Digitally Enhanced Inter-modulation Distortion Compensation in Wideband Spectrum Sensing. Han Yan and Prof. Danijela Cabric Nov.

Transcription

1 Digitally Enhanced Inter-modulation Distortion Compensation in Wideband Spectrum Sensing Han Yan and Prof. Danijela Cabric Nov.9 th 016 1

2 Challenges of Wideband Spectrum Sensing Rx Signal LNA LO Front-end LPF AGC ADC PSD Estimation DSP Thresholding Decision IMD Interference in Front-end PSD with Ideal LNA PSD with Practical LNA PSD (dbm/hz) Interferers PSD (dbm/hz) IMD Sensing threshold Freq. (MHz) Freq. (MHz) Wideband spectrum sensing Increases chances of finding unoccupied bands Vulnerable to imperfection in the front-end IMD: inter-modulation distortion D. Markovic / Slide

3 Addressing IMD in Wideband Sensing Rx Signal LNA LO Front-end LPF AGC ADC PSD Estimation DSP Thresholding Decision Reconfigurable RF Filtering Current-mode Amplifier Mixer-first Architecture DSP compensation Various ways to address IMD in spectrum sensing RF solutions: increases front-end cost & thermal noise Digital solutions A. Ghaffari, E. A. M. Klumperink and B. Nauta, "Tunable N-Path Notch Filters for Blocker Suppression: Modeling and Verification," in IEEE Journal of Solid-State Circuits, vol. 48, no. 6, pp , June 01. D. Murphy et al., "A Blocker-Tolerant, Noise-Cancelling Receiver Suitable for Wideband Wireless Applications," in IEEE Journal of Solid-State Circuits, vol. 47, no. 1, pp , Dec. 01. C. Andrews and A. C. Molnar, "A Passive Mixer-First Receiver With Digitally Controlled and Widely Tunable RF Interface," in IEEE Journal of Solid-State Circuits, vol. 45, no. 1, pp , Dec D. Markovic / Slide

4 Outline System Model RF Front-end Nonlinearity Model State-of-the-Art: LMS Based Compensation Performance Analysis of LMS Based Method Compensation via Iterative IMD Reconstruction Simulation Results Conclusions & Future Works D. Markovic / Slide 4 4

5 RF Front-end Nonlinearity Model RF Signal LNA Non-ideal Front-end LO ADC LPF AGC DSP Decision Baseband Signal rd order AM-AM model of LNA y t x t x t v t ( ) = 1 ( ) + ( ) + ( ) RF signal at 1: PSD N j fbit ( ) = e zi ( t) e i= 0 x t Focus on three components f = f f b0 b b1 Spectra in z () t z () t 1 SOI z () t 0 PSD Signal at : Spectra in y( t) ( ) j fbit = e 1 zi t e i= 0 + e ( ) ( ) + ( ) j f z0 t z1 t z t e * j ( fb1 fb) t + e z1 ( t) z( t) e + e + + e z( t) z( t) + z ( t) + terms at other freq. + vt ( ) b0 j fb1t z1( t) z1( t) z( t) e j f 1 e t t b f b1 f b f b0 f b1 f b f b0 D. Markovic / Slide 5 5

6 Equivalent Baseband Model RF Signal LNA Non-ideal Front-end LO ADC LPF AGC DSP Decision Baseband Signal rd order AM-AM model of LNA y t x t x t v t ( ) = 1 ( ) + ( ) + ( ) RF signal at (equivalent baseband model): Signal at f 0 * ( ) Dominant IMD z0( t) = 1z0 ( t) + z0( t) z1( t) + z( t) + z ( t) z1 ( t) + v0( t) PSD Spectra in Signals at f 1 and f z () t z () t 1 ( ) ( ) z1( t) = 1z1( t) + z1( t) z1( t) + z ( t) z( t) = 1z ( t) + z( t) z( t) + z1( t) f 1 f f 0 z () t 0 D. Markovic / Slide 6 6

7 LMS Based Digital Compensation LMS based Compensation: Digital IMD reconstruction zƹ IMD n = β z 1 ǁ [n] zǁ [n] 1 LMS filter based on instantaneous samples w n = w n 1 + μ Ƹ n e[n] z IMD Error signal in adaptation e n = zǁ 0 n w[n 1] zƹ IMD n Output of ADC BPF@f1 z[ n] 1 z [ n] z [ n] 0 (.) (.) * zˆ IMD[ n] w[n] en ][ ] e[n] E. Rebeiz, A. Shahed Hagh Ghadam, M. Valkama and D. Cabric, "Spectrum Sensing Under RF Non-Linearities: Performance Analysis and DSP-Enhanced Receivers," in IEEE Transactions on Signal Processing, vol. 6, no. 8, pp , April, 015. D. Markovic / Slide 7 7

8 LMS Based Digital Compensation LMS based Compensation: Digital IMD reconstruction zƹ IMD n = β z 1 ǁ [n] zǁ [n] 1 LMS filter based on instantaneous samples w n = w n 1 + μ Ƹ n e[n] z IMD Output of ADC BPF@f1 z[ n] 1 z [ n] (.) (.) * + zˆ IMD[ n] w[n] - Error signal in adaptation e n = zǁ 0 n w[n 1] zƹ IMD n + - E. Rebeiz, A. Shahed Hagh Ghadam, M. Valkama and D. Cabric, "Spectrum Sensing Under RF Non-Linearities: Performance Analysis and DSP-Enhanced Receivers," in IEEE Transactions on Signal Processing, vol. 6, no. 8, pp , April, 015. D. Markovic / Slide 8 8

9 Outline System Model Performance Analysis of LMS Based Method Compensation via Iterative IMD Reconstruction Simulation Results Conclusions D. Markovic / Slide 9 9

10 Analysis: LMS based Compensation Linear model ( ) z [ n] = w [ n] z [ n] z [ n] + z [ n] + z [ n] + z [ n] + v [ n] 0 * vn [ ] zˆ [ n] IMD where 0 w [ n 1 1 ] = ( z n z1 n ) ( z1 n z n ) 1 + [ ] + [ ] 1 [ ] [ ] The residual IMD power in steady-state LMS ( 1 z v ) Power of SOI and AWGN 1 IMD q = IMD Power of IMD 1 1. Exact expression are written as func. of nd 4 th order statistic of z 1 [n] and z [n]. Reference: Ali H. Sayed, Adaptive filters, John Wiley & Sons, 011. D. Markovic / Slide 10 10

11 Validation: LMS Based Compensation Wideband: 0MHz BW, OFDM & MHz BW, 4-QAM Narrowband: MHz BW, 4-QAM MHz BW, 4-QAM Simulation settings LNA Gain 5 db (β 1 = 56.) LNA IIP LNA NF ADC Rate SOI Power -10 dbm (β = 7497.) 4 db 00 MHz 0 db SNR Without compensation Interferers with higher BW result in higher residual IMD power D. Markovic / Slide 11 11

12 Outline System Model Performance Analysis of LMS Based Method Compensation via Iterative IMD Reconstruction Simulation Results Conclusions D. Markovic / Slide 1 1

13 Proposed Method Sample-wise IMD reconstruction: Solve z 1 n, z n from ( ) ( ) z1[ n] = 1z1[ n] + z1[ n] z1[ n] + z [ n] z[ n] = 1z[ n] + z[ n] z[ n] + z1[ n] Output of ADC BPF@f1 z () t 1 z () t IMD COMP. + - Compute IMD based on z 1 [n] and z [n] Subtract from received signal D. Markovic / Slide 1 1

14 Iterative Computation of IMD Only amplitude matters in the previous equations (0) (0) Initialization: solve for z e,1 and ze, Iterations: ze,1 = z + z z + z ze, z z z z = + + z = z + z z = z + z Error evaluation step = z 1z z + z + z = z z z + z + z ( ) 1 e,1 e,1 e,1 e, ( ) 1 e, e, e,1 e, ( ) (0) (0) e,1 1 e,1 e,1 ( ) (0) (0) e, 1 e, e, ( ) ( ) ( i) ( i 1) ( i 1) ( i 1) ( i 1) 1 e,1 e,1 e,1 e,1 e, ( ) ( ) ( i) ( i 1) ( i 1) ( i 1) ( i 1) e, 1 e,1 e, e,1 e, z z Ignore them updating step = z + ( i) ( i 1) ( i) e,1 e,1 1 = z + ( i) ( i 1) ( i) e, e, D. Markovic / Slide 14 14

15 Outline System Model Performance Analysis of LMS Based Method Compensation via Iterative IMD Reconstruction Simulation Results Conclusions D. Markovic / Slide 15 15

16 Residual IMD power v.s. Iteration Number -5dBm Interferers -0dBm Interferers -5dBm Interferers -0dBm Interferers Few iterations if interferers power is not very strong IMD from interferers with wider BW and stronger power requires more iterations D. Markovic / Slide 16 16

17 Performance Comparison: Measured PSD Interferer #1, : MHz BW, 4-QAM Interferer #1: 0MHz BW, OFDM Interferer #: MHz BW, 4-QAM Interferer Power (after LNA) Simulation settings - dbm LNA Gain 5 db (β 1 = 56.) LNA IIP LNA NF ADC Rate LMS TRN Proposed Ite. -10 dbm (β = 7497.) 4 db 00 MHz 1,000 Samples 10 Iterations FFT size 819 D. Markovic / Slide 17 17

18 Performance Comparison: Residual IMD LMS based Compensation Proposed method The proposed method significantly reduces residual IMD power Without compensation D. Markovic / Slide 18 18

19 Spectrum Sensing Performance Proposed method Spectrum sensing settings: Energy detection with 500 Samples Threshold set for P FA = 0.05 LMS based method Computational complexity 1 : Without compensation Interferers Power Ave. Operation per Sample LMS Proposed -5 dbm dbm dbm dbm LMS uses 1,000 samples training beforehand, proposed method uses minimum number of iterations for P D = 0.9 D. Markovic / Slide 19 19

20 Summary & Future Works Analyzed performance of LMS-based digital compensation Explained the degradation with wideband interferers Proposed a novel compensation method Reduces IMD power Provides higher Pd in spectrum sensing Future research works Study compensation sensitivity due to estimation error of β 1 and β Study potential memory effect of wideband LNA D. Markovic / Slide 0 0

21 References A. Ghaffari, E. A. M. Klumperink and B. Nauta, "Tunable N-Path Notch Filters for Blocker Suppression: Modeling and Verification," in IEEE Journal of Solid-State Circuits, vol. 48, no. 6, pp , June 01. D. Murphy et al., "A Blocker-Tolerant, Noise-Cancelling Receiver Suitable for Wideband Wireless Applications," in IEEE Journal of Solid-State Circuits, vol. 47, no. 1, pp , Dec. 01. C. Andrews and A. C. Molnar, "A Passive Mixer-First Receiver With Digitally Controlled and Widely Tunable RF Interface," in IEEE Journal of Solid-State Circuits, vol. 45, no. 1, pp , Dec E. Rebeiz, A. Shahed Hagh Ghadam, M. Valkama and D. Cabric, "Spectrum Sensing Under RF Non-Linearities: Performance Analysis and DSP-Enhanced Receivers," in IEEE Transactions on Signal Processing, vol. 6, no. 8, pp , April15, 015. Sayed, Ali H. Adaptive filters. John Wiley & Sons, 011. D. Markovic / Slide 1 1

22 Thanks! Question?