AD8315
Control Loop Dynamics
where VSLP is the volts-per-decade slope from Equation 1, having
a value of 480 mV/decade, and T is an effective time constant for
the integration, being equal to 4.15 kW ¥ CFLT/1.35; the resistor
value comes from the setpoint interface scaling Equation 4 and
the factor 1.35 arises because of the voltage gain of the buffer.
So the integration time constant can be written as:
In order to understand how the AD8315 behaves in a complete
control loop, an expression for the current in the integration
capacitor as a function of the input VIN and the setpoint voltage
VSET must be developed. Refer to Figure 3.
V
SET
I
= V
/4.15k
SET
SET
SETPOINT
INTERFACE
(8)
T = 3.07 CFLT in ms, whenC is expressed in nF
3
V
SET
To simplify our understanding of the control loop dynamics,
begin by assuming that the power amplifier gain function actu-
ally is linear-in-dB. Also use voltages to express the signals at
the power amplifier input and output, for the moment. Let the
RF output voltage be VPA and its input be VCW. Further, to
characterize the gain control function, this form is used:
FLTR
VAPC
7
RFIN
1
LOGARITHMIC
RF DETECTION
SUBSYSTEM
1.35
4
V
I
I
IN
DET
ERR
C
FLT
I
= I
LOG (V /V )
DET
SLP 10 IN Z
Figure 3. Behavioral Model of the AD8315
APC /VGBC
)
VPA = GOVCW 10(V
(9)
First, the summed detector currents are written as a function of
the input:
where GO is the gain of the power amplifier when VAPC = 0 and
V
GBC is the gain-scaling. While few amplifiers will conform so
(3)
IDET = ISLP log10 (VIN /VZ )
conveniently to this law, it provides a clearer starting point for
understanding the more complex situation that arises when the
gain control law is less ideal.
where IDET is the partially filtered demodulated signal, whose
exact average value will be extracted through the subsequent
This idealized control loop is shown in Figure 4. With some
manipulation, it is found that the characteristic equation of this
system is:
integration step; ISLP is the current-mode slope and has a value
of 115 mA per decade (that is, 5.75 mA/dB); VIN is the input in
volts-rms; and VZ is the effective intercept voltage, which, as
previously noted, is dependent on waveform but is 316 mV rms
(–70 dBV) for a sine wave input. Now the current generated by
the setpoint interface is simply:
(VSET
V
GBC )/VSLP –VGBC log10 kG VCW /VZ
(
)
O
V
APC(s) =
(10)
1+ sTO
(4)
ISET = VSET /4.15 kW
where k is the coupling factor from the output of the power
amplifier to the input of the AD8315 (e.g., ¥ 0.1 for a “20 dB
coupler”), and TO is a modified time constant (VGBC /VSLP)T.
The difference between this current and IDET is applied to the
loop filter capacitor CFLT. It follows that the voltage appearing on
this capacitor, VFLT, is the time integral of the difference current:
This is quite easy to interpret. First, it shows that a system of
this sort will exhibit a simple single-pole response, for any power
level, with the customary exponential time domain form for
either increasing or decreasing step polarities in the demand
level VSET or the carrier input VCW. Second, it reveals that the
final value of the control voltage VAPC will be determined by
several fixed factors:
(5)
VFLT (s) = (ISET – IDET )/sCFLT
V
SET /4.15 kW – ISLP log10 (VIN /VZ )
=
(6)
sCFLT
The control output VAPC is slightly greater than this, since the
gain of the output buffer is ¥1.35. Also, an offset voltage is
deliberately introduced in this stage; this is inconsequential
since the integration function implicitly allows for an arbitrary
constant to be added to the form of Equation 6. The polarity is
such that VAPC will rise to its maximum value for any value of
VSET greater than the equivalent value of VIN. In practice, the
VAPC t = • = VSETVGBC /V
– log10 kG V /VZ
(11)
(
)
(
)
(
)
SLP
O
CW
Example
Assume that the gain magnitude of the power amplifier runs
from a minimum value of ¥0.316 (–10 dB) at VAPC = 0 to ¥100
(40 dB) at VAPC = 2.5 V. Applying Equation 9, GO = 0.316 and
V
APC output will rail to the positive supply under this condition
V
GBC = 1 V. Using a coupling factor of k = 0.0316 (that is, a
30 dB directional coupler) and recalling that the nominal value of
SLP is 480 mV and VZ = 316 V for the AD8315, first calculate
unless the control loop through the power amplifier is present.
In other words, the AD8315 seeks to drive the RF power to its
maximum value whenever it falls below the setpoint. The use
of exact integration results in a final error that is theoretically
zero, and the logarithmic detection law would ideally result in a
constant response time following a step change of either the
setpoint or the power level, if the power-amplifier control
function were likewise linear-in-dB. This latter condition is
rarely true, however, and it follows that in practice, the loop
response time will depend on the power level, and this effect can
strongly influence the design of the control loop.
V
the range of values needed for VSET to control an output range
of 33 dBm to –17 dBm. This can be found by noting that, in
the steady state, the numerator of Equation 7 must be zero,
that is:
(12)
VSET = VSLP log10 (kVPA /VZ )
when VIN is expanded to kVPA, the fractional voltage sample of
the power amplifier output. Now, for +33 dBm, VPA = 10 V rms,
this evaluates to:
Equation 6 can be restated as:
(13)
VSET (max) = 0.48 log10 (316 mV /316 mV ) = 1.44V
VSET -VSLP log10 (VIN /VZ )
For a delivered power of –17 dBm, VPA = 31.6 mV rms:
V
APC(s) =
(7)
sT
(14)
VSET (min) = 0.48 log10 (1mV/316 mV ) = 0.24V
–10–
REV. B