Sep 17, 2014 - Implementation of cognitive channelization on a GNU Radio/USRP framework. ... Denote cumulative interference as pi + ni . If H1 is known then ...
An All-GNU Radio Software-defined Radio Transceiver for All-Spectrum Cognitive Channelization George Sklivanitis,† Emrecan Demirors,‡ , Adam Gannon,† Dimitris A. Pados,† Stella N. Batalama,‡ Tommaso Melodia,† and John D. Matyjas? † SUNY ‡ Northeastern
at Buffalo, Department of Electrical Engineering University, Department of Electrical and Computer Engineering ? Air Force Research Laboratory/RGIF, Rome, NY
September 17, 2014
State University of New York at Buffalo
1 / 26
Outline
Motivation Basic Idea Testbed Architecture & Design Experimental Results Conclusions
State University of New York at Buffalo
2 / 26
Motivation
Highly occupied spectrum bands + Exponential growth in traffic. Underutilization of the device’s available/accessible bandwidth. Practical co-existence of cognitive secondary and primary stations. Hardware radios are application specific. Innovation comes from PHY. Need for reconfigurable, agile, intelligently-flexible autonomous radios. Key question Are we efficiently utilizing the available spectrum resources?
State University of New York at Buffalo
3 / 26
Cognitive Radio Principles
Primary/Secondary user setup. SU transmissions over gray or white spaces (underlay, overlay, interweave). Satisfy QoS constraints at the PT.
State University of New York at Buffalo
4 / 26
Outline
Motivation Basic Idea Testbed Architecture & Design Experimental Results Conclusions
State University of New York at Buffalo
5 / 26
In this presentation ...
4 Implementation of cognitive channelization on a GNU Radio/USRP framework. SU and PU coexist in both frequency and time. (grey spaces transmissions). SU utilizes a code channel that exhibits minimum interference with PU.
4 Technical implementation challenges of real-time reconfigurability for channelization (code-domain).
State University of New York at Buffalo
6 / 26
System Setup
PT
PR H3
GigE
GigE
H4
H2
ST GigE
H1
SR GigE
H Feedback
Figure : Primary transmitter-receiver PT/PR and secondary transmitter/receiver ST/SR pairs. All signals propagate over independent multipath Rayleigh fading channels.
State University of New York at Buffalo
7 / 26
Problem Formulation - Signal Model PU/SU transmitted signal: xk (t) =
J−1 X
p bk (i) Ek dk (t − iT )e j(2πfc t+φk ) ,
k = 1, 2 for PU/SU.
i=0
bk (i) ∈ {±1}, binary antipodal information symbols. k = 1, 2 for primary/secondary user respectively. Ek : transmitted energy per bit. P dk (t) = L−1 l=0 sk (l)gT (t − lTd ),
where
sk (l) ∈
√1 {±1}. L
gT (·): SRRC pulse-shaping filter. φk : carrier phase relative to the carrier frequency fc .
State University of New York at Buffalo
8 / 26
Problem Formulation - Signal Model (cnt’d) Received baseband signal after carrier demodulation: r (t) =
J−1 X 2 X
bk (i)
i=0 k=1 N−1 X
0 hk,n dk (t − iT − nTd − τk )e −j(2π∆fk t) + n(t),
×
k = 1, 2
n=0 0 where hk,n =
√
Ek hk,n e −j(2πfc nTd )+γk , and γk = 2πfc τk − φk .
hk,n : independent zero-mean complex Gaussian channel coefficients. {∆fk }: carrier frequency offsets between any TX-RX pair. τk = κk Td : propagation delays w.r.t ST for κk ∈ {0, 1, . . . L − 1}. n(t): CWGN. State University of New York at Buffalo
9 / 26
Reconfigurable Channelization: Optimal Waveform Design Denote secondary user’s signal of interest as b1 H1 s1 . Denote cumulative interference as pi + ni . If H1 is known then, wmaxSINR = arg maxw
E{|wH (b1 H1 s1 )|2 } E{|wH (p + n)|2 }
= (Rp + σn2 IN+L−1 )−1 H1 s1 is the linear filter maximizing the SINR at the output of the SR. Let RI+N = Rp + σn2 IN+L−1 , then the maximum SINR attained is −1 SINRmax = s1 T HH 1 RI+N H1 s1 .
State University of New York at Buffalo
10 / 26
Reconfigurable Channelization: Optimal Waveform Design (cnt’d) Now consider SINRmax as a function of waveform s1 . Then, T H −1 sopt 1 = arg maxs1 s1 H1 RI+N H1 s1 maximizes the SINR at the output of the maximum SINR filter. Define
4 −1 e = M HH 1 RI+N H1 ,
where q1 , q2 , . . . , qL denote its eigenvectors with corresponding eigenvalues λ1 ≥ λ2 ≥ · · · ≥ λL . Then, sopt is the eigenvector that corresponds to the maximum 1 eigenvalue λ1 . State University of New York at Buffalo
11 / 26
Implementation Challenges Unknown chip timing CFO’s {∆fk }
I&D operation will lead to SNR loss.
No cooperation assumed between PU-SU. Maximal-SINR waveform design for SU (H1 and RI+N are unknown). Spread-spectrum receiver design. Frame Detection. CFO estimation/compensation. Symbol time synchronization. Maximum SINR RAKE filtering.
State University of New York at Buffalo
12 / 26
Outline
Motivation Basic Idea Testbed Architecture & Design Experimental Results Conclusions
State University of New York at Buffalo
13 / 26
Indoor Testbed Deployment
USRP N-210 + RFX-2400 daughtercards.
Transceiver 2 secondary receiver TX: fc
2 RX: f c1
Data channel at fc1 = 2.48GHz.
TX TX: fc
Control/feedback channel at fc2 = 2.42GHz.
State University of New York at Buffalo
1 RX: f c2
primary transmitter
Transceiver 1 secondary transmitter
14 / 26
Transmitter Design Message passing and stream tagging features exploited.
State University of New York at Buffalo
15 / 26
Transmitter Design (cnt’d)
Transmitter blocks: packet assembly blocks (e.g., packet header, stream CRC32). burst message generator (i.e., vector pdu). spreading block for modulating transmitter’s bits in a waveform.
Message passing blocks allowed us dynamic adaptation of the ST to the received optimal waveform. Feedback waveform is communicated by SR using already available GMSK modulation GNU Radio blocks.
State University of New York at Buffalo
16 / 26
Receiver Design
State University of New York at Buffalo
17 / 26
Receiver Design (cnt’d) Frame acquisition: Plateau detection based on unmodulated bits. CFO estimation/correction: T
c = ∆f
1 2πL TTds
(P−1)L Td −1
∠
Xs i=0
Td . r [i] r i + L Ts ∗
Channel estimation: PX AC −1 H −1 H 1 b h1 = (S1 S1 ) S1 yi b1∗ (P + i), PAC i=0
where yi = b1 (P + i)S1 h1 + pi + ni , i = 0, . . . , J − P − 1 and S1 is the channel-processed code matrix. Maximum SINR RAKE filtering: wRAKE−MVDR
b −1 S1 hb1 R = , (S1 hb1 )H R−1 S1 hb1 4
State University of New York at Buffalo
b= R
PAC −1 1 X yi yiH . PAC i=0
18 / 26
GNU Radio Technical Details: Stream Tags/Meta-data
Meta-data are used to tag steams of samples. Meta-data examples: tx sob, tx time, and tx eob ⇒ Burst data transmissions with precise timing. PMTs can carry arbitrary amount and type of information. Tags are associated with samples through an absolute counter.
State University of New York at Buffalo
19 / 26
GNU Radio Technical Details: Asynchronous Message Passing
Sample streams are unidirectional (downstream connections only). Messages can now be exchanged in both directions. Do not rely on buffers that operate synchronously between blocks. Queues used to pass messages between blocks.
State University of New York at Buffalo
20 / 26
Outline
Motivation Basic Idea Testbed Architecture & Design Experimental Results Conclusions
State University of New York at Buffalo
21 / 26
Cognitive channelization vs. fixed channelization Adaptive vs. fixed channelization compared with respect to postfiltering SINR 50
45
SINR in dB
40
primary user’s transmissions start
35
30
25
20 Experiment 1(adaptive channelization) Experiment 2(fixed channelization) 15 0
5
10
15
20
25
30
Time
Figure : Pre-detection SINR at the secondary receiver. State University of New York at Buffalo
22 / 26
Cognitive channelization vs. fixed channelization (cnt’d) Adaptive vs. fixed channelization compared with respect to Bit-Error-Rate
1
10
Bit- Error- Rate
Experiment 1(adaptive channelization) Experiment 2(fixed channelization)
2
10
primary user’s transmissions start
3
10
0
5
10
15
20
25
30
Time
Figure : Bit-error-rate at the secondary receiver. State University of New York at Buffalo
23 / 26
Outline
Motivation Basic Idea Testbed Architecture & Design Experimental Results Conclusions
State University of New York at Buffalo
24 / 26
Conclusions
Designed and implemented an SDR testbed for cognitive channelization evaluation. Implemented a multi-user, spread-spectrum receiver, operating in a multipath fading, indoor-lab environment. Demonstrated optimal waveform design and channelization benefits in a three-node deployment.
State University of New York at Buffalo
25 / 26
THANK YOU! Questions? Contact: {gsklivan, pados} buffalo.edu
State University of New York at Buffalo
26 / 26
