Reconfigurable Intelligent Surfaces:

A Signal Processing Perspective

Emil Björnson

Visiting professor, KTH, Sweden

Associate professor, Linköping
University, Sweden
 Outline

Introduction
• Reconfigurable intelligent surface (RIS)
• Vision for how to improve wireless communications

Developing a system model

• Basic signals and systems theory
• Application to model RIS systems
• Optimization of RIS for communication

Spatial channel structure

• Reduce number of parameters to learn

Summary
2
 INTRODUCTION

3
 Physics of Wireless Signal Propagation

• Electromagnetic travel at speed of light

• Spreads out in all directions

• Friis’ propagation formula:

𝐴
Receive power = Transmit power 1
4𝜋𝑑 !
" !
Example: 𝐴 = , 𝜆 = 0.1 m (3 GHz)
#

0.005% received at 1 m (−43 dB)

0.00005% received at 10 m (−63 dB)

4
Only a tiny fraction of transmit power is received!
 No Direct Path: Even Larger Propagation Losses

Wall penetration:
− 20 dB or more Extra paths

5
 Shaping the Signal Scattering Towards the Receiver
Other names:
Intelligent reflecting surface
Software-controlled metasurface
Reconfigurable
intelligent surface
(RIS)

Reconfigurable: Properties can be changed

Intelligent: Real-time programmable/controllable
Surface: Two-dimensional array of scatterers
6
 Reconfigurable Intelligent Surface (RIS)
One element User 2

Passive patch

Switch
(e.g., diod)

Programmable
controller

Pattern of impedances
Approximate shape of another object
Sub-wavelength-sized elements
7
Transmitter User 1
 Vision: Controllable propagation Mitigate shadow fading

RIS as a whole can control

• Directivity of scattered signal
• Signal absorption
• Change polarization

Improved indoor coverage

8
 DEVELOPING A SYSTEM MODEL

9
 Introduction to Signals and Systems

Linear-time invariant: $#
𝑥!" (𝑡) Impulse response 𝑦!" (𝑡) = ℎ!" ∗ 𝑥!" 𝑡 = 0 ℎ!" 𝑢 𝑥!" 𝑡 − 𝑢 𝑑𝑢
ℎ"! (𝑡) #
𝑋!" 𝑓
𝑌!" (𝑓)

• Communication channels are systems/filters: 0 𝑓) 𝑓

𝐵
𝑥!" (𝑡)

𝑦!" 𝑡 = 𝜌% 𝑥!" 𝑡 − 𝜏% + 𝜌& 𝑥!" 𝑡 − 𝜏& + 𝜌' 𝑥!" 𝑡 − 𝜏'

Impulse response: ℎ"! 𝑡 = 𝜌% 𝛿 𝑡 − 𝜏% + 𝜌& 𝛿 𝑡 − 𝜏& + 𝜌' 𝛿 𝑡 − 𝜏'

10
 Complex Baseband Representation

• Communication theory is developed for the baseband

Real passband Complex baseband

𝑋!" 𝑓 𝑋 𝑓
𝑌!" (𝑓) 𝑌(𝑓)
Downshifting

0 𝑓) 𝑓 0 𝑓) 𝑓

𝐵 𝐵/2
# $%$( &# ∗ %$%$( ( $%$( &( ∗ %$%$(
• Connection: 𝑋!" 𝑓 = , 𝑌!" 𝑓 =
' '

Linear-time invariant: $#
𝑥(𝑡) Impulse response 𝑦(𝑡) = 0 ℎ 𝑢 𝑥 𝑡 − 𝑢 𝑑𝑢
#
ℎ(𝑡)

11
Downshifted channel: ℎ 𝑡 = ℎ89 𝑡 𝑒 :;!<=!>
 Analyzing Reconfigurable Intelligent Surface
End-to-end impulse response:
)

ℎ"# 𝑡 = ℎ$,"# 𝑡 + ( 𝑏&,"# ∗ 𝜗&,"#;+! ∗ 𝑎&,"# (𝑡)

&'(

Conventional channel models

Controlled by RIS using 𝜃% , … , 𝜃*

12
 Varactor diods
How Will an RIS Element Filter the Signal?
Prototype for 5.8 GHz band Example: Patch with bias voltage 𝑉
Reflection coefficient:
𝑍+ 𝑉 − 𝑍,
Dielectric
20 rows 𝑍+ 𝑉 + 𝑍, substrates

Reference: X. Pei, H. Yin, L. Tan, L. Cao, Z. Li, K. Wang, K. Zhang, E.

Björnson, “RIS-Aided Wireless Communications: Prototyping, Adaptive
55 columns Beamforming, and Indoor/Outdoor Field Trials,” arXiv:2103.00534
Phase response [deg]

Amplitude response

13
Frequency [GHz] Frequency [GHz]
 How To Transmit Data?
2.5

• Pulse amplitude modulation: 2

𝑥 1 =2
𝑚 1.5
𝑥 𝑡 = '𝑥 𝑚 𝑝 𝑡 −
𝐵 1
) 𝑥 0 =1
0.5

• Transmit discrete sequence: 𝑥 𝑚 , 𝑚 = integer 0

Use a pulse-form 𝑝(𝑡) satisfying the Nyquist criterion: -0.5

𝑥 2 = −1
) -1
𝑝 = 0 for integer 𝑚 ≠ 0 and non-zero for 𝑚 = 0 -2 -1 0 1 2 3 4
*

• Example: 𝑝 𝑡 = 𝐵sinc 𝐵𝑡

Sampling of received signal 𝑦 𝑡 = 𝑥(𝑡):

𝑘 𝑘 𝑘−𝑚
𝑦 =𝑥 = I𝑥 𝑚 𝑝 = 𝑥[𝑘]
𝐵 𝐵 𝐵
G
14
 Reception with Channel and Noise

• Received signal (with Gaussian noise):

𝑦 𝑡 = ℎ ∗ 𝑥 𝑡 + 𝑤(𝑡)
1. Filter using 𝑝 𝑡 = 𝐵sinc 𝐵𝑡 :
𝑚
𝑧 𝑡 = 𝑝 ∗ 𝑦 𝑡 = I𝑥 𝑚 𝑝 ∗ ℎ ∗ 𝑝 𝑡− + 𝑝 ∗ 𝑤 (𝑡)
𝐵
G

2. Sample received signal:

𝑘 𝑘−𝑚 𝑘
𝑧 = (𝑥 𝑚 𝑝 ∗ ℎ ∗ 𝑝 + 𝑝∗𝑤
𝐵 𝐵 𝐵
,

Call it 𝑧[𝑘] Effective pulse Complex Gaussian

function noise 𝐶𝑁(0, 𝑁, )

Narrowband channel: ℎ ≈ constant C δ(𝑡 − 𝜏) in the band, Nyquist criterion satisfied

𝑧[𝑘] = constant C 𝑥 𝑘 + Gaussian noise
15
 Putting the Pieces Together: Narrowband Channels

• Direct channel: ℎ+,!" 𝑡 = 𝜌𝛿 𝑡 − 𝜏+ → ℎ+ 𝑡 = 𝜌𝑒 %-'.$( / 𝛿 𝑡 − 𝜏+

• Related to element 𝑛: 𝑎0,!" 𝑡 = 𝛼0 𝛿 𝑡 − 𝜏0,1 → 𝑎0 𝑡 = 𝛼0 𝑒 %-'.$( / 𝛿 𝑡 − 𝜏0,1

𝜗0,!";+! 𝑡 = 𝛾0 𝛿 𝑡 − 𝜏3- → 𝜗0;+! 𝑡 = 𝛾0 𝑒 %-'.$( / 𝛿 𝑡 − 𝜏3-

𝑏0,!" 𝑡 = 𝛽0 𝛿 𝑡 − 𝜏0," → 𝑏0 𝑡 = 𝛽0 𝑒 %-'.$( / 𝛿 𝑡 − 𝜏0,"

End-to-end discrete-time system model:

𝑧𝑘
R
= 𝜌𝑒 :;!<=!N- + I 𝛼O 𝛽O 𝛾O 𝑒 :;!<=! N.,/SN0. SN.,1
𝑥[𝑘] + Noise
OPQ

Direct path Joint amplitude loss Joint delay

16
Tunable!
 OPTIMIZING COMMUNICATION
PERFORMANCE

17
 Maximizing Performance Without a Direct Path

Received signal without direct path:

R
𝑦 = I 𝛼O 𝛽O 𝛾O 𝑒 :;!<=! N.,/SN0. SN.,1
⋅ signal + noise
OPQ

Signal processing problem: Minimum positive delay solution:

Maximize the signal-to-noise ratio 𝜏+! = max 𝜏,,2 + 𝜏,,# − 𝜏&,2 + 𝜏&,#
,

$ ' $ '

! 𝛼! 𝛽! 𝛾! 𝑒 %&'()4 *5,6+*75 +*5,8

≤ ! 𝛼! 𝛽! 𝛾! ≈ 𝑁 ' 𝛼𝛽𝛾
!"# !"#

Cauchy–Schwarz inequality
(𝜏&,2 + 𝜏+! + 𝜏&,# = constant)
18
Example: Synthesizing Surface Shapes

Beamforming:
Toward point/direction

1: Normal reflection 2: Signal focusing

19
 Maximizing Performance With a Direct Path

Received signal with direct path:

*
𝑦= 𝜌𝑒 !"#$%!&- + I 𝛼' 𝛽' 𝛾' 𝑒 !"#$%! &.,/+&0. +&.,1
⋅ signal + noise
'()

Maximize channel gain:

$ ' $ '

𝜌𝑒 %&'()4*9 + ! 𝛼! 𝛽! 𝛾! 𝑒 %&'()4 *5,6+*75 +*5,8

≤ 𝜌 + ! 𝛼! 𝛽! 𝛾!
!"# !"#

Minimum positive delay solution:

Achieved when:
𝜏',- + 𝜏.. + 𝜏',/ = 𝜏0 integer
𝜏+! = 𝜏$ − 𝜏&,2 + 𝜏&,# +
𝑓3
20
 Basic Performance Benefit
Case 1: −100 dB
8 Case 2: −75 dB
Transmitter Receiver
Case 2

6 −75 dB −75 dB
RIS

4 Transmit power: 10 mW per 20 MHz

Case 1
0
0 200 400 600 800 1000

RIS is particularly helpful

21 when direct path is relatively weak
 Experimental Validation Transmitter
Freq: 5.8 GHz
BW: 20 MHz
Power: 13 dBm

Baseline: −66 dBm

TX
antenna

RX
antenna
50 m

USRP

RIS: −39 dBm

RIS
Reference: X. Pei, H. Yin, L. Tan, L. Cao, Z. Li, K. Wang, K. Zhang, E.
Björnson, “RIS-Aided Wireless Communications: Prototyping, Adaptive
Beamforming, and Indoor/Outdoor Field Trials,” arXiv:2103.00534
22 Baseline: Copper plate
 Reconfigurability is Complicated?
The RIS is blind!

2b) Solve optimization problem

23
 SPATIAL CHANNEL STRUCTURE

24
Element 𝑛 at
Array Response Vector
location 𝒖' ∈ ℝ6
Incoming/outgoing plane wave determined by
• Azimuth angle 𝜑 ∈ −𝜋, 𝜋
. .
• Elevation angle 𝜃 ∈ − ' , '

Channel vector for one plane wave

constant 1 𝑎 𝜑, 𝜃

𝑒 4𝒌 6,+ " 𝒖# cos 𝜃 cos 𝜑

𝑎 E.
𝜑,Björnson,
𝜃 = L. Sanguinetti, , 𝒌“Rayleigh
𝜑, 𝜃 = Fading cosModeling
𝜃 sin 𝜑and
89
⋮ for Reconfigurable
Channel Hardening :Intelligent Surfaces,”
6,+ " 𝒖$ sin 𝜃
25 𝑒 4𝒌 Communications
IEEE Wireless Letters, To appear.
 Sparse Multipath Channel
RIS with 𝑁 elements

Channel vector for L plane wave

8
,𝜃
Wave from 𝜑( ( 𝒉 = I 𝑐7 1 𝑎 𝜑7 , 𝜃7
7()

Wave
Number of parameters
from
𝜑8 , 𝜃 𝑁 complex parameters in 𝒉
8
or 3𝐿 real parameters

Wave from 𝜑; , 𝜃;

Sparse channels
𝐿 is small or one path is much stronger than all other
26
 Sparse Channels with RIS
Recall the channel gain:
Element-wise multiplication
*
𝜌𝑒 !"#$%!&- + I 𝛼' 𝛽' 𝛾' 𝑒 !"#$%! &.,/+&0. +&.,1

'()
9
𝛽)𝑒 !"#$%!&<,1 𝛼) 𝑒 !"#$%!&<,/ 𝛾)𝑒 !"#$%!&0<
= 𝜌𝑒 !"#$%!&- + ⋮ ⨀ ⋮ ⋮
𝛽* 𝑒 !"#$%!&=,1 𝛼* 𝑒 !"#$%!&=,/ 𝛾* 𝑒 !"#$%!&0=

Sparse channel from RIS? Sparse channel to RIS? Controllable vector

Unknowns without sparsity: 𝑁 + 1

Unknowns with sparsity: Much fewer?

Slow variations for adjacent elements:

27
Similar phase shifts in RIS
 Reconfigure RIS by Exploiting Channel Sparsity
Basic configuration algorithm

20 rows

55 columns

Two states per element: ±90 degrees

• Flip one row or column, pick best option

• Sweep horizontally, then vertically

2b) Solve optimization problem

28
 SUMMARY

29
Reconfigurable Intelligent Surfaces for Wireless Communications

Use RIS technology to:

• Control propagation environment
• Example: Improve signal-to-noise ratio
A signal processing problem
• Learn propagation channels
• Determine how to configure the RIS

Speculations about the future

• A cost and energy efficient relay?
• Suitable when there is sparsity
• “Easier” to channel estimate
• Sensitive propagation
• Particularly useful above 100 GHz?
30
 Podcast:

YouTube Videos

31
 Key References
Overview papers
1. E. Björnson, H. Wymeersch, B. Matthiesen, P. Popovski, L. Sanguinetti, E. de Carvalho, “A
Signal Processing Perspective on Reconfigurable Intelligent Surfaces With Wireless
Applications”, Available on arXiv:2102.00742.
2. E. Björnson, Ö. Özdogan, E. G. Larsson, “Reconfigurable Intelligent Surfaces: Three Myths
and Two Critical Questions,” IEEE Communications Magazine, 2020.

Channel modeling
3. Ö. Özdogan, E. Björnson, E. G. Larsson, “Intelligent Reflecting Surfaces: Physics,
Propagation, and Pathloss Modeling,” IEEE Wireless Commun. Letters, 2020.
4. E. Björnson, L. Sanguinetti, “Power Scaling Laws and Near-Field Behaviors of Massive
MIMO and Intelligent Reflecting Surfaces,” IEEE O. J. Commun. Soc. 2020.
Prototyping and field trials
5. X. Pei, H. Yin, L. Tan, L. Cao, Z. Li, K. Wang, K. Zhang, E. Björnson, “RIS-Aided Wireless
Communications: Prototyping, Adaptive Beamforming, and Indoor/Outdoor Field Trials,”
Available on arXiv:2102.00742

32 Questions?