Dpd/Clgc Configuration - Analog Devices ADRV9001 User Manual

Reference Manual
DIGITAL PREDISTORTION (DPD)
achieved, the gain in the compression region is adjusted (increased) to match the gain of the lower linear region so that the overall gain is
independent of the input or output average power level. This has the advantage of having CLGC focusing on compensating only the PA gain
variation mainly due to temperature change. Furthermore, by setting a proper gain target, CLGC can also help to monitor and limit the transmit
power level to keep the amplifier output power from rising beyond the linearization capable power limit of the PA. Therefore, it is crucial always
to enable CLGC while DPD is active.
As the first step of CLGC, set up a target transmit gain. The target transmit gain can be measured through the ADRV9001 using the "CLGC
Loop Open" method. The detailed steps of measuring the target gain are discussed later. After the measurement, the ADRV9001 provides an
unfiltered and filtered transmit gain value. Based on these, further adjust the value and set a proper gain target, then close the loop and start
the ADRV9001 CLGC algorithm to continuously track the gain variation based on the determined gain target.
When DPD is active, the PA gain is defined as the gain in the linear region of the AM-AM curve, as shown in
reference gain, data samples are usually selected in the upper linear region and below the compression region, as indicated in red in
188. Note: This same gain plus relative phase is used by the DPD to scale the y(t) loopback data to match the predistorted transmit x(t)
data. The PA gain derived from data between these bounds represents the actual gain in the linear region of the amplifier while excluding
the distorted gain at the upper end due to compression. However, after DPD has converged, the gain in the compression region increases to
match the gain in the lower region. When DPD is off, it must not correct the compression at the top region. Hence, it is necessary to include all
samples for integration to estimate the total power, including the compressed region, to define the transmit gain. Define the region to calculate
the gain through API configurations.
Similar to the DPD algorithm, the CLGC algorithm requires the time alignment between the transmit x(t) and loopback y(t) for data capture.
Measure the delay and provide it to the ADRV9001, which is essential for wideband profiles. When the DPD is enabled, the same delay
measurement serves both the DPD and CLGC algorithm.
Note: The ADRV9001 CLGC algorithm has a limit to track gain variations not exceeding ±3 dB (this must accommodate most types of PA), and
for each CLGC iteration, the maximum gain adjustment is limited to ±0.5 dB to prevent the DPD algorithm becoming unstable.

DPD/CLGC CONFIGURATION

To use the integrated DPD/CLGC properly and ensure optimal performance, properly configure the DPD/CLGC parameters. Configure through
the ADRV9001 TES or SDK. The configuration consists of two sets of DPD/CLGC parameters:"preinitial calibration" parameters, as they should
be configured before performing initial calibration when the device is at the "STANDBY" state, and"post initial calibration" parameters, as they
must be configured after performing initial calibration when the device is at the "CALIBRATED" state. The following subsections explain these
DPD/CLGC parameters in detail.
DPD/CLGC Preinitial Calibration Parameters Configuration
To properly set the preinitial calibration parameters of the DPD/CLGC, there should be a general understanding of the DPD model used in the
x n = ∑ ψ d n − l
device. The following equations describe the DPD model.
T − 1
t = 0 t
ψ d n − l = ∑ b
7
t t i = 0
where:
)|) is the function implemented by the LUT for tap "t" .
T is the total number of taps in the DPD model.
ψ (|d(n – l
t t
l and k are part of the hardware model representing the amplitude and data delay, respectively.
t t
Optionally include/exclude each individual power term in ψ
a are coefficients estimated by the coefficients calculation engine and used to generate the LUTs by the DPD actuator. For b
or 1 for including to better model the PA.
t, l t , i
a
t, l t , i
, the subscript t represents the index for the tap, l
ADRV9001 only supports 0 to 7th order power term in the function ψ
analog.com
d n − k
t t
a d n − l
i
t, l t , i t, l t , i t
(|d(n – l
t
represents the amplitude delay, and i represents the order of the power term. The
t
)|) by controlling the corresponding b
t
(|d(n – l )|).
t t
ADRV9001
Figure 188. To estimate this
Figure
t, l t , i
, setting it to either 0 for excluding
t, l t , i
Rev. 0 | 213 of 351
and