AD8307
2
3
4
A V
IN
AV
A V
A V
IN
IN
IN
motion of VX resulting from the temperature variation of EK. Do
this by adding an offset with the required temperature behavior.
V
V
LIM
IN
A/0
A/0
A/0
A/0
The precise temperature shaping of the intercept positioning offset
results in a log amp having stable scaling parameters, making it a
true measurement device, for example, as a calibrated received
signal strength indicator (RSSI). In this application, one is more
interested in the value of the output for an input waveform that
is invariably sinusoidal. Although the input level can
alternatively be stated as an equivalent power, in dBm, be sure
to work carefully. It is essential to know the load impedance in
which this power is presumed to be measured.
gm
gm
gm
gm
gm
I
OUT
Figure 26. Log Amp Using A/0 Stages and Auxiliary Summing Cells
The chief advantage of this approach is that the slope voltage
can now be decoupled from the knee voltage EK = 2 kT/q, which
is inherently PTAT. By contrast, the simple summation of the
cell outputs would result in a very high temperature coefficient
of the slope voltage given in Equation ꢀ. To do this, the detector
stages are biased with currents (not shown) which are rendered
stable with temperature. These are derived either from the
supply voltage (as in the ADꢀ0ꢀ and ADꢀ08) or from an
internal band gap reference (as in the ADꢀ40 and AD8307).
This topology affords complete control over the magnitude and
temperature behavior of the logarithmic slope, decoupling it
completely from EK.
In RF practice, it is generally safe to assume a reference impedance
of 50 Ω in which 0 dBm (1 mW) corresponds to a sinusoidal
amplitude of 31ꢀ.2 mV (223.ꢀ mV rms). The intercept can likewise
be specified in dBm. For the AD8307, it is positioned at −84 dBm,
corresponding to a sine amplitude of 20 μV. It is important to bear
in mind that log amps do not respond to power, but to the voltage
applied to their input.
The AD8307 presents a nominal input impedance much higher
than 50 Ω (typically 1.1 kΩ low frequencies). A simple input
matching network can considerably improve the sensitivity of
this type of log amp. This increases the voltage applied to the
input and thus alters the intercept. For a 50 Ω match, the
voltage gain is 4.8 and the entire dynamic range moves down by
13.ꢀ dB (see Figure 35). Note that the effective intercept is a
function of waveform. For example, a square wave input reads
ꢀ dB higher than a sine wave of the same amplitude and a
Gaussian noise input 0.5 dB higher than a sine wave of the same
rms value.
A further step is needed to achieve the demodulation response,
required when the log amp is to convert an alternating input
into a quasi-dc baseband output. This is achieved by altering the
gm cells used for summation purposes to also implement the
rectification function. Early discrete log amps based on the
progressive compression technique used half-wave rectifiers.
This made post-detection filtering difficult. The ADꢀ40 was the
first commercial monolithic log amp to use a full wave rectifier,
a practice followed in all subsequent Analog Devices types.
These detectors can be modeled as being essentially linear gm
cells, but producing an output current independent of the sign
of the voltage applied to the input of each cell. That is, they
implement the absolute value function. Since the output from
the later A/0 stages closely approximates an amplitude
symmetric square wave for even moderate input levels (most
stages of the amplifier chain operate in a limiting mode), the
current output from each detector is almost constant over each
period of the input. Somewhat earlier detector stages produce a
waveform having only very brief dropouts, while the detectors
nearest the input produce a low level, almost sinusoidal
waveform at twice the input frequency. These aspects of the
detector system result in a signal that is easily filtered, resulting
in low residual ripple on the output.
OFFSET CONTROL
In a monolithic log amp, direct coupling between the stages is
used for several reasons. First, this avoids the use of coupling
capacitors, which typically have a chip area equal to that of a
basic gain cell, thus considerably increasing die size. Second, the
capacitor values predetermine the lowest frequency at which the
log amp can operate; for moderate values, this can be as high as
30 MHz, limiting the application range. Third, the parasitic
(backplate) capacitance lowers the bandwidth of the cell, further
limiting the applications.
However, the very high dc gain of a direct-coupled amplifier
raises a practical issue. An offset voltage in the early stages of
the chain is indistinguishable from a real signal. For example, if
it were as high as 400 μV, it would be 18 dB larger than the
smallest ac signal (50 μV), potentially reducing the dynamic
range by this amount. This problem is averted by using a global
feedback path from the last stage to the first, which corrects this
offset in a similar fashion to the dc negative feedback applied
around an op amp. The high frequency components of the
signal must be removed to prevent a reduction of the HF gain in
the forward path.
INTERCEPT CALIBRATION
All monolithic log amps from Analog Devices include accurate
means to position the intercept voltage VX (or equivalent power for
a demodulating log amp). Using the scheme shown in Figure 2ꢀ,
the basic value of the intercept level departs considerably from that
predicted by the simpler analyses given earlier. However, the
intrinsic intercept voltage is still proportional to EK, which is PTAT
(Equation 5). Recalling that the addition of an offset to the output
produces an effect that is indistinguishable from a change in the
position of the intercept, it is possible to cancel the left-right
In the AD8307, this is achieved by an on-chip filter, providing
sufficient suppression of HF feedback to allow operation above
Rev. C | Page ±2 of 24
ADI [ ADI ]
