Energy Efficient Group-Sort QRD Processor With On-Line Update for MIMO Channel Pre-Processing

Abstract

This paper presents a Sorted QR-Decomposition (SQRD) processor for 3GPP LTE-A system. It achieves energy-efficiency by co-optimizing techniques, such as heterogeneous processing, reconfigurable architecture, and dual-supply voltage operation. At algorithm level, a low-complexity hybrid decomposition scheme is adopted, which switches, depending on the energy distribution of spatial channels, between the traditional brute-force SQRD and a proposed group-sort QR-update strategy. A reconfigurable vector processor is accordingly developed to support the adaptive processing with high hardware efficiency. Furthermore, on-chip power management technique is also integrated to obtain real-time power-saving by adapting the voltage supply based on the instantaneous workload. As a proof-of-concept, we implemented the processor using a 65 nm CMOS technology and conducted post-layout simulation. The proposed SQRD processor occupies 0.71 core area and has a throughput of up to 69 MQRD/s. Compared to the brute-force approach, an energy reduction of 10 61.8% is achieved.

Low latency QRD algorithm for future communication

“…8, the number of fractional bits is set to 13, while the total WL changes from The QRD architectures in [5,7,8,9] are based on GR algorithm, and [6,10] are based on Hybrid algorithm. Lin JS [5] proposed a two-phase QRD architecture, which incorporated a single exchange of the column and arrow into the general QRD procedure to achieve a better NR performance of 138.5 M. Zhang C [6] designed a hardware architecture based on hybrid GR algorithm to increase the parallelism, which achieves a good NR performance of 138 M and relatively low NPL of 87 ns. As demonstrated in Table I, one common architecture of GR is its high NR performance deriving from its high working frequency.…”

“…(4) (5), RSR can be expressed as (6), where a 2 , a 1 and a 0 depend on the expansion point x 0 and x 0 2 ð0:5; 2. Thus the value set of a 2 , a 1 , a 0 corresponding to various expansion points of x 0 in interval ð0:5; 2 can be stored in a look-up table (LUT), and the specific value of a 2 , a 1 , a 0 in (6) is chosen from LUT according the expansion point of x 0 nearest to x.…”

“…The majority of QRD algorithms are mainly based on three original algorithms -Householder transformation (HT) algorithm [3], modified Gram-Schmidt (MGS) algorithm [4], and Givens rotation (GR) algorithm [5,6,7,8,9]. As the HT algorithm is too complex to be implemented in hardware, the existing QRD architectures are mainly based on GR algorithm and MGS algorithm in MIMO systems.…”

Adaptive resource scheduling for energy efficient QRD processor with DVFS

“…To evaluate the proposed scheme, we mapped the algorithms on a vector-oriented reconfigurable platform [4]. As previously discussed, the platform is implemented in 65 nm ST microelectronics CMOS technology and measured with a library set of different supply voltages in Common P ower F ormat (CPF) low power design flow.…”

“…QR decomposition (QRD) [1] is essential for various MIMO detection algorithms, like sphere decoder [2]. To tackle the high computing cost of QRD, several algorithms [3], [4] have been proposed. These low complexity implementations rely on approximations of QRD by leveraging wireless channel feature.…”