1: \begin{abstract}

2: \glspl{ris} allow to control the propagation environment in wireless networks by properly tuning multiple reflecting elements.

3: %These planar surfaces are composed of multiple reflecting elements which can be properly tuned to steer the reflected signal towards the intended direction.

4: Traditionally, RISs have been realized through single connected reconfigurable impedance networks, in which each RIS element is independently controlled by an impedance connected to ground.

5: In a recent work, this architecture has been extended by realizing more efficient RISs with group and fully connected reconfigurable impedance networks.

6: %These new architectures are more efficient than the traditional single connected.

7: However, impedance networks tunable with arbitrary precision are hard to realize in practice.

8: %For single connected architectures, a well established assumption is to consider a finite number of possible phase shifts at each RIS element, but a general solution for group and fully connected architectures is still missing.

9: In this paper, we propose a practical RIS design strategy based on reconfigurable impedance networks with discrete values.

10: Besides, we address the problem of how to group the RIS elements in group connected architectures.

11: We optimize single, group, and fully connected architectures considering finite-resolution elements, and we compare them in terms of received signal power.

12: Through Monte Carlo simulations, supported by theoretical justifications, we show that only a few resolution bits per reconfigurable impedance are sufficient to achieve the performance upper bound. 

13: In particular, while four resolution bits are needed to reach the upper bound in single connected architectures, only a single resolution bit is sufficient in fully connected ones, simplifying significantly the future development of these promising RIS architectures.

14: \end{abstract}

15: