REV. B
AD8309
–7–
THEORY OF OPERATION
The AD8309 is an advanced IF signal processing IC, intended
for use in high performance receivers, combining two key func-
tions. First, it provides a large voltage gain combined with pro-
gressive compression, through which an IF signal of high dynamic
range is converted into a square-wave (that is, hard limited)
output, from which frequency and phase information modulated
on this input can be recovered by subsequent signal processing.
For this purpose, the noise level referred to the input must be
very low, since it determines the detection threshold for the receiver.
Further, it is often important that the group delay in this ampli-
fier be essentially independent of the signal level, to minimize
the risk of amplitude-to-phase conversion. Finally, it is also desir-
able that the amplitude of the limited output be well defined and
temperature stable. In the AD8309, this amplitude can be con-
trolled by the user, or even completely shut off, providing greater
flexibility.
The second function is to provide a demodulated (baseband)
output proportional to the decibel value of the signal input,
which may be used to measure the signal strength. This output,
which typically runs from a value close to the ground level to a
few volts above ground, is called the Received Signal Strength
Indication, or RSSI. The provision of this function requires the
use of a logarithmic amplifier (log amp). For this output to be
suitable for measuring signal strength, it is important that its
scaling attributes are well controlled.
These are the logarithmic slope, specified in mV/dB, and the
intercept, often specified as an equivalent power level at the
amplifier input, although a log amp is inherently a voltage-
responding device. (See further discussion, below). Also
important is the law conformance, that is, how well the RSSI
approximates an ideal function. Many low quality log amps
provide only an approximate solution, resulting in large errors in
law conformance and scaling. All Analog Devices log amps are
designed with close attention to matters affecting accuracy of
the overall function.
In the AD8309, these two basic signal-processing functions are
combined to provide the necessary voltage gain with progressive
compression and hard limiting, and the determination of the
logarithmic magnitude of the input (RSSI). This combination is
called a log limiting amplifier. A good grasp of how this product
works will avoid many pitfalls in their application.
Log-Amp Fundamentals
The essential purpose of a logarithmic amplifier is to reduce a
signal of wide dynamic range to its decibel equivalent. It is thus
primarily a measurement device. The logarithmic representation
leads to situations that may be confusing or even paradoxical.
For example, a voltage offset added to the RSSI output of a log
amp is equivalent to a gain increase ahead of its input.
When all the variables expressed as voltages, then, regardless of
the particular structure, the output can be expressed as
V
OUT
= V
Y
log (V
IN
/V
X
) (1)
where V
Y
is the “slope voltage.” V
IN
is the input voltage, and V
X
is the “intercept voltage.” The logarithm is usually to base-10,
which is appropriate to a decibel-calibrated device, in which
case V
Y
is also the “volts-per-decade.” It will be apparent from
(1) that a log amp requires two references, here V
X
and V
Y
, that
determine the scaling of the circuit. The absolute accuracy of a
log amp cannot be any better than the accuracy of its scaling
references. Note that (1) is mathematically incomplete in rep-
resenting the behavior of a demodulating log amp such as the
AD8309, where V
IN
has an alternating sign. However, the basic
principles are unaffected.
Figure 19 shows the input/output relationship of an ideal log
amp, conforming to Equation (1). The horizontal scale is loga-
rithmic, and spans a very wide dynamic range, shown here as
over 120 dB, that is, six decades of voltage or twelve decades of
input-referred power. The output passes through zero (the
“log-intercept”) at the unique value V
IN
= V
X
and becomes
negative for inputs below the intercept. In the ideal case, the
straight line describing V
OUT
for all values of V
IN
would con-
tinue indefinitely in both directions. The dotted line shows that
the effect of adding an offset voltage V
SHIFT
to the output is to
lower the effective intercept voltage V
X
.
V
OUT
5V
Y
4V
Y
3V
Y
2V
Y
V
Y
–2V
Y
V
OUT
= 0
LOG V
IN
V
SHIFT
LOWER INTERCEPT
V
IN
= 10
–2
V
X
–40dBc
V
IN
= 10
2
V
X
+40dBc
V
IN
= 10
4
V
X
+80dBc
V
IN
= V
X
0dBc
Figure 19. Ideal Log Amp Function
Exactly the same modification could be achieved raising the gain
(or signal level) ahead of the log amp by the factor V
SHIFT
/V
Y
.
For example, if V
Y
is 400 mV/decade (that is, 20 mV/dB, as for
the AD8309), an offset of 120 mV added to the output will
appear to lower the intercept by two tenths of a decade, or 6 dB.
Adding an offset to the output is thus indistinguishable from
applying an input level that is 6 dB higher.
The log amp function described by (1) differs from that of a
linear amplifier in that the incremental gain DV
OUT
/DV
IN
is a
very strong function of the instantaneous value of V
IN
, as is
apparent by calculating the derivative. For the case where the
logarithmic base is e, it is easy to show that
V
V
V
V
OUT
IN
Y
IN
=
(2)
That is, the incremental gain of a log amp is inversely propor-
tional to the instantaneous value of the input voltage. This re-
mains true for any logarithmic base. A “perfect” log amp would
be required to have infinite gain under classical “small-signal”
(zero-amplitude) conditions. This demonstrates that, whatever
means might be used to implement a log amp, accurate HF
response under small signal conditions (that is, at the lower end
of the full dynamic range) demands the provision of a very high
gain-bandwidth product. A wideband log amp must therefore use
many cascaded gain cells each of low gain but high bandwidth.
For the AD8309, the gain-bandwidth (–10 dB) product is
52,500 GHz.
REV. B
AD8309
–8–
As a consequence of this high gain, even very small amounts of
thermal noise at the input of a log amp will cause a finite output
for zero input, resulting in the response line curving away from
the ideal (Figure 19) at small inputs, toward a fixed baseline.
This can either be above or below the intercept, depending on
the design. Note that the value specified for this intercept is
invariably an extrapolated one: the RSSI output voltage will never
attain a value of exactly zero in a single supply implementation.
Voltage (dBV) and Power (dBm) Response
While Equation 1 is fundamentally correct, a simpler formula is
appropriate for specifying the RSSI calibration attributes of a
log amp like the AD8309, which demodulates an RF input. The
usual measure is input power:
V
OUT
= V
SLOPE
(P
IN
P
0
) (3)
V
OUT
is the demodulated and filtered RSSI output, V
SLOPE
is the
logarithmic slope, expressed in volts/dB, P
IN
is the input power,
expressed in decibels relative to some reference power level and
P
0
is the logarithmic intercept, expressed in decibels relative to
the same reference level.
The most widely used convention in RF systems is to specify
power in decibels above 1 mW in 50 , written dBm. (However,
that the quantity [P
IN
– P
0
] is simply dB). The logarithmic
function disappears from this formula because the conversion
has already been implicitly performed in stating the input in
decibels.
Specification of log amp input level in terms of power is strictly
a concession to popular convention: they do not respond to
power (tacitly “power absorbed at the input”), but to the input
voltage. In this connection, note that the input impedance of the
AD8309 is much higher that 50 , allowing the use of an im-
pedance transformer at the input to raise the sensitivity, by up
to 13 dB.
The use of dBV, defined as decibels with respect to a 1 V rms sine
amplitude, is more precise, although this is still not unambiguous
complete as a general metric, because waveform is also involved
in the response of a log amp, which, for a complex input (such
as a CDMA signal) will not follow the rms value exactly. Since
most users specify RF signals in terms of power—more specifi-
cally, in dBm/50 —we use both dBV and dBm in specifying
the performance of the AD8309, showing equivalent dBm levels
for the special case of a 50 environment.
Progressive Compression
High speed, high dynamic range log amps use a cascade of
nonlinear amplifier cells (Figure 20) to generate the logarithmic
function from a series of contiguous segments, a type of piece-
wise-linear technique. This basic topology offers enormous gain-
bandwidth products. For example, the AD8309 employs in its
main signal path six cells each having a small-signal gain of
12.04 dB (×4) and a –3 dB bandwidth of 850 MHz, followed by
a final limiter stage whose gain is typically 18 dB. The overall
gain is thus 100,000 (100 dB) and the bandwidth to –10 dB
point at the limiter output is 525 MHz. This very high gain-
bandwidth product (52,500 GHz) is an essential prerequisite to
accurate operation under small signal conditions and at high
frequencies: Equation (2) reminds us that the incremental gain
decreases rapidly as V
IN
increases. The AD8309 exhibits a loga-
rithmic response over most of the range from the noise floor of
–91 dBV, or 28 µV rms, (or –78 dBm/50 ) to a breakdown-
limited peak input of 4 V (requiring a balanced drive at the
differential inputs INHI and INLO).
A
V
X
STAGE 1 STAGE 2 STAGE N –1 STAGE N
V
W
A A A
Figure 20. Cascade of Nonlinear Gain Cells
Theory of Logarithmic Amplifiers
To develop the theory, we will first consider a somewhat differ-
ent scheme to that employed in the AD8309, but which is sim-
pler to explain, and mathematically more straightforward to
analyze. This approach is based on a nonlinear amplifier unit,
which we may call an A/1 cell, having the transfer characteristic
shown in Figure 21. We here use lowercase variables to define
the local inputs and outputs of these cells, reserving uppercase
for external signals.
The small signal gain V
OUT
/V
IN
is A, and is maintained for
inputs up to the knee voltage E
K
, above which the incremental
gain drops to unity. The function is symmetrical: the same drop
in gain occurs for instantaneous values of V
IN
less than –E
K
.
The large signal gain has a value of A for inputs in the range
–E
K
<
V
IN
< +E
K
, but falls asymptotically toward unity for very
large inputs.
In logarithmic amplifiers based on this simple function, both the
slope voltage and the intercept voltage must be traceable to the
one reference voltage, E
K
. Therefore, in this fundamental analy-
sis, the calibration accuracy of the log amp is dependent solely on
this voltage. In practice, it is possible to separate the basic refer-
ences used to determine V
Y
and V
X
. In the AD8309, V
Y
is trace-
able to an on-chip band-gap reference, while V
X
is derived from
the thermal voltage kT/q and later temperature-corrected by a
precise means.
Let the input of an N-cell cascade be V
IN
, and the final output
V
OUT
. For small signals, the overall gain is simply A
N
. A six-
stage system in which A = 5 (14 dB) has an overall gain of
15,625 (84 dB). The importance of a very high small-signal ac
gain in implementing the logarithmic function has already been
noted. However, this is a parameter of only incidental interest in
the design of log amps; greater emphasis needs to be placed on
the nonlinear behavior.
SLOPE = A
SLOPE = 1
OUTPUT
AE
K
0
E
K
INPUT
A/1
Figure 21. The A/1 Amplifier Function
Thus, rather than considering gain, we will analyze the overall
nonlinear behavior of the cascade in response to a simple dc
input, corresponding to the V
IN
of Equation (1). For very small
inputs, the output from the first cell is V
1
= AV
IN
; from the
second, V
2
= A
2
V
IN
, and so on, up to V
N
= A
N
V
IN
. At a certain
value of V
IN
, the input to the Nth cell, V
N–1
, is exactly equal to
the knee voltage E
K
. Thus, V
OUT
= AE
K
and since there are N–1
cells of gain A ahead of this node, we can calculate that V
IN
=
E
K
/A
N–1
. This unique point corresponds to the lin-log transition,
REV. B
AD8309
–9–
labeled on Figure 22. Below this input, the cascade of gain
cells is acting as a simple linear amplifier, while for higher values
of V
IN
, it enters into a series of segments which lie on a logarith-
mic approximation.
Continuing this analysis, we find that the next transition occurs
when the input to the (N–1)th stage just reaches E
K
, that is,
when V
IN
= E
K
/A
N–2
. The output of this stage is then exactly
AE
K
. It is easily demonstrated (from the function shown in
Figure 21) that the output of the final stage is (2A–1)E
K
(la-
beled on Figure 22). Thus, the output has changed by an
amount (A–1)E
K
for a change in V
IN
from E
K
/A
N–1
to E
K
/A
N–2
,
that is, a ratio change of A.
V
OUT
LOG V
IN
0
RATIO
OF A
E
K
/A
N–1
E
K
/A
N–2
E
K
/A
N–3
E
K
/A
N–4
(A-1) E
K
(4A-3) E
K
(3A-2) E
K
(2A-1) E
K
AE
K
Figure 22. The First Three Transitions
At the next critical point, labeled , the input is A times larger
and V
OUT
has increased to (3A–2)E
K
, that is, by another linear
increment of (A–1)E
K
. Further analysis shows that, right up to
the point where the input to the first cell reaches the knee volt-
age, V
OUT
changes by (A–1)E
K
for a ratio change of A in V
IN
.
Expressed as a certain fraction of a decade, this is simply log
10
(A).
For example, when A = 5 a transition in the piecewise linear
output function occurs at regular intervals of 0.7 decade (log10(A),
or 14 dB divided by 20 dB). This insight allows us to immedi-
ately state the “Volts per Decade” scaling parameter, which is
also the “Scaling Voltage” V
Y
when using base-10 logarithms:
V
Linear Change inV
Decades Change inV
AE
A
Y
OUT
IN
K
==
( –)
log ( )
1
10
(4)
Note that only two design parameters are involved in determin-
ing V
Y
, namely, the cell gain A and the knee voltage E
K
, while
N, the number of stages, is unimportant in setting the slope of
the overall function. For A = 5 and E
K
= 100 mV, the slope
would be a rather awkward 572.3 mV per decade (28.6 mV/dB).
A well designed practical log amp will provide more rational
scaling parameters.
The intercept voltage can be determined by solving Equation
(4) for any two pairs of transition points on the output function
(see Figure 22). The result is:
V
E
A
X
K
NA
=
+(/[–])11
(5)
For the example under consideration, using N = 6, V
X
evaluates
to 4.28 µV, which thus far in this analysis is still a simple dc
voltage.
A/0
SLOPE = 0
SLOPE = A
E
K
AE
K
0
OUTPUT
INPUT
Figure 23. A/0 Amplifier Functions (Ideal and tanh)
Care is needed in the interpretation of this parameter. It was
earlier defined as the input voltage at which the output passes
through zero (see Figure 19). Clearly, in the absence of noise
and offsets, the output of the amplifier chain shown in Figure 20
can only be zero when V
IN
= 0. This anomaly is due to the finite
gain of the cascaded amplifier, which results in a failure to main-
tain the logarithmic approximation below the “lin-log transition”
(Point in Figure 22). Closer analysis shows that the voltage
given by Equation (5) represents the extrapolated, rather than
actual, intercept.
Demodulating Log Amps
Log amps based on a cascade of A/1 cells are useful in baseband
(pulse) applications, because they do not demodulate their input
signal. Demodulating (detecting) log-limiting amplifiers such as
the AD8309 use a different type of amplifier stage, which we
will call an A/0 cell. Its function differs from that of the A/1 cell
in that the gain above the knee voltage E
K
falls to zero, as shown
by the solid line in Figure 23. This is also known as the limiter
function, and a chain of N such cells is often used alone to
generate a hard limited output, in recovering the signal in FM
and PM modes.
The AD640, AD606, AD608, AD8307, AD8309, AD8313 and
other Analog Devices communications products incorporating a
logarithmic IF amplifier all use this technique. It will be appar-
ent that the output of the last stage cannot now provide a loga-
rithmic output, since this remains unchanged for all inputs
above the limiting threshold, which occurs at V
IN
= E
K
/A
N–1
.
Instead, the logarithmic output is generated by summing the
outputs of all the stages. The full analysis for this type of log amp
is only slightly more complicated than that of the previous case.
It can be shown that, for practical purpose, the intercept voltage
V
X
is identical to that given in Equation (5), while the slope
voltage is:
V
AE
A
Y
K
=
log ( )
10
(6)
An A/0 cell can be very simple. In the AD8309 it is based on a
bipolar-transistor differential pair, having resistive loads R
L
and
an emitter current source I
E
. This amplifier limiter cell exhibits
an equivalent knee-voltage of E
K
= 2kT/q and a small-signal
gain of A = I
E
R
L
/E
K
. The large signal transfer function is the
hyperbolic tangent (see dotted line in Figure 23). This function
is very precise, and the deviation from an ideal A/0 form is not
detrimental. In fact, the “rounded shoulders” of the tanh func-
tion beneficially result in a lower ripple in the logarithmic con-
formance than that which would be obtained using an ideal A/0
function. A practical amplifier chain built of these cells is differ-
ential in structure from input to final output, and has a low

AD8309ARUZ

Mfr. #:
Manufacturer:
Analog Devices Inc.
Description:
Logarithmic Amplifiers 5-500 MHz 100 dB w/ Limiter Output
Lifecycle:
New from this manufacturer.
Delivery:
DHL FedEx Ups TNT EMS
Payment:
T/T Paypal Visa MoneyGram Western Union