digital radio front-end strategies provide game-changing ... · pdf filesince the output power...
TRANSCRIPT
Introduction
Explosive growth in cellular data usage is dictating communication infrastructure evolution,
pushing for increased capacity and reduction in cost and environmental impact. Heteroge-
neous networks – coexisting macro/pico/femto cells, along with advanced receivers and
transmitters to maximize spectral effi ciency can provide the required boost in capacity. At the
same time, improving power effi ciency will be a key consideration for next-generation radio
architectures.
As wireless service providers and base station manufacturers aggressively push towards
deployment of small cell base stations, they are faced with challenges to meet viable business
and network performance models: approximately a 10x reduction of power consumption,
size and cost compared to the traditional macro base station. A lot of focus has been given
to single-chip base station SoCs, along with optimized small cell software as a key strategy
for achieving these objectives, and rightfully so as the SoC architecture and software plays a
signifi cant factor in small cell cost, power, size and performance. The digital radio front ends
of macro base stations are typically on separate boards, sometimes even in separate enclo-
sures. Since the output power for small cells is signifi cantly lower than for traditional macro
base stations, one approach could be to scale down the macro design (perhaps also reducing
functionality) to achieve acceptable performance at reduced size and cost.
However, as we make the paradigm shift from a macro to a small cell base station, there
are additional opportunities for further integration of the digital radio front end technology
and the baseband processing executed by the SoC, as well as portions of the radio front end
with the analog RF circuitry. Newer interface standards offer more optimized board layout and
interconnect options, offering additional optimization. Finally, by leveraging the radio front end
power amplifi er linearization techniques, even further gains can be made in small cell solution
power reductions without compromising performance.
Abstract
It is no secret that small cell base stations
are expected to be a major lifesaver in the
wireless data deluge by helping to provide
significant capacity gains as part of 3G and
4G wireless heterogeneous networks. While
the industry standards and algorithms are
mostly in place for these heterogeneous net-
works to function, power, performance and
cost hurdles must be met before small cell
base station solutions manifest into practi-
cal reality. Base station manufacturers tend
to focus their attention on the performance
and attributes of the small cell baseband
System on Chips (SoCs). The baseband SoC
silicon and software does contribute to a sig-
nificant portion of the small cell solution per-
formance, however the digital radio front end
portion of the design can have a substantial
impact as well and is often overlooked. This
white paper focuses on the digital front end
portion of the small cell base station and
delves into design factors that play a sig-
nificant role in achieving the performance
and power targets demanded by small cells.
Digital Radio Front-End strategies provide game-changing benefi ts for small cell base stations
Hardik Gandhi,Radio IP Development Manager,
Debbie Greenstreet,Strategic Marketing Director,
Joe Quintal,Senior Applications Engineer,
Texas Instruments
W H I T E P A P E R
Digital Radio Front-End strategies provide game-changing benefi ts for small cell base stations May 2013
2 Texas Instruments
Traditional macro base station architectures can roughly be divided into three functional categories: control
processing, baseband (BB) processing and radio front end. While some level of integration of these func-
tions is happening today, the control, baseband and some of the radio front end functions of a macro base
station are typically on a single board, with the bulk of the radio front end often times on a separate board,
and maybe even in a separate enclosure, as depicted in Figure 1. Macro base-station radios often have
dozens of integrated circuits partitioned along functional boundaries, with baseband and control processing
in a leading-edge process node digital CMOS SoC, digital front end technologies in digital CMOS ASICs or
ASSPs, data-converters in CMOS/BiCMOS technologies with 1–8 converters per device, and RF up and down
conversion functions like I/Q modulators and mixers, buffers, attenuators, etc. in typically separate devices
with optimal noise and linearity performance.
When small cell base stations became an integral topic of heterogeneous network planning several years
ago, the success strategy hinged on the assumption that they would be much smaller and lower power and
hence, employ a single-board solution with much fewer numbers of discrete components compared to macro
cells. Since then, as operators and the 3GPP organization continue to hone the small cell requirements and
algorithms, the performance requirements and complexity has grown, providing further challenges in meeting
those form factor and power expectations.
While macro cells typically range in output power from anywhere between 10W to 60W or more, the
small cell umbrella covers a variety of applications with different output power ranges (transmit power at the
antenna):
• Indoor femto cells: <200mW Pout, often Power over Ethernet (PoE) requirements
• Enterprise femto cells: 125–250mW Pout, often PoE+ requirements
From macro cells
to small cells
Figure 1: Macro base-station radios often have dozens of integrated circuits partitioned along functional boundaries
Digital Radio Front-End strategies provide game-changing benefi ts for small cell base stations May 2013
3Texas Instruments
• Outdoor pico cells: 1–2W Pout
• Micro cells: 2–5W Pout and larger
To meet area and cost targets in small cells, the stringent performance requirements imposed on analog/
RF components are often relaxed, with the expectation that evolution of digital impairment correction
algorithms would help recover any loss in performance. For example, in order to lower size, the transceiver
front ends would use surface or bulk acoustic wave (SAW or BAW) duplexers, which are much smaller and
cheaper than the ceramic/cavity duplexers normally used in macro cells. But they add signifi cantly to post-PA
RF power loss. Due to this the RF PAs for small cells often need to operate at 2–3dB higher Pout than what
the antenna actually radiates, pushing them further into non-linear operating regions, in turn requiring linear-
ization algorithms to ensure compliance with spectral emissions and modulation accuracy limits.
The fact that a vast majority of, perhaps all, small cell base stations will be deployed with enclosures that
have no special provisions for air cooling makes reducing every watt of power consumption (and resulting
heat dissipation) ever so important. Along with the motivation to reduce power, cost and area, small cells
also need to support newer features like network listening modes for spectral sniffi ng and synchronization,
and be more frequency agile to be able to support different 3G/4G frequency bands and signal bandwidths
with as few variants of the design as possible. These requirements drive a rethink of traditional architectures,
potentially evolving into a confi guration as depicted in Figure 2.
Next-generation small cell architectures can achieve these goals by taking advantage of advances in many
areas:
• Aggressive power optimization solutions
•• PA linearization techniques like digital pre-distortion (DPD) in combination with crest factor reduc-
tion (CFR) techniques help lower overall system power consumption signifi cantly, enabling more
PA choices – cheaper PAs and less expensive board and enclosure designs, which help minimize
BOM cost
Figure 2: Small cell base station architectures strive for single-board solutions with minimal number of discrete components
Digital Radio Front-End strategies provide game-changing benefi ts for small cell base stations May 2013
4 Texas Instruments
•• PA Linearization techniques also provide margin for system performance – extending the reach
and performance of a small cell solution (bigger cell size, less interference) which maximizes
system effi ciency
• Flexible IF architectures and optimized interfaces
•• For traditional high-performance macro base stations, a costly heterodyne design has until now
remained the de facto radio topology. A direct-conversion, or homodyne architecture, where the RF
signal is directly down-converted to a BB signal or vice versa without any intermediate frequency
stages has many attractive features. It is also referred to as a zero IF architecture. A zero-IF (or
very-low-IF) approach enables dramatically reducing component count and thus footprint and cost
of radio transceivers. Reducing the number of parts also simplifi es the supply chain and manufac-
turing and improves yield. Zero-IF architectures provide a great deal of fl exibility in frequency plan-
ning and allow multi-mode and/or multiband operations with minimal changes to baseband digital
and analog circuitry. Zero IF is not a new concept: radio designers have used these architectures
for low-end handsets and some base-station designs. Historically, many of the benefi ts offered by
a zero-IF architecture have been offset by a series of problems – I/Q distortion, DC offset and LO
leakage being some of the most critical. Until recently, these technical barriers in receiver design
have prohibited its practical implementation for high-performance base stations. But with advances
in digital SoCs, novel, adaptive digital compensation techniques can be implemented cost-effectively
to mitigate these problems, and allow radio designers to fully exploit all the advantages this archi-
tecture choice brings to the table.
•• Newer interface standards, like JESD204B, help optimize board-level layout, reduce power
consumption and enable faster time to market compared to traditional parallel LVDS/LVCMOS
interfaces.
• Higher levels of integration
•• Aggressive integration of traditionally discrete digital-/analog-processing components into area,
cost and power-optimized System-on-Chips. As shown in Figure 2, all of the high throughput
digital processing could be combined in one optimized leading-edge digital process node BB SoC,
combining control, baseband and radio processing, including some or all elements of the digital
radio front end processing. All of the data conversion and RF up/down conversion processing
could be combined in another radio SoC in a process node suited for analog/RF performance, with
elements of the digital front end processing included to provide required analog/RF impairment
correction.
Every watt of power saved with the above techniques translates into a proportional reduction in operational
expenditure for the carrier (OPEX) as well as into proportional reduction in cooling (enclosure) costs, and may
provide a wider choice in component selection and resulting reduction in capital expenditure (CAPEX). With
Digital Radio Front-End strategies provide game-changing benefi ts for small cell base stations May 2013
5Texas Instruments
the sheer number of small cell deployments projected, even an 10W power saving per small cell base sta-
tion (typical savings with CFR/DPD for a 0.5W 2×2 small cell) using advanced technologies can translate to
signifi cant savings in lifetime operating expenses (~$1M/year for a single dense metro with 70k small cells
deployed), as well as reducing the environmental impact of vast scale deployment.
Digitally assisted RF transceiver architectures utilizing the techniques described in this paper take us a
step closer to an all-digital radio and help next-generation base-station transceivers meet some of the newer
challenges as they emerge. These novel technologies enable the small cell architecture to take on a single-
board solution as depicted in Figure 2, and are discussed in more detail in the rest of the paper.
The air interface for cellular base-station radios requires essential digital, analog and RF signal-processing
components to prepare the modulated samples for transmission, or extract modulated data from received
signals at the antenna, and is functionally partitioned as depicted in Figure 3.
Beyond providing an interface (LVDS/LVCMOS or JESD204A/B SerDes) to analog-to-digital and digital-
to-analog data converters, the digital front end blocks perform a variety of critical functions which can be
roughly classifi ed into two categories:
1. Channelization and re-sampling functions – these are mandatory signal-processing functions to
be performed for any type of base station [micro or macro, Time Division Duplex (TDD) or Frequency
Division Duplex (FDD), 3G or 4G]
• Carrier fi ltering to comply with spectral emission masks and spectral leakage requirements, includ-
ing root-raised-cosine fi ltering (RRC) and/or linear channel equalization.
• Tuning and channel aggregation/distribution – essential for multi-carrier and/or multi-standard
base stations.
• Gain, phase, delay adjustments and power-measurement functions
Figure 4 on the following page shows a fl exible digital up/down converter block (DUC/DDC) that performs
these manda tory channelization and resampling functions. Programmable fi nite impulse response (FIR) fi lters,
Figure 3: Functional partitioning of cellular base station processing components
AntennaAnalogDFELayer 1Layer 2Layer 3 Transport
Radio
Interface
Backhaul
Interface
RF CarriersSamplesFramesPackets
Digital Front End
technologies overview
Digital Radio Front-End strategies provide game-changing benefi ts for small cell base stations May 2013
farrow-based resampling fi lters, cascaded-integrator-comb (CIC) fi lters and numerically controlled oscillators
(NCOs) and mixers are the essential signal processing blocks that perform these functions.
2. Power amplifi er (PA) linearization and RF impairment correction functions – these are
optional in the true sense, but often mandated by system effi ciency and cost requirements. Every
Analog/RF component suffers from some impairment (group delay, non-linear distortion, gain/phase
imbalance) at its optimal operating point (best effi ciency, best dynamic range, best noise fi gure) that
can be corrected by digital pre- or post-processing. Some of the key algorithms for impairment cor-
rection include:
• Crest Factor Reduction – required to limit signal peak to average power ratios to reduce PA
peak power and linearity requirements and hence system cost
• Digital Pre-Distortion – required to improve system linearity to allow PAs to be operated more
effi ciently, reducing both system cost (CAPEX) and operating expenses (OPEX)
• I/Q distortion and DC offset/LO-leakage correction – essential to enable zero-IF system
architectures, which in turn help reduce system cost and improve fl exibility.
Fourth-generation multi-carrier communication systems based on orthogonal frequency division multiplex-
ing (OFDM) as well as third-generation code-division multiple-access (CDMA) based systems exhibit signals
with high peak-to-average ratios (PARs), also known as crest factors. The non-constant envelope-modulation
techniques employed in such systems have stringent Error Vector Magnitude (EVM) requirements. This re-
quires a highly linear PA amplitude and phase response, often resulting in an increased PA back-off (driving
the PAs at lower output power levels) to maintain acceptable linearity to meet spectral mask and modulation
accuracy requirements. PAs are most power effi cient at close to peak drive levels. A high PA back-off results
in a drastic reduction in PA effi ciency.
6 Texas Instruments
Crest Factor
Reduction
Figure 4: Block diagram of an implementation of essential channelization and re-sampling functions
Symbol rateBase bandI/Q Data
To CFRorFrom RX
Small cellsto/from
L1/L2processingMacro cells
to/fromCPRI/OBSAI
Up sampledcompositeAntennastreams
Digital Radio Front-End strategies provide game-changing benefi ts for small cell base stations May 2013
Crest-factor-reduction techniques strive to reduce the peak-to-average power ratios of transmit signals by
deliberately introducing noise into the signals within the EVM and Adjacent Channel Power Ratio (ACPR) limits
imposed by the standards. Multi-carrier (and even single-carrier LTE) signals can have a peak-to-average
power ratio (PAR) as high as 12dB. The application of CFR can reduce signal peaks by as much as 4–6dB,
with acceptable in-band EVM degradation, allowing the PA to operate at higher input/output power levels
(resulting in more effi ciency) while maintaining linearity at the output of the PA.
Even for the smallest category of small cells with 100–200mW antenna output power, the 4–6dB decrease
in PA peak power through the use of CFR enables a roughly 2–4× difference in PA size and power effi ciency,
and a corresponding reduction in cost. This should be a key trade-off when doing power budgeting to meet
small cell power over Ethernet (PoE) limits.
Figure 5 shows CFR processing for a typical multi-carrier 3G test model signal. The Complementary Cumu-
lative Distribution Function (CCDF) plot on the right shows the signal sample distribution before and after CFR.
As can be seen, the signal peak-to-average power ratio is reduced from over 11dB to under 7dB. CFR can
reduce the PAR even more, but would be limited by the EVM limits set for the standard. The plot on the left
shows a snippet of the time domain waveform around a signal peak. A cancellation pulse scaled to the ap-
propriate gain and phase is applied to the signal to reduce the peak amplitude to below a set threshold. The
cancellation pulse needs to have an appropriate spectral shape suited to limit most of the noise in-band so as
not to violate spectral mask requirements.
CFR is usually applied in baseband at low sample rates suitable for digital signal processing. But the PAs
operate on signals up-converted to RF frequencies. This up-conversion process will generate signal peaks
that may not be visible at low baseband sample rates. For optimal peak cancellation, CFR signal processing
7Texas Instruments
Figure 5: Crest-factor-reduction performance example
CFR InputCFR Output
CFR Input (blue)CFR Output (red)Cancellation (green)
CFR Threshold
Cancelled Peaks
Digital Radio Front-End strategies provide game-changing benefi ts for small cell base stations May 2013
has to apply special interpolation techniques to estimate the RF signal peaks and apply fractionally adjusted
cancellation pulses to the signal. Also special care needs to be taken to accommodate multiple overlapping
peaks to limit over cancellation.
RF power amplifi ers achieve maximum power effi ciency near the saturation point of the PA (often listed as
Psat or P1dB or P3dB in PA datasheets). Due to the inherent nature of LDMOS/GaN/GaAs power transistors,
the PAs are most nonlinear near Psat (as shown in Figure 6 below), introducing severe distortion effects into
the transmitted signals. As shown in Figure 7 on the following page, these effects manifest themselves as
in-band distortion (degraded Error Vector Magnitude – EVM) and increased out-of-band spectral re-growth
(degraded Adjacent Channel Leakage Ratio – ACLR). Minimum EVM and ACLR requirements are defi ned by
regulatory bodies and OEMs need to add suffi cient margin on top of these requirements to allow for perfor-
mance variations with temperature and time (component aging). Very often the PA drive levels are backed off
such that the signal falls within the linear region of the PA to avoid these distortion effects. But at these back-
off regions, the PA power effi ciency is extremely poor. Linearization techniques need to be employed to oper-
ate the PA close to its saturation region, where it offers maximum output power and best power effi ciency.
A variety of power amplifi er linearization techniques such as RF feed-forward, RF feedback, RF/IF pre-
distortion and post-distortion have been proposed and implemented over the years. Of these, adaptive Digital
Pre-Distortion (DPD) schemes have proven to be the most effi cient and cost effective compared to traditional
Figure 6: Typical RF power amplifi er response
Input Power
Out
put P
ower
PA transistor characteristics cause signal distortion in and beyond this region because the output power drops below the ideal (linear) curve
Without CFR/DPD, the PA must be operated in this region to avoid distortion
Ideal (linear) PA response
Typical PA responseP3dB
P1dB
Optimal PA power efficiency close to the saturation points (P1dB/P3dB)
CFR reduces signal peaks to concentrate the signal power within a limited region –allows pushing the PA output power higher
DPD allows the PA to be operated closer to its saturation region to maximize efficiency -while still meeting Spectral Mask and Modulation Accuracy requirements
8 Texas Instruments
Digital
Pre-Distortion
Digital Radio Front-End strategies provide game-changing benefi ts for small cell base stations May 2013
analog/RF linearization techniques. Effi ciencies of traditional LDMOS Class AB power amplifi ers widely in use
today when operated under a back-off condition range from 3–10%. However, with crest factor reduction
and adaptive digital pre-distortion techniques, the effi ciencies can be improved by 3 to 5 times. Newer PA
topologies such as advanced Doherty, or Class AB with drain modulation, in combination with digital pre-
distortion and newer GaN or GaAs power transistors, can achieve effi ciencies over 50%.
For small cell power amplifi ers, DPD allows the PA to operate linearly closer to Psat, improving system ef-
fi ciency. DPD also enables more PA choices – cheaper PAs and less expensive board design minimizes BOM
cost, and provides margin for system performance – extending the reach and performance of a small cell
base station (bigger cell size, less interference). As discussed later in this whitepaper, DPD can provide from
~1W/antenna to ~20W/antenna system power savings for small cells ranging from 100mW to 2W antenna
output power with exponential increase in savings beyond that.
The idea behind Digital Pre-Distortion (DPD) is to cascade a non-linear system prior to the PA, which
provides an inverse response to the PA such that the cascaded system has a linear response. As shown in
Figure 8 on the following page, f(.) is the inverse of g(.) such that y is a linear representation of x (with the
desired amplifi cation factor). RF power amplifi ers have complex non-linear models, made more complicated
by the presence of memory effects. Memory effects refer to the bandwidth-dependant nonlinear behavior
often exhibited by RF PAs. These encompass envelope memory effects and frequency response memory
effects. Envelope memory effects are primarily a result of thermal hysteresis and electrical properties inherent
to PAs. Frequency memory effects are due to the variations in the frequency spacing of the transmitted signal
and are characterized by shorter time constants. These memory effects and non-linearity in general changes
over time and temperature and requires a real time adaptive DPD correction.
The plots in Figure 9 on page 11 show amplitude and phase responses of a typical RF PA, where with DPD
the resulting outputs (in blue) are clearly much more linear compared to the original PA outputs (in red). Typi-
cally, the harder a PA is driven to maximize its effi ciency, the more severe the non-linear distortion and the
Figure 7: Effects of RF PA non-linearities
Non-Linear distortion causes out of band spectral leakage (ACPR degradation)
Non-Linear and linear distortion causes in-band error (EVM degradation)
9Texas Instruments
Digital Radio Front-End strategies provide game-changing benefi ts for small cell base stations May 2013
wider the bandwidth a PA has to support, the more severe the memory effects are, making the need for DPD
even more pressing.
The DPD non-linearity order, memory depth and adaptation rates are often PA and signal dependent, and
a commercially viable small cell DPD solution has to envision all usage scenarios and draw reasonable trade-
offs between performance and hardware and software computational requirements.
Direct conversion or zero-IF radio architectures are a popular choice for small cell transceiver design since
they use fewer components and are easier to integrate comparing with the conventional heterodyne archi-
tectures. But those benefi ts also come with several well-known impairments, namely I/Q imbalance and DC
offset.
In a direct conversion transceiver, a quadrature modulator (QM) implements I/Q modulation and RF upcon-
version, while a quadrature demodulator (QDM) realizes I/Q demodulation and RF downconversion. Because
of the limitation in analog circuit precision, the quadrature carriers used in QM and QDM cannot have exactly
the same amplitude and a perfect 90-degree phase difference that are essential for accurate signal conver-
sion. Similarly, the analog reconstruction fi lters on the I and Q paths may not match exactly. These imperfec-
tions are called I/Q imbalance, which causes cross talk between I and Q channels and creates undesired
images of the original signal. In addition, because of limited isolation in analog components, some of the local
oscillator (LO) power leaks into the RF output, which creates LO spikes in transmitted signal and DC offset in
10 Texas Instruments
DC offset and
I/Q distortion
compensation
Figure 8: A model for PA linearization via digital pre-distortion
Typical PA responseDPD models a response inverse to the PA response
Input Signal Pre-Distorted Signal
PA Output Signal
DPD cancels out PA non-linearity and helps eliminate the distortion effects (ACPR and EVM degradation) from the PA output signal.
Digital Radio Front-End strategies provide game-changing benefi ts for small cell base stations May 2013
Figure 9: Power amplifi er memory effects
received signal. For traditional high IF heterodyne architectures these unwanted images can be fi ltered out by
analog/RF fi lters. But for zero-IF architectures these unwanted distortion images fall very close to or right on
top of the signal bands of interest, degrading signal SNR/EVM and/or violate spectral emissions limits, and
are very expensive (or often impossible) to eliminate via analog fi ltering. Digital pre- and post-compensation
techniques have to be employed to compensate for these distortions. Further complicating matters, these
distortion components are often frequency-dependent and may vary over time as components age, requiring
adaptive cancellation techniques.
Most existing compensation techniques treat transmit and receive I/Q imbalances separately. To compen-
sate for I/Q imbalance in the transmitter, the receive side is either assumed to have dedicated feedback loops
or digital demodulators with perfect quadrature carriers. To compensate for I/Q imbalance in the receiver,
the transmit side is usually assumed to be free of I/Q imbalance. For optimal zero-IF transceivers, the above
limitations are often impractical, resulting in the transmitter and receiver distortion components coexisting
and overlapping with each other, making extraction of the respective distortion compensation coeffi cients
ever more complicated.
The push for ever more compact small cell enclosures to minimize visual impact and installation costs
necessitates highly condensed transceiver board layouts. Traditional data-converter interfaces based on
parallel LVDS or LVCMOS signaling protocols require considerable board area and a great deal of analysis
and special layouts to minimize skew across data bits and optimize the setup/hold times relative to the clock
traces. Board bring up with high-speed parallel interfaces is a big challenge and often gates system/software
bring-up and delays time to market.
JESD204 is a new standard that defi nes a serial communications link between data converters (ADCs
and DACs) and other devices such as FPGAs, DSPs, ASICs and clocking devices. Similar to other more
well-known signaling protocols like PCIe or CPRI, this SerDes-based interface highly simplifi es the digital
11Texas Instruments
JESD204
Digital Radio Front-End strategies provide game-changing benefi ts for small cell base stations May 2013
data interface between devices. With the clock embedded in the data stream and embedded algorithms to
optimize sampling of the data bits, this simplifi es the routing between devices because there are much fewer
lanes on the PCB, and it simplifi es system design and board bring-up – no setup/hold time margin across as
many as 16 LVDS pairs with one data clock to worry about.
This new standard reduces the number of I/Os and thus pin count of devices allowing for smaller pack-
ages, and offers a fl exible and scalable solution to accommodate different data traffi c needs (e.g., multiple
ADCs on one JESD differential pair).
JESD204A (ratifi ed in 2008) and JESD204B (ratifi ed in 2011) both provide support for multiple lanes
per converter or multiple converters per lane. JESD204B supports data rates up to 12.5 Gbps compared to
3.125 Gbps for JESD204A. In addition, the JESD204B Subclass 1 operating mode provides support for ac-
curate synchronization across multiple converters. Multiple transmitters and receivers can get synchronized
in order to obtain a deterministic latency across multiple devices. This requires the use of an external system
reference signal (also known as SysRef) for synchronization. SysRef signals and device clocks need to be
distributed with matched length to all devices in order for the internal “local multi frame clocks (LMFC)” to
be synchronized properly. This ensures that the SysRef signal gets processed at the same instant across all
devices. But the JESD204B traces don’t have to be length matched. This provides a great deal of fl exibility
in board layouts while still maintaining synchronization and deterministic latency across devices. The spatial
benefi ts can be clearly seen in Figure 10.
Current macro cell system architectures often employ discrete DFE solutions with built-in linearization func-
tions (see reference 1 for an example). With macro cells typically designed for 10W–60W or higher antenna
output power, the 1–2W per antenna linearization cost is a small fraction of the overall system power budget
and is dwarfed by the power savings linearization brings to the table.
For small cell applications, with these discrete linearization solutions, the break-even PA output power at
which linearization starts adding net benefi t is high. But with high levels of integration and digital technology
Shrinking cost of
linearization
12 Texas Instruments
Dual 14bit Data converter• DDR LVDS• 16 diff pairs(14 data + clock + sync)
Dual 14bit Data converter• JESD204B• 4 diff pairs @ 3.1Gbit or• 2 diff pairs @ 6.2Gbit
Figure 10: Comparison of sample layouts with traditional LVDS and newer JESD interfaces
scaling, benefi t of linearization may be seen even for the lowest class of small cell applications. Many techno-
logical advances contribute to the shrinking cost of linearization, some of them being:
• High levels of integration in fi ner lithography SoCs:
As advanced DFE technologies were evolving over the last decade or so, low investment cost (but
high power and high production cost) ASIC/ASSP or FPGA solutions were effective given the pace
of change. But as DFE technologies have become more mature, and with built in fl exibility in newer
DFE architectures, integration with other high-throughput baseband and control-processing functions
in deep sub-micron process nodes becomes more attractive, and allows for signifi cant reduction in
power and cost for the same functionality compared to discrete solutions. As an example and shown in
Figure 11, over 4× reduction in power and 7× reduction in cost (die area) can be seen as we progress
from stand-alone DFE solutions in 90nm to integrated DFE solutions in 28nm.
• More power and area effi cient interfaces between devices:
Higher amounts of integration, and new high-speed serial interfaces can help with substantial reduc-
tion in board area and cost, eliminating what would have been seen as barriers to implementing
advanced high-sample-rate and adaptive algorithms. As we saw earlier, with JESD204B a 4 to 8×
reduction in the number of interface signals can be seen compared to traditional LVDS interfaces.
• Exploiting synergies enabled due to integration:
Various resources can be time shared between linearization and other functions required for small cell
operation – like sharing the feedback path with network listening for clock synchronization or spectral
monitoring, or sharing a DSP processor for L1/L2 processing as well as DPD and I/Q offset compensa-
tion adaptation. This helps reduce the number of components on the board signifi cantly, and makes
optimal use of all available resources.
13Texas Instruments
Figure 11: Reduction in power and area with semiconductor technology scaling
0
20
40
60
80
100
0 20 40 60 80 100
Are
a
Power
Technology Scaling90nm
65nm
40nm28nm
Digital Radio Front-End strategies provide game-changing benefi ts for small cell base stations May 2013
Lowering the bar
The theoretical analysis shown in the graph in Figure 12 is based on the following assumptions:
• 2× increase in PA effi ciency with CFR (typical 3dB PAR reduction and a linear PA effi ciency curve)
• 3× increase in PA effi ciency with DPD
• 4× drop in Linearization power from discrete to integrated DFE.
From this simplistic analysis it is apparent that with integrated DFE, benefi ts of linearization can be seen
even with output power as low as 150mW (region highlighted in the fi rst gray circle), whereas with discrete
DFE, the break even power at which linearization adds value is around 0.5W (region highlighted in second
gray circle).
As seen from the graph in Figure 12, even at 150mW output power, greater than 2W power saving per
transmitter can be realized by including integrated CFR and DPD in the solution – Especially important
when trying to meet PoE (power over Ethernet) requirements.
To validate the benefi ts of linearization for the lowest class of small cells, a variety of small cell power ampli-
fi ers from different vendors were evaluated in the lab with and without linearization, and the results are
encapsulated in Figure 13 on the following page. A 4-carrier W-CDMA test model signal was used for this
analysis. The PAs were tuned to output 200mW of transmit power, which after accounting for the 2–3dB post
PA loss due to fi lters/duplexers would correspond to a 100–125mW power level at the antenna. Per antenna
cost of integrated linearization (CFR/DPD datapath power, additional analog/RF power for requisite feedback
and differences in sampling/interface rates and power consumed in executing the adaptation algorithms) was
computed using 28-nm benchmarks and factored into the above analysis.
PA biasing and/or drain voltage can be tweaked to trade-off effi ciency and linearity. For the above
experiments, without linearization, the PAs were biased to optimal linearity to meet spectral mask (ACPR)
14 Texas Instruments
Figure 12: Theoretical analysis of linearization benefi ts
0.0
2.0
4.0
6.0
8.0
10.0
12.0
14.0
16.0
0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00 1.10 1.20 1.30 1.40 1.50 1.60
Tota
l Pow
er C
onsu
mpt
ion
-PA
+ Li
near
izat
ion
(W)
Output Power (W)Power Consumptionwithout CFR/DPD (W)
Power Consumptionwith CFR, no DPD (W)
Power Consumptionwith discrete CFR & DPD (W)
Power Consumptionwith integrated CFR & DPD (W)
Digital Radio Front-End strategies provide game-changing benefi ts for small cell base stations May 2013
requirements. With linearization, the PAs were biased to optimal effi ciency, and the linearization algorithms
provided required ACPR improvement to meet spectral mask requirements.
As can be summarized from the results in Figure 13, even for this low end of small cell applications, with
CFR/DPD one can select a PA with > 3× better PA effi ciency (>27% compared to <8% for a PA without DPD,
or <7% for a PA without CFR/DPD). After factoring in the cost of linearization, 2.3× improvement in sys-
tem effi ciency is realized – while transmitting the same output power and meeting ACPR requirements. For a
2×2 single-band quad carrier MIMO 3G small cell with 100–125mW output power per antenna, inclusion of
linearization would provide >3.4W of system power savings, which may be critical to meet PoE budgets.
For higher output power levels, benefi ts of linearization are even more dramatic as can be seen from
Figure 14 on the following page.
Note that the analysis in Figure 14 shows only a comparison of power consumption with or without DPD,
with the assumption that the value CFR brings to the table for all classes of small cells is signifi cant enough
Figure 13: Experimental analysis of CFR/DPD benefi ts
15Texas Instruments
Digital Radio Front-End strategies provide game-changing benefi ts for small cell base stations May 2013
that its inclusion in any small cell solution is obviously warranted. If comparing results with neither CFR nor
DPD, against results with CFR and DPD, the projected power saving will be signifi cantly higher. CFR and
DPD are both key to optimal linearization – one without the other may give you less than half the benefi ts,
especially at higher output power levels.
It’s also worth mentioning that PA non-linear behavior and memory effects become worse as signal
bandwidth increases, leading to decreasing ACPR. Most PA datasheets provide performance results using
single-carrier W-CDMA test model (5MHz) or LTE (10MHz) data. But small cell deployment scenarios call for
a wide range of signal bandwidths to be supported, up to 40MHz occupied and beyond, sometimes with non-
contiguous carrier placements like in LTE rel10 and 11 with inter- or intra-band carrier aggregation, which
increases the edge-to-edge signal spread to be supported. The expanding signal bandwidth requirements
make the need for state-of-the-art linearization performance even more pressing, which may not be seem
obvious from reading PA datasheets.
Integrated digital front-end radio technology blocks are key additions to next-generation TI SoCs based on the
KeyStone II architectures, like the optimized dual-mode, dual-band small cell solution (TCI6630K2L) shown in
Figure 15 on the following page. Some of the key elements are briefl y described below.
• DDUC – Multiple digital up/down converter modules (shown as DDUCs in Figure 15) provide support
for a variety of signal types (W-CDMA, LTE-5MHz, LTE-10MHz, LTE-20MHz, etc.) with single- or multi-
band frequency confi gurations, with fl exibility that allows re-confi guring the BTS from single-mode
3G/4G to mixed-mode 3G/4G and vice versa. Also supported is both inter- and intra-band carrier
16 Texas Instruments
State-of-the-art DFE
capabilities integrated
in next-generation
TI KeyStone II SoCs
Figure 14: Experimental analysis of DPD benefi ts – Exponential power savings at higher output powers
0.0
5.0
10.0
15.0
20.0
25.0
30.0
0.10 0.50 1.00
Pow
er C
onsu
mpt
ion
PA+L
inea
rizat
ion
(W)
Antenna Output Power (W)(Assuming 3dB post-PA loss)
Power Consumptionwith CFR only, no DPD(Average 8% PA Efficiency)(W)Power Consumptionwith CFR/DPD(Including PA+Linearization power)(W)
200mW @ PA / 100mW @ Antenna
With CFR/DPD:~16% System
Efficiency~1.3W power
savings/antenna
1W @ PA / 0.5W @ AntennaWith CFR/DPD:~21% System
Efficiency~7.5W power
savings/antenna
2W @ PA / 1W @ AntennaWith CFR/DPD:~34% System
Efficiency~19W power
savings/antenna
Digital Radio Front-End strategies provide game-changing benefi ts for small cell base stations May 2013
aggregation with wide carrier separation fl exibility. As heterogenous network (Het-Net) strategies
evolve over the next few years, fl exibility to switch between carrier types, signal bandwidths and
frequency bands is of utmost importance.
• CFR – State-of-the-art algorithms in the TI CFR modules include advanced features like multiple
stages of peak cancellation with provisions to estimate fractional peaks and limit over cancellation,
automatic estimation of the CFR cancellation pulse shapes based on signal spectral content monitor-
ing, dynamic threshold adjustments and automatic gain control loops. As can be seen from the results
in Figure 16 below, optimal CFR cancellation and resultant signal PAR needs to be traded off against
system EVM budget, and a CFR algorithm that offers minimal EVM degradation with maximum PAR
reduction will be key to meeting system power and linearity constraints.
• DPD – An advanced datapath employing an optimized Volterra model to implement the PA pre-inverse
is an integral part of the TI DPD module. The Volterra coeffi cients are often adapted iteratively using a
17Texas Instruments
Figure 15: Block diagram of a KeyStone II-based TI baseband SoC with integrated digital radio functionality
TeraNet
Multicore Navigator
Security AccelerationPacAir and IP
Packet AccelerationPac
++**<<<<C66x DSPC66x DSP
+*<<C66x DSP
++**<<<<C66x DSPC66x DSP
+*<<C66x DSPARM A15ARM A15ARM A15
SystemServices
PowerManager
Debug
SystemMonitor
EDMAPktDMA
011100100010001111
DDR3/3L 72b 1333
1G Ethernet Switch
++**<<<<
C66x DSPC66x DSP
+*<<
C66x DSP
++**<<<<
C66x DSPC66x DSP
+*<<
C66x DSP**
2MB
1MB L2 1MB L2per C66x Core
Radio AccelerationPac
ARM A15
28 nm
Multicore SharedMemory Controller
2MB L3
011100100010001111
Front End Dig. RadioBit Rate
NANDNAND USB3USB3 SPISPI 1GbE1GbE
EMIF and IO High Speed SerDes Lanes
CPRI/OBSAICPRI/OBSAI JESD204BJESD204B
USIMUSIM
I2CI2C UARTUART GPIOGPIO
PCIePCIe
Single-carrier LTE20:
Peak PAR (db) 11.01 7.76 7.39 7.01 6.62 6.21
EVM (%) 0.254 2.19 2.88 3.75 4.8 6.09
Dual-carrier LTE20:
Peak PAR (db) 10.66 7.77 7.4 7.03 6.64 6.28
Carrier 0 EVM (%) 0.551 2.44 3.07 3.87 4.87 6.07
Carrier 1 EVM (%) 0.538 2.43 2.86 3.87 4.8 6.06
Figure 16: Example TI CFR results (PAR vs. EVM) for single- and dual-carrier LTE-20 signals
Digital Radio Front-End strategies provide game-changing benefi ts for small cell base stations May 2013
variety of least-squares type algorithms (Conjugate Gradient, Kalman, etc.) which can be implemented
in software on the high-performance fl oating-point DSP or ARM® cores, with additional hardware ac-
celeration options available for faster iteration times.
Figure 17 shows the PA output spectrum without DPD (red) and after DPD correction has been applied
(blue) for a representative small cell PA biased to optimal effi ciency. As can be seen from this fi gure, TI’s
high-performance DPD enables the PA output spectrum to meet spectral mask requirements while achieving
highest obtainable power effi ciency for a given PA.
• TXRX – Novel joint I/Q distortion and DC offset/LO-leakage correction techniques like the one showed
in Figure 18 below and implemented in the TI TXRX modules (and accompanying integrated analog/
RF transceiver solution – AFE750x) can be used to jointly separate the transmit and receive side I/Q
18 Texas Instruments
Figure 18: A joint TX/RX I/Q distortion cancellation technique
Figure 17: Example TI DPD performance results for a small cell PA
Pre-DPD PA Output Spectrum –Violates Spectral Mask
Post-DPD PA Output Spectrum -Meets Spectral Mask With Margin
Adjacent Channel ACLR limit:
-45dBcAlternate Channel
ACLR limit: -50dBc
Digital Radio Front-End strategies provide game-changing benefi ts for small cell base stations May 2013
imbalance by introducing phase rotations in the analog domain. Since the phase rotations have differ-
ent effects on transmit and receive side I/Q imbalance, it enables using advanced digital algorithms to
separate the distortion effects and be able to pre-compensate (on the transmitter) or post-compensate
(on the receiver). Using real-time adaptive blind- or calibration-based least squares algorithms with
frequency dependent or frequency-independent compensation, optimal signal SNR and emissions-
mask compliance can be achieved. Figure 19 shows example results using such an algorithm, where
the I/Q distortion and DC offset images that would have violated spectral emissions requirements (or
degraded signal SNR if falling in-band underneath an adjacent carrier) are effectively eliminated down
to the system noise fl oor.
In addition to the above, TI’s digital radio modules include other functions like front-end and back-end
Automatic Gain Control loops (AGCs) that help maximize data-converter effi ciency and reduce dynamic range
required in baseband processing, transmit/receive equalizers to compensate for analog/RF fi lter droop and
phase distortion effects and digital protection functions to limit signal excursions to prevent damage to the
PA and associated circuitry. These are welcome additions to any BTS transceiver design, providing signifi cant
performance boost and RF/analog component cost reduction at the expense of low-cost integrated digital
logic.
State-of-the-art TI discrete and integrated data-converter, RF, clocking devices and BB SoCs now support
JESD204B subclasses 0 and 1 interfaces for optimal board designs, and enabling rapid system bring-up.
Last, but not least, integrated and tested, production-ready platform software from TI supporting not just
the baseband processing, but the digital radio processing, such as DDC/DUC, CFR and DPD libraries, as well
as analog/RF control enables rapid integration of system components and quick ramp to production.
As shown in Figure 20 on the following page, TI’s baseband SoC (TCI6630K2L) with integrated digital
front-end technologies, closely coupled with TI’s integrated radio transceiver solution (AFE750x) and other
clocking, power and RF devices from TI, with production ready software, enables a high-performance small
19Texas Instruments
Figure 19: Example I/Q distortion and DC offset cancellation results
Digital Radio Front-End strategies provide game-changing benefi ts for small cell base stations May 2013
cell solution with optimized power consumption meeting PoE requirements, with a low BOM cost, fast time to
market, and fl exibility to support evolving Het-Net strategies.
• Integrated DFE solutions bring the benefi ts of linearization to lower output power systems; the use of
CFR/DPD for PA output as low as 200mW (100–125mW at the antenna) seems not only viable, but
mandatory to meet stringent PoE requirements.
•• >3.4W total system power savings with CFR+DPD (compared to a solution with no CFR, no DPD)
for a 2×2 indoor small cell with a 25.5W overall system power budget seems very compelling.
• Power savings increase exponentially with higher output powers. Taking into account higher post-PA
losses of 2–3dB for small cells, savings can be much more.
•• >15W saved with DPD (compared to a solution with CFR but no DPD) for a 2×2 system at 1W
output power per PA (0.5–0.625W output power at the antenna)
•• >38W saved with DPD (compared to a solution with CFR but no DPD) for a 2×2 system at 2W
output power per PA (1–1.25W output power at the antenna)
• Dynamic nature of LTE signals, coupled with multi-mode and multiband requirements and the need
to extract the last ounce of system power savings make the need for best-in-class CFR and DPD
algorithms in integrated base-stations ever more compelling.
20 Texas Instruments
Conclusion
Figure 20: TI’s small cell system solution
AFE750x
JESD204B
JESD204B
2xA15
DD
UC
/ C
FR /
DP
D/T
XR
X
Network CoProcessor
GigESwitch
PTP, Sync
GigEPHY
WIFI
DDR3
DDR3
EMIF
Flash
Acc
eler
atio
n P
acs
I2C
UART
USIM USB
4xC66x
Clocking
PMU
GPS
SPI
DAC
ET
AFE750x
2x TX RF
TRX
2x TX Dig 2x DAC
2x RX RF2x RX Dig 2x ADC
RX RFRX Dig ADC
Auxiliary Receiver
2x2 TX and RX
ET
2x TX RF
TRX
2x TX Dig 2x DAC
2x RX RF2x RX Dig 2x ADC
RX RFRX Dig ADC
Auxiliary Receiver
2x2 TX and RX
PoE AVS
CombinerDuplexer
HPALNA
CombinerDuplexer
HPALNA
PCIe
TCI6630K2L
Important Notice: The products and services of Texas Instruments Incorporated and its subsidiaries described herein are sold subject to TI’s standard terms and conditions of sale. Customers are advised to obtain the most current and complete information about TI products and services before placing orders. TI assumes no liability for applications assistance, customer’s applications or product designs, software performance, or infringement of patents. The publication of information regarding any other company’s products or services does not constitute TI’s approval, warranty or endorsement thereof.
SPRY236© 2013 Texas Instruments Incorporated
References
• Flexibility of any linearization solution is key towards reuse across multiple classes of base stations,
and to tackle future advances in PA technology – Could there be in the near future a PA available that
can simultaneously transmit multiple signal bands spread over 100s of MHz of spectrum with accept-
able effi ciency and cost? If so, a linearization solution which is capable of supporting such a PA is vital.
• TI’s fl exible small cell SoCs with best-in-class integrated-linearization solutions enable
optimum system cost and power consumption across different classes of small cells, and
provide an easy upgrade path to satisfy future evolutionary needs.
1. H. Gandhi, W. Abbott “A digital signal processing solution for PA linearization and RF impairment cor-
rection for multi-standard wireless transceiver systems,” in Proc. IEEE European Microwave Confer-
ence. Sept. 2010.
21Texas Instruments
IMPORTANT NOTICE
Texas Instruments Incorporated and its subsidiaries (TI) reserve the right to make corrections, enhancements, improvements and otherchanges to its semiconductor products and services per JESD46, latest issue, and to discontinue any product or service per JESD48, latestissue. Buyers should obtain the latest relevant information before placing orders and should verify that such information is current andcomplete. All semiconductor products (also referred to herein as “components”) are sold subject to TI’s terms and conditions of salesupplied at the time of order acknowledgment.
TI warrants performance of its components to the specifications applicable at the time of sale, in accordance with the warranty in TI’s termsand conditions of sale of semiconductor products. Testing and other quality control techniques are used to the extent TI deems necessaryto support this warranty. Except where mandated by applicable law, testing of all parameters of each component is not necessarilyperformed.
TI assumes no liability for applications assistance or the design of Buyers’ products. Buyers are responsible for their products andapplications using TI components. To minimize the risks associated with Buyers’ products and applications, Buyers should provideadequate design and operating safeguards.
TI does not warrant or represent that any license, either express or implied, is granted under any patent right, copyright, mask work right, orother intellectual property right relating to any combination, machine, or process in which TI components or services are used. Informationpublished by TI regarding third-party products or services does not constitute a license to use such products or services or a warranty orendorsement thereof. Use of such information may require a license from a third party under the patents or other intellectual property of thethird party, or a license from TI under the patents or other intellectual property of TI.
Reproduction of significant portions of TI information in TI data books or data sheets is permissible only if reproduction is without alterationand is accompanied by all associated warranties, conditions, limitations, and notices. TI is not responsible or liable for such altereddocumentation. Information of third parties may be subject to additional restrictions.
Resale of TI components or services with statements different from or beyond the parameters stated by TI for that component or servicevoids all express and any implied warranties for the associated TI component or service and is an unfair and deceptive business practice.TI is not responsible or liable for any such statements.
Buyer acknowledges and agrees that it is solely responsible for compliance with all legal, regulatory and safety-related requirementsconcerning its products, and any use of TI components in its applications, notwithstanding any applications-related information or supportthat may be provided by TI. Buyer represents and agrees that it has all the necessary expertise to create and implement safeguards whichanticipate dangerous consequences of failures, monitor failures and their consequences, lessen the likelihood of failures that might causeharm and take appropriate remedial actions. Buyer will fully indemnify TI and its representatives against any damages arising out of the useof any TI components in safety-critical applications.
In some cases, TI components may be promoted specifically to facilitate safety-related applications. With such components, TI’s goal is tohelp enable customers to design and create their own end-product solutions that meet applicable functional safety standards andrequirements. Nonetheless, such components are subject to these terms.
No TI components are authorized for use in FDA Class III (or similar life-critical medical equipment) unless authorized officers of the partieshave executed a special agreement specifically governing such use.
Only those TI components which TI has specifically designated as military grade or “enhanced plastic” are designed and intended for use inmilitary/aerospace applications or environments. Buyer acknowledges and agrees that any military or aerospace use of TI componentswhich have not been so designated is solely at the Buyer's risk, and that Buyer is solely responsible for compliance with all legal andregulatory requirements in connection with such use.
TI has specifically designated certain components as meeting ISO/TS16949 requirements, mainly for automotive use. In any case of use ofnon-designated products, TI will not be responsible for any failure to meet ISO/TS16949.
Products Applications
Audio www.ti.com/audio Automotive and Transportation www.ti.com/automotive
Amplifiers amplifier.ti.com Communications and Telecom www.ti.com/communications
Data Converters dataconverter.ti.com Computers and Peripherals www.ti.com/computers
DLP® Products www.dlp.com Consumer Electronics www.ti.com/consumer-apps
DSP dsp.ti.com Energy and Lighting www.ti.com/energy
Clocks and Timers www.ti.com/clocks Industrial www.ti.com/industrial
Interface interface.ti.com Medical www.ti.com/medical
Logic logic.ti.com Security www.ti.com/security
Power Mgmt power.ti.com Space, Avionics and Defense www.ti.com/space-avionics-defense
Microcontrollers microcontroller.ti.com Video and Imaging www.ti.com/video
RFID www.ti-rfid.com
OMAP Applications Processors www.ti.com/omap TI E2E Community e2e.ti.com
Wireless Connectivity www.ti.com/wirelessconnectivity
Mailing Address: Texas Instruments, Post Office Box 655303, Dallas, Texas 75265Copyright © 2013, Texas Instruments Incorporated