MGMSK Links and Iterative Decoding

1
MGMSK Links and Iterative Decoding

• Recent Advances in Modulation and Coding for
• Bandwidth-efficient Power-efficient Wireless
  Links

2
Outline

• Introduction
• Modulation Background
• Coding Background
• Bandwidth- and Power-efficient Signaling
• M-ary GMSK and Iterative Decoding
• Future Areas for Investigation

3
Introduction

• Spun-off from Motorola GEG in 1985
• Signal processing for U.S. Government
  applications
• Restricted-access RD for US intelligence
  community
• SIGINT and LPI wireless communications
• Expanded into commercial wireless in 1995
• Flexible silicon-based broadband wireless modem
  technology
• End-of-year 2000 80 employees 15 M in
  revenue
• Continuing to serve U.S. government
• Advanced RD for DARPA and other agencies
• Making available superior COTS gear for armed
  service applications

4
Motivation for Research

• GMSK is a key worldwide wireless standard
• Constant-envelope permits use of highly efficient
  high-power amps (HPAs)
• Relatively bandwidth-efficient, for a
  constant-envelope waveform
• GMSK exhibits significant deficiencies
• Makes poor use of error-correction coding
• Capacity-limited by
• GMSK has potential for significant improvement

5
Modulation Background

• Ideal bandwidth-efficiency
• Root-Nyquist (RN) signaling
• Ideal power-efficiency
• Constant-envelope (CE) signaling
• Practical problems
• High peak-to-average-power-ratio (PAPR) of RN
  signals
• Poor bandwidth-efficiency of CE signals

6
Modulation Background RN Signals

• Bandwidth-efficiency evolutionary path

(1990)
(1995)
(2000)
7
RN Signals Waste HPA Peak-Power

• Low-a RN signals exhibit large magnitude
  excursions
• Peak-to-average-power ratio (PAPR) increases of
  several dB
• PAPR forces HPA backoff, reducing average signal
  power

8
Modulation Background CE Signals
Constant Radial Velocity
High AM
Staggered Transitions
Reduce AM
MSK SMSK
OQPSK
QPSK

• Power-efficiency evolutionary paths

MSK plus Partial Response
Multi-baud Gaussian Pulse
Sinusoidal Radial Velocity
GMSK
SFSK
TFM
9
CE Signals Waste Spectral Bandwidth

• Constant-envelope constrains signal trajectories
• Increased higher-derivative energy results in
  expanded spectrum
• At 45dB levels associated with spectral masks,
  3-fold bandwidth penalty

GMSK
10
Primary Modulation Challenges

• Reconcile bandwidth-efficiency with
  power-efficiency!
• How can RN signals be made more power-efficient?
• How can CE signals be made more
  bandwidth-efficient?
• Which is the best path forward?
• Power-efficient RN signals are practical
• CERN (constrained-envelope RN) technology has
  now been demonstrated
• CERN virtually eliminates PAPR penalty for
  bandwidth-efficient RN signals
• But . . . CERN requires linearized HPAs
• This excludes the highest efficiency (class C-F)
  HPAs
• A better solution is needed for use with class
  C-F HPAs

11
Coding Background TCM

• Trellis-coded modulation (TCM)
• Ungerboecks (NTC1977) vision of combined
  modulation/coding
• Euclidean metric replaced Hamming distance
• Use higher-order constellation to convey parity
  information for coding
• No increase in symbol rate (i.e. bandwidth)
  needed for coding gain
• Pragmatic TCM flexible single-engine Viterbi
  decoder
• Coset coding primary and secondary tiling of
  Euclidean space
• Completely flexible modulation order and code
  rate combinations

12
Coding Background Iterative Decoding

• Coding technology has substantially advanced in
  recent years
• Forney (66) concatenated serial trellis
  decoders
• a posteriori probabilities (APP) from inner
  decoder to outer decoder
• interleaving between the concatenated decoders
• Bahl, Cocke, Jelinek and Raviv (74) simple APP
  decoder architecture
• Hagenauer (93) MAP (maximum APP) ID for (block)
  product codes
• Berrou, Glavieux and Thitimajshima (93) turbo
  parallel-trellis (RSC) ID
• MacKay (95 ) low-density parity-check (LDPC) ID
• Divsalar, Pollara, Benedetto and Montorsi (97)
  serial-trellis ID
• Coding gain within 1 dB of Shannons limit is
  practical
• Only remaining issue comparative complexity of
  the 4 major approaches
• Challenge exploit coding to increase wireless
  link capacity

13
BW- and Power-efficiency Objectives

• Fully exploit forward-error-correction (FEC)
  coding
• No cost-competitive modern wireless link can
  afford to ignore FEC
• A higher-order modulation is needed to convey
  parity structure
• Inter-symbol interference (ISI) should be
  minimized to simplify decoding
• Fully exploit high-efficiency HPA technology
• Modern nonlinear (class C-F) HPAs offer
  near-ideal power-added efficiency
• This implies use of constant-envelope signals

14
MGMSK and Iterative Decoding (ID)

• GMSK can be extended
• to higher-order constellations
• to eliminate inter-symbol interference
• to be differentially-encoded (to support mobile
  or fixed links)
• Iterative Decoding (ID) can be extended
• to reduce implementation complexity
• to reduce decoder power-consumption
• Integrated transmitter technology can be extended
• to achieve higher symbol rates (gt1 Msps) with
  high efficiency
• to enhance HPA output power efficiency, as much
  as practical

15
MGMSK Link Model
16
MGMSK Modulator Architecture
To FM Modulator
17
Spectral-efficiency of MGMSK

• MGMSK is 50 of GMSKs spectral compactness
• MGMSK spectrum is twice as wide as GMSK, for all
  values of M
• Spectral compactness is not a valid measure of
  link capacity potential
• MGMSK is more spectrally-efficient than GMSK
• MGMSK can be spaced more closely than GMSK for
  same BER loss
• GMSK suffers from high sensitivity1 to
  adjacent-channel interference (ACI)

1 O Andrisiano and N. Ladisa, On the Spectral
Efficiency of CPM Systems over Real Channels in
the Presence of Adjacent Channel and Cochannel
Interference A Comparison between Partial and
Full Response Systems, IEEE Trans. On Vehic.
Tech., vol. 39, pp. 89-100, May 1990.
18
MGMSK Demodulation Options

• Differentially-coherent demodulation
• Simplest and most robust GMSK demodulator
• Least SNR-efficient
• Coherent demodulation with differential decoding
• Nearly 3 dB better SNR than differentially-coheren
  t demodulation
• Far more sensitive to channel dynamics and
  platform velocity
• Adaptively coherent demodulation
• Automatically select better approach
  coherent/diffcoherent demodulation
• Max-likelihood sequence estimation
• Likely prohibitive complexity for high capacity
  links

19
MGMSK BER Results

(a) Standard GMSK ( b) Compensated
4GMSK at 1/2 symbol rate ( c)
Compensated 8GMSK with rate 2/3 pragmatic TCM
20
Iterative Decoding (ID) SCCC

• BER-vs-EbNo Characteristic
• rate 2/3 8PSK
• differentially decoded
• Relative to GMSK
• x2 data capacity, same power

21
Status and Required Future Efforts

• SiCOM developed a viable MGMSK architecture
• MGMSK supports arbitrary signal order
• MGMSK eliminates ISI due to transmitting
  stretched pulses (BT 0.3)
• SiCOM developed a viable ID architecture
• Lower implementation complexity than JPLs recent
  breakthrough
• MGMSK/ID combination offers improved wireless
  links
• We need a high-efficiency high-power integrated
  transmitter
• We need prototype transceivers to demonstrate
  capability