A Modular Pragmatic Architecture for Multiuser MIMO with Array-Fed RIS

Abstract

We propose a power- and hardware-efficient, pragmatic, modular, multiuser/multibeam
array-fed RIS architecture particularly suited to operate in very high frequency bands
(high mmWave and sub-THz), where channels are typically sparse in the beamspace and line-of-sight (LOS) is required to achieve an acceptable received signal level.
The key module is an active multi-antenna feeder (AMAF) with a small number of active antennas placed in the near field of a RIS with a much larger number of passive controllable reflecting elements. We propose a pragmatic approach to obtain a steerable beam with high gain and very low sidelobes. Then, K independently controlled beams can be achieved by stacking K of such AMAF-RIS modules. Our analysis takes in full account: 1) the near-end crosstalk (NEXT) between the modules, 2) the far-end crosstalk (FEXT) due to the sidelobes; 3) a thorough energy efficiency
comparison with respect to conventional active arrays with the same beamforming performance.
Overall, we show that the proposed architecture is very attractive in terms of
spectral efficiency, ease of implementation (hardware complexity), and energy efficiency.

Full text

A Modular Pragmatic Architecture for Multiuser
MIMO with Array-Fed RIS
Krishan K. Tiwari, Giuseppe Caire
Technische Universit¨
at Berlin, 10587 Berlin, Germany
Email addresses: lastname@tu-berlin.de
Abstract—We propose a power- and hardware-efficient, prag-
matic, modular, multiuser/multibeam array-fed RIS architecture
particularly suited to operate in very high frequency bands (high
mmWave and sub-THz), where channels are typically sparse in
the beamspace and line-of-sight (LOS) is required to achieve an
acceptable received signal level. The key module is an active
multi-antenna feeder (AMAF) with a small number of active
antennas placed in the near field of a RIS with a much larger
number of passive controllable reflecting elements. We propose
a pragmatic approach to obtain a steerable beam with high
gain and very low sidelobes. Then, Kindependently controlled
beams can be achieved by stacking Kof such AMAF-RIS
modules. Our analysis takes in full account: 1) the near-end
crosstalk (NEXT) between the modules, 2) the far-end crosstalk
(FEXT) due to the sidelobes; 3) a thorough energy efficiency
comparison with respect to conventional active arrays with the
same beamforming performance. Overall, we show that the
proposed architecture is very attractive in terms of spectral
efficiency, ease of implementation (hardware complexity), and
energy efficiency.
Index Terms—Reflective intelligent surface (RIS), reflectarrays,
multiuser MIMO, mmWave and sub-THz communications, line
of sight MIMO.
I. INT ROD UC TI ON
Wireless communication in the millimeter wave (mmWave)
and sub-THz frequency bands has garnered significant at-
tention recently due to the promise of high data rates and
ultra-low latency [1]. At these frequencies, traditional wide-
angle antennas and non-line-of-sight (NLOS) propagation are
inadequate, prompting the use of large aperture antenna arrays
for highly directional line-of-sight (LOS) propagation. This
enables applications like wireless fronthaul, fixed point-to-
multipoint wireless access (FWA), and LOS multiuser MIMO
with highly directional beams [2]. However, the complexity
and efficiency of large beamsteering active arrays remain
problematic. To address this, innovative antenna configurations
like Reflectarrays and Reflective Intelligent Surfaces (RIS)
have been explored [3]–[5]. RIS, in particular, has been studied
to modify wireless channels, but its effectiveness in the far
field is limited by signal strength unless the RIS size is
impractically large, even for indoor applications [6].
From the Friis transmission equation [7, eq. (1)],
Prx
Ptx
=AtxArx
(dλ)2,(1)
where λ,d,P, and Adenote the carrier wavelength, dis-
tance, power, and antenna (effective) aperture, respectively
and suffixes rx and tx denote receiver and transmitter. For
fixed antenna size and decreasing λ, extreme power efficiency
<latexit sha1_base64="3MUkKiCoiAzY5Y40sOOBicuGr2A=">AAAB7XicbVDLSgNBEOyNrxhfUY9eBoPgKeyKr2PQi8cI5gHJEnons8mY2Z1lZlYIS/7BiwdFvPo/3vwbJ8keNLGgoajqprsrSATXxnW/ncLK6tr6RnGztLW9s7tX3j9oapkqyhpUCqnaAWomeMwahhvB2oliGAWCtYLR7dRvPTGluYwfzDhhfoSDmIecorFSs4siGWKvXHGr7gxkmXg5qUCOeq/81e1LmkYsNlSg1h3PTYyfoTKcCjYpdVPNEqQjHLCOpTFGTPvZ7NoJObFKn4RS2YoNmam/JzKMtB5Hge2M0Az1ojcV//M6qQmv/YzHSWpYTOeLwlQQI8n0ddLnilEjxpYgVdzeSugQFVJjAyrZELzFl5dJ86zqXVYv7s8rtZs8jiIcwTGcggdXUIM7qEMDKDzCM7zCmyOdF+fd+Zi3Fpx85hD+wPn8AY7BjyI=</latexit>
↵
<latexit sha1_base64="3MUkKiCoiAzY5Y40sOOBicuGr2A=">AAAB7XicbVDLSgNBEOyNrxhfUY9eBoPgKeyKr2PQi8cI5gHJEnons8mY2Z1lZlYIS/7BiwdFvPo/3vwbJ8keNLGgoajqprsrSATXxnW/ncLK6tr6RnGztLW9s7tX3j9oapkqyhpUCqnaAWomeMwahhvB2oliGAWCtYLR7dRvPTGluYwfzDhhfoSDmIecorFSs4siGWKvXHGr7gxkmXg5qUCOeq/81e1LmkYsNlSg1h3PTYyfoTKcCjYpdVPNEqQjHLCOpTFGTPvZ7NoJObFKn4RS2YoNmam/JzKMtB5Hge2M0Az1ojcV//M6qQmv/YzHSWpYTOeLwlQQI8n0ddLnilEjxpYgVdzeSugQFVJjAyrZELzFl5dJ86zqXVYv7s8rtZs8jiIcwTGcggdXUIM7qEMDKDzCM7zCmyOdF+fd+Zi3Fpx85hD+wPn8AY7BjyI=</latexit>
↵
<latexit sha1_base64="ZwKkaJcdU3hHlnhMyVD4tdz55QM=">AAAB6HicbVDLTgJBEOzFF+IL9ehlIjHxRHaNryPRi0dI5JHAhswODYzMzm5mZo1kwxd48aAxXv0kb/6NA+xBwUo6qVR1p7sriAXXxnW/ndzK6tr6Rn6zsLW9s7tX3D9o6ChRDOssEpFqBVSj4BLrhhuBrVghDQOBzWB0O/Wbj6g0j+S9Gcfoh3QgeZ8zaqxUe+oWS27ZnYEsEy8jJchQ7Ra/Or2IJSFKwwTVuu25sfFTqgxnAieFTqIxpmxEB9i2VNIQtZ/ODp2QE6v0SD9StqQhM/X3REpDrcdhYDtDaoZ60ZuK/3ntxPSv/ZTLODEo2XxRPxHERGT6NelxhcyIsSWUKW5vJWxIFWXGZlOwIXiLLy+TxlnZuyxf1M5LlZssjjwcwTGcggdXUIE7qEIdGCA8wyu8OQ/Oi/PufMxbc042cwh/4Hz+AOl5jQY=</latexit>
x
<latexit sha1_base64="tmEvmksfb9cWDAa5xB1ckWUhUbQ=">AAAB6HicbVDLSgNBEOz1GeMr6tHLYBA8hV3xdQx68ZiAeUCyhNlJbzJmdnaZmRXCki/w4kERr36SN//GSbIHTSxoKKq66e4KEsG1cd1vZ2V1bX1js7BV3N7Z3dsvHRw2dZwqhg0Wi1i1A6pRcIkNw43AdqKQRoHAVjC6m/qtJ1Sax/LBjBP0IzqQPOSMGivVx71S2a24M5Bl4uWkDDlqvdJXtx+zNEJpmKBadzw3MX5GleFM4KTYTTUmlI3oADuWShqh9rPZoRNyapU+CWNlSxoyU39PZDTSehwFtjOiZqgXvan4n9dJTXjjZ1wmqUHJ5ovCVBATk+nXpM8VMiPGllCmuL2VsCFVlBmbTdGG4C2+vEya5xXvqnJZvyhXb/M4CnAMJ3AGHlxDFe6hBg1ggPAMr/DmPDovzrvzMW9dcfKZI/gD5/MH6v2NBw==</latexit>
y
<latexit sha1_base64="I6lx7kJ3J2ctPwbk+TFgr5/RVcU=">AAAB6HicbVDLTgJBEOzFF+IL9ehlIjHxRHaNryPRi0dI5JHAhswODYzMzm5mZk1wwxd48aAxXv0kb/6NA+xBwUo6qVR1p7sriAXXxnW/ndzK6tr6Rn6zsLW9s7tX3D9o6ChRDOssEpFqBVSj4BLrhhuBrVghDQOBzWB0O/Wbj6g0j+S9Gcfoh3QgeZ8zaqxUe+oWS27ZnYEsEy8jJchQ7Ra/Or2IJSFKwwTVuu25sfFTqgxnAieFTqIxpmxEB9i2VNIQtZ/ODp2QE6v0SD9StqQhM/X3REpDrcdhYDtDaoZ60ZuK/3ntxPSv/ZTLODEo2XxRPxHERGT6NelxhcyIsSWUKW5vJWxIFWXGZlOwIXiLLy+TxlnZuyxf1M5LlZssjjwcwTGcggdXUIE7qEIdGCA8wyu8OQ/Oi/PufMxbc042cwh/4Hz+AOyBjQg=</latexit>
z
<latexit sha1_base64="5coW10tvP99hmi468qu016uVHwc=">AAAB6nicbVDLSgNBEOyNrxhfUY9eBoPgKeyKr2PQi8dIzAOSJcxOepMhs7PLzKwQlnyCFw+KePWLvPk3TpI9aLSgoajqprsrSATXxnW/nMLK6tr6RnGztLW9s7tX3j9o6ThVDJssFrHqBFSj4BKbhhuBnUQhjQKB7WB8O/Pbj6g0j+WDmSToR3QoecgZNVZqNPpev1xxq+4c5C/xclKBHPV++bM3iFkaoTRMUK27npsYP6PKcCZwWuqlGhPKxnSIXUsljVD72fzUKTmxyoCEsbIlDZmrPycyGmk9iQLbGVEz0sveTPzP66YmvPYzLpPUoGSLRWEqiInJ7G8y4AqZERNLKFPc3krYiCrKjE2nZEPwll/+S1pnVe+yenF/Xqnd5HEU4QiO4RQ8uIIa3EEdmsBgCE/wAq+OcJ6dN+d90Vpw8plD+AXn4xvWg42F</latexit>
S1
<latexit sha1_base64="UaDg3IGQlJ82jWzM0HaGSLHskZs=">AAAB6nicdVDLSgNBEOyNrxhfUY9eBoPgadndPL0FvXiMxMRAsoTZyWwyZPbBzKwQlnyCFw+KePWLvPk3ziYRVLSgoajqprvLizmTyrI+jNza+sbmVn67sLO7t39QPDzqyigRhHZIxCPR87CknIW0o5jitBcLigOP0ztvepX5d/dUSBaFt2oWUzfA45D5jGClpXZ76AyLJcu8aNScqoMs07LqTrmWEadeccrI1kqGEqzQGhbfB6OIJAENFeFYyr5txcpNsVCMcDovDBJJY0ymeEz7moY4oNJNF6fO0ZlWRsiPhK5QoYX6fSLFgZSzwNOdAVYT+dvLxL+8fqL8hpuyME4UDclykZ9wpCKU/Y1GTFCi+EwTTATTtyIywQITpdMp6BC+PkX/k65j2jWzelMpNS9XceThBE7hHGyoQxOuoQUdIDCGB3iCZ4Mbj8aL8bpszRmrmWP4AePtEzr/jco=</latexit>
S2
<latexit sha1_base64="Z6E38nck0khI7CtIlntpBZhcRxU=">AAAB8HicdVDLSgMxFM3UV62vqks3wSK4GmamT3dFNy4rWFtph5JJM21okhmSjFiGfoUbF4q49XPc+Tdm2goqeuDC4Zx7ufeeIGZUacf5sHIrq2vrG/nNwtb2zu5ecf/gRkWJxKSNIxbJboAUYVSQtqaakW4sCeIBI51gcpH5nTsiFY3EtZ7GxOdoJGhIMdJGuu3jMcGT9H42KJYc+6xR86oedGzHqXvlWka8esUrQ9coGUpgidag+N4fRjjhRGjMkFI914m1nyKpKWZkVugnisQIT9CI9AwViBPlp/ODZ/DEKEMYRtKU0HCufp9IEVdqygPTyZEeq99eJv7l9RIdNvyUijjRRODFojBhUEcw+x4OqSRYs6khCEtqboV4jCTC2mRUMCF8fQr/Jzee7dbs6lWl1DxfxpEHR+AYnAIX1EETXIIWaAMMOHgAT+DZktaj9WK9Llpz1nLmEPyA9fYJiqSQ7A==</latexit>
ˇx
<latexit sha1_base64="2kvpDSSBx/q9XafIxFmx+w24fQE=">AAAB8HicdVDLSgMxFM3UV62vqks3wSK4GmamT3dFNy4r2Ie0Q8mkmTY0yQxJRihDv8KNC0Xc+jnu/BszbQUVPXDhcM693HtPEDOqtON8WLm19Y3Nrfx2YWd3b/+geHjUUVEiMWnjiEWyFyBFGBWkralmpBdLgnjASDeYXmV+955IRSNxq2cx8TkaCxpSjLSR7gZ4QvA0nc2HxZJjXzRqXtWDju04da9cy4hXr3hl6BolQwms0BoW3wejCCecCI0ZUqrvOrH2UyQ1xYzMC4NEkRjhKRqTvqECcaL8dHHwHJ4ZZQTDSJoSGi7U7xMp4krNeGA6OdIT9dvLxL+8fqLDhp9SESeaCLxcFCYM6ghm38MRlQRrNjMEYUnNrRBPkERYm4wKJoSvT+H/pOPZbs2u3lRKzctVHHlwAk7BOXBBHTTBNWiBNsCAgwfwBJ4taT1aL9brsjVnrWaOwQ9Yb5+MKZDt</latexit>
ˇy
<latexit sha1_base64="fiHvfNtQpZk+jF7ivOH83SwUQCE=">AAAB8HicdVDLSgMxFM3UV62vqks3wSK4GmamT3dFNy4rWFtph5JJM21okhmSjFCHfoUbF4q49XPc+Tdm2goqeuDC4Zx7ufeeIGZUacf5sHIrq2vrG/nNwtb2zu5ecf/gRkWJxKSNIxbJboAUYVSQtqaakW4sCeIBI51gcpH5nTsiFY3EtZ7GxOdoJGhIMdJGuu3jMcGT9H42KJYc+6xR86oedGzHqXvlWka8esUrQ9coGUpgidag+N4fRjjhRGjMkFI914m1nyKpKWZkVugnisQIT9CI9AwViBPlp/ODZ/DEKEMYRtKU0HCufp9IEVdqygPTyZEeq99eJv7l9RIdNvyUijjRRODFojBhUEcw+x4OqSRYs6khCEtqboV4jCTC2mRUMCF8fQr/Jzee7dbs6lWl1DxfxpEHR+AYnAIX1EETXIIWaAMMOHgAT+DZktaj9WK9Llpz1nLmEPyA9fYJja6Q7g==</latexit>
ˇz
<latexit sha1_base64="fDp9Q3Hu8/8EJL9Nk2n8BR7BLHU=">AAAB6HicbVDLSgNBEOz1GeMr6tHLYBA8hV3xdQx68ZiAeUCyhNlJbzJmdnaZmRXCki/w4kERr36SN//GSbIHTSxoKKq66e4KEsG1cd1vZ2V1bX1js7BV3N7Z3dsvHRw2dZwqhg0Wi1i1A6pRcIkNw43AdqKQRoHAVjC6m/qtJ1Sax/LBjBP0IzqQPOSMGivVh71S2a24M5Bl4uWkDDlqvdJXtx+zNEJpmKBadzw3MX5GleFM4KTYTTUmlI3oADuWShqh9rPZoRNyapU+CWNlSxoyU39PZDTSehwFtjOiZqgXvan4n9dJTXjjZ1wmqUHJ5ovCVBATk+nXpM8VMiPGllCmuL2VsCFVlBmbTdGG4C2+vEya5xXvqnJZvyhXb/M4CnAMJ3AGHlxDFe6hBg1ggPAMr/DmPDovzrvzMW9dcfKZI/gD5/MH0TmM9g==</latexit>
h
<latexit sha1_base64="bv9Vpcvv4c5kBq6M/JBHcolZFzo=">AAACHnicbVDLSgNBEJz1bXxFPXoZjIKnsCtGPQpePCoYFZIQeie9yeDszDLTK4YlX+LFX/HiQRHBk/6Nk5iDrzoVVd10dcWZko7C8COYmJyanpmdmy8tLC4tr5RX1y6cya3AujDK2KsYHCqpsU6SFF5lFiGNFV7G18dD//IGrZNGn1M/w1YKXS0TKYC81C7XmmlsbgvQ3VyB5Q4FGctNwqUmtOiIG82ph7xrTa47PFOgcdAuV8JqOAL/S6IxqbAxTtvlt2bHiDxFTUKBc40ozKhVgCUpFA5KzdxhBuIautjwVEOKrlWM3hvwba90eOJzJUYTH6nfNwpIneunsZ9MgXrutzcU//MaOSWHrULqLCfU4utQkitOhg+74h1pfR+q7wkIK31WLnpgQfhqXMmXEP1++S+52K1G+9Xa2V7laGtcxxzbYJtsh0XsgB2xE3bK6kywO/bAnthzcB88Bi/B69foRDDeWWc/ELx/AjG0ow0=</latexit>
angular sector of interest on the ground plane
<latexit sha1_base64="wKZXi4ZuFz08djeyNt4AF7l8Ul4=">AAAB8XicbVC5TgMxEJ0NVwhXgJLGIiBRRbuIq4xEA104cohkFXkdb2LFa69sLyJa5S9oKECIlr+h429wki0g4UkjPb03o5l5QcyZNq777eQWFpeWV/KrhbX1jc2t4vZOXctEEVojkkvVDLCmnAlaM8xw2owVxVHAaSMYXI79xiNVmklxb4Yx9SPcEyxkBBsrPbSjQD6lt9d3o06x5JbdCdA88TJSggzVTvGr3ZUkiagwhGOtW54bGz/FyjDC6ajQTjSNMRngHm1ZKnBEtZ9OLh6hQ6t0USiVLWHQRP09keJI62EU2M4Im76e9cbif14rMeGFnzIRJ4YKMl0UJhwZicbvoy5TlBg+tAQTxeytiPSxwsTYkAo2BG/25XlSPy57Z+XTm5NS5SCLIw97sA9H4ME5VOAKqlADAgKe4RXeHO28OO/Ox7Q152Qzu/AHzucPjTOQwg==</latexit>
RIS
<latexit sha1_base64="mu1wjUd0i4MGrLQtiQgqqSPqr6g=">AAAB8nicbVDLSgNBEJyNrxhfUY9eBqPgKeyKr2OCIF6ECOYBmyXMTibJkJmdZaZXDEs+w4sHRbz6Nd78GyfJHjSxoKGo6qa7K4wFN+C6305uaXlldS2/XtjY3NreKe7uNYxKNGV1qoTSrZAYJnjE6sBBsFasGZGhYM1weD3xm49MG66iBxjFLJCkH/EepwSs5LdlqJ7S6l31ZtwpltyyOwVeJF5GSihDrVP8ancVTSSLgApijO+5MQQp0cCpYONCOzEsJnRI+sy3NCKSmSCdnjzGx1bp4p7StiLAU/X3REqkMSMZ2k5JYGDmvYn4n+cn0LsKUh7FCbCIzhb1EoFB4cn/uMs1oyBGlhCqub0V0wHRhIJNqWBD8OZfXiSN07J3UT6/PytVjrI48ugAHaIT5KFLVEG3qIbqiCKFntErenPAeXHenY9Za87JZvbRHzifP/CqkPM=</latexit>
AMAF
Fig. 1: 3D geometry, conceptual (not to scale).
can be achieved. To achieve such gains at moderate hardware
complexity, we propose the usage of RIS in a reflectarray
configuration fed [8] by a small active multiantenna feeder
(AMAF). In this paper, building on our previous 2D single
module work [5], we present a full-dimensional (3D) multi-
beam, multi-user model with planar RIS and AMAF arrays
for terrestrial mmWave/sub-THz picocell base station applica-
tions, where the RIS is mechanically downtilted pointing to
the picocell centroid on the ground, see Fig. 1. Recall that
the aperture efficiency of a planar patch array is much higher
(≈90% due to a more uniform aperture distribution [9, Fig. 6])
than that of “legacy” feed horn (51%for optimum pyramidal
horns). In addition, the small size of the AMAF further
alleviates the aperture blockage in the front illumination of
reflectarrays. This motivates us to consider the AMAF instead
of a horn as in traditional reflectarrays. Our mathematical
model encompasses also the back illumination [10] which has
a reflected geometry with respect to the front illumination.
Whether front or back illumination is preferable depends on
the blocking effect of the front feeder versus the power waste
due to material absorption of the propagation through the
metasurface [10]. These aspects are design/material dependent
and go beyond the scope of this paper.
II. TH E AMAF-RIS MO DU LE :DE SI GN P RINCIPLES
As in Fig. 1, we define two coordinate systems. S1 has
its origin on the ground plane x-y. S2 has its origin in the
2024 IEEE 25th International Workshop on Signal Processing Advances in Wireless Communications (SPAWC)
556
2024 IEEE 25th International Workshop on Signal Processing Advances in Wireless Communications (SPAWC) | 979-8-3503-9318-7/24/$31.00 ©2024 IEEE | DOI: 10.1109/SPAWC60668.2024.10694442
Authorized licensed use limited to: Technische Universitaet Berlin. Downloaded on October 27,2024 at 19:27:03 UTC from IEEE Xplore. Restrictions apply.

center of the RIS, positioned at (0,0, h)with respect to the
S1 system, and is downtilted by a rotation of −αin the z-
y plane. Let i= (1,0,0)T,j= (0,1,0)T,k= (0,0,1)T
denote the three versors of S1 in the coordinate system S1,
and ˇ
i= (1,0,0)T,ˇ
j= (0,1,0)T,ˇ
k= (0,0,1)Tdenote the
three versors of S2 in the coordinate system S2. A S1 point
p=pxi+pyj+pzkcan be expressed in S2 Cartesian
coordinates ˇ
p= ˇpxˇ
i+ ˇpyˇ
j+ ˇpzˇ
kby
ˇ
p=

1 0 0
0 cos(α)−sin(α)
0 sin(α) cos(α)
(p−(0,0, h)T).(2)
<latexit sha1_base64="UaDg3IGQlJ82jWzM0HaGSLHskZs=">AAAB6nicdVDLSgNBEOyNrxhfUY9eBoPgadndPL0FvXiMxMRAsoTZyWwyZPbBzKwQlnyCFw+KePWLvPk3ziYRVLSgoajqprvLizmTyrI+jNza+sbmVn67sLO7t39QPDzqyigRhHZIxCPR87CknIW0o5jitBcLigOP0ztvepX5d/dUSBaFt2oWUzfA45D5jGClpXZ76AyLJcu8aNScqoMs07LqTrmWEadeccrI1kqGEqzQGhbfB6OIJAENFeFYyr5txcpNsVCMcDovDBJJY0ymeEz7moY4oNJNF6fO0ZlWRsiPhK5QoYX6fSLFgZSzwNOdAVYT+dvLxL+8fqL8hpuyME4UDclykZ9wpCKU/Y1GTFCi+EwTTATTtyIywQITpdMp6BC+PkX/k65j2jWzelMpNS9XceThBE7hHGyoQxOuoQUdIDCGB3iCZ4Mbj8aL8bpszRmrmWP4AePtEzr/jco=</latexit>
S2
<latexit sha1_base64="Z6E38nck0khI7CtIlntpBZhcRxU=">AAAB8HicdVDLSgMxFM3UV62vqks3wSK4GmamT3dFNy4rWFtph5JJM21okhmSjFiGfoUbF4q49XPc+Tdm2goqeuDC4Zx7ufeeIGZUacf5sHIrq2vrG/nNwtb2zu5ecf/gRkWJxKSNIxbJboAUYVSQtqaakW4sCeIBI51gcpH5nTsiFY3EtZ7GxOdoJGhIMdJGuu3jMcGT9H42KJYc+6xR86oedGzHqXvlWka8esUrQ9coGUpgidag+N4fRjjhRGjMkFI914m1nyKpKWZkVugnisQIT9CI9AwViBPlp/ODZ/DEKEMYRtKU0HCufp9IEVdqygPTyZEeq99eJv7l9RIdNvyUijjRRODFojBhUEcw+x4OqSRYs6khCEtqboV4jCTC2mRUMCF8fQr/Jzee7dbs6lWl1DxfxpEHR+AYnAIX1EETXIIWaAMMOHgAT+DZktaj9WK9Llpz1nLmEPyA9fYJiqSQ7A==</latexit>
ˇx
<latexit sha1_base64="2kvpDSSBx/q9XafIxFmx+w24fQE=">AAAB8HicdVDLSgMxFM3UV62vqks3wSK4GmamT3dFNy4r2Ie0Q8mkmTY0yQxJRihDv8KNC0Xc+jnu/BszbQUVPXDhcM693HtPEDOqtON8WLm19Y3Nrfx2YWd3b/+geHjUUVEiMWnjiEWyFyBFGBWkralmpBdLgnjASDeYXmV+955IRSNxq2cx8TkaCxpSjLSR7gZ4QvA0nc2HxZJjXzRqXtWDju04da9cy4hXr3hl6BolQwms0BoW3wejCCecCI0ZUqrvOrH2UyQ1xYzMC4NEkRjhKRqTvqECcaL8dHHwHJ4ZZQTDSJoSGi7U7xMp4krNeGA6OdIT9dvLxL+8fqLDhp9SESeaCLxcFCYM6ghm38MRlQRrNjMEYUnNrRBPkERYm4wKJoSvT+H/pOPZbs2u3lRKzctVHHlwAk7BOXBBHTTBNWiBNsCAgwfwBJ4taT1aL9brsjVnrWaOwQ9Yb5+MKZDt</latexit>
ˇy
<latexit sha1_base64="fiHvfNtQpZk+jF7ivOH83SwUQCE=">AAAB8HicdVDLSgMxFM3UV62vqks3wSK4GmamT3dFNy4rWFtph5JJM21okhmSjFCHfoUbF4q49XPc+Tdm2goqeuDC4Zx7ufeeIGZUacf5sHIrq2vrG/nNwtb2zu5ecf/gRkWJxKSNIxbJboAUYVSQtqaakW4sCeIBI51gcpH5nTsiFY3EtZ7GxOdoJGhIMdJGuu3jMcGT9H42KJYc+6xR86oedGzHqXvlWka8esUrQ9coGUpgidag+N4fRjjhRGjMkFI914m1nyKpKWZkVugnisQIT9CI9AwViBPlp/ODZ/DEKEMYRtKU0HCufp9IEVdqygPTyZEeq99eJv7l9RIdNvyUijjRRODFojBhUEcw+x4OqSRYs6khCEtqboV4jCTC2mRUMCF8fQr/Jzee7dbs6lWl1DxfxpEHR+AYnAIX1EETXIIWaAMMOHgAT+DZktaj9WK9Llpz1nLmEPyA9fYJja6Q7g==</latexit>
ˇz
<latexit sha1_base64="wKZXi4ZuFz08djeyNt4AF7l8Ul4=">AAAB8XicbVC5TgMxEJ0NVwhXgJLGIiBRRbuIq4xEA104cohkFXkdb2LFa69sLyJa5S9oKECIlr+h429wki0g4UkjPb03o5l5QcyZNq777eQWFpeWV/KrhbX1jc2t4vZOXctEEVojkkvVDLCmnAlaM8xw2owVxVHAaSMYXI79xiNVmklxb4Yx9SPcEyxkBBsrPbSjQD6lt9d3o06x5JbdCdA88TJSggzVTvGr3ZUkiagwhGOtW54bGz/FyjDC6ajQTjSNMRngHm1ZKnBEtZ9OLh6hQ6t0USiVLWHQRP09keJI62EU2M4Im76e9cbif14rMeGFnzIRJ4YKMl0UJhwZicbvoy5TlBg+tAQTxeytiPSxwsTYkAo2BG/25XlSPy57Z+XTm5NS5SCLIw97sA9H4ME5VOAKqlADAgKe4RXeHO28OO/Ox7Q152Qzu/AHzucPjTOQwg==</latexit>
RIS
<latexit sha1_base64="k11aXVxtue3ceF+8Q+UD3RSospI=">AAAB63icbVDLSgNBEOyNrxhfUY9eBoPgKeyKr2PQi8cIxgSSJcxOZrND5rHMzAoh5Be8eFDEqz/kzb9xNtmDJhY0FFXddHdFKWfG+v63V1pZXVvfKG9WtrZ3dveq+wePRmWa0BZRXOlOhA3lTNKWZZbTTqopFhGn7Wh0m/vtJ6oNU/LBjlMaCjyULGYE21zq6UT1qzW/7s+AlklQkBoUaParX72BIpmg0hKOjekGfmrDCdaWEU6nlV5maIrJCA9p11GJBTXhZHbrFJ04ZYBipV1Ji2bq74kJFsaMReQ6BbaJWfRy8T+vm9n4OpwwmWaWSjJfFGccWYXyx9GAaUosHzuCiWbuVkQSrDGxLp6KCyFYfHmZPJ7Vg8v6xf15rXFTxFGGIziGUwjgChpwB01oAYEEnuEV3jzhvXjv3se8teQVM4fwB97nDyLsjlE=</latexit>
⇢
<latexit sha1_base64="CSyXaWxx6UzyxtLAdkiwM8Z5mag=">AAAB63icbVBNS8NAEJ3Ur1q/qh69LBbBU0nEqseiF48V7Ae0oWy2m2bp7ibsboQS+he8eFDEq3/Im//GTZuDtj4YeLw3w8y8IOFMG9f9dkpr6xubW+Xtys7u3v5B9fCoo+NUEdomMY9VL8CaciZp2zDDaS9RFIuA024wucv97hNVmsXy0UwT6gs8lixkBJtcGiQRG1Zrbt2dA60SryA1KNAaVr8Go5ikgkpDONa677mJ8TOsDCOcziqDVNMEkwke076lEguq/Wx+6wydWWWEwljZkgbN1d8TGRZaT0VgOwU2kV72cvE/r5+a8MbPmExSQyVZLApTjkyM8sfRiClKDJ9agoli9lZEIqwwMTaeig3BW355lXQu6t5VvfFwWWveFnGU4QRO4Rw8uIYm3EML2kAggmd4hTdHOC/Ou/OxaC05xcwx/IHz+QMWyI5J</latexit>

<latexit sha1_base64="RnyzIg4Iy64zhhoBANN9EcBOI5Y=">AAAB7XicbVDLSgNBEJyNrxhfUY9eBoPgKeyKr2PQi8cI5gHJEmYnvcmY2Z1lplcIS/7BiwdFvPo/3vwbJ8keNLGgoajqprsrSKQw6LrfTmFldW19o7hZ2tre2d0r7x80jUo1hwZXUul2wAxIEUMDBUpoJxpYFEhoBaPbqd96Am2Eih9wnIAfsUEsQsEZWqnZxSEg65UrbtWdgS4TLycVkqPeK391+4qnEcTIJTOm47kJ+hnTKLiESambGkgYH7EBdCyNWQTGz2bXTuiJVfo0VNpWjHSm/p7IWGTMOApsZ8RwaBa9qfif10kxvPYzEScpQszni8JUUlR0+jrtCw0c5dgSxrWwt1I+ZJpxtAGVbAje4svLpHlW9S6rF/fnldpNHkeRHJFjcko8ckVq5I7USYNw8kieySt5c5Tz4rw7H/PWgpPPHJI/cD5/AKc3jzI=</latexit>
✓
Fig. 2: Spherical coordinates definition for the RIS coordinate
system S2.
For the S2 spherical coordinates in Fig. 2, the RIS boresight
direction ˇ
jcorresponds to the angles ϕ= 0, θ = 0, and this
is a line tilted down and impinging the picocell centroid on
the ground plane (S1 x-y plane) by an angle α(see Fig. 1).
Hence, a point of ˇ
pin S2 with spherical coordinates (ρ, ϕ, θ)
has ˇpx=ρsin ϕcos θ, ˇpy=ρcos ϕcos θ, ˇpz=ρsin θ. This,
along with (2), maps any S1 ground plane point to a far-field
direction (ϕ, θ)and a range ρwith respect to the RIS. In the
following, all lengths are normalized by λ/2.
A. RIS Array and Far-Field Array Response
A planar wavefront impinging a standard rectangular array
(SRA) RIS at an angle (ϕ, θ)has normal vector given by
n(ϕ, θ) = (sin ϕcos θ, cos ϕcos θ, sin θ)T.(3)
The planar wave propagation phase delay (complex phasor
term) of the (n, m)element of the RIS is given by
an,m(ϕ, θ) = exp −jπ ˇ
pT
n,mn(ϕ, θ).(4)
where ˇ
pn,m is the position of the (n, m)RIS elements in the
system S2. We use axisymmetric model for the RIS and the
AMAF patch elements (having a 3 dB beamwidth of 90◦and
the power gain of 6 dBi) given by Gpatch(θ, ϕ) = 4 cos2(ψ),
where ψis the angle of the direction (ϕ, θ)with respect to the
patch broadside, i.e., the S2 y-axis. Noticing that cos(ψ) =
n(ϕ, θ)·(0,1,0)T= cos ϕcos θ, we have
Gpatch(ϕ, θ) = 4 (cosϕcos θ)2.(5)
Assuming a RIS excitation by the AMAF such that the
complex signal at each element is un,m, the RIS can further
impose a phase rotation wn,m =ejµn,m for each n, m element.
Hence, the resulting far-field radiation pattern as a function of
the angle direction (ϕ, θ)is given by
G(ϕ, θ) = Gpatch(ϕ, θ)
Nx−1
X
n=0
Nz−1
X
m=0
wn,mun,m a∗
n,m(ϕ, θ)
2
.
(6)
We also notice that, without loss of generality, we can
incorporate the phase of un,m into wn,m. Hence, without
loss of generality, we can replace un,m by its magnitude
|un,m|. For convenience, we define the “tapered” RIS weights
˜wn,m(ϕ, θ) = |un,m|wn,m. Collecting {an,m(ϕ, θ)}and
{˜wn,m}into two Np×1vectors a(ϕ, θ)and ˜
w, (6) can be
compactly written as
G(ϕ, θ) = 4 (cos ϕcos θ)2a(ϕ, θ)H˜
w
2.(7)
B. AMAF-RIS Illumination and Beam-Steering
We focus now on the illumination, i.e., how to obtain
a suitable (complex) amplitude profile {un,m}induced by
the AMAF on the RIS surface. The AMAF is formed by
Na=NhNvactive elements arranged in a Nh×NvSRA
and placed at a distance Ffrom the RIS with Np=NxNz
elements arranged in a Nx×NzSRA. For convenience, we
enumerate the RIS and the AMAF elements row by row using
indices k∈ {0, . . . , Np−1}and ℓ∈ {0, . . . , Na−1},
respectively. Letting rk,ℓ denote the distance between the k-
th RIS element and the ℓ-th AMAF element, and letting
(φk,ℓ, ϑk ,ℓ)the angle at which they see each other with respect
to their own normal (boresight) direction, narrowband AMAF-
RIS propagation matrix T∈CNp×Nahas (Friis formula and
distance dependent phase term based) entries
Tk,ℓ =pEA(φk,ℓ , ϑk,ℓ)ER(φk,ℓ , ϑk,ℓ )
2πrk,ℓ
e−jπrk,ℓ ,(8)
where EA(φ, ϑ) = ER(φ, ϑ) = Gpatch(φ, ϑ)in this work.
Consider the Singular Value Decomposition (SVD) of T=
USVHwhere U∈CNp×Npand V∈CNa×Naare unitary
matrices and S∈CNp×Nais a diagonal matrix containing
ordered singular values σ1, σ2, . . , σNa. Letting uℓ,vℓdenotes
the ℓth columns U,V, respectively, any AMAF weight vector
b∈CNacan be written as b=PNa
ℓ=1 βℓvℓwith the transmit
power normalization PNa
ℓ=1 |βℓ|2= 1. This results in the
complex amplitude profile u=Tb =PNa
ℓ=1 σℓβℓuℓ. As an
effective pragmatic (maximum power transfer) choice of the
AMAF weight vector, we let b=v1, referred to as principal
eigenmode (PEM) design. This results in the “template” RIS
weight vector ˜
w0=σ1u1⊙e−j∠u1with all elements in R+,
corresponding to a beam pointing in the boresight direction of
the RIS.1For example, Fig. 3shows the template beam ground
footprint for a 16x16 RIS center fed by a 2x2 AMAF at a 20m
high base station mast with α= 37.37◦as detailed in Section
IV. The PEM design yields the RIS amplitude taper (13.92
dB) shown in Fig. 4, yielding -35 dB sidelobes and 27.5 dBi
1Here ∠u1is the vector of phases of u1,exp is applied componentwise,
and ⊙is elementwise product.
2024 IEEE 25th International Workshop on Signal Processing Advances in Wireless Communications (SPAWC)
557
Authorized licensed use limited to: Technische Universitaet Berlin. Downloaded on October 27,2024 at 19:27:03 UTC from IEEE Xplore. Restrictions apply.

gain. Beamsteering of the template beam to a desired direction
(ϕ0, θ0)is obtained by imposing a linear phase gradient in the
form w=a(ϕ0, θ0)⊙˜
w0.
Remark 1: It is important to notice that the AMAF weight
vector depends only on the AMAF-RIS geometry of the
basic module, and not on the beam-steering. In a multi-
beam multiuser setting, only the steering depends on the user
channels and can be adapted by changing the RIS phases while
the AMAF configuration remains fixed. ♢
Fig. 3: Ground footprint (coordinate system S1) of the RIS
PEM spotbeam pointing at the RIS boresight (coordinate
system S2), for a 16 ×16 RIS with 2×2AMAF at distance
F= 6.
Fig. 4: RIS PEM amplitude profile for a 16 ×16 RIS with
2×2AMAF at distance F= 6.
III. MULTI -BE AM M ULT IU SE R MIMO
Following the“One Stream Per Subarray” (OSPS) approach
to multiuser MIMO [11], we consider stacking K > 1AMAF-
RIS basic modules with minimal separation of length λ/2(i.e.,
unit separation in normalized distance) for maximum space
utilization.
The global matrix T∈CKNp×K Nafor the stacked modules
can be written as a K×Kblock matrix where each block
Ti,j ∈CNp×Namodels the propagation between the j-th
AMAF and the i-th RIS and is calculated from (8). We define
also Wi=diag(wi), for the i-th RIS phase-only weight
vector (not inclusive of the amplitude taper as in ˜
w) and
bj∈CNa×1to be the j-th AMAF weight vector. Defining
the block-diagonal matrices W=diag(Wi:i= 1, . . . , K)
and B=diag(bj:j= 1, . . . , K), the global transmission
matrix N∈CKNp×Kfrom the Kbaseband antenna ports
(each driving one AMAF) and the RIS reflecting elements is
given by N=WTBwith (i, j)-th Np×1blocks
[N]i,j =WiTi,j bj.(9)
Since the Kmodules are identical, we have that T1,1=· · · =
TK,K . Our pragmatic design uses the PEM approach for each
module, i.e.,bj=v1for all j= 1, . . . , K. Hence, the diagonal
blocks Nj,j (in isolation) produce radiation patterns as seen
before, i.e., independently steered versions of the basic weight
vector ˜
w0. The off-diagonal blocks Ni,j for i=jcapture the
effect of the near-end crosstalk (NEXT) between the modules.
Fortunately, given the good tapered profile of the PEM design
(see Fig. 4), it turns out that the NEXT is essentially negligible
even for unit separation between the modules.
We consider a multiuser MIMO scenario where the base
station (BS) is equipped with KAMAF-RIS modules and
serves multiple users located in the coverage area (see Fig. 1).
The downlink scheduler chooses groups of Kusers to be
served on the same time slot by spatial multiplexing. The
resulting LOS baseband channel matrix H∈CK×Kbetween
the KBS antenna ports and the K(far-field) users is given
by
H=AHN=AHWTB,(10)
where
A= 2 [cosϕ1cosθ1a(ϕ1, θ1), . . . , cosϕKcosθKa(ϕK, θK)]
(11)
is the KNp×Kmatrix containing the steering vectors whose
elements are given by (4) from the overall stacked RIS array
to the Kusers, where each user kis seen at an angle (ϕk, θk)
with respect to the RIS S2 coordinate system. Each k-th steer-
ing vector is weighted by pGpatch(ϕk, θk) = 2 cos ϕkcos θk
due to the RIS element directivity. The off-diagonal terms in
Hcapture the far-end crosstalk (FEXT) due to the sidelobes
of the AMAF-RIS beams. Ideally, we want the matrix Hto
be strongly diagonal-dominant which allows us to dispense
with (digital) baseband signal processing techniques such as
zero-forcing. This can be achieved by a) our proposed PEM
design which yields very low sidelobes, and 2) by scheduling
sets of K“compatible” users, which in the LOS MU-MIMO
case means users with sufficient angular separation in az-
imuth and/or elevation. Notice that the selection of compatible
(nearly mutually orthogonal) sets of users in MU-MIMO is a
common practice, as currently implemented in 802.11ax MU-
MIMO mode (e.g., see [12] and references therein).
The achievable communication rate of user kunder Gaus-
sian single-user capacity achieving codebooks and treating
multiuser interference as noise is given by
Rk= log2 1 + |Hk,k |2PRF
W N0/Lk+PK
j=1,j=k|Hk,j |2PRF !,
(12)
bits per complex signal dimension, where Hk,j is the (k, j)-th
element of H,PRF is the total AMAF output RF power, N0is
the complex baseband AWGN power spectral density, Wis the
channel bandwidth, and Lk= (λ/(4πρk))2is the free-space
2024 IEEE 25th International Workshop on Signal Processing Advances in Wireless Communications (SPAWC)
558
Authorized licensed use limited to: Technische Universitaet Berlin. Downloaded on October 27,2024 at 19:27:03 UTC from IEEE Xplore. Restrictions apply.

pathloss due to distance ρk(in meters) between user kand
the BS. In the LOS condition, His deterministic, and hence
any standard synchronization (carrier frequency, timing, and
phase) at the user receivers can easily achieve (almost) ideal
coherent detection.
IV. MU -M IM O EX AM PL E
We present a case study where the RIS and the AMAF are
SRAs of size Nx=Nz= 16 and Nh=Nv= 2, respectively.
We choose an empirically optimum focal length F= 6, chosen
to strike a good balance between beam directivity, loss in the
AMAF-RIS structure (captured by the value of singular-value
σ1), and height of the sidelobes. Qualitatively, if Fis too small
only a central portion of the RIS is illuminated by the AMAF,
i.e., the elements of the RIS away from the center play no
role in beamforming. If Fis too large, then a large fraction
of the RF power radiated by the AMAF is lost in space (and
creates significant NEXT in the multi-module stacked array),
since the solid angle covered by the RIS is too small and also
higher FEXT due to a smaller RIS taper. In the case at hand,
F= 6 with the PEM beamforming yields the nice tapered
amplitude profile shown in Fig. 4, resulting in the template
beam with footprint in Fig. 3.
We consider K= 4 stacked modules height h= 20m on the
ground, serving a sector on the ground S1 x-y plane with range
between rmin = 10m to rmax = 100m, azimuth ϕfrom −60◦
to 60◦. The 10m and 100m ground distances correspond to
the downlook angles of αmax =acot (rmin/h) = 63.43◦and
αmin =acot (rmax/h) = 11.30◦, respectively, with respect
to the S2 origin. Therefore, we choose the RIS mechanical
downtilt angle αto be the arithmetic mean, i.e. α= 37.37◦,
for the optimum element factor utilization in the elevation.
This downtilt angle causes the RIS normal vector to intercept
the ground at distance r= 26.19 m, as shown in Fig. 3.
In order to ensure a minimum 0 dB2SNR to any user in the
picocell, we consider the link budget of Table I. The RF feed
power from the AMAF, PRF =PT/G(ϕ, θ)where G(ϕ, θ)
is given in (7). At the cell edge, G(ϕ= 60◦, θ = 26.06◦) =
20.7 dBi, yielding a required AMAF RF power of PRF =
40.7 dBm −20.7 dBi = 20 dBm.
TABLE I: Example system specifications.
Specification Value Specification Value
Carrier freq. (GHz) 100 Receive noise pow. (dBm) -72
Cell range (m) 10 to 100 Receive SNR (dB) 0
Azimuth span (ϕ) +/-60◦Receive signal power (dBm) -72
Bandwidth W(GHz) 5 Path Loss Lmax (dB) 112.7
Thermal noise pow. (dBm) -77 EIRP PT(dBm) 40.7
Rx NF (dB) 5 RIS size (Nx×Nz) 16 x 16
We consider K= 4 downlink data streams serving 4
users randomly distributed with azimuth ϕ∈[−60◦,60◦]and
range r∈[10m,100m]. The scheduler chooses the Kusers
a minimum azimuth angle separation of 15◦, corresponding
2Notice that an SNR of 0 dB corresponds to a channel capacity equal to
1 per complex dimension, which is approachable in practice using QPSK
modulation with powerful binary LDPC coding of rate (slightly less than)
1/2. Hence, such a system is quite realistic also from a practical viewpoint.
to the -20 dB beam contour of the template beam. Fig. 5
shows a snapshot (random realization) of 4 user positions and
the corresponding ground beam footprints with ideal beam
steering (i.e., by pointing the beams towards the corresponding
users angles).
Fig. 5: Ground footprints: an example set of 4 PEM spot
beams, with perfect beam pointing onto the 4 UEs.
0 1 2 3 4 5 6
Rate (bits/s/Hz)
0
0.2
0.4
0.6
0.8
1
CDF
1U, pp
1U, s.d. = 2.5 deg
4U, pp
4U, s.d. = 2.5 deg
Fig. 6: Rate CDFs for the beam pointing case, with perfect
pointing (pp) and with Gaussian beam pointing errors.
Fig. 6shows the achievable rate CDF for the case of a
single user (1U), and 4 users (4U), with perfect beam pointing
(pp), and independent Gaussian beam pointing errors in both
ϕand θwith standard deviations of 2.5◦. We find that the rate
degradation due to beam pointing errors is tolerable because
the beam footprints have a smooth contour, which provides
some robustness to pointing errors. We see also that there is no
practical degradation between the single user and the 4 users
case, indicating that the FEXT/NEXT multiuser interference
is effectively negligible. This means that any further hybrid
precoding would yield no improvement at the cost of a much
higher computational complexity in the baseband.
0123456
Rate (bits/s/Hz)
0
0.2
0.4
0.6
0.8
1
CDF
bits
3 bits
2 bits
1 bit
Fig. 7: Rate CDFs for the beam pointing case, with perfect
pointing and RIS quantized phase-shifters.
In Fig. 7we show the rate CDF for the same system
scenario, with no pointing errors and quantized RIS phase-
2024 IEEE 25th International Workshop on Signal Processing Advances in Wireless Communications (SPAWC)
559
Authorized licensed use limited to: Technische Universitaet Berlin. Downloaded on October 27,2024 at 19:27:03 UTC from IEEE Xplore. Restrictions apply.

shifters. We used simple scalar quantization of the continuous
phases, without any further complicated optimization. For 3
or more quantization bits, the rate CDF is essentially identical
to the unquantized case. Nevertheless, even for just 1 bit
quantization, i.e., restricting the RIS phase shifts to ±1,
the per-user rates range from ≈0.5to ≈5bits/symbol,
corresponding to ≈2.5to ≈25 Gb/s (with W= 5 GHz
as in Table I).
V. POWER EFFICIENCY ANALYS IS
For the same planar array dimension and beamforming
radiation pattern of the proposed AMAF-RIS architecture
(Arch.1), we consider a baseline architecture consisting of
a constrained-fed active array (Arch.2), where each element
has its own dedicated power amplifier (PA), see Fig. 8. For
Arch.2 we neglect the power loss incurred by the beamforming
network from the antenna port to the array elements (i.e., to
the signals before the per-antenna PAs) because this operates
on low-level signals and impacts essentially only the noise
figure, which we assume here to be ideal. This assumption
is favorable to Arch.2. Nevertheless, we shall see that Arch.1
(the proposed one) is still very competitive from the energy
efficiency viewpoint.
Fig. 8: Arch.2 block diagram. Beamforming network (BFN)
consists of amplitude and phase shifters.
From the previously computed link budget in Section IV,
we target 0 dB SNR for a cell edge user, given by SNRrx =
PRFG(ϕ= 60◦, θ = 26.06◦)/(LmaxW N0). Plugging in
the values from Section IV, we get PRF = 20 dBm. For
Arch.1, the AMAF weights, v1= [0.5,0.5,0.5,0.5]. Hence,
the maximum AMAF PA output power is given by P(1)
pa−max =
max|v1i|2PRF =−6dB + 20 dBm = 14 dBm = 25.1mW.
We assume that all the PAs in the (AMAF) array are de-
veloped in the same semiconductor technology, and are all
biased with the same DC power dictated by the maximum
requested RF power. Considering Indium Phosphide (InP) PAs
with efficiency η= 0.3[13], [14], the Arch.1 DC power
consumption, P(1)
DC =NaP(1)
pa−max/η = 4 ×25.1mW/0.3 =
0.33 W. Likewise, for Arch.2, P(2)
pa−max = max|u1i|2PRF =
−14.65 dB + 20 dBm = 5.34 dBm = 3.42 mW. Thus, Arch.2
DC power P(2)
DC =NpP(2)
pa−max/η = 2.92 W. We see that the
proposed architecture is almost 10x times more power efficient
than the baseline. VI. CON CL US IO NS
We proposed a novel multiuser multibeam architecture with
over-the-air active array-based feeding (AMAF) and RIS-
based beam steering suited to very high frequency bands. The
scheme is based on a fundamental “module” formed by an
AMAF-RIS pair, with fixed geometry, and can accommodate
any suitable number Kof independently steered data streams
by stacking such module in a larger array. We demonstrated a
design example with (2×2) active antennas at the AMAF and
16 ×16 passive elements at the RIS. We also demonstrated
that there is no dramatic performance loss with practical
hardware constraints such as quantized RIS phase shifters (3
bits or more) and beam pointing errors. Our pragmatic design
approach is very simple, does not require complicated on-line
optimization (unlike most analog-digital multiuser precoding
approaches), and can be easily applied to different combina-
tions of AMAF and RIS. The proposed architecture has low
hardware complexity (very small number of PAs, simple active
beamforming network), and achieves large energy efficiency
gains with respect to the baseline active array design with the
same beamforming capability.
ACK NOW LE DG ME NT
The work of G. Caire was supported by BMBF Germany in
the program of “Souver¨
an. Digital. Vernetzt.” Joint Project 6G-RIC
(Project IDs 16KISK030).
REF ER EN CE S
[1] T. S. Rappaport et al., “Wireless Communications and Applications
Above 100 GHz: Opportunities and Challenges for 6G and Beyond,”
IEEE Access, vol. 7, pp. 78 729–78 757, Jun. 2019.
[2] A. Adhikary et al., “Joint Spatial Division and Multiplexing—The
Large-Scale Array Regime,” IEEE Trans. on Inf. Theory, vol. 59, no. 10,
pp. 6441–6463, Jun. 2013.
[3] Z. Zhou et al., “Hardware-Efficient Hybrid Precoding for Millimeter
Wave Systems With Multi-Feed Reflectarrays,” IEEE Access, vol. 6, pp.
6795–6806, Jan. 2018.
[4] V. Jamali et al., “Intelligent Surface-Aided Transmitter Architectures for
Millimeter-Wave Ultra Massive MIMO Systems,” IEEE Open J. of the
Commun. Society, vol. 2, pp. 144–167, Jan. 2021.
[5] K. K. Tiwari et al., “RIS-Based Steerable Beamforming Antenna with
Near-Field Eigenmode Feeder,” in ICC 2023 - IEEE Int. Conf. on
Commun., May 2023, pp. 1293–1299.
[6] M. Sadeghian et al., “RIS in Indoor Environments: Benchmarking
Against Ambient Propagation,” in 2023 57th Asilomar Conf. on Signals,
Syst., and Comput., Oct. 2023, pp. 127–133.
[7] H. Friis, “A Note on a Simple Transmission Formula,” Proceedings of
the IRE, vol. 34, no. 5, pp. 254–256, May 1946.
[8] K. K. Tiwari et al., “Power Transfer between Two Antenna Arrays in
the Near Field,” in IEEE 99th Veh. Technol. Conf. (VTC2024-Spring),
June 2024, pp. 1–6.
[9] R. Munson, “Conformal Microstrip Antennas and Microstrip Phased
Arrays,” IEEE Trans. on Antennas and Propag., vol. 22, no. 1, pp. 74–
78, Jan. 1974.
[10] S. Venkatesh et al., “A high-speed programmable and scalable terahertz
holographic metasurface based on tiled CMOS chips,” Nature Electron.,
vol. 3, no. 12, pp. 785–793, Dec. 2020.
[11] X. Song et al., “Fully-/Partially-Connected Hybrid Beamforming Archi-
tectures for mmWave MU-MIMO,” IEEE Trans. on Wireless Commun.,
vol. 19, no. 3, pp. 1754–1769, Mar. 2020.
[12] K. Wang et al., “Scheduling and Resource Allocation in 802.11ax,” in
IEEE INFOCOM 2018-IEEE Conf. on Comput. Commun., Apr. 2018,
pp. 279–287.
[13] H. Wang et al. Power Amplifiers Performance Survey 2000-
Present. [Online]. Available: https://ideas.ethz.ch/research/surveys/
pa-survey.html
[14] J. F. Buckwalter et al., “Fundamental Limits of High-Efficiency Silicon
and Compound Semiconductor Power Amplifiers in 100-300 GHz
Bands,” ITU J. on Future and Evolving Technol., vol. 2 (2021), no.
7 - Terahertz Commun., pp. 39–50, Oct. 2021.
2024 IEEE 25th International Workshop on Signal Processing Advances in Wireless Communications (SPAWC)
560
Authorized licensed use limited to: Technische Universitaet Berlin. Downloaded on October 27,2024 at 19:27:03 UTC from IEEE Xplore. Restrictions apply.