AD8362
The network can readily be scaled to other frequencies by
varying the product LC, while keeping the ratio L/C constant to
preserve a 50 Ω input impedance. Table 4 provides some spot
values; these take into account the reactive ZIN of the AD8362.
CHOOSING THE RIGHT VALUE FOR CHPF AND
CLPF
The AD8362’s 3.5 GHz variable gain amplifier includes an offset
cancellation loop, which introduces a high-pass filter effect in
its transfer function. The corner frequency, fHP, of this filter
must be below that of the lowest input signal in the desired
measurement bandwidth frequency to properly measure the
amplitude of the input signal. The required value of the external
capacitor is given by
Table 4. Suggested Components for Narrow-Band 50 Ω
Match
Frequency (MHz)
L (nH)
21850
10925
4370
2185
1093
437
220
100
40
C1 (pF)
2230
1115
446
223
112
45
22
10
3.9
C2 (pF)
2765
1383
553
276
138
55
27
12
4.7
1
2
5
10
20
50
100
200
500
CHPF = 200μF fHP
fHP in Hz
(12)
Thus, for operation at frequencies down to 100 kHz, CHPF
should be 2 nF.
In the standard connections for the measurement mode, the
VSET pin is tied to VOUT. For small changes in input amplitude
(a few decibels), the time-domain response of this loop is
essentially linear with a 3 dB low-pass corner frequency of
nominally fLP = 1/(CLPF × 1.1 kΩ). Internal time delays around
this local loop set the minimum recommended value of this
capacitor to about 300 pF, making fLP = 3 MHz.
This coupling method can be used down to much lower
frequencies than shown in Table 4 simply by multiplying the
1 MHz component values proportionally. The effects of the
reactive components of the AD8362’s inputs above 500 MHz
may require some fine tuning of the suggested values. In the
gigahertz region, the input coupling is usually more effectively
implemented using a balun.
For operation at lower signal frequencies, or whenever the
averaging time needs to be longer, use
UNCERTAINTIES IN RIN AND POWER CALIBRATION
CLPF = 900μF fLP
fLP in Hz
(13)
In all the cases where a 50 Ω to 200 Ω transformation is
implemented, the voltage gain is only nominally ×2 (6 dB). This
ideal is impaired by the fact that the input resistances of the
AD8362 are not precise; variations of 20% can be expected
from lot to lot. Therefore, it is necessary to use a calibration step
whenever an accurate value for the power intercept, PZ, must be
established.
When the input signal exhibits large crest factors, such as a
WCDMA signal, CLPF must be much larger than might at
first seem necessary. This is due to the presence of significant
low frequency components in the complex, pseudo-random
modulation, which generates fluctuations in the output of
the AD8362.
When driven differentially, a significant improvement in
intercept accuracy can be achieved by shunting the 200 Ω
resistance from INHI to INLO with a 66.5 Ω resistor to set the
differential input resistance to 50 Ω. Assuming a tolerance of
20% for the basic RIN and 1% for the chip resistor, the net
input resistance could exhibit an error of 2.5%. The resulting
error in PZ (and thus in the absolute power measurement) may
vary from −0.26 dB to +0.21 dB.
USE OF NONSTANDARD TARGET VOLTAGES
An external connection between VREF and VTGT sets up
the internal target voltage, that is, the rms voltage that must
be provided by the VGA to balance the AGC feedback loop. In
the default scheme, the VREF of 1.25 V positions this target to
0.06 × 1.25 V = 75 mV. In principle, however, VTGT may be
driven by any voltage in the range −4 V to +4 V (the sign is
ignored) to alter this target, either in a fixed or dynamic way.
These precautions regarding input impedance do not apply
when the input is presented in voltage form, as is often the case
at low frequencies, or when the source impedance is low
compared to 200 Ω. For example, when using a feedback
amplifier as an impedance buffer ahead of the input, as in the
example in Figure 61, the loss at the interface at moderate
frequencies is negligible.
For example, if this pin is supplied from VREF via a simple
resistive attenuator of 1 kΩ:1 kΩ, the output required from the
VGA is halved (to 37.5 mV rms), which moves the nominal
intercept to −73 dBV. Under these conditions, the effective
headroom in the signal path that drives the squaring cell is
doubled. In principle, this doubles the peak crest factor that may
be handled by the system.
If VTGT is reduced too far, the accuracy and stability of the
intercept are compromised. The currents generated by the
transconductance mode squaring cells become smaller by the
square of the ratio. Thus, a factor of 5 reduction in VTGT
Rev. B | Page 24 of 36