Text preview for : 5991-1463EN Simulating Envelope Tracking with Advanced Design System Software - Application Note c20 part of Agilent 5991-1463EN Simulating Envelope Tracking with Advanced Design System Software - Application Note c20 Agilent 5991-1463EN Simulating Envelope Tracking with Advanced Design System Software - Application Note c20140909 [27].pdf
Back to : 5991-1463EN Simulating En | Home
Simulating
Envelope Tracking with
Keysight Advanced Design System
Application Note
1.0 Introduction
Modern modulated signals often have high peak-to-average One way to bolster low efficiency in the back-off region is
power ratios (PAPR). Use of such high PAPR signals results to use Envelope Tracking (ET), a technique that allows the
in power amplifiers (PAs) having to operate at a large amplifier's drain bias to track the magnitude of the input
back-off when stringent linearity requirements must be signal envelope. When the input signal envelope is low, the
satisfied. Operating a PA at relatively high output power drain bias can be reduced so the amplifier operates closer
back-off ensures the signal is not greatly distorted when its to its optimal efficiency point.
envelope excursion is near its peak. However, the greater
the amount of back-off, the lower the PA's efficiency. If we could dynamically adjust the drain bias to lower
values when the input signal is lower, we can operate the
A power amplifier with a fixed drain bias may have high power amplifier at a higher level of efficiency (Param1 is
power-added efficiency (PAE), but only at a high output the drain bias), as shown if Figure 2.
power (the red trace in Figure 1.) The power amplifier with
the PAE-versus-output power characteristic shown in this More information about Envelope Tracking is available at:
figure would have to be operated in back off (thus with low www.open-et.org
PAE) to avoid distorting the signal.
Table of Contents
1.0 Introduction .................................................................................................................................................. i
2.0 Characterizing the Power Amplifier ........................................................................................................ 4
3.0 Examining the Modulated RF Signal ...................................................................................................10
4.0 Simulating the Power Amplifier with a Fixed Drain Bias ................................................................13
5.0 Implementing Envelope Tracking Using Behavioral Model Components ....................................14
6.0 Using Ptolemy Sources and Sinks to Simulate EVM and Drain Efficiency ..................................16
7.0 Using the 89600 VSA Software to Display Results ..........................................................................22
8.0 Modeling Effects That Influence Distortion ........................................................................................23
9.0 Summary ................................................................................................................................... Back cover
Figure 1. This power amplifier
must be operated in back off
(where PAE is low) to avoid
compressing and distorting a
high-PAPR signal.
Figure 2. Higher PAE can be
attained if the PA's drain bias
can be adjusted dynamically
along with the input signal
magnitude.
2
This application note details a method for applying ET network. The simulation setups in these examples don't
to a PA. The method employs Keysight Technologies' show or necessarily correspond to how you would imple-
Advanced Design System (ADS) software to perform the ment an envelope tracking system. They do show different
necessary simulations. Keysight's 89600 Vector Signal techniques to efficiently simulate such systems if you have
Analysis (VSA) software can be used to display the a model of the power amplifier, however. The model could
simulation results. This example, along with another one be a transistor-level schematic or some sort of behavioral
highlighting the application of ET to a PA specifically model, including X-parameters*.
designed for narrow-band modulation, is available on
the Keysight EDA Knowledge Center at: http://edocs. Figure 3 shows the basic block diagram for modeling an
soco.keysight.com/display/eesofkcads/Applying+en envelope tracking system used in the examples. The shap-
velope+tracking+to+Improve+Efficiency. In the latter ing is applied to keep the amplifier operating at a constant
example, the presence of wideband modulation causes gain or gain compression even though its drain bias is
the PA to exhibit memory effects likely due to the bias being modulated.
Apply Shaping
Sample
Input Signal
Power Modulate
Drain
Voltage
Load
Modulated
Input Signal
Figure 3. A block diagram for simulating Envelope Tracking.
3
2.0 Characterizing the Power Ampliier
2.1 Testing the Power Ampliier for Memory Effects
Prior to simulating ET on an amplifier, you should check for The easiest way to test for memory effects is to run a
memory effects (assuming the model you are using would two-tone harmonic balance simulation (MemoryEffectTest_
show them) and run simulations to generate the shaping Freq_Delta_Sweep_HB) in which the frequency spacing is
functions. swept between the input tones (Figure 4). If the amplitudes
of the lower and upper third-order intermodulation distor-
Before applying ET, it is first useful to test the PA for tion sidebands are not equal then the amplifier is exhibiting
memory effects that may be caused by the behavior of memory effects.
the active device(s) or by the bias network. This can be
done using the harmonic balance test available in the 1. In this case, the example of a PA designed for wideband
Memory Effect Testing folder of the example. If an ampli- modulation, the simulation results shown in Figure 5
fier has memory effects, its output is not just a function of indicate that memory effects are minimal. This is true even
its input at the current instant in time. It will also depend when the frequency spacing between the two input signals
on what the input was previously. is 5 MHz. Consequently, the amplifier should behave fairly
well (even with the modulation bandwidth of the input
signal at 5 MHz).
Figure 4. Simulation setup for a two-tone harmonic balance simulation.
4
Figure 5. Simulation results for the example of a PA designed for wideband modulation.
5
2.2 Generating Data for a Shaping Table to Maintain Constant Gain
While not absolutely necessary, it should be useful to see The setup for this simulation is shown in Figure 6.
how a PA's power-added efficiency (PAE) varies with out- It was generated by selecting, from any schematic,
put power, with the drain bias voltage swept as a param- DesignGuide>Amplifier>1-Tone Nonlinear
eter, prior to applying ET. This should give you an idea of Simulations>Spectrum, Gain, Harmonic Distortion at
the potential performance improvement you can attain. X-dB Gain Compression (w/PAE) vs. 1 Param. Here, the
The HB1ToneGComp1swp schematic in the 2. Simulations drain bias voltage is being swept to see how gain and
to Generate Shaping Functions folder is used to simulate gain compression vary. The gate bias voltage has been set
the PA as a function of drain bias voltage. This schematic to -1.5 V. Looking at the simulation in the 0. Device I-V
and data display came from an updated ADS Amplifier Curves folder provides input on exactly how the gate bias
DesignGuide available at http://edocs.soco.keysight. affects the PA's performance.
com/pages/viewpage.action?pageId=161047818. (This
updated Amplifier DesignGuide will be in ADS 2012.)
Figure 6. Simulation setup for characterizing the PA.
6
The HB1ToneGComp1swp data display, Spectrum, Gain, The ADS Gain and Gain Compression plot (Figure 8) shows
Harmonics tab shows, among other things, a PAE versus how these parameters vary as a function of output power
fundamental output power plot (Figure 7). As evidenced by and drain bias. An example of this plot, with equations
the plot curves, for lower output powers the PAE increases and listing columns added to the HB1ToneGComp1swp
as the drain bias is decreased. Consider, for example, the data display in the example, is shown in Figure 9. The
fourth curve from the left (drain bias 3 V) in the plot (for plot shows what the drain bias (HB.Param1, the swept
the gray curves from left to right, the drain bias increases parameter in the simulation) should be for a particular
from 1.5 V to 6 V in 0.5 V steps.) It indicates that if a bias available source power to maintain a constant gain (speci-
of 3 V is used when the output power is near 21 dBm, the fied by Desired_Gain). A new feature for this plot that's
PAE would be about 34 % versus about 14% with a fixed available in ADS 2012 is the interpolate_y_vs_x() function
drain bias of 6 V. (Figure 9). This function is found in the utility_fun.ael file
inside the example PA's workspace. (If using a release prior
to ADS 2012, you will have to place this .ael file in your
$HOME/hpeesof/expressions/ael directory and re-start
ADS.)
Figure 7. Plot of PAE versus output power, showing increasing drain bias from left to right.
7
As an example, consider that when the drain bias
(HB.Param1) is 2 V, at 12 dB gain the output power
(Interpolated_Pout_at_Desired_Gain) is 12.068 dBm.
The available source power (Interpolated_Pavs_dBm_
at_Desired_Gain) is 0.05 dBm. The data from the listing
columns can be typed into a GMDIF file (figure 10), which
then becomes the shaping table when running an ET
simulation. Note that a shaping table may be generated for
a different gain just by setting Desired_Gain to a different
value.
Figure 8. Plot of gain and gain compression. Figure 10. Shaping table as a GMDIF file.
Figure 9. Using the interpolate_y_vs_x() function to obtain constant gain data.
8
2.3 Generating Data for a Shaping Table to Maintain Constant Gain Compression
In the same HB1ToneGComp1swp data display file, the The gain values at the 1.5 dB gain compression points (the
X-dB Gain Compression Data tab shows the gain and gain pink dots in Figure 11) can be used to shape the amplifier's
compression data, but also presents the data at a specified drain bias voltage. Figure 12 shows the data that would be
level of gain compression. In Figure 11, for example, 1.5 dB used in the shaping table.
of gain compression (relative to the maximum gain point)
has been specified.
Figure 11: Interpolated values at 1.5 dB of gain compression.
Figure 12. Shaping table data to maintain constant gain
compression.
9
3.0 Examining the Modulated RF Signal
One final task that may be useful to perform prior to simu- example. In this case, the behavioral model amplifier was
lating an amplifier with ET is to examine the test signal. removed, the RLoad parameter on the EVM sink was set to
The source used in all the ADS Ptolemy co-simulations in DefaultRLoad (1e18) instead of 50 ohms, and a TimedSink
this example is shown in the LTE_Source_Test schematic was added to record the voltage across the 50-ohm load
in the 3. Testing the LTE Signal folder (Figure 13). resistor so the power from the source could be observed
as a function of time. Additionally, the FramesToMeasure
The setup employed to simulate the test signal comes variable was reduced to 1, which is the lowest number
from the ADS example: LTE/LTE_FDD_UL_Tx_wrk/ allowed by the EVM sink.
LTE_UL_TxEVM. Many other modulated signals are
available in ADS's Wireless Libraries, so a signal from Figure 14 shows part of the LTE_Source_Test data display.
a different example could also be used. Note that the
schematic in Figure 13 has been modified from the original
Figure 13. Setup for examining
the ADS Ptolemy co-simulation
source.
Figure 14. The LTE_
Source_Test data
display.
10
The test signal has bursts of different power levels, lead- the markers can be moved to their locations by setting
ing to a high PAPR. Moving the markers on the display time=7msec on one marker readout and time=8msec on
shows power data within a particular burst. Alternatively, the other, as shown in Figure 15.
Figure 15. Setting marker locations is one way to see power data within a particular burst.
11
4.0 Simulating the Power Ampliier with a Fixed Drain Bias
There are a number of different ways to set up and run As an example, the LTE_Signal_w_Fixed_Bias schematic
simulations of the PA with a fixed drain bias. One way is to in the 4. Simulations with Fixed Bias folder is used to
employ a source that uses data generated from a different simulate the PA with an LTE signal and a fixed drain bias
Ptolemy simulation, such as the one shown above. The (Figure 16). This is the simplest simulation setup, however;
advantage of this technique is that simulations can be the modulated signal must be either generated separately
run for as long or as short as desired, without the user or obtained from another source (e.g., someone else). It is
having to know anything about setting up and running also possible to use I and Q data. In this case, the signal
Ptolemy co-simulations. Another method employs Ptolemy was generated from an ADS LTE example, as described in
co-simulation. While this technique has the advantage of the Envelope_Tracking_Sim document found at http://
providing specification-compliant measurement results, the edocs.soco.keysight.com/display/eesofkcads/Applyi
overall simulation time may be longer. One other technique ng+envelope+tracking+to+Improve+Efficiency. When
involves simulation with the Keysight 89600 VSA software generating the modulated signal, users can specify QPSK,
being used to post-process the results. Here, users can 16QAM, 64QAM, or some combination of these modulation
view simulation results using the same 89600 VSA soft- schemes. Note that the statistics of this source signal are
ware that is used to make physical measurements. shown in the LTE_UL_TxSpectrum data display file in the
3. Testing the LTE Signal, Statistics of a short LTE signal
folder.
Figure 16. Simulating the power amplifier with a fixed drain bias.
12
The simulation runs in about 8 seconds. However, since The above simulation could be re-run with the VtDataset
the simulation can be as long or short as the user wants, source reading data from the LTE_Source_Test dataset.
the actual run time will differ depending on the set stop This dataset may have been removed from the example
time. Figure 17 shows the simulation results and indicates to save space, in which case it can be re-created by re-
a pretty low mean PAE (only 15.7 %), which signals the running a simulation from the LTE_Source_Test schematic.
need for ET. Notice that the AM-to-AM and AM-to-PM The time point in the dataset where the data starts can be
distortion curves look a little "fuzzy." This is because the changed via the Toffset parameter.
amplifier is exhibiting slight memory effects.
Figure 17. Simulation results.
13
5.0 Implementing Envelope Tracking Using Behavioral Model Components
For a clearer understanding of how ET is implemented in ADS, begin by downloading and reviewing the Envelope_
Tracking_Sim document.
http://edocs.soco.keysight.com/display/eesofkcads/Applying+envelope+tracking+to+Improve+Efficiency.
5.1 Applying Envelope Tracking to the Ampliier
To apply ET to the same amplifier used in Section 4.0, the The above fixed bias and ET simulations were re-run with
LTE_Signal_w_ET_Bias_12_dB_Gain schematic in the Gain=2.5 on the LTE source. The results with fixed bias
5. Envelope Tracking with Constant Gain folder is used are shown in Figure 20. The results with ET are shown in
(Figure 18). Doing so improves the PAE by 4.9%, raising Figure 21.
it from 15.7% to 20.6 %, as shown in Figure 19. The black
curve in the figure represents the PAE with a fixed drain
bias.
Figure 18. Schematic for using ET to adjust the drain bias.
14
Figure 19. Simulation results after applying ET.
Figure 20. Fixed bias simulation with Gain=2.5 on the LTE source.
15
Figure 21. ET simulation with Gain=2.5 on the LTE source.
Notice that when using ET, there is only a slight improve- As the below data indicates, different simulation results
ment in PAE, about 24% versus 21.5%. As shown in Figure are obtained depending on the amplitude of the input sig-
22, that improvement increases further (about 22.5% ver- nal and the shaping table. Although not shown here, how
sus 15.7%), when an 11-dB constant gain shaping table is the amplifier is biased will also affect the performance.
used instead of a 12-dB constant gain (the gain parameter
on the source was set back to 2.)
Figure 22. ET simulation using a shaping table to maintain 11 dB constant gain.
16
6.0 Using Ptolemy Sources and Sinks to Simulate EVM and Drain Eficiency
Using Ptolemy co-simulation, it is possible to obtain specification-compliant error-vector-magnitude (EVM) data and drain
efficiency data.
6.1 Simulations with a ixed bias
As an example, the LTE_Cosim_Fixed_Bias_EVM sche- bias (Figure 23). The setup for this co-simulation came
matic in the 4. Simulations with Fixed Bias folder shows from the ADS example, $HPEESOF_DIR/examples/LTE/
a Ptolemy co-simulation of the amplifier with a fixed LTE_FDD_UL_Tx_wrk.
Figure 23. Ptolemy co-simulation schematic.
17
Some important things to note about this simulation setup.