev. 0 | Page 9 of 16
1.00UFS
3001.000
S11 DIFF
1.000
201.000 0.654 U –36.340 DEG
1001.000 0.594 U –92.533 DEG
1901.000 0.531 U –94.448 DEG
201.000 0.800 U –17.218 DEG
2001.000 0.564 U –58.167 DEG
06059-017
S11 SE
Figure 15. Input S11
1.00UFS
3001.000
1.000
201.000 0.947 U +170.736 DEG
1001.000 0.569 U +58.257 DEG
1901.000 0.597 U –69.673 DEG
201.000 0.905 U +157.308 DEG
2001.000 0.663 U –39.468 DEG
06059-018
S22 DIFF
S22 SE
Figure 16. Output S22
ADL5391
Rev. 0 | Page 10 of 16
GENERAL DESCRIPTION
BASIC THEORY
The multiplication of two analog variables is a fundamental
signal processing function that has been around for decades.
By convention, the desired transfer function is given by
W = αXY/U + Z (1)
where:
X and Y are the multiplicands.
U is the multiplier scaling factor.
α is the multiplier gain.
W is the product output.
Z is a summing input.
All the variables and the scaling factor have the dimension of volts.
In the past, analog multipliers, such as the
AD835, were
implemented almost exclusively with a Gilbert Cell topology
or a close derivative. The inherently asymmetric signal paths
for X and Y inevitably create amplitude and delay imbalances
between X and Y. In the ADL5391, the novel multiplier core
provides absolute symmetry between X and Y, minimizing
scaling and phasing differences inherent in the Gilbert Cell.
The simplified block diagram of the ADL5391 shows a main
multiplier cell that receives inputs X and Y and a second
multiplier cell in the feedback path around an integrating
buffer. The inputs to this feedback multiplier are the difference
of the output signal and the summing input, W − Z, and the
internal scaling reference, U. At dc, the integrating buffer
ensures that the output of both multipliers is exactly 0, therefore
(W Z)xU = XY, or W = XY/U + Z (2)
By using a feedback multiplier that is identical to the main
multiplier, the scaling is traced back solely to U, which is
an accurate reference generated on-chip. As is apparent in
Equation 2, noise, drift, or distortion that is common to both
multipliers is rejected to first-order because the feedback
multiplier essentially compensates the impairments generated
in the main multiplier.
The scaling factor, U, is fixed by design to 1.12 V. However, the
multiplier gain, α, can be adjusted by driving the GADJ pin with
a voltage ranging from 0 V to 2 V. If left floating, then α = 1 or
0 dB, and the overall scaling is simply U = 1 V. For VGADJ = 0 V,
the gain is lowered by approximately 4 dB; for VGADJ = 2 V,
the gain is raised by approximately 6 dB.
Figure 5 shows the
relationship between α(V/V) and VGADJ.
The small-signal bandwidth from the inputs X, Y, and Z to
the output W is a single-pole response. The pole is inversely
proportional to α. For α = 1 (GADJ floating), the bandwidth is
about 2 GHz; for α > 1, the bandwidth is reduced; and for α < 1,
the bandwidth is increased.
All input ports, X, Y, and Z, are differential and internally
biased to midsupply, V
POS
/2. The differential input impedance is
500 Ω up to 100 MHz, rolling off to 50  at 2 GHz. All inputs
can be driven in single-ended fashion and can be ac-coupled. In
dc-coupled operation, the inputs can be biased to a common
mode that is lower than V
POS
/2. The bias current flowing out of
the input pins to accommodate the lower common mode is
subtracted from the 50 mA total available from the internal
reference V
POS
/2 at the VREF pin. Each input pin presents an
equivalent 250  dc resistance to V
POS
/2. If all six input pins sit
1 V below V
POS
/2, a total of 6 × 1 V/250 Ω = 24 mA must flow
internally from VREF to the input pins.
Calibration
The dc offset of the ADL5391 is approximately 20 mV but
changes over temperature and has variation from part to part
(see
Figure 4). It is generally not of concern unless the ADL5391
is operated down to dc (close to the point X = 0 V or Y = 0 V),
where 0 V is expected on the output (W = 0 V). For example,
when the ADL5391 is used as a VGA and a large amount of
attenuation is needed, the maximum attenuation is determined
by the input dc offset.
Applying the proper voltage on the Z input removes the W
offset. Calibration can be accomplished by making the appropriate
cross plots and adjusting the Z input to remove the offset.
Additionally, gain scaling can be adjusted by applying a dc
voltage to the GADJ pin, as shown in
Figure 5.
BASIC CONNECTIONS
Multiplier Connections
The best ADL5391 performance is achieved when the X, Y, and
Z inputs and W output are driven differentially; however, they
can be driven single-ended. Single-ended-to-differential
transformations (or differential-to-single-ended transformations)
can be done using a balun or active components, such as the
AD8313, the AD8132 (both with operation down to dc), or the
AD8352 (for higher drive capability). If using the ADL5391
single-ended without ac coupling capacitors, the reference
voltage of 2.5 V needs to be taken into account. Voltages above
2.5 V are positive voltages and voltages below 2.5 V are negative
voltages. Care needs to be taken not to load the ADL5391 too
heavily, the maximum reference current available is 50 mA.
ADL5391
Rev. 0 | Page 11 of 16
Matching the Input/Output
The input and output impedances of the ADL5391 change over
frequency, making it difficult to match over a broad frequency
range (see
Figure 15 and Figure 16). The evaluation board is
matched for lower frequency operation, and the impedance
change at higher frequencies causes the change in gain seen in
Figure 6. If desired, the user of the ADL5391 can design a
matching network to fit their application.
Wideband Voltage-Controlled Amplifier/Amplitude
Modulator
Most of the data for the ADL5391 was collected by using it as a
fast reacting analog VGA. Either X or Y inputs can be used for
the RF input (and the other as the very fast analog control),
because either input can be used from dc to 2 GHz. There is a
linear relationship between the analog control and the output of
the multiplier in the VGA mode.
Figure 6 and Figure 7 show the
dynamic range available in VGA mode (without optimizing the
dc offsets).
The speed of the ADL5391 in VGA mode allows it to be used as
an amplitude modulator. Either or both inputs can have
modulation or CW applied. AM modulation is achieved by
feeding CW into X (or Y) and adding AM modulation to the Y
(or X) input.
Squaring and Frequency Doubling
Amplitude domain squaring of an input signal, E, is achieved
simply by connecting the X and Y inputs in parallel to produce
an output of E
2
. The input can be single-ended, differential, or
through a balun (frequency range and dynamic range can be
limited if used single ended).
When the input is a sine wave Esin(ωt), a signal squarer behaves
as a frequency doubler, because
[]
()
(
t
E
tE
2
ω=ω 2cos1
2
)sin(
2
)
(3)
Ideally, when used for squaring and frequency doubling, there is
no component of the original signals on the output. Because of
internal offsets, this is not the case. If Equation 3 were rewritten
to include theses offsets, it could separate into three output
terms (Equation 4).
[
]
[]
[]
++ω+ω
=+ω×+ω
2
)sin(2)cos(2
2
)sin()sin(
2
2
2
E
OFSTOFSTtEt
E
OFSTtEOFSTtE
(4)
where:
The dc component is OFST
2
+ E
2
/2.
The input signal bleedthrough is 2Esin(ωt)OFST.
The input squared is E
2
/2[cos(2ωt)].
The dc component of the output is related to the square of both
the offset (OFST) and the signal input amplitude (E). The offset
can be found in
Figure 4 and is approximately 20 mV. The
second harmonic output grows with the square of the input
amplitude, and the signal bleedthrough grows proportionally
with the input signal. For smaller signal amplitudes, the signal
bleedthrough can be higher than the second harmonic
component. As the input amplitude increases, the second
harmonic component grows much faster than the signal
bleedthrough and becomes the dominant signal at the output.
If the X and Y inputs are driven too hard, third harmonic
components will also increase.
For best performance creating harmonics, the ADL5391 should
be driven differentially.
Figure 17 shows the performance of the
ADL5391 when used as a harmonic generator (the evaluation
board was used with R9 and R10 removed and R2 = 56.2 Ω). If
dc operation is necessary, the ADL5391 can be driven single
ended (without the dc blocks). The flatness of the response over
a broad frequency range depends on the input/output match.
The fundamental bleed through not only depends on the
amount of power put into the device but also depends on
matching the unused differential input/output to the same
impedance as the used input/output.
Figure 18 shows the
performance of the ADL5391 when driven single ended
(without ac coupling capacitors), and
Figure 19 shows the
schematic of the setup. A resistive input/output match were
used to match the input from dc to 1 GHz and the output from
dc to 2 GHz. Reactive matching can be used for more narrow
frequency ranges. When matching the input/output of the
ADL5391, care needs to be taken not to load the ADL5391 too
heavily; the maximum reference current available is 50 mA.
15
–65
–60
–55
–50
–45
–40
–35
–30
–25
–20
10 100 200 300 400 500 600 700 800 900 1000
GAIN (dBm)
FREQUENCY (MHz)
SECOND HARMONIC GAIN
BLEEDTHRU GAIN
THIRD HARMONIC GAIN
06059-026
Figure 17. ADL5391 Used as a Harmonic Generator

ADL5391ACPZ-R7

Mfr. #:
Manufacturer:
Analog Devices Inc.
Description:
Multipliers / Dividers RF IF Multiplier
Lifecycle:
New from this manufacturer.
Delivery:
DHL FedEx Ups TNT EMS
Payment:
T/T Paypal Visa MoneyGram Western Union

Products related to this Datasheet