|
Modeling
3G/WCDMA/HSDPA Handset Transmit System
By Chris W. Liu, Staff Systems Engineer,
RFMD®
Wideband Code Division Multiple Access (WCDMA)
has been chosen as one of the standards for the third generation
(3G) cellular systems that are being deployed in various
parts of the world. System performances have been well defined
in 3GPP standards. However, these system specifications
need to be translated down to circuit block level for designers
to execute circuit implementation. To optimize the overall
system performance, system modeling is the essential tool
to allocate system budgets with accurate analysis, and therefore,
minimize design iterations and reduce the time to market.
Model of Transmit System
Figure 1 shows the proposed system model
of a linear direct conversion transmit system (DCT) for
the WCDMA system. This model consists of a number of subsystems
or circuit blocks: Digital Signal Generator, Digital to
Analog Converters (DAC) and Reconstruction Filters, Modulator,
Pre-amplifier and Power Amplifier, as well as RF Front-end
components.
The role of the signal generator is to receive data symbols
from the baseband processor, and then create ideal WCDMA
digital I/Q signals. The digital I/Q signals are then converted
to analog format by the DAC and reconstruction filter. The
analog I/Q signals are then fed into a modulator, where
the complex WCDMA signal is up-converted and modulated on
a RF carrier. The RF signal is amplified by the power amplifier,
and then delivered to the antenna through the front-end
components such as the duplexer, isolator, coupler, as well
as RF switches, etc.
There are many sources of distortion in the circuits, such
as noise, nonlinear transfer function, saturation limits,
etc. Now that the functionality of each circuit is quite
different, the main sources of distortion are different
as well. The system simulation focuses mainly on major sources
of each block. A system model has been developed in time
domain using Matlab since it allows a more accurate description
of the non-idealities of the blocks on the top of a complex
digital signal. The distortions of each subsystem are quantized
by passing an ideal test signal through the system model
so that the performance of each subsystem is simulated and
the optimum parameters can be found.
The system impacts are evaluated by computing
three key system parameters: Error Vector Magnitude (EVM),
Adjacent Channel Power Ratio (ACPR) and Spectral Emission
Mask (SEM). Based on the simulation results of EVM, ACPR
and SEM, the trade-offs and optimized parameters of each
subsystem are made. The 3GPP 25.101 concludes the system
requirements for WCDMA User Equipments (UE). Several key
transmit parameters of Band-1 are summarized in Tables
1 and 2. In practice, some margin
is usually added to guarantee the performance over all conditions.
Spectrum Emission Mask (SEM) is one of the
tough requirements for WCDMA system design. The SEM of 3GPP
specs are summarized in Table 2 and also
illustrated in Figure 2 for better view.
Though out of band spurious requirements from 3GPP are not
shown in this paper, they are important for system design
as well.
Test Signal Generation
The WCDMA signal is created in three main steps: spreading,
scrambling and pulse shaping. During spreading, every symbol
is transformed into a number of chips based on spreading
factor.
The complex scrambling is used not only to provide differentiation
among users, but also to distribute the power evenly between
I channel and Q channel because power levels at multiple
I/Q multiplexed channels may be different. Next, both I
channel and Q channel signals go though pulse shaping and
then combined into a complex chip stream.
Verifying uplink performance, 3GPP 34.121 defines a number
of test cases, referred to as uplink Reference Measurement
Channel (RMC). The most often used RMC is the RMC 12.2 kbps;
its channel configurations are listed in Table 3,
where DPCCH stands for Dedicated Physical Control Channel,
and DPDCH stands for Dedicated Physical Data Channel.
For data service of High Speed Downlink Packet Access (HSDPA),
High Speed Dedicated Physical Control Channel (HS_DPCCH)
is introduced in uplink. The uplink signal with HS_DPCCH
has much higher Peak to Average Ratio (PAR), and therefore
tightens requirements on the WCDMA transmit system, especially
power amplifiers. Because of the high PARs of a signal with
HS_DPCCH, 3GPP standards allow Maximum Power Reduction (MPR)
for HSDPA application shown in Table 4.
The MPR is based on Cubic Metric (CM), which is defined
in 3GPP 25.101 Release 7.0, shown in Equation (1).
Evaluating the transmitter’s performance
to meet HSDPA requirements, it is essential to have a proper
test signal with HS_DPCCH channel. The desired test signal
that could represent the worst test case shall have a property
of CM=1. With CM=1, the test signal has MPR of zero, thus
no power back off is allowed. The chosen test signal shown
in Table 5 consists of three channels:
DPCCH, DPDCH and HS_DPCCH.
The comparison of Complementary Cumulative Distribution
Function (CCDF) curves between the test signal and RMC 12.2
is also plotted in Figure 3. The much higher
PAR of the test signal is observed clearly in the plot.
Although PAR is high, the CM of the test signal is 1 (CM=1)
so that no MPR is allowed (MPR=0). The test signal is created
as an ideal reference signal with minimum EVM and ACPR.
In the following, the test signal will pass
through each subsystem so that the distortions can be evaluated
by computing the distorted signal at output of each subsystem.
DAC and Reconstruction Filter
The bandwidth of reconstruction filter and the resolution
of the DAC have direct impacts on system EVM, ACPR. The
EVM and ACPR as a function of the number bits in the DAC
are simulated and shown in Figure 4. The
level of quantization noise is proportional to the resolution
of the DAC. The simulation shows that increasing number
of bits greater than 8 has insignificant improvements of
EVM and ACPR.
The simulation results of ACPR and EVM versus the bandwidth
of the reconstruction filter are plotted in Figure
5. The ACPR is good overall since the resolution
of DAC is fixed at 9 bits. The EVM is degrading with decreasing
filter bandwidth, as expected, since both noise and useful
information are filtered out at same time. With 3.5 MHz
bandwidth and 9 bits resolution, 2 percent EVM and -54 dBc
ACPR are achieved.
Modulator
Imbalance of I/Q channels is one of the major issues in
the IQ modulator. Gain imbalance between I and Q channels
results in different amplitudes. Ideally, the Q channel
should differ 90 degrees in phase compared to I channel,
but phase imbalance adds errors on the phase. With gain
and phase imbalances in the system, an interference component
appears at the image frequency in frequency domain. The
image suppression (IS) can be calculated from I/Q imbalance
using Equation (2).
where AI is relative amplitude imbalance, PI is phase
imbalance in radians.
The impact of the Image Suppression is simulated and plotted
in Figure 6. It shows that EVM performance
is good enough when the image suppression is better than
40 dB.
Phase noise of Local Oscillator (LO) or Phase Locked Loop
(PLL) is a form of noise energy around the center carrier
in the frequency domain. In time domain, the phase noise
is defined as the random timing fluctuation in an oscillator
period. The integrated phase errors can be used to model
phase noise in time domain. The less phase errors, the better
EVM is achieved, as shown in Figure 7.
It is important to have a low pass filter to remove out
of band noise above certain cut-off frequency. The bandwidth
of loop filter determines the output noise of the PLL so
that it definitely affects the ACPR performance. ACPR and
EVM against the loop filter’s bandwidth are simulated
and plotted in Figure 8. The bandwidth
below 300 KHz gives good ACPR performance with reasonable
EVM.
Power Amplifier
The power amplifier (PA) is the most non-linear device in
a transmit system. The WCDMA modulation scheme results in
a fluctuating envelope of modulated signal. The amplitude
of the test signal is plotted in Figure 9,
and shows high peak to peak transitions between chips. The
large envelope variations with fast transition set very
stringent linearity requirements for the power amplifier.
The nonlinearity of the PA is modeled in the form of Amplitude-to-Amplitude
conversion (AM/AM) and Amplitude-to-Phase (AM/PM) conversion,
which reflect the amplifier gain and phase nonlinearity
as a function of the input power, respectively. The AM/AM
and AM/PM coefficients are extracted from measurement data,
namely input power, output power and phase, and then mathematically
modeled into amplitude and phase paths of the output signal.
The desired PA is to have linear relations between input
power and output power, and constant phase performance over
all power levels. The AM/AM and AM/PM performance of the
desired PA is shown in Figure 10, where
AM/AM is plotted in blue, and the pink line is desired phase
response. However, the PA design objectives are to make
optimal performance, especially between efficiency and linearity.
A PA working at class A has best linear transfer function
but suffers from low efficiency. In order to achieve good
efficiency, PA is usually designed in Class AB or Class
B at peak power level with acceptable sacrifice of linearity
performance. The nonlinear AM/AM (orange) and AM/PM (green)
curves in Figure 10 shows that the amplitude
and phase compressions occur at high power levels, where
the PA is close to saturation. In addition to compression
at high power level, PA also has gain expansion characteristics
at low power levels, which is not shown in the graph. The
PA’s nonlinear transfer characteristics are the root
cause of degradation of ACPR, EVM and SEM performance.
In this paper, a RF3266 WCDMA Band-I power
amplifier from RFMD is used in the simulation. The extracted
AM/AM and AM/PM curves of RF3266 are plotted in Figure
11.
The simulation shows that the degradations of ACPR at maximum
output power are reasonable, from -54 dBc at the input of
the PA down to -42 dBc at the output of the PA, but still
have 9 dB margin to system specification of -33 dBc. The
RF spectrum of the output signal is computed and plotted
along with SEM requirements (red line) in Figure
12, which shows spectral emission mask is satisfied
with margins.
Front-End Components
The last stage of the system model is the RF front-end section
including isolators, duplexers, and couplers, as well as
RF switches, etc. The loss of RF front-end components is
critical since the attenuated RF power needed to be compensated
by increasing RF power of the power amplifier, which result
in deteriorations of system efficiency and heat dissipation.
In addition, the gain ripples of the duplexers across band
cause degradation of EVM. The gain ripples of a duplexer
are modeled using S-parameters which can be measured by
a vector network analyzer. Once the S-parameters are collected,
gain and ripples are calculated from S-parameters. The S11
and S21 of a duplexer in Band I are measured and plotted
in Figure 13.
EVM suffers from the gain ripples of the
duplexer by almost 3 percent, while ACPR does not experience
any degradation due to the nature of high linearity of passive
devices. Note that the input impedance of a duplexer usually
does not match to 50 ohm and 2:1 input VSWR is typical for
a commercial duplexer. Carefully matching the impedances
between output of a power amplifier and input of a duplexer
is essential but very challenging.
Overall System Performance
The total system performance is computed at antenna port.
EVM with 13 % and ACPR with 41.2 dBc have been achieved
for overall system. Degradation of EVM and ACPR performance
at each stage are summarized in Figure 14
(a) and (b). The system
budget of each subsystem at normal operating conditions
is derived and optimized. EVM budget can be allocated as
2% at DAC and reconstruction filter, 9% at modulator, 9.7
% at output of PA, and 13% at antenna port. Similarly, the
budget if ACPR is allocated as -54 dBc at DAC and reconstruction
filter, -53 dbc at modulator, 42 dBc at PA and antenna port.
The margins are added on top of system specification to
ensure system performance over all conditions.
Summary
The technique of systematic modeling a WCDMA transmit system
has been presented. Simulation results indicate the modeling
technique gives accurate results in a fast and efficient
manner. This method could also be extended to model other
transmit systems with different architectures with minimum
modifications.
Acknowledgement
The author would like to thank colleagues at RFMD, especially
Brian Roberts, Steve Egolf, Hong Jiang and Hannes Rahn,
for information, reviews and feedbacks.
Reference
1. www.rfmd.com
2. 3GPP TS 25.101, “User Equipment (UE) Radio Transmission
and Reception (FDD)”
3. 3GPP TS 25.213, “Spreading and modulation (FDD)”
4. 3GPP TS 34.121, “Conformance specification”
About the Author
Chris W. Liu received his B.Sc from Tianjin University,
China and M.Sc from University of Montreal, Canada. Chris
has more than 15 years experience in the wireless, RF and
microwave communication industry. Currently he is a staff
systems engineer at RFMD. Chris Liu can be reached by email
at cliu@rfmd.com.
RFMD®
www.rfmd.com
TXTLINX.COM 70
Email this article
to a friend!
|
