Académique Documents
Professionnel Documents
Culture Documents
260621774, syed.hussain3@mail.mcgill.ca
1
Outline:
INTRODUCTION
TYPES OF BEAMFORMING
ADAPTIVE ANALOG BEAMFORMING ALGORITHMS
MODIFIED STRUCTURE AND EXTENDED ALGORITHMS
RESULTS
CONCLUSION
2
Introduction:
Beamforming
Spatial filtering depends on the Weiner filter theory, i.e., to compute statistical
estimate of the unknown random signal using a random signal that contains
some information of it.
3
Introduction:
Applications Description
Geophysical Seismic Arrays for Oil Exploration and Detection of Under Ground
Exploration Nuclear Test
4
Types:
Types of Beamforming
I. Digital Beamforming :
Also known as Multiport Beamforming. Multi-port Beamforming Architecture
Beamformer can access signals on all antenna-elements.
For digitally sampled data, approaches like conjugate-gradient and Least-square
are used for better convergence speed of beamforming algorithms.
Not very economical as one RF downconverter and ADC has to be used at each
antenna element.
5
Types:
6
Adaptive Analog Beamforming Algorithms:
Assumptions
𝑄−1
𝒙 𝑡 = ℎ𝑞 𝑖𝑞 𝑡 𝒔 𝜃𝑞 + 𝑛(𝑡)
𝑞=0
where,
𝑖𝑞 (𝑡) is the transmitted signal from 𝑞 𝑡ℎ user
ℎ𝑞 is the channel coefficient
𝒔(𝜃𝑞 ) is the steering vector
𝑛(𝑡) is the AWGN noise.
9
Adaptive Analog Beamforming Algorithms:
10
Adaptive Analog Beamforming Algorithms:
unbiased MSE gradient estimate, i.e., 2 Rw for a single receiver system is given as
𝐾
1
𝒈1 𝒘 = 𝑓1 (𝒘+ , 𝑖)𝜹(𝑖)
𝛾𝐾
𝑖=1
where, 𝑓1 𝒘+ , 𝑖 is the instantaneous array output power .
𝒘+ = 𝒘 + 𝛾𝜹 𝑖 ,
𝒘− = 𝒘 − 𝛾𝜹 𝑖 ,
1
𝒈2 𝒘 ′ = σ𝐾 𝑑 𝑖 − 𝒘𝐻 𝐻
+ 𝒙 𝑖 − 𝒘− 𝒙 𝑖 𝜹 𝑖
𝛾𝐾 𝑖=1
12
Adaptive Analog Beamforming Algorithms:
𝒘+ = 𝒘 + 𝛾𝜹 𝑖 ,
1
𝒈2 𝒘 ′ = σ𝐾 𝑑 𝑖 − 𝒘𝐻 𝐻
+𝒙 𝑖 − 𝒘 𝒙 𝑖 𝜹 𝑖
𝛾𝐾 𝑖=1
Dual Receiver Perturbation System
With Reference Receiver
• Properties of perturbation sequence
• Orthogonality
• Odd symmetry
13
Adaptive Analog Beamforming Algorithms:
ii. dmr-based algorithm [3] :
𝑃 𝒘 ≝𝐸 𝒘𝐻 𝒙 𝑖 𝒙𝐻 𝑖 𝒘 = 𝒘𝐻 𝑹𝒘
𝛼 𝒘 ≝ 𝐸 𝑦 𝑖 𝑑∗ 𝑖 = 𝒘𝐻 𝒛
1
𝑃 𝒘
𝑚 = 𝐾 σ𝑚𝐾
𝑖= 𝑚−1 𝐾+1 𝑦 𝑖 𝑦∗ 𝑖 = 𝒘
𝐻
𝑚𝒓ො
1
𝛼ො 𝒘𝑚 = 𝐾 σ𝑚𝐾
𝑖= 𝑚−1 𝐾+1 𝑦(𝑖)𝑑
∗ (𝑖)
= 𝒘𝐻
𝑚𝒛ො
14
Adaptive Analog Beamforming Algorithms:
where,
≝ 𝒘𝐻 ⊗ 𝒘,
𝒘 ⊗ is the kronecker-product
𝐽 𝑀
𝑑𝑚𝑟 = 𝜆መ 𝑚 𝒒
𝑹 ෝ𝐻
ෝ𝑚 𝒒 ෝ𝐻
ෝ𝑚 𝒒
𝑚+ 𝒒 𝑚 𝛼𝑑𝑚𝑟
𝑚=1 𝑚=𝐽+1
𝐽 𝑀
1 1
−1
𝑹 𝑑𝑚𝑟 = 𝒒 ෝ𝐻
ෝ𝑚 𝒒 𝑚 + 𝒒 ෝ𝐻
ෝ𝑚 𝒒 𝑚
𝜆መ 𝑚 𝛼𝑑𝑚𝑟
𝑚=1 𝑚=𝐽+1
15
Modified Structure:
𝒓 = (𝑾
෪𝐻 )−1 𝑷
Dotted lines in the matrices show 2 ADCs are used, each samples half of the array.
16
Extended Algorithm:
(𝑢)
𝑃𝑢 (𝒘𝑚 ) = E 𝑦𝑢 𝑖 𝑦𝑢∗ 𝑖 ,
𝑚𝐾
1
= | 𝑦𝑢 𝑖 𝑦𝑢∗ 𝑖 | ,
𝐾
𝑘= 𝑚−1 𝐾+1
𝑚𝐾
1 𝑢 𝐻 (𝑢)
= | 𝒘𝑚 𝒙𝑢 𝑖 × 𝒙𝐻
𝑢 (𝑖) 𝒘𝑚 |,
𝐾
𝑘= 𝑚−1 𝐾+1
where,
𝑢 = 1, … , # of ADCs, 𝑚 = 1, … , (𝑀Τ𝐷)2 for each 𝑢,
𝒘(𝑢) is the 𝑀Τ𝐷 long sub-vector (𝑤𝑏 , … , 𝑤𝑗(𝑢) ) of w with 𝑏𝑢 , 𝑗𝑢 being the indices of
𝑢
the first and last element, respectively of 𝑢𝑡ℎ ADC, i.e., 𝒘 = [0, … , 0, 𝒘 𝑢 , 0, … , 0]𝑇
{
𝑏𝑢 − 1
17
Extended Algorithm:
where,
𝑢 = 1, … , # of ADCs, 𝑐 = 𝑀 Τ𝐷 − 1 , 𝑧𝑢 = (𝑢 − 1) 𝑀2 Τ𝐷 + 𝑐
𝑃1 (𝒘1 ) 𝑃1 (𝒘7 ) 𝑃1 (𝒘13 ) 𝑃2 (𝒘22 ) 𝑃2 (𝒘28 ) 𝑃2 (𝒘28 )
∆1 = 𝑃1 (𝒘2 ) 𝑃1 (𝒘8 ) 𝑃1 (𝒘14 ) ∆2 = 𝑃2 (𝒘23 ) 𝑃2 (𝒘29 ) 𝑃2 (𝒘29 )
𝑃1 (𝒘3 ) 𝑃1 (𝒘9 ) 𝑃1 (𝒘15 ) 𝑃2 (𝒘24 ) 𝑃2 (𝒘30 ) 𝑃2 (𝒘36 )
2
Since the 𝑀2 Τ𝐷 estimates of main-diagonal partitions reuse (𝑀Τ𝐷) weights D
times, therefore the weight structure of 1st estimates of each main-diagonal
partition, i.e., 𝒘1 and 𝒘22 can be given as
𝒘1= [𝑤𝑎 , 𝑤𝑏 , 𝑤𝑐 , 0,0,0]𝑇 𝒘22= [0,0,0, 𝑤𝑎 , 𝑤𝑏 , 𝑤𝑐 ]𝑇
18
Extended Algorithm:
where,
𝑢, 𝑣 = 1, … , # of ADCs s. t. u ≠ 𝑣 𝑚 = 1, … , 𝑀2 Τ𝐷 ,
19
Extended Algorithm:
where,
𝑢, 𝑣 = 1, … , # of ADCs, 𝑧𝑢 = (𝑢 − 1) (𝑀𝐷 + 𝑀Τ𝐷 − 1) + (𝑀/𝐷)
𝑃1,2 (𝒘4 ) 𝑃1,2 (𝒘10 ) 𝑃1,2 (𝒘16 ) 𝑃2,1 (𝒘19 ) 𝑃2,1 (𝒘25 ) 𝑃2,1 (𝒘31 )
∆1,2 = 𝑃1,2 (𝒘5 ) 𝑃1,2 (𝒘11 ) 𝑃1,2 (𝒘17 ) ∆2,1 = 𝑃2,1 (𝒘20 ) 𝑃2,1 (𝒘26 ) 𝑃2,1 (𝒘32 )
𝑃1,2 (𝒘6 ) 𝑃1,2 (𝒘12 ) 𝑃1,2 (𝒘18 ) 𝑃2,1 (𝒘21 ) 𝑃2,1 (𝒘27 ) 𝑃2,1 (𝒘33 )
20
Extended Algorithm:
In case, if single antenna-element is left out, say 𝑎 𝑡ℎ element, then it will be directly
input to 𝑑 𝑡ℎ ADC and (2𝑀 − 1) mean-power estimates out of 𝑀2 (𝐷 − 1)Τ𝐷 can be
found by
(𝑑)
𝑃𝑑 (𝑤𝑚 ) = E 𝑦𝑑 𝑖 𝑦𝑑∗ (𝑖) ,
𝑚𝐾
1
= | (𝑦𝑑 𝑖 |2 ,
𝐾
𝑘= 𝑚−1 𝐾+1
𝑚𝐾
1 𝑑 ∗
= |𝑤𝑚 𝑥𝑑 𝑖 × 𝑥𝑑∗ 𝑖 𝑤𝑚 |2 ,
𝐾
𝑘= 𝑚−1 𝐾+1
where,
𝑎=𝑀, 𝑑 = 𝐷, 𝑚 = 1, … , 2M − 1
21
Extended Algorithm:
Mean-power estimates of the 𝑎𝑡ℎ element of 𝑑 𝑡ℎ ADC in vector form can be written as
T
𝑃3 𝒘7 , 𝑃3 𝒘14 , 𝑃3 𝒘21 , 𝑃3 𝒘28 , 𝑃3 𝒘35 , 𝑃3 𝒘42 , 𝑃3 𝒘43 , …
∆3 =
…𝑃3 𝒘44 , 𝑃3 𝒘45 , 𝑃3 𝒘46 , 𝑃3 𝒘𝟒7 , 𝑃3 𝒘48 , 𝑃3 𝒘29
Thus, mean-power matrix for the case of 6 antenna-elements and 2 ADCs can be
given as
∆𝑢 𝑢=1 ∆𝑢,𝑣 𝑢 = 2, 𝑣 = 1
∆𝑢,𝑣 𝑢 = 1, 𝑣 = 2 ∆𝑢 𝑢=2
22
Extended Algorithm:
Cross-correlation values are computed separately for each ADC and then estimates
are concatenated for all M elements, i.e., 𝒛 = [𝒛1 , … , 𝒛 𝐷 ] 𝑇
(𝑢)
𝛼ො 𝑢 𝒘𝑚 = 𝐸 [𝑦𝑢 (𝑖)𝑑 ∗ (𝑖)],
𝑢 𝐻
= 𝐸 𝒘𝑚 𝒙𝑢 (𝑖)𝑑 ∗ 𝑖 ,
𝑚𝐾
1 𝑢 𝐻
= 𝒘𝑚 𝒙𝑢 𝑖 𝑑 ∗ 𝑖 ,
𝐾
𝑘= 𝑚−1 𝐾+1
𝑢 𝑢 𝐻 𝑢
𝒛 𝑢 = (𝑾𝑚 𝑾𝑚 )−1 𝑾𝑚 𝜶
ෝ𝑢
where,
𝑢 = 1, … , # of ADCs, 𝑚 = 1, … , 𝑀2
23
Extended Algorithm:
(𝑢)
𝒘+ 𝒘(𝑢) , 𝑖 = 𝒘(𝑢) + γ𝜹(𝑢) 𝑖 ,
𝐾 𝐾
′
1 1
𝒈1 𝒘 = 𝑑 𝑖 −𝑦 𝑖 𝜹 𝑖 = | 𝑑 𝑖 − 𝒘𝐻
+ 𝒙(𝑖)|𝜹(𝑖) ,
𝛾𝐾 𝛾𝐾
𝑖=1 𝑖=1
𝒔0
𝒘 𝑖 + 1 = 𝑷𝟎 𝒘 𝑖 − 𝛼𝒈 𝒘 𝑖 +
𝑀
24
Results :
Performance parameter
• switching behavior of analog weights ~ 𝑇𝑤
• Quantization behavior ~ B-bit phase shifter & amplitude control element
step-size 𝛼= 0.0025.
25
Results :
26
Results :
27
Results :
28
Results :
29
Results :
30
Results :
31
Results :
32
Conclusion :
Algorithm in [3] overcomes the need for these requirements with even better
performance which makes it, a cost effective and promising candidate for high data
rate applications.
33
References
34