Introduction to Envelope Tracking G J Wimpenny Snr Director Technology, Qualcomm UK Ltd
Envelope Tracking Historical Context EER first proposed by Leonard Kahn in 1952 to improve efficiency of SSB transmitters ET offers similar efficiency enhancement to EER but has fewer drawbacks Technique not widely adopted for many years due to difficulty of implementation, particularly for wide bandwidth signals In the last few years the implementational issues have been overcome and ET is now widely used to improve PA efficiency in Cellular Handsets 2
Envelope Tracking Overview Modern high spectral efficiency wireless communications standards have high Peak to Average Power Ratio (PAPR) e.g. 4G/LTE, 802.11ac WiFi 3 Conventional fixed supply Power Amplifier has to use supply voltage high enough to support Peak Power, but is only energy efficient at the peaks. Most of the time the voltage is much higher than needed, resulting in high PA heat dissipation In ET, the PA supply voltage is dynamically adjusted to the instantaneous amplitude of the signal, resulting in high PA efficiency at all times
PA Efficiency Curves: ET vs Fixed supply The efficiency of a Fixed Supply PA is much lower at average Pout than at peak Pout. ET PA efficiency is only slightly lower at average Pout The difference in average PAPR between fixed supply and ET PAs increases with signal PAPR 4
Wireless standard trends Standard Launched Typical Carrier BW (MHz) Typical Spectral Efficiency (bps/hz) Approx PAPR(dB) 1G cellular AMPS 1983 0.03 0.046 0 2G cellular GSM 1991 0.2 0.17 0 Digital TV DVB-T 1997 8 0.55 8 3G cellular WCDMA FDD 2001 5 0.51 7 2.75G cellular GSM + EDGE 2003 0.2 0.33 3.5 Wi-Fi IEEE 802.11a/g 2003 20 0.9 9 WiMAX IEEE 802.16d 2004 20 1.2 8.5 Wi-Fi IEEE 802.11n 2007 20 2.4 9 3.5G cellular HSDPA 2007 5 2.88 8 Digital TV DVB-H 2007 8 0.28 8 4G cellular LTE 2009 20 16 5-10 WiFi IEEE 802.11ac 2012 80 14 10 4G cellular LTE-A 2013 20 30 5-10 5G cellular 5G-NR 2018 100-5-10 Increasing Spectral Efficiency Increasing PAPR Decreasing Conventional PA efficiency Need for PA efficiency enhancement techniques such as ET continues to grow Envelope Tracking Implementation difficulty increases with carrier bandwidth 5
ET System Elements Envelope detection: most accurate if performed in digital domain Envelope shaping: Determines relationship between RF power and PA supply voltage Envelope Amplifier: High BW, Low Noise, High efficiency Amplifier used to generate PA supply voltage ET PA: ET can be applied to standard fixed supply PA. Improved performance possible by optimising PA for ET operation PA operates in polar mode at high instantaneous power and linear mode at low instantaneous power Delay Alignment: ET requires accurate (~ns) timing alignment between envelope and RF paths. Most accurate / repeatable if performed in digital domain 6
Envelope Tracking Signal Processing Differential Analogue Envelope Interface BW of Envelope path needs to be 2-3x RF bandwidth Delay alignment between RF and Envelope paths at PA (interfaces B and C) is essential to achieve good RF linearity (ACLR / EVM). Sub-sample timing alignment ( <ns) required for wideband signals 7 Analogue imperfections in both Envelope and RF paths must be corrected Gain / DC Offset in Envelope path Gain vs frequency in RF path
Envelope Processing Basics 1.5 Swing Range Optimise efficiency of combined modulator /PA Prevent gross PA nonlinearity due IV curve knee Envelope Shaping Control envelope bandwidth Optimise efficiency Can be used to linearise PA Timing Alignment Timing error leads to memory effect (AM-PM) Fine adjustment necessary (~1ns) 1 0.5 0-0.5-1 1.5-1.5 1 0.5 0-0.5-1 -1.5 8
ET High Power operation Minimum PA supply voltage (Vmin) set by Envelope shaping table Determined by PA technology e.g GaAs HBT Vmin chosen to ensure the PA never operates in its highly non-linear region At high power, PA supply voltage swings between Min voltage defined in shaping table up to Max voltage supported by PA As power is backed off, Vmax falls but Vmin remains unaltered i.e ET swing range reduces 9
ET Lower Power operation ET generally used for highest 10dB operating power range At lower power there is insufficient swing range to significantly improve PA efficiency Static power dissipation of ET modulator starts to outweigh PA efficiency benefit Average Power Tracking (APT) typically used at lower powers PA DC supply voltage varied with average slot power 10
ET PA = 3 Port Device Vsupply RFin ET PA RFout Envelope Tracking PA can be considered to be a 3 port network RF Input, Supply Input, RF output PA Gain, Phase, Efficiency influenced by PA Device technology PA Circuit design (matching, biasing) Instantaneous Supply Voltage 3D Characterisation of PA surfaces allows ET system performance to be predicted Gain, Phase, Efficiency vs Instantaneous (Pin,Vsupply) 11
ET PA + Shaping Table = 2 Port Device Vsupply RFin ET PA RFout Envelope Shaping table defines Rfin to Vsupply mapping Once shaping is defined the ET subsystem is reduced to a 2 port network Envelope Shaping determines key ET PA black box metrics Linearity (AM/AM and AM/PM) Efficiency Gain System Linearity metrics for a wide range of waveforms can then be derived 12 Allows EVM / ACLR / Efficiency trade-offs to be explored PA / system memory effects are not captured but good predicted vs measured results can still be obtained for low-med BW waveforms
Characterising the ET PA Test methodology PA current measurement Supply impedance Supply bandwidth requirements ET Efficiency prediction ET Linearity prediction Parameters measured Swept CW testing Bench PSU Low (decoupling Capacitor) Low (Bench PSU) Poor, due to PA die heating Poor, due to PA die heating Gain (AM:AM), Efficiency Pulsed RF /DC testing Instrumentation grade current probe, ~5 us resolution Low (decoupling Capacitor) Low (Bench PSU) Good, if short pulses (~10 us, 10% duty cycle). Fair Gain (AM:AM), Efficiency Dynamic supply modulation Challenging high BW with high common mode voltage current sense Requires low impedance dynamic supply (no decoupling) High (~60 MHz BW) V. Good V. Good Gain (AM:AM), Phase (AM:PM), Efficiency 13
PA Surface Characterisation setup Supply Modulator: High BW / Low output impedance Dynamic measurement of: Gain, Phase, Efficiency PA Supply Current: High BW / High CM rejection 14
Gain (db) Example 3D surface model of ET PA Trajectory across surfaces set by choice of shaping table shown by black lines Gain Surface Phase Surface Efficiency Surface 15
Alternative 2D views of 3D ET Gain Surface 3D view 2D Shaping table view (Colour = Gain) 2D Waterfall view Gain vs Power parameterised by Vsupply 16
Gain Efficiency ET Shaping Table - Optimum Efficiency vs Isogain Optimum Efficiency Shaping Isogain shaping ET PA shows gain expansion Kink in AM/AM characteristic is very difficult to linearise high BW required Flat gain characteristic No DPD required if ET PA has low AM/PM 17
Isogain Shaping Families Higher Gain Low swing range Low Efficiency Lower Gain High swing range High Efficiency 18
Ideal ET PA Gain characteristic Unlike traditional PAs, low voltage gain reduction is desirable as it allows isogain operation with high gain compression at high voltages (= high efficiency) Gain peaking with fixed voltage supply is unimportant these points are not visited when in ET mode Flat gain at lowest voltage is required (PA operates in linear mode at low voltages) Best ET efficiency achieved if high peak efficiency can be maintained at low supply voltages 19
ET PA AM/PM optimisation APT optimised PA (Flat AM/PM with fixed supply) ET optimised Phase Characteristic (Flat AM/PM with isogain shaping) Different Phase optimisation required for ET PA Best ET PA performance achieved by designing PA for ET to meet ET performance objectives from the outset, rather than by adapting an APT PA 20
ET Noise and Distortion mechanisms in ET When operating in ET mode a PA is in compression and acts as a mixer 90% Supply Sensitivity Noise and distortion on ET supply mixes with RF generating unwanted sidebands Supply noise transfer is significantly higher than for fixed supply or APT PA Supply noise transfer is controlled by PA compression level set by ET shaping table To maintain full ET efficiency benefit, supply noise needs to sufficiently low to avoid need to back off compression level Some distortion mechanisms can be corrected at system level using DPD, others 21 cannot Correctable distortion: Frequency Response, ET modulator Output Impedance Non-correctable noise/distortion: Switcher noise, Slew rate limiting, Thermal noise
Supply Sensitivity Determines Supply Noise RF conversion Partial derivative of PA RF output voltage wrt PA supply voltage S = V rfout V supply Ideal AM mixer has S=1 (100% supply sensitivity) ET PA in hard compression S ~ 1, Fixed Supply Linear PA S ~ 0.2-0.3 dbm 0 SoftPlot Measurement Presentation Trace A dbm 0 SoftPlot Measurement Presentation Trace A -10-20 -30-40 40MHz Supply Spectrum -10-20 -30-40 RF Spectrum 40MHz 1 2-1 -50-50 -60-70 2-1 1-60 -70 5 6-5 3 4-3 -80-80 -90-90 22-100 Start: 0 Hz Stop: 200.0000 MHz Res BW: 100 khz Vid BW: 300 khz Sweep: 200.00 ms 17/11/2010 18:44:43 FSEB 30 40MHz test tone added to Envelope Amplifier O/P (whilst amplifying 5MHz WCDMA signal) Mkr Trace X-Axis Value Notes 1 Trace A 39.2786 MHz -64.56 dbm 2-1 Trace A 801.6032 khz -22.25 dbm{sum} Average DC drain voltage 2.62V Skyworks 174 isogain 25dB 28.8dBm Avg DC drain volts = 2.62V peak = 3.9V Trace A 40MHz 0dBm injected tone Measured 40MHz injected tone level Measurement Parameter Value 3.84 MHz Calculated RF sideband level for ideal AM modulator Channel bandwidth Channel spacing 40.00 MHz -65.53 dbm Measured RF sideband level On- channel power Adjacent channel power (channel -1) 4.45 db PA Supply Sensitivity (db) 0.09 Adjacent channel power (channel +1) 17.3mV rms -49.6dBC -51dBC -1.4dB PA Supply Sensitivity (%) 85% -100 Start: 1.8500 GHz Stop: 2.0500 GHz Res BW: 100 khz Vid BW: 300 khz Sweep: 200.00 ms 17/11/2010 18:36:58 FSEB 30 Skyworks 174 ET Iso gain 25dB 5MHz WCDMA 28.8dBm Trace A Corresponding RF sidebands Mkr Trace X-Axis Value Notes 1 Trace A 1.9474 GHz -37.61 dbm 2-1 Trace A 1.9522 GHz 8.54 dbm{sum} 3 Trace A 1.9875 GHz -72.81 dbm 4-3 Trace A 4.8096 MHz -41.89 dbm{sum} 5 Trace A 1.9073 GHz -74.51 dbm 6-5 Trace A 4.8096 MHz -42.89 dbm{sum} Injected CW tone (0dBm @ 40MHz) V V rf rf V V env env V V V V rf rf env env
Envelope Tracking Modulator Requirements Hybrid Switch-mode / Linear-mode ET Modulator 23
ET CFR Via Shaping Table Very simple hardware Configurable soft clipping provides simple CFR for free ET much more amenable to open loop correction than fixed supply PA operates in compression over most of envelope cycle Moderate linearity performance Moderate bandwidth capability No direct phase correction relies on PA designed for low ET AM/PM No memory correction high intrinsic ET modulator performance needed 24
ET and Digital PreDistortion Envelope Detection Envelope Shaping Envelope Amplifier r DPD only needed to clean up residual AM/AM errors and correct ET PA s AM/PM IQ CFR DPD PA (if required) Memoryless DPD ET PA Subsystem CFR applied to source waveform (both Envelope and RF are clipped) 25
Thank You Questions