Pro-VIZOR: Process Tunable Virtually Zero Margin Low Power Adaptive RF for Wireless Systems Presented by: Shreyas Sen June 11, Paper 27.3, DAC 08. Shreyas Sen, Vishwanath Natarajan, Rajarajan Senguttuvan, Abhijit Chatterjee Abhijit Chatterjee Georgia Institute of Technology
Goal Traditional wireless circuits Designed for worst case environmental variations and worst case process corners. Involves significant built in design margin. Motivation Waste of performance (i.e. power) under favorable condition. SAVE POWER by adapting to environment and process.
Process Tunable Virtually Zero Margin RF Baseband Receiver Transmitter Baseband Transmitter Receiver Base Station (BS) Mobile Station (MS) Save Power Worst Case System Design Adaptive System Design Varying channel Process adaptability
Outline Environment Adaptability in Transceivers - Adaptive RF receivers. - Adaptive RF transmitters. Process Adaptability -Process Sensing Using Test. -Tuning for process variation. Experimental Results
Outline Environment Adaptability in Transceivers - Adaptive RF receivers. - Adaptive RF transmitters. Process Adaptability -Process Sensing Using Test. -Tuning for process variation. Experimental Results
VIZOR: Receiver Adaptation metric is a measure of the quality of the received signal.
Baseband & RF Baseband & RF Transmission Reception Channel BER = Modulation Demodulation N e = Number of errors R = Data rate t = Test time Adaptation metric selection Bit Error Rate (BER) Bit Error Rate (BER) Cannot be calculated in real-time Error Vector Magnitude (EVM) Error Vector Magnitude (EVM) 1 2 e I Q Error in the received symbol V2V2 V1V1 e = V1V1 - V2V2 Can be calculated in real-time
EVM vs. BER Channel effects Adaptation Metric: EVM vs BER Hence EVM is chosen as the real-time adaptation metric. Multi-path effects Interference Path loss Noise Multi-path effects Interference Path loss Noise
VIZOR operation -Operation close to error threshold VIZOR operation -Operation close to error threshold Normal operation Save power VIZOR Operation : Constellation EVM=22% EVM=8% EVM=35% EVM=14% VIZOR operation Channel 1 Channel 2
Receiver hardware and Power Savings
Tuning Knob 2 Performance Tuning knob 1 Minimum power, maximum EVM locus Increasing performance P Tuning knob: Vdd, Vbias of different RF block, ADC wordsize Power vs. Performance locus
Outline Environment Adaptability in Transceivers - Adaptive RF receivers. - Adaptive RF transmitters. Process Adaptability -Process Sensing Using Test. -Tuning for process variation. Experimental Results
VIZOR: Transmitter RF PA Drain Bias Gate Bias Power Management DAC Mixer Baseband DSP PAR Reduction Control Law Channe l Quality Dynamic Adaptive
Companded Signal; PAR reduction = 5.6 dB OFDM: High Peak to Average Ratio (PAR). Makes Power Amplifier (PA) inefficient due to high back off required. 3 dB back off halves a PA efficiency. Pin Pout PA characteristics Original Signal P1dB 5.6 dB Companding and PAR reduction
Compliance to BER threshold PAR Reduction BER moderate channel BER good channel EVM moderate channel EVM good channel µ : Companding Factor
PAR reduction under different channels 64 QAM BPSK 5%14%35% 16 QAMQPSK *EVM= + EVM threshold 16 QAM =12.5 % + EVM threshold QPSK=33 % Good Channel EVM=5.15 % Moderate Channel EVM=8.5 % 6.3 dB 4 dB Moderate Channel EVM=24.4 % 7.25 dB 5.6 dB Good Channel EVM=15.5 % *From SNR boundaries (IEEE) + From BER threshold
Transmitter Power Savings 3X Pdc static PA
Outline Environment Adaptability in Transceivers - Adaptive RF receivers. - Adaptive RF transmitters. Process Adaptability -Process Sensing Using Test. -Tuning for process variation. Experimental Results
Minimum power, maximum EVM locus Increasing performance P PROCESS Effect of process variations: run RF BIST to pick the “Right” locus under variation. Effect of Process variation Performance Tuning Knob 2 Tuning knob 1
Process Sensing *V. Natarajan et. al. “ACT: Adaptive Calibration test….” VTS 08
Outline Environment Adaptability in Transceivers - Adaptive RF receivers. - Adaptive RF transmitters. Process Adaptability -Process Sensing Using Test. -Tuning for process variation. Experimental Results
Process tuning Design Phase Production Test/Tuning Phase Minimum power & Maximum EVM locus for N p instances Process Sensing & Process Tuning Process Sensing & Process Tuning Locus corresponding to the BEST MATCH (LMS) DUT N p process instances For each process instance Perform environmental Adaptation during run time Process 1 Process 2 Process N p Performance Gain, IIP3, P/ V dd, P/ V b.. Gain, IIP3, P/ V dd, P/ V b Process adaptation metric for N p instances Performance Measure TX process Measure RX process Find BEST MATCH with available N p process instances Measure Power sensitivities
With tuning Without tuning Channel Index Receiver Power (W) Power Savings through process tuning
Outline Environment Adaptability in Transceivers - Adaptive RF receivers. - Adaptive RF transmitters. Process Adaptability -Process Sensing Using Test. -Tuning for process variation. Experimental Results
VIZOR system VIZOR TX and RX
Runtime experimental results: Receiver EVMincreasing OFDM signal Power consumption decreasing
System in Operation
Environment Adaptability allows transceiver operation with very less built in design margin. (i.e. Virtually Zero Margin) - Saves significant power under favorable environmental condition. - 3X transmitter and 4X receiver power savings could be achieved. Process Adaptability makes this adaptation near optimal even under process variation. Conclusion Significant increase in battery life.
Questions ?
Environment Adaptability: Overview Baseband Receiver Transmitter Baseband Transmitter Receiver Base Station (BS) Mobile Station (MS) 1 bit Encoded in MAC Header 0: EVM > EVM threshold 1: EVM < EVM threshold Power saved in MS increasing battery life Worst Case System Design Adaptive System Design Varying channel conditions
VIZOR Optimizer Zero-margin operation – Save more power under favorable conditions (good channel)!! Identify tunable Parameters LNA supply LNA bias Mixer supply Mixer bias ADC word size Set EVM threshold for satisfactory operation Generate different Channels Interference Multi-path Noise Optimization Optimal values of tunable parameters for Different channel conditions Different modulations (data rates)
PAR reduction and compliance to BER threshold
Estimate transmitter specs Tune the transmitter – Hardware + software knobs -Process variation, non-idealities Estimate Receiver specs Transmitter spec estimation Receiver spec estimation BASEBAND DSP Rx Module Embedded sensor Loopback hardware Tuning control Tx Module Tuned output Baseband response Process Sensing: Adaptive Calibration Test
Control law: LS error between golden and actual Hardware knobs: Power supply, bias control Software knobs: Reverse distortion polynomials Control law Test input System under consideration Knob Controller Software knobs Tuned output Actual response Golden Hardware knob control Envelope Detector Appendix: ACT: Tuning technique
Transmitter parameter estimation Receiver parameter estimation without transmitter tuning Receiver parameter estimation after transmitter tuning Process estimation accuracy