12
5/6/99 HDSL2 Modem Modeling and Simulation 1 HDSL2 Modem Modeling and HDSL2 Modem Modeling and Simulation Simulation Patrick Jackson Reza Koohrangpour
5/6/99 HDSL2 Modem Modeling and Simulation 1
HDSL2 Modem Modeling andHDSL2 Modem Modeling andSimulationSimulation
Patrick Jackson
Reza Koohrangpour
5/6/99 HDSL2 Modem Modeling and Simulation 2
HDSL2 OverviewHDSL2 Overview
l High-bit-rate Digital Subscriber Line - 2ndGeneration– Symmetric 1.544 Mbps– Applications
• T1 replacement• Telecommuting• Internet access
– Advantages Over T1• Much easier to deploy
– Bridge taps OK– 12,000 feet range– Spectral compatibility
• Single twisted pair
5/6/99 HDSL2 Modem Modeling and Simulation 3
Project OverviewProject Overview
l Ptolemy Simulation– Trade-off analysis
• Highly configurable• Extensible
– SDF domain
l Reference Design– HDSL2– Other modems
l Ptolemy Contributions
5/6/99 HDSL2 Modem Modeling and Simulation 4
Startup ModeStartup Mode
l Equalize channel– Reverse distortions– Adaptive filters in receiver– Training sequence
l Communicate parameters back totransmitter– Equalizer coefficients– Convolutional encoder polynomial
5/6/99 HDSL2 Modem Modeling and Simulation 5
Startup Mode SimulationStartup Mode Simulation
l Configurable– Channel model– Adaptive equalizer
l Passes parameters to data mode through files
5/6/99 HDSL2 Modem Modeling and Simulation 6
Data ModeData Mode
l Operating Mode– Sending and receiving data
l Additional Components– Transmitter
• Trellis encoder• Tomlinson precoder
– Receiver• Viterbi decoder
5/6/99 HDSL2 Modem Modeling and Simulation 7
Trellis EncoderTrellis Encoder
l Trellis Encoder– Adds redundancy– Output depends on state and input (512 states)– 3 bits in - 16-PAM pulse out
ConvolutionalEncoder
Serial toParallel
X(2)
Bit-to-LevelMapper
X(1)
X(0)
5/6/99 HDSL2 Modem Modeling and Simulation 8
Programmable EncoderProgrammable Encoder
X1(m-1)
a1a0
TsTsTs
a2
...
...
...a19 a20
b1
Y0(m)
Y1(m)
X1(m-20)X1(m-19)X1(m)
b0 b2 b19 b20
l Convolutional Encoder– Ts = delay of one symbol time– ⊕ = binary exclusive-OR– ⊗ = binary AND
5/6/99 HDSL2 Modem Modeling and Simulation 9
Trellis DecoderTrellis Decoder
l Viterbi Algorithm– Maximum likelihood sequence detection
• Calculate difference between receivedcodes and codes leading to next possiblestates
• Choose minimal difference sequence
– Window length• Sequence depth before decision• Trade-off: complexity,latency/gain
5/6/99 HDSL2 Modem Modeling and Simulation 10
Data Mode SimulationData Mode Simulation
5/6/99 HDSL2 Modem Modeling and Simulation 11
Simulation ParametersSimulation Parameters
5/6/99 HDSL2 Modem Modeling and Simulation 12
ConclusionConclusion
l Contributions– Configurable/extensible simulation– Reference design– Ptolemy
• Programmable Viterbi decoder• Tomlinson Precoder
l Limitations and Future Work– Startup-to-data mode sequence
• Mixed FSM, BDF, and SDF
– Timing recovery• Mixed BDF and SDF
