GC5328 Low-Power Wideband Digital Predistortion Transmit Processor
Check for Samples: GC5328
1
FEATURES
•Integrated DUC, CFR, and DPD Solution•Supports Direct Interface to TI High-Speed
•20-MHz Max. Signal Bandwidth, Based on Max.
DPD Clock of 200 Mhz, Fifth-Order Correction
•DUC: Up to 12 CDMA2000/TDSCDMA, 4
W-CDMA, 2–10 MHz or 1–20 MHz OFDMA
Carriers
•CFR: Typically Meets 3GPP TS 25.141 < 6.5 dB
PAR, < 8.5 dB PAR for OFDMA Signals
•DPD: Short-Term Memory Compensation,
Typical ACLR Improvement > 20 dB
•GC5328IZER PBGA Package, 23 mm × 23 mm
•1.2-V Core, 1.8-V HSTL, 3.3-V I/O
•2.5-W Typical Power Consumption
•TMS320C6727 DPD Optimization Software
Data Converters
APPLICATIONS
•3 GPP (W-CDMA) Base Stations
•3 GPP2 (CDMA2000) Base Stations
•WiMAX, WiBRO, and LTE (OFDMA) Base
Stations
•Multicarrier Power Amplifiers (MCPAs)
DESCRIPTION
The GC5328 is a lower-power version of the GC5322 wideband digital predistortion transmit processor. The
GC5328 includes a digital upconverter (DUC) block, a crest factor reduction (CFR) block, a digital predistortion
(DPD) block, feedback (FB) block, and capture buffer (CB) blocks.
The GC5328 GPP block receives the interleaved IQ data from the baseband input. The individual IQ channels
are gain-adjusted in the GPP and routed to the DUC. The GPP and DUC can be bypassed to input a combined
IQ signal. The DUC provides three stages of interpolation and a complex mixer. There are two DUC blocks. The
output from the DUC blocks is combined in the sum chain. Each of the 1 to 12 DUC channels can be summed,
and the composite signal can be scaled.
1
Please be aware that an important notice concerning availability, standard warranty, and use in critical applications of Texas
Instruments semiconductor products and disclaimers thereto appears at the end of this data sheet.
PRODUCTION DATA information is current as of publication date.
Products conform to specifications per the terms of the Texas
Instruments standard warranty. Production processing does not
necessarily include testing of all parameters.
The CFR block has four serial stages of peak detection and cancellation. The CFR block cancellation filter can
be programmed as real or complex. The CFR peak-reduced output is routed to the Farrow resampler. The
Farrow resampler resamples the CFR output to the DPD clock rate. The Farrow resampler block also has a
complex mixer for composite carrier frequency offset.
The DPD subsystem has a circular limiter, nonlinear DPD correction, and a transmit equalizer. The DPD
correction can reduce the follow-on circuitry distortion products. The DPD output is sent to the BUC. The BUC
provides a post-DPD interpolation, and also provides a complex mixer for frequency offset. The DAC interface
converts the BUC signal output to the interleaved IQ or parallel IQ output signals for the DAC5682Z or DAC5688.
The CB block captures the selected internal reference signal, and the feedback block in two up to 4K capture
buffers. The signal capture can be based on an externally timed event (standard capture buffer), delay after a
timed event, or signal statistics (smart capture buffer). Normally the DPD input and feedback output are selected.
The capture buffers are stored and read by the microprocessor.
The FB block receives the LVDS ADC information and performs signal processing to downconvert the received
signal to 0IF. The FB block also has a feedback-path receive equalizer.
The GC5328 is a flexible transmit sector processor that includes a digital upconverter (DUC) block, a crest factor
reduction (CFR) block, and a digital predistortion (DPD) block and its associated feedback chain. The GC5328
processes composite input bandwidths of up to 20 MHz and processes DPD expansion bandwidths of up to 100
MHz. By reducing both the peak-to-average ratio (PAR) of the input signals using the CFR block and linearizing
the power amplifier (PA) using the DPD block, the GC5328 reduces the costs of multicarrier PAs (MCPA) for
wireless infrastructure applications. The GC5328 applies CFR and DPD while a separate microprocessor (a
Texas Instruments TMS320C6727 DSP) is used to optimize performance levels and maintain target PA
performance levels.
By including the GC5328 in their system architecture, manufacturers of BTS equipment can realize significant
savings on power amplifier bill of materials (BOM) and overall operational costs due to the PA efficiency
improvement. The GC5328 meets multicarrier 3G performance standards (PCDE, composite EVM, and ACLR) at
PAR levels down to 6.5 dB and improves the ACLR, at the PA output, by 20 dB or more. The GC5328 integrates
easily into the transmit signal chain between baseband processors (such as the Texas Instruments
TMS320C64x™ DSP family) and TI high-performance data converters.
A typical GC5328 system application would include the following transmit-chain components:
•ADS6149 14-bit, 250-Msps ADC or ADS5517 11-bit 200-Msps (feedback path)
•AMC7823 analog monitoring and control circuit with GPIO and SPI
BASEBAND INTERFACE
The GC5328 baseband interface block accepts baseband signals over an interleaved parallel interface at a data
rate of up to 70 MHz. The input interface supports up to 12 separate baseband carriers. The baseband interface
sends the interleaved IQ data to the DUC, or in DUC bypass to the sum chain, with up to 35-Mhz composite BW.
The baseband interface has 18-bit data (top16) BBData[15.0], BBFrame, and two additional (bottom two data)
MFIO(18,19).
The baseband clock input is a CMOS, low-jitter clock.
GAIN/PILOT INSERTION/AntCal INSERTION/POWER METER
Baseband gain can be applied on a per-carrier basis to control the individual channel power accurately through
the system. A UMTS pilot sequence at a programmable gain can be added for antenna calibration. Each
individual baseband channel has an integrated I2+ Q2power accumulator. The baseband power meters have a
common integration counter and interval counter for all channels. The GPP block has an IPDL detection and
control section to select one of four CFR memories when IPDL autoselection is used. Normally, IPDL 0 is
manually selected.
DIGITAL UPCONVERTERS (DUCs)
The GC5328 DUC block has interpolation filters, programmable delays, and complex mixers for each channel.
There are two DUC blocks within the GC5328. The sum chain after the DUC channel combines the DUC channel
streams or the bypass stream and sends the data to the CFR block. Each DUC can operate in one wide, two
medium, or six CDMA channels. Each DUC has a PFIR for spectral shaping, a CFIR for interpolation and image
rejection, and a bulk interpolation CIC.
The 2 DUCs can support:
•(6 channel/DUC mode) up to 12 – 1.23(8) Mhz CDMA, 1xEVDO, or TDSCDMA carriers
•(2 channel/DUC mode) up to 4 – WCDMA or LTE-5 carriers
•(1 channel/DUC mode) up to 2 – Wibro, Wimax, LTE 10 carriers
•(1 channel/DUC mode) 1 – Wimax or LTE20 carrier
Users can specify the filter characteristics of the DUC. The filters are the programmable finite impulse response
(PFIR), compensating finite impulse response (CFIR), and cascade integrator comb (CIC) filters. Users can also
specify the center frequencies of each carrier with a resolution of 0.25 mHz. Additional controls available in the
DUCs include bulk and fractional-time delay adjustments, phase adjustments, and equalization. The maximum
DUC output bandwidth is 40 MHz.
The GC5328 CFR block selectively reduces the peak-to-average ratio (PAR) of wideband digital signals. There
are four peak detection cancellation sections in series in the CFR block. Each stage compares the estimated
peak at the stage input with the target, and subtracts a scaled cancellation peak from the signal. There are 24
cancellers pooled among the four stages. The CFR interpolation filter must have at least 1.6× bandwidth, typical
is 2× BBClock to signal bandwidth.
There are four canceller memories and an update shadow memory that can be used for the auto-IPDL UMTS
select cancellation filter. The shadow memory allows the user to update one of the four filter banks during
operation. The CFR block has a composite RMS meter that can select the CFR input or output for monitoring.
The CFR block for WCDMA reduces TM1, TM3 signals for four adjacent carriers to 6.5 db PAR within the 3GPP
limit. The Wimax 10 reduction for two adjacent carriers is to 8.5 db PAR. TDSCDMA and CDMA performance is
limited by the carrier allocations and carrier coding.The CFR processing complex BW is limited to 62.5% of the
baseband clock rate.
FRACTIONAL FARROW RESAMPLER (FR)
The fractional resampler block takes the peak-reduced composite signals from CFR and resamples this through
fractional interpolation to the DPD processing rate. The user-programmable Farrow resampler supports
upsampling rates from 1× to 64×, with 16-bit precision on the interpolation ratio. After the fractional interpolation,
a complex mixer is available to provide a composite carrier IF offset frequency. A peak I or Q monitor is
provided.
DIGITAL PREDISTORTION (DPD)
The DPD block provides predistortion for up to Nth-order nonlinearities, and can correct multiple orders and
lengths of PA memory effects. The circular hard limiter provides a circular clipper that limits the
magnitude-squared value to –6 dbFS. This is optimized for hardware, and for the allowed gain expansion in the
nonlinear DPD correction.
The DPD has an RMS power meter, and a peak I or Q monitor.
The predistortion is performed for the nonlinear correction in the DPD section. The linear correction is performed
in the Tx equalizer. The predistortion correction terms are computed by an external processor (TMS320C6727
DSP) based on capture buffer information and the DPD software.
The DSP sets up the condition for collecting capture buffer data, retrieves the captured data over the EMIF bus,
and then performs calculations to compute the error and corrections to be used for the transmit path.
The host interface controls the mode of operation of the software in the TI DSP. TI provides a base delivery of
'C6727 software to GC5328 customers that achieves a typical ACLR improvement of 20 dB or more when
compared to a PA without DPD.
DPD CLOCK INPUT
The DPD clock input is an LVDS, low-jitter clock.
BULK UPCONVERTER (BUC)
The bulk upconverter block can interpolate the DPD block output by 1×, 1.5×, 2×, or 3× with a complex output.
The BUC interpolation blocks of 2 and 1.5 can provide 1×, 2×, or 3× interpolation for complex signals. The 1.5×
interpolation after DPD is performed by interpolating by 3 in the BUC and decimating by 2 in the OFMT block.
The BUC mixer can translate the composite IQ predistorted Tx output if the BUC Interpolation is > 1. Note: the
BUC interpolation of 1, 1.5, or 2 is recommended.
OUTPUT FORMATTER AND DAC INTERFACE (OFMT)
The output format and DAC interface presents the GC5328 output in the proper format for the different output
interfaces. The output formatter supports a test pattern for testing the DAC5682Z interface. The two output
interfaces supported for the GC5328 are:
The feedback path has two LVDS input ports. The A port is preferred (it has better timing). The external ADC
Input is converted or processed to generate a complex signal. The feedback equalizer has eight complex taps as
a receive equalizer. The feedback path has a mixer to translate the complex IF to the 0IF reference. The ADC
feedback rate is at the same rate as the DPD clock (fS). The typical feedback is fS/4, fS3/4 (m), or fS5/4 IF. The
feedback equalizer can provide (m) inverted spectral output, if needed.
The FB complex mixer translates the frequency of the complex input signal to 0IF. The feedback path has the
capability for nonlinear correction with a lookup table. TI ADCs that connect to the feedback path are the SDR
type ADS5444, DDR type ADS5445 (6149, 5517), DDR with reversed data phase ADSC217. The ADC feedback
path has modified connections for shared feedback path operation (see GC5325 schematic, User's Guide, in
References ). The GC5328 simplifies timing by providing a FIFO for each ADC port.
NOTE
There are eight LVDS data lanes and 1 LVDS clock lane. If the ADC has < 8 LVDS
data lanes the MSB of the ADC is connected to LVDS lane 7 (MSB) of the A feedback
port.
MICROPROCESSOR (MPU) INTERFACE
The MPU interface is designed to interface with external memory interface (EMIF) ports on TI DSPs operating in
asynchronous mode. It consists of a 16-bit bidirectional data bus, a 10-bit address bus, and RDB, WRB, OEB,
and CEB control signals. The CEB and OEB signals to the GC5328 require additional logic outside the
TMS320C6727B; see Table 1.
6727 DSP EMIFGC5328NOTES
EM_D[15.0]UPDATA[15.0]
EM_A[8.0]UPADDR[9.1]
EM_BA[1]UPADDR[0]
EM_CS2CEBNote: DSP HD[22.20] are used for logic for multiple chip-select, inverted outputs.
EM_RWBOEBInvert RWB send to OEB
EM_WEBWRB
EM_OEBRDB
AXRO[7]InterruptNote: DSP [HD22.20] can also be used with a multiplexer to select GC5328 interrupt.
The GC5328 has two capture buffers of 4096 complex words. The capture buffers are normally used to capture
the Tx reference signal and the feedback output signal. Capture buffer A can capture:
•The TX reference from the DPD after the circular hard limiter
•The feedback output; this represents the waveform as seen by the PA.
•The error output
•Testbus(31:16)
•QRD error output
Capture buffer B can capture:
•The TX reference from the DPD after the circular hard limiter
•The feedback output; this represents the waveform as seen by the PA.
•The error output
•Testbus(15:0)
Standard capture mode – The capture buffers can be armed to collect the 4K complex samples after a
programmable delay following a sync event.
Smart capture mode – There are two trigger conditions that combine the number of samples greater than a
threshold; these are used to find a number of peak events while the transmit signal is above a threshold. In this
case, the magnitude and magnitude-squared of the signal are compared against a threshold and counted. If the
capture buffer finds the trigger condition, the capture logic captures the programmed capture-buffer depth after
the trigger. This is a combination of DSP software and the GC5328 hardware.
Capture buffer A has a special mode to source data for diagnostic testing.
The DSP host-interface software has a function to select and get capture-buffer data.
The complex data is then passed from the GC5328 to the EMIF bus, to the DSP, and
back to the host processor.
The DSP host software has a signal-power monitoring function. This uses the
capture-buffer data to perform special monitoring, power measurement, and error
measurements.
There are special DSP software PA protection modes that use the capture buffer to
determine the DPD correction applied to the signal, the error between the DPD
reference input and the feedback signal. The capture buffers are also used in the
initial bulk delay and fractional delay alignment.
INPUT SYNCS AND OUTPUT SYNC
The GC5328 features multiple user-programmable input syncs. There are three syncs sampled with the BBClock,
(A, B, and C), and the Sync D,DC as an LVDS sync sampled by the DPD clock. Internally, the GC5328 can also
generate timed and software-controlled syncs. The sync A input is required for the GC5328 hardware to initialize.
It should ideally be the start of the frame or frame downlink. The output sync is a test signal used for debugging.
The input syncs can be used to trigger:
•Power measurements
•DUC channel delay, dither, and mixer-phase alignment
•Initializing/loading the DUC, feedback, equalizer, LUTs, etc.
•Feedback path tuner alignment
•Capturing and sourcing of data through SCBs
NOTE
The sync A external synchronization should match the customer Tx frame (total Tx
period – i.e., 5 ms).
See the baseband interface figure, these synchronization signals must meet the
timing of the BBClk.
POWER METERS AND PEAK I–or–Q MONITORS
There are three integrated I2+ Q2power meters in the GC5328:
•GPP – each baseband input channel
•CFR – the CFR input or output, and which antenna stream (0, 1)
•DPD – the input to the DPD nonlinear correction after the DPDL gain, and which antenna stream (0, 1)
There are several peak I or Q monitors within the GC5328.
•FRW– The resampled combined IQ interleaved input to the DPD
•DPD – The input to the DPD nonlinear correction after the DPDL gain
•DPD – After the nonlinear correction in DPD, and separately after the linear correction in DPD
•FDBK – There is a peak monitor at the output of the feedback path.
NOTE
The DSP host software has a HW POWER meter setup and Get(Monitor) function to
configure and get data from the integrated I2+Q2values.