Date post: | 02-Jan-2016 |
Category: |
Documents |
Upload: | kathlyn-palmer |
View: | 219 times |
Download: | 3 times |
DCPR-12/26/2004
MIT Lincoln Laboratory
DCPR RRC Pulse Shaping to Increase Capacity
Ryan Shoup, John Taylor, Bob Wezalis, & Josh Model
Aug 2004
MIT Lincoln Laboratory2
Rationale / Overview
• Root Raised Cosine (RRC) Review– Spectrally efficient waveform– Implementation
• General RRC considerations wrt DCPR– Envelope variation– Performance considerations
• Timing recovery• Matched filter demodulation
• General considerations when increasing DCPR capacity
• Laboratory demonstration of the Current modulation scheme with Root Raised Cosine (RRC) Filtering
– Concept– Laboratory demonstration
MIT Lincoln Laboratory3
GOES Data Collection System (Today)
• Communications link designed to relay information gathered from data collection platforms (DCPs) located throughout Western Hemisphere
• 400 KHz bandwidth allocated for the the GOES DCS communications link– Multiple Access system
• 200 FDMA channels @ 1500 Hz• 33 FDMA channels @ 3000 Hz
– Three data rates supported per channel• 100 bps BPSK modulation @ 1500 Hz• 300 bps 8-PSK TCM modulation @ 50 dBm (max) and 1500 Hz• 1200 bps 8-PSK TCM modulation @ 53 dBm (max) and 3000 Hz
– Channels alternate between two satellites
• DCP Messages comprise of a header + information
• Desirable to modify system to support growth in DCPs during the GOES-R era
MIT Lincoln Laboratory4
DCPR 8-PSK Performance
1.E-08
1.E-07
1.E-06
1.E-05
1.E-04
1.E-03
1.E-02
1.E-01
1.E+00
2 3 4 5 6 7 8 9 10 11 12
SNR per Bit (EB/N0)
BE
R
Trellis Encoded DCPR 8-PSK BER (Theoretical/Asymptotic)
Trellis Encoded DCPR 8-PSK BER (Simulated)
Uncoded QPSK BER (Theoretical)
MIT Lincoln Laboratory5
RRC
• Root Raised Cosine (RRC) Filtering implemented in software/firmware
• Spectrally efficient pulse shape– Pulse shape well suited to accommodate modest amount of growth to system
• Capacity increase of 2x
• Implementation in transmitter relatively easy via digital filtering– FPGA, D/A, and LPF– Components probably in many transmitters already anyhow– Commercial ICs also available to perform RRC filtering
• Pulse shape parameters– Rolloff coefficient ()
• 0 1 • Bandwidth required as • BER performance sensitivity as
– Number of coefficients (N)• Bandwidth required as N • BER performance as N
• Wide use of RRC– Cellular telephony– Satellite
MIT Lincoln Laboratory6
Notional RRC Transmitter
Sensor
Notional DCPR transmitter
TCM D/ADCPR MessageFormatter
UpconverterFIR Filter
BasebandRRC Waveform
LPF
• RRC filtering would be performed on data collected from sensor
• FIR Filter easily performed on small low cost CPLD/FPGA or software via microprocessor
• Only low speed D/A (~ 15 KHz) needed
• Analog LPF removes D/A [sampling] images created at the sample frequency
HPA
MIT Lincoln Laboratory7
Notional RRC Receiver
Notional DCPR 8-PSK Receiver
TrellisDecoder
Downconverter LPF A/D FIR Filter
Peak Matched FilterOutput Detector
DownSample
PhaseCompensation
Digital PLL
• Rx employs RRC matched filter detection for best performance
• Rx employs software/circuitry to detect/track the timing of peak outputs of the matched filter
• A digital PLL can be used with an easily [software] configurable bandwidth
• Trellis decoder remains same as used today
MIT Lincoln Laboratory8
RRC Implementation Tradeoffs
• Practical baseband spectrum approaches theoretical as Coeffs
• 100 Coefficients results in near ideal baseband spectrum
MIT Lincoln Laboratory9
RRC Implementation Tradeoffs
• Practical baseband spectrum more closely resembles theoretical as
MIT Lincoln Laboratory10
RRC Considerations: Envelope Variation
• Although average power same, instantaneous power varies more than that of a waveform that occupies more bandwidth
– Effect measured or quantified by the “peak to average ratio”
• Ideally, transmit HPA will be linear over full range of instantaneous power– If effect not mitigated, may potentially require larger transmitter power
amplifiers
• Effect of envelope variation mitigated by:– Use of higher value of roll off coefficient
– Use of coding to reduce required EB/N0
– High Power Amplifier linearization techniques
MIT Lincoln Laboratory11
0 1 2 3 4 5 6 7 8 9 100
0.2
0.4
0.6
0.8
1
Instantaneous Peak Powerto Average Power Ratio
Pro
babi
lity
Standard-compliant filter
RRC, alpha = 1.0
RRC, alpha = 0.1
Legend
~ 3 to 4 dB
~ 6 dB~ 7.25 dB
8-PSK Peak to Average
Instantaneous RRC 8-PSK power can be upto 2-4 dB higherthan notional standards-compliant filter
MIT Lincoln Laboratory12
DCPR 8-PSK TCM Asymptotic BER• Optimal implementation
requires that CDA employ matched filter demodulation
– Filter matched to transmitted RRC waveform
• Without matched filter, loss can be on the order of 1 dB
10 -8
10 -7
10 -6
10 -5
10 -4
10 -3
10 -2
Pro
bab
ility
of
Bit
Err
or
3 4 5 6 7 8 9 10
SNR per bit (EB/N0)
Ideal
Integrate and Dump
RRC Considerations: Matched Filtering
No matched filter at receiver degrades BER
MIT Lincoln Laboratory13
3 4 5 6 7 8 9 1010 -8
10 -7
10 -6
10 -5
10 -4
10 -3
10 -2
Pro
bab
ility
of
Bit
Err
or
DCPR 8-PSK TCM Asymptotic BER
Ideal
Few Coeffs
• With matched filter at the receiver, only relatively few coefficients needed for near ideal BER
RRC Considerations: Matched Filtering
SNR per bit (EB/N0)
MIT Lincoln Laboratory14
RRC Considerations: Timing Recovery
• Ideal BER performance requires matched filter output sampled at appropriate time
– Error results in BER degradation due to ISI and SNR loss
– Performance degradation function of timing error and
1.2 dB0.2 dB
2.2 dB
Example: RRC sensitivityto timing error
Loss due to ISIIncurred fromtiming error1.E-07
1.E-06
1.E-05
1.E-04
1.E-03
1.E-02
9 10 11 12 13 14 15 16 17
SNR per Bit
BE
R
'Uncoded 8-PSK (Theoretical) RRC, a=0.1, Timing Error = 5% (Simulated)
RRC, a=0.99, Timing Error = 5% (Simulated) RRC, a=0.35, Timing Error = 5% (Simulated)
Performance Loss:
MIT Lincoln Laboratory15
DCPR Capacity Considerations
• DCP Transmit Power Levels– Keep at current levels to minimize changes to DCPs
• Desirable to avoid need for new antennas, larger power amplifiers, etc.• Ideally DCP power levels even reduced to avoid issues with
instantaneous power variation associated with RRC waveforms
• Satellite Power Levels– Desire to keep signal power levels through satellite only moderately
greater than that associated with GOES NOP series • Power limitations on AGC circuitry and amplifier
– Transmit Power Levels• Need to ensure compliance with PFD requirements
• Neg 154 dBW per m2 per 4 KHz
• Current levels ~ 10 dB lower when channels fully loaded
• Frequency tolerance– Need to consider frequency tolerance specification when adding additional
FDMA channels
MIT Lincoln Laboratory16
8-PSK TCM with RRC Filtering
• Apply RRC filtering to the current modulation– Roll off coefficient: = 1.0– Theoretical Bandwidth required = 300 Hz
• Subdivide the current 1500 Hz DCPR channel into two channels doubling the number of 300 bps channels available
– Allocate 750 Hz per channel• More than necessary to accommodate actual (non-ideal) waveform, frequency
tolerance specification, and some guard band
• Additional channels decrease power per channel
– If AGC limits power levels, then EB/N0 at the ground receiver decreases– Capability for additional channels function of specified minimum G/T at ground site
• Increasing min ground station G/T by 3 dB would accommodate a doubling of capacity while maintaining satellite power levels to those seen today
• Only necessary for DRGS sites, as Wallops G/T far exceeds specified min DCPR min receive G/T
• Draft version of GOES-R IRD calls for G/T increase as well as satellite downlink EIRP increase
MIT Lincoln Laboratory17
RRC 8-PSK TCM Demo
• Benchtop demonstration to demonstrate 8-PSK TCM RRC BER performance
• Demonstration details– Data sequence (PRN) generation and Trellis encoding performed in a
Xilinx FPGA– RRC filtering and upconversion performed by R&S Signal Generator
(SMIQ)– Noise added at RF via Carrier to Noise Generator (CNG)– Downconversion done using mini-circuits RF component mixer and HP
signal generator (LO)– “Labjack” used to A/D baseband I/Q signals– PC microprocessor receiver performs timing recovery, phase recovery,
matched filter demodulation, trellis decoding, and measures BER• Frequency recovery not needed as HP LO frequency locked to SMIQ
• Observed BER performance ~ 1 dB from theoretical
• Caveat: Transmit amplifiers operated in linear region
MIT Lincoln Laboratory18
RRC 8-PSK TCM Demo Block Diagram
XilinxFPGA
R&SSMIQ
Data (300 bps)Clock
CNG
+70 MHz
Mixer
HPSig Gen
InterfaceBoard
FrequencyReference
Labjack PC
Baseband I/Q
I&Q A/D samples
MIT Lincoln Laboratory19
RRC 8-PSK TCM Demo Photo
MIT Lincoln Laboratory20
RRC 8-PSK TCM Demo Photo
MIT Lincoln Laboratory21
RRC 8-PSK TCM Demo Spectrum
300 BPS 8-PSK Spectra (Measured)
-120
-110
-100
-90
-80
-70
-60
-50
-40
-30
-20
-500 -450 -400 -350 -300 -250 -200 -150 -100 -50 0 50 100 150 200 250 300 350 400 450 500
Frequency Relative to Center (Hz)
Mag
nit
ud
e (d
B)
MIT Lincoln Laboratory22
RRC 8-PSK TCM Demo BER
RRC 8-PSK Trellis Coding
1.00E-07
1.00E-06
1.00E-05
1.00E-04
1.00E-03
1.00E-02
1.00E-01
2 3 4 5 6 7 8 9 10
SNR Per Bit (EB/N0)
Pro
bab
ilit
y o
f B
it E
rro
r
DCPR TCM (Simulated)
S/W Rx TCM 8-PSKPerformance (Measured)
MIT Lincoln Laboratory23
Summary
• RRC filtering effective means to increase capacity of DCPR system to accommodate future growth
• RRC filtering introduces issues– Power levels– Transmit filtering– Timing recovery– Need to perform Matched Filtering
• At DCPR data rates, transmit filtering, matched filtering, phase/frequency recovery easily implemented in low cost FPGAs, or microprocessors