NTU Confidential
The Effect of Delay in CFO Tracking Loop
The Effect of Delay in CFO Tracking Loop
in Multi-Carrier Systems: Simulation and
in Multi-Carrier Systems: Simulation and
Preliminary Analysis
Preliminary Analysis
Presenter: Pin-Hsun Lin
Advisor: Prof. Tzi-Dar Chiueh Date: Nov. 3rd, 2003
2
NTU Confidential
Outline
Outline
• Model transformation• Main factors affects the tracking performance with simulation
• Preliminary knowledge about Markov chain • Some preliminary analysis
3
NTU Confidential
Motivation
Motivation
• In the 802.11a project the time domain CFO tracking is said to be unstable since there’s a large delay (FFT block)
• Find out how the delay affects the burst
communication and how to solve the problem caused by the effect efficiently.
4
NTU Confidential
Transformation from multi-carrier to
Transformation from multi-carrier to
single carrier model
single carrier model
sm,p FFT CFO Estimator LF AWGN ej2πεp PD LF Noise Delay 1 or 2 ICI(Δε(m)) NCO NCO CFO Time domain CFO Tracking Loop
Approximation of Time- domain CFO Tracking Loop
Xm,p N’m,p φ(ε,m) CFO φ(ε,m)^ φ(Δε,m) Sample rate Symbol rate 1/Xm,p
5
NTU Confidential
Main factors affect the tracking
Main factors affect the tracking
accuracy
accuracy
• Constellation size • SNR
• Number of delay in the tracking loop
6
NTU Confidential
The effect of constellation size and SNR
The effect of constellation size and SNR
Constellat ion (QAM) Angle (degre e) 16 16.87 64 7.69 256 3.7
7
NTU Confidential
Selection of the lower bound of the
Selection of the lower bound of the
acquisition accuracy acquisition accuracy SNR π L 1 2
• The variance of the correlation type
estimation is lower bounded by: [3]
Correlation length=16 samples Correlation length=32 samples Correlation length=64 samples
L is the correlation length
L samples
CP L samples
correlate
8
NTU Confidential
The delay effect to convergence:
The delay effect to convergence:
simulation in the absence of noise
simulation in the absence of noise
Parameter value
Constellation
size 64 QAM CFO Initial error 5
degree/sec Noise variance 0 ISI 0 FFT size (N) 64 Guard interval length (Ng) 16 2-symbol delay
No delay Number of symbol
Number of symbol Number of symbol
9
NTU Confidential
The delay effect to convergence :
The delay effect to convergence :
simulation model with noise
1 0
NTU Confidential
The delay effect to convergence :
The delay effect to convergence :
simulation result with noise
simulation result with noise
SNR
1 1
NTU Confidential
Simulation results under different
Simulation results under different
constellation size, acquisition accuracy, and
constellation size, acquisition accuracy, and
delay
1 2
NTU Confidential
Summary
Summary
• When a high speed communication system which requires short preambles (bad
acquisition) and large size of constellation
(low tolerance of constellation rotation) is
considered, a fast tracking mechanism is required.
NTU Confidential
Fundamental knowledge for Analysis
1 4
NTU Confidential
The Markov property and Markov
The Markov property and Markov
chain (1)[8]
chain (1)[8]
• Using only transition probability and initial condition
can describe the whole process
• A discrete time Markov chain is a Markov process whose state space is finite or countable set and whose time
index is 0,1,2…
• In digital systems, finite word length makes the loop’s behavior as a finite state Markov chain.
1 5
NTU Confidential
The Markov property and Markov
The Markov property and Markov
chain (2)[8]
chain (2)[8]
• What kind of stochastic difference equation obey the Markov property? [7]
– Simple linear model
– Autoregressive (AR) model – Autoregressive (ARMA) model
• The 2nd order loop without delay generates
Markov-2 random sequence
• The 2nd order loop with K delay generates
1 6
NTU Confidential
The transition probability and
The transition probability and
transition matrix (1)
transition matrix (1) [8][8]
• A simple 1-step state transition diagram and transition matrix: p1,2 1 2 3 4 P2,3 p2,4 p3,3 p4,4
Absorbing state Transient state
Starting state destination state
1 7
NTU Confidential
The transition probability and
The transition probability and
transition matrix (2) transition matrix (2) p12,24 1,2 p23,33 p24,44 p4,4 p14,44 2,3 2,4 3,3 4,4 1,4 p12,23 p33,33 p44,44
•A 2-step state transition diagram and transition matrix:
1,1 1,2 1,3 1,4 2,1 2,2 2,3 2,4 3,1 3,2 3,3 3,4 4,1 4,2 4,3 4,4 1,1 1,2 1,3 1,4 2,1 2,2 2,3 2,4 3,1 3,2 3,3 3,4 4,1 4,2 4,3 4,4 Previous state
This state This state
next state
42x42 transition matrix
• This kind of transition matrix can describe 1st order loop
with 1 delay or 2nd order loop without delay
• A 2nd order loop with 1 and 2 delay need 3 and 4-step
1 8
NTU Confidential
Mean time to converge of the loop (1)
Mean time to converge of the loop (1)
t (sample) phase error acceptable phase error t=0 t=1 t=2 t=3 t=n t=n+1 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 Time to converge
1 9
NTU Confidential
Mean time to converge of the loop (2)
Mean time to converge of the loop (2)
The transition probability can be rearranged as:
Transient state
Absorption state
The mean time to converge can be proved as [8][8]:
NTU Confidential
Some Preliminary Analysis
2 1
NTU Confidential
The distribution of the Phase detector
The distribution of the Phase detector
output (1)
output (1)
Sum of N-1 independent R.Vs given Φ
2 2
NTU Confidential
The distribution of the Phase detector
The distribution of the Phase detector
output (2)
output (2)
From [4], the pdf of the arg{.} term can be represented by:
2 3
NTU Confidential
The stochastic difference equation of
The stochastic difference equation of
the loop (1)
the loop (1)
The difference equation of the loop filter:
Conversion to phase error difference equation:
which is an autoregressive moving average (ARMA) processes of order (4,2)
2 4
NTU Confidential
The stochastic difference equation of
The stochastic difference equation of
the loop (2)
the loop (2)
The ARMA process can be transformed into Markov-2 [7] by:
2 5
NTU Confidential
Conclusion
Conclusion
• Using the single carrier representation we can easily map the CFO tracking loop to the well known phase-locked loop (PLL).
• Based on the PLL model, we can find a high speed communication system requires a fast tracking
mechanism via simulation.
• Apply Liapounoff CLT to the sum of noise and ICI term, the output of PD can be derived as the Rician phasor. • Using Markov property of the stochastic difference
equation of the loop the mean time to converge can be derived.