Modeling and Refining Heterogeneous Systems With SystemC-AMS: Application to WSN

1
Modeling and Refining HeterogeneousSystems With
SystemC-AMS Application to WSN

• M. Vasilevski
• F. Pecheux, N. Beilleau, H. Aboushady
• K. Einwich

Laboratory LIP6 University Pierre and Marie
Curie, Paris 6, France Fraunhofer IIS/EAS,
Dresden, Germany
March 2008
2

• Issues
• SystemC-AMS Language
• Models of Computation
• SDF Behavioral Description
• SDF Multi-rates
• RF and AMS Modeling
• AMS Models
• RF Models
• Wireless Sensor Network Node
• Conclusion

3

1. Issues Mixed Systems Design

SystemC Verilog VHDL
Matlab Verilog-A VHDL-AMS Spice-RF
Matlab Verilog-A VHDL-AMS Spice
A/D Converter
Microcontroller
RF Transceiver
4

• Issues
• SystemC-AMS Language
• Models of Computation
• SDF Behavioral Description
• SDF Multi-rates
• RF and AMS Modeling
• AMS Models
• RF Models
• Wireless Sensor Network Node
• Conclusion

5
2.a Models of Computation
SystemC-AMS
SystemC
Synchronous Data Flow
Linear Network

• Models of computation
• Conservative Linear network
• Synchronous Data Flow

DE, MoCs (CP,FSM, etc)
Other Modeling Formalism
LN Modeling Formalism
SDF Modeling Formalism
Other Solver
LN Solver
Synchronisation Layer
SystemC Simulation Kernel
6
2.b SDF Behavioral Description
SCA_SDF_MODULE(B)
B
SCA_SDF_INltdoublegt
Input
Output
SCA_SDF_OUTltdoublegt
Behaviour
void sig_proc( )
A
C
7
2.c SDF Multi-Rates
Cluster
Tin
Tout
A
B
C
1
2
1
3
2
1
16 kHz
8 Hz
48 kHz
24 kHz
8

• Issues
• SystemC-AMS Language
• Models of Computation
• SDF Behavioral Description
• SDF Multi-rates
• RF and AMS Modeling
• AMS Models
• RF Models
• Wireless Sensor Network Node
• Conclusion

9
3.a AMS models Integrator
SCA_SDF_MODULE (integrator) sca_sdf_in lt
double gtin sca_sdf_out lt double gtout
double f sca_vector lt double gtNUM,DEN,S
sca_ltf_nd ltf1 void set_coeffs(double A)
DEN (0) 0.0 DEN (1) 1.0 NUM (0)
A void sig_proc() out.write( ltf1(NUM
, DEN, S, in.read())) SCA_CTOR
(integrator)
In/Out ports
Other Attributes
Initialisation method
Signal processing method
10
3.a AMS models Decimator
Decimator
2
2
2
11

• Issues
• SystemC-AMS Language
• Models of Computation
• SDF Behavioral Description
• SDF Multi-rates
• RF and AMS Modeling
• AMS Models
• RF Models
• Wireless Sensor Network Node
• Conclusion

12
3.b RF models
Power gain
NF
Rin
Rout
IIP3
Na
input
output
Rout
a1xa3x³
Rin
13
3.b RF models IIP3 and Noise Figure Test
FFT BW 120kHz
Power Gain 10 dB
Input amplitude -16.02 dBm
IIP3 10 dBm
NF 30 dB
14
3.b RF models Baseband Equivalent
X(t) DC I1cos(wt) I2cos(2wt) I3cos(3wt)
Q1cos(wt) Q2cos(2wt) Q2cos(3wt)
DC
I2
I3
I1
xBB(t)
0
w
2w
3w
Q1
Q2
Q3
15
3.b RF models Baseband Equivalent
Implementation
class BB double DC,I1,I2,I3,
Q1,Q2,Q3 ... BB operator(BB x)const BB
z(this-gtDCx.DC, this-gtI1x.I1,
this-gtI2x.I2, this-gtI3x.I3,
this-gtQ1x.Q1, this-gtQ2x.Q2,
this-gtQ3x.Q3) return z ...
SCA_SDF_MODULE (adder) sca_sdf_in lt double
gtinI sca_sdf_in lt double gtinQ sca_sdf_out lt
double gtout ... void sig_proc ()
out.write (inI.read()
inQ.read()) ...
SCA_SDF_MODULE (adder) sca_sdf_in lt BB gtinI
sca_sdf_in lt BB gtinQ sca_sdf_out lt BB
gtout ... void sig_proc () out.write
(inI.read() inQ.read()) ...
16

• Issues
• SystemC-AMS Language
• Models of Computation
• SDF Behavioral Description
• SDF Multi-rates
• RF and AMS Modeling
• AMS Models
• RF Models
• Wireless Sensor Network Node
• Conclusion

17

1. Wireless Sensor Network Node

• Wireless sensor network for environmental and
  physical monitoring
• Temperature, vibration, pressure, motion,
  polluants

18

1. Wireless Sensor Network Node

SystemC-AMS
SystemC
ATMEGA128 8 bits
A/D Converter
Microcontroller
RF Transceiver
2nd order OSR64 10 bits RZ feedback
Application Binary File
QPSK fc2.4GHz
decimator
2.4 MHz
8.53 MHz
2.4 GHz
19

1. Wireless Sensor Network Node

RF QPSK 2.4 GHz
20

1. Wireless Sensor Network Node Results

Noisy channel
DC offset
RF Simulation (2.4 GHz) SC-AMS classical simulation SC-AMS BB eq. RF simulation
1000 bits transmission 63.0s 0.036s
DC offset 19.9s 0.018s
Frequency offset 24.9s 0.022s
Phase mismatch 44.4s 0.031s
Frequencyoffset
Phase mismatch
21

1. Wireless Sensor Network Node Results

Settings Simulation Matlab SystemC-AMS SystemC-AMS
ADC alone OSR64 10 bits 8.53MHz 161024 pts 1.6 s 0.9 s 0.9 s
RF alone 2.4 GHz 10e3 bits 10e7 pts RF 150.7 s classic BB
RF alone 2.4 GHz 10e3 bits 10e7 pts RF 150.7 s 63.0 s 0.036s
2-nodestransmission Same settings 10e3 bits - 181.7 s 181.7 s
22
Conclusion

• Advantages to use SystemC-AMS
• Digital and Analog-Mixed Signal systems
  simulation
• Interface with SystemC
• Simulations very fast
• C based
• Polymorphism
• Easy to refine components with C inheritance
  ability
• Generic declaration of components with C
  templates
• Easy software programmer contribution
• Example of a free FFT library used for IIP3 test.