Design and Performance Trade-offs in Parallelized RF SDR - - PowerPoint PPT Presentation
Design and Performance Trade-offs in Parallelized RF SDR Architecture Ville Saari 1 arssinen 2 Sami Kiminki Aarno P Antti Immonen 2 Vesa Hirvisalo Jussi Ryyn anen Tommi Zetterman Aalto University Aalto University Nokia Research Center
Sami Kiminki Ville Saari1 Aarno P¨ arssinen2 Vesa Hirvisalo Jussi Ryyn¨ anen Antti Immonen2 Tommi Zetterman
Aalto University Aalto University Nokia Research Center Computer Science and Eng. Micro- and Nanosciences
1currently with EPCOS Nordic 2currently with Renesas Mobile Europe
Design and Performance Trade-offs in Parallelized RF SDR Architecture Sami Kiminki — 2011-06-02 9%
◮ Parallel RF SDR platform for LTE and WLAN (UE) ◮ Holistic RF platform design
◮ effect on RX and TX front-end filters
◮ Multi-radio opportunities
◮ we show system-level data throughput opportunities by RF
resource sharing (simulation results)
Design and Performance Trade-offs in Parallelized RF SDR Architecture Sami Kiminki — 2011-06-02 18%
A D A D A D D A D A Filter bank Duplexers Band # ISM-band filtering Option 2
Figure: Archetype of a parallel multi-standard RF transceiver. The number of RX and TX pipes may be varied.
Design and Performance Trade-offs in Parallelized RF SDR Architecture Sami Kiminki — 2011-06-02 27%
◮ We analyse how the RX noise floor is accumulated from
different sources
◮ receiver noise figure ◮ TX signal spilling to RX frequencies due to TX
non-linearities
◮ TX signal transferred to RX noise due to RX non-linearities
◮ By setting the allowed desensitization threshold, we can
determine RX and TX filter requirements
◮ Focus on a difficult case
◮ WLAN 2.4-GHz (2400–2483.5 MHz) ◮ LTE Band 7 (TX: 2500–2570 MHz, RX: 2620–2690 MHz)
Design and Performance Trade-offs in Parallelized RF SDR Architecture Sami Kiminki — 2011-06-02 36%
−10 −5 5 10 20 25 30 35 40 45 Receiver IIP3 [dBm] Attenuation Requirement [dB]
LTE WLAN
Figure: TX blocker attenuation requirement vs. IIP3 of the receiver to achieve 1 dB sensitivity loss
Design and Performance Trade-offs in Parallelized RF SDR Architecture Sami Kiminki — 2011-06-02 45%
2300 2400 2500 2600 2700 2800 −70 −60 −50 −40 −30 −20 −10 Frequency [MHz] Attenuation Requirement [dB] LTE RX WLAN LTE TX
Figure: TX stopband attenuation requirement for LTE band 7 for sensitivity losses of 1 dB and 3 dB in 2.4-GHz WLAN receiver
Design and Performance Trade-offs in Parallelized RF SDR Architecture Sami Kiminki — 2011-06-02 55%
◮ Non-dedicated RF resources ◮ Two approaches for sharing
◮ more performance in the average case (high-end) ◮ less hardware for same functionality (low-end)
◮ Sharing requires favourable conditions, e.g., discontinuous
modes in use
◮ high-end approach is currently more feasible
Design and Performance Trade-offs in Parallelized RF SDR Architecture Sami Kiminki — 2011-06-02 64%
◮ The fundamental idea
◮ when one radio does not need full HW capabilities, use
spare resources to boost another radio
◮ all radios maxed out is not the common case
◮ Some techniques
◮ semi-static scheduling: SISO vs MIMO ◮ dynamic scheduling: fine-grain traffic shaping
i.e., “TDM of chip resources”
Design and Performance Trade-offs in Parallelized RF SDR Architecture Sami Kiminki — 2011-06-02 73%
◮ Assume discontinuous modes
◮ LTE: DRX ◮ WLAN: powersave
◮ The idea
◮ LTE reserves the resources first ◮ WLAN uses what is left ◮ PS-Poll enables fine-grained traffic shaping for RX
◮ In experiments, we assume
◮ bandwidths are 20 MHz (≈ 150 Mbps) ◮ device has control on SISO vs MIMO ◮ MCS & RI feedback
Design and Performance Trade-offs in Parallelized RF SDR Architecture Sami Kiminki — 2011-06-02 82%
20 40 60 80 100 120 140 160 20 40 60 80 100 120 140 160
WLAN Throughput [Mbps] LTE Throughput [Mbps]
Sharing overhead Opportunity area (co-op LTE) Opportunity area (non-co-op LTE) Guaranteed area
Figure: Performance estimation for 2 shared receivers
Design and Performance Trade-offs in Parallelized RF SDR Architecture Sami Kiminki — 2011-06-02 91%
20 40 60 80 100 120 140 160 20 40 60 80 100 120 140 160
WLAN Throughput [Mbps] LTE Throughput [Mbps]
Sharing overhead Opportunity area (co-op LTE) Guaranteed area
Figure: Performance estimation for 3 shared receivers
Design and Performance Trade-offs in Parallelized RF SDR Architecture Sami Kiminki — 2011-06-02 100%
◮ RF systems must be designed as a whole, not only
per-protocol
◮ e.g., additional filter requirements
◮ Parallel SDR approach brings new opportunities
◮ Resource sharing for better system-level throughputs ◮ Is there even a fundamental reason for dedicated RF pipes? ◮ Cognitive radio connection:
Don’t share only the spectrum, share the resources too
◮ Resource sharing calls for further protocol work
◮ We want better flexibility and predictability ◮ ongoing in-device coexistence work helps in this
◮ Ultimately, we’d like to see general-purpose RF platforms
◮ think CPUs, GPGPUs, FPGAs, . . .
Design and Performance Trade-offs in Parallelized RF SDR Architecture Sami Kiminki — 2011-06-02 109%
Design and Performance Trade-offs in Parallelized RF SDR Architecture Sami Kiminki — 2011-06-02 118%
Figure: A resource schedule for LTE and WLAN on 2 RX + 2 TX
Design and Performance Trade-offs in Parallelized RF SDR Architecture Sami Kiminki — 2011-06-02 127%
20 40 60 80 100 2 4 6 8 10
Range (%) Desensitization (dB)
Range loss 10% 30% Range loss
Relation of desensitization and range loss Visualization for 1 dB and 3 dB desensitization
Figure: Desensitization as range loss in free space
Design and Performance Trade-offs in Parallelized RF SDR Architecture Sami Kiminki — 2011-06-02 136%
time Resource allocation block Resource allocation block because of allocation gap Defer sending PS-Poll here ACK DATA Wait until free to send PS-Poll ACK DATA Wait until free to send PS-Poll ACK DATA Wait until free to send PS-Poll
RX TX
Design and Performance Trade-offs in Parallelized RF SDR Architecture Sami Kiminki — 2011-06-02 145%
◮ External PNG slides