RF Harmonic Oscillators Integrated in Silicon Technologies

Transcription

1 RF Harmonic Oscillators Integrated in Silicon Technologies Pietro Andreani Dept. of Electrical and Information Technology (EIT) Lund University, Sweden SSCS Distinguished Lecture Toronto, Ontario, Canada Friday, June 9 th, 2017

2 Overview o Popular harmonic oscillators Phase noise Architectures for low 1/f 3 phase noise o Design techniques for very wide frequency tuning range RF CMOS VCOs 2 of 59

3 LC resonator We begin with an inductor-capacitor resonator R L R C L C L C L C R w = 0 1 LC w» 0 1 LC Z ( w ) 0 = R 3 of 59

4 Building a harmonic oscillator Tank losses are compensated by an active negative resistance in parallel to the tank L C R R active Ractive R < R Q = w0rc = w L 0 4 of 59

5 Colpitts oscillator Classical embodiment of active negative resistance with only one active device V DD V out L R V DD V out V B C 1 I tank t I B C 2 I tank 5 of 59

6 Cross-coupled differential-pair oscillator Extremely popular oscillator family V DD I B I B Operation in class-b V + V - t V - V DD N1 N2 V + I B A diff 2 = IR B p t 6 of 59

7 Class-B with double switch pair V DD I B -I B t V + V - V + V - V DD I B A diff 4 = IR B p Compared to single-switch-pair: double amplitude, but also double V DD (no swing above V DD ) t 7 of 59

8 Real oscillations o Phase uncertainty grows with time Caused by various noise sources jitter o Jitter increases without bound in a free-running oscillator o In the frequency domain, the oscillator displays phase noise dbc Phase noise, in dbc/hz f 0 f 8 of 59

9 Why bother? Phase noise in transceiver is important for at least three reasons: q In a receiver, it can downconvert large nearby signals on top of the desired signal q In a transmitter, it can increase the noise floor in the receive band q In both, it can directly corrupt the phase information in the signal Not seldom, the phase noise of the VCO is the bottleneck for the whole radio performance 9 of 59

10 Phase noise LTI approach L C R R act 2 ( 1- LC) 1 w Y( w) = G+ jwc+ - Gact = Ł jwlł jwl Z ( w w) ( +D ) ( w w) j w w L w L +D»»-j ( 1- Dw 0 +D LC) 2 w w0l w0l 1 w0 Z( w0 +D w) = = R = R Dw Dw 2 R 2 2QDw w w of 59

11 Phase noise from tank losses i 2 n L C v i n n 0 Z ( 0 ) 4kBTG R w w = w +D w = = 4kBTR Df Df Ł 2QDwł Ł2Q Dwł q Both amplitude and phase noise, but amplitude noise is rejected q Thus, phase noise is defined as half the above expression, normalized to the output signal power (in db below the carrier per Hertz, dbc/hz): L 2 v 2 2k 1 n BTR D = 10log10 = 10log Apk 2 Ł Apk 2 Ł2Q Ł ł ( w) w0 Dwł 2 ł 11 of 59

12 Leeson s equation L L ( Dw) ( w) 10log 2k TR Ø 1 B 0 D = 10 F Œ 2 1+ œ Apk 2 2Q Dw Ł w 2 ø Œº Ł ł œßł 1+ Ł Up-converted 1/f noise (30dB/dec, 1/f 3 region) Dw 1 f Dw 3 ł Thermal noise (20 db/dec, 1/f 2 region) Ideal L( Dw ) Noise floor Dw 1 f 3 w0 2Q log ( Dw ) 12 of 59

13 Hajimiri and Lee s theory of phase noise d ( t t) - C L V ( t) No phase noise out Conversion of noise into phase noise is timedependent LTV phase noise analysis needed! Maximum phase noise V out DV V out DV d ( t -t) d ( t -t) Hajimiri and Lee, JSSC Feb of 59

14 Impulse sensitivity function (ISF, G) q Current noise source i ( f ) is weighed by associated G i ( f ) n n effective current noise i ( f) = i ( f) G ( f) n eff n i, n ( f = w t 0 ) q ISF is dimensionless, frequency- and amplitude independent, with period 2p: c G = + n ( f) c cos( nf f ) n= 1 n 14 of 59

15 Phase noise expression ( ) If i is a (cyclo)stationary white current noise source, n f its contribution to 1/f 2 phase noise is L ( w) D = 10log Ł 2 i 2 n, eff, rms 2 2 ( CA ) ( Dw) pk ł ( ) i f n + ( + f( )) Apk cos wt t - 15 of 59

16 Example of ISF LC oscillators V out = A pk cos( f) ( ) i f n + V - out f G i n =-sin ( f ) f Hajimiri and Lee, JSSC Feb. 98; Andreani and Wang, JSSC Nov of 59

17 A particularly simple case Parallel RLC resonator again phase noise from tank losses: L i i n, eff, rms n i, rms D = 10log = 10log Ł 2( CApk ) ( Dw) ł Ł 2( CApk ) ( Dw) ł ( w) G n i 2 n L C = 10log Ł = 10log Ł 2 4k TG 12 B ( CApk ) ( Dw) 2kBTR 1 2 Apk 2Ł2Q 2 2 ł w0 Dwł 2 ł Leeson with F=1 recovered without any ad hoc assumptions! 17 of 59

18 Phase noise from MOS pair ( f) = 4 g ( f) i ( f) = i ( f) G ( f) i k T g 2 N1 B n mn, 1 N1, eff N1 N1 ( f) G» N1 1 I N2 I N1 V + V - N1 N2 I B -F F Two commutations in one oscillation period 18 of 59

19 L Total phase noise Ø kt ( ) ( ) ø B D = 10log Œ 1+ g n œ Œº RAtankC Dw œß ( w) Tank F MOS pair = + 1 g n q Transistors appear only through channel noise factor g n q Transistor phase noise always proportional to tank noise (60% from tank, 40% from MOS pair, if g n = 2/3) q This is because: 1) transistor noise is proportional to commutation time, 2) which is inversely proportional to the oscillation amplitude, 3) which is proportional to the tank parallel resistance q A simple-minded LTI analysis would yield very wrong predictions (i.e., MOS phase noise increases with MOS g m ) 19 of 59

20 MOS phase noise invariance ( f) = 4 g ( f) i k T g 2 N1 B n mn, 1 F = n 1 2IB A b tank n 2 bn Two effects balance each other: 1) Larger MOS produces more noise during current commutation, and 2) Larger MOS allows a faster commutation 1 bn Result: the two areas are identical -F -F n n,2 b F n,2 b F n Andreani et al., JSSC May of 59

21 Alternative phase noise analysis Resonator V Y + - I G M Transconductor Matrix-based Fourier-series LTV approach, starting from r I ur = YV and uur di = G M uuur dv All quantities are functions of w 0, 2w 0,, nw 0 Pepe and Andreani, TCAS-I, Feb of 59

22 Results of new phase noise analysis o Rigorous analysis under very broad hypotheses G M pure transconductance; Y linear; G M noise proportional to G M via g o Phase noise from G M always in proportion of g:1 to phase noise from Y, independently of resonator and transconductor nature o Phase noise expressions as functions of V and Y o Closed-form, explicit phase noise expressions if Q is high General case of Y resonating at multiples of w 0 22 of 59

23 Double-switch pair vs. single-switch pair Double-switch (DS) pair oscillator Single-switch (SS) pair oscillator A DS 4 = IR B p A SS 2 = IR B p V + V - V + V - I B What phase noise difference should we expect? I B 23 of 59

24 DS pair vs. SS pair phase noise L L DS SS ( w) 2 2kBTR 1 w0 10log 1 2 ADS Q Dw gn + g p D = + Ł 2Ł2 ł Ł 2 łł 2k TR 1 2 Ł ASS 2 Ł2Q B 0 ( D w) = 10log ( 1+ g ) w Dwł o 60% from tank, 40% from transistors If g n = g p = 2/3 2 n ł o If I B,DS = I B,SS and g n = g p A = 2 A fi L = L -6dB DS SS DS SS (!) Andreani and Fard, JSSC Dec of 59

25 DS vs. SS MOS noise Area (SS) = 2 Area (DS) SS 4 DS transistors make as much noise as 2 SS transistors! DS -F n, SS -F n, DS F n, DS F n, SS 25 of 59

26 Figure of merit (FoM) Phase noise normalized to power consumption, oscillation frequency, and frequency offset 2 ( D ) ( Dw) DC[ mw ] 2 w0 w -3 2Q h FoM = = 10 L P k T F High tank Q crucial for high FoM B power efficiency noise factor Andreani et al., JSSC Dec. 2006; Garampazzi et al., JSSC Jul. 2015; Murphy et al., ISSCC of 59

27 SS vs DS PN and FoM with fixed V dd DS oscillation limit SS oscillation limit Phase Noise FoM DS max FoM DS SS SS max FoM I B 4I B Liscidini et al., ISSCC 2012, JSSC Mar of 59

28 Current bias resistive source Resistor V + V - N 1 N 2 I B Very simple bias circuitry No 1/f noise generation Low-Z source Significant upconversion of 1/f noise from N 1 -N 2 Abidi and Ismail, ISSCC of 59

29 Current bias MOS source Tail current source V + N 1 N 2 V - V B High-Z source Less upconversion of 1/f noise from N 1 -N 2 Own 1/f noise generation I B 29 of 59

30 Impact of parasitic tail capacitance C tail,par o C tail,par + cross-coupled MOS entering linear region MOS contribution to phase noise increases, even by a large amount o C tail,par good for filtering HF noise from bias o But, increase of 1/f noise upconversion 30 of 59

31 Overview o Popular harmonic oscillators With some results about phase noise Architectures for low 1/f 3 phase noise o Design techniques for very wide frequency tuning range RF CMOS VCOs 31 of 59

32 Possible solution noise filter L tail C tail,par C BIG o Noise filter: C tail,par resonates with L tail at 2w 0 MOS switches see high-z at 2w 0 o C BIG filters tail noise and acgrounds L tail o C BIG includes C DB of MOS tail long and large MOS, low 1/f noise o Drawbacks: narrow-band, C tail,par must be known with some precision, extra L tail Hegazi, Sjöland, Abidi, JSSC Dec of 59

33 Dramatic performance improvement o 0.35μm CMOS o 1.2GHz, 3.5mA, 2.5V o L tail = 10nH, C BIG = 40pF o FoM = 196dBc/Hz o TR? 33 of 59

34 More on tail filter C BIG o Many variations on the same basic theme o Extremely popular! L tail C tail,par Hegazi, Sjöland, Abidi, JSSC Dec of 59

35 A recent variation 7mW o Tail/top resonance with Xrfm o Very low PN and great FoM (up to 195.6dBc/Hz) o kHz 1/f 3 corner Garampazzi et al., JSSC July /2 35 of 59

36 Alternative to tail resonance o Design tank for differential resonance at w 0 and common-mode resonance at 2w 0 2w w 0 Also here, the resonance must track the 0 resonance two capacitor banks Babaie et al., RFIC 2013, JSSC Mar. 2015; Shahmohammadi et al., ISSCC 2015; Murphy et al., ISSCC of 59

37 Class-F 2 (or, here, F 2,3 ) oscillator o GHz; 1V, 10-12mW o Low 1/f 3 corner (60-130kHz) o Very good FoM (~191dBc/Hz) at very low PN 1MHz) o Very low V DD pushing (12-23MHz/V) Shahmohammadi et al., ISSCC of 59

38 Implicit common-mode resonance o GHz; 0.7V, 0.5mW o Low 1/f 3 corner (200kHz) o Great FoM ( dBc/Hz) Murphy et al., ISSCC of 59

39 A totally different approach class-c I N1 I N2 I w0» I bias V + V - N 2 N 1 I B C tail o C I w0» tail turns class-b into class-c: p optimal differential Colpitts oscillator 2 Ibias o Ideally, 3.9dB lower phase noise for the same bias current o Also here, C tail filters off highfrequency noise from tail, and includes tail C DB long and large MOS, low 1/f noise Mazzanti and Andreani, JSSC Dec of 59

40 Original prototype o 4.90GHz < f c < 5.65GHz o 1V, 1.4mW o 193.5dBc/Hz<FoM<196dBc/Hz f osc = 5.52GHz f osc = 4.90GHz 40 of 59

41 Design issues in class-c CMOS VCO V + V - V B C tail o Diff-pair must not enter the linear region (otherwise, large PN penalty) shift of MOS DC gate voltage V B (which may be generated with feedback loop) o RC bias should not load tank (transformer feedback possible) o Nevertheless, higher maximum oscillation amplitude in the ideal class-b CMOS oscillator o Class-C very attractive for BJT VCOs Mazzanti and Andreani, JSSC Dec. 2008; Fanori and Andreani, JSSC July of 59

42 Class-C VCO in SiGe BJT process o GHz; 3.3V, 5.5mA o PN = 1MHz o FoM =189dBc/Hz o Lower PN (~2dB) possible with Colpitts, but degraded FoM Padovan et al., TCAS-I Feb of 59

43 Colpitts VCO in SiGe BiCMOS process o GHz; 4.0V, 17.5mA o PN = 1MHz (best) o FoM =188dBc/Hz Boscolo et al., to be presented at ESSCIRC of 59

44 Overview o Popular harmonic oscillators With some results about phase noise Architectures for low 1/f 3 phase noise o Design techniques for very wide frequency tuning range RF CMOS VCOs 44 of 59

45 VCOs in modern radios I o Carrier aggregation requires several harmonic VCOs Active at the same time Should not pull one another o Band proliferation favors VCOs with a very wide tuning range (TR) Wider than 1 octave is particularly attractive 45 of 59

46 VCOs in modern radios II o VCO with 8-shaped tank inductor Much less sensitive to external magnetic fields Generates itself a vanishing magnetic field Slightly lower Q acceptable Often used M. Nilsson et al., ISSCC of 59

47 Very-Wide-TR VCOs I o Two or more VCOs with overlapping TRs Saves power, costs area Very popular choice in real-life products Hadjichristos et al., ISSCC of 59

48 Very-Wide-TR VCOs II o Large switchable C in parallel to small L floating switches power wasted at low frequencies, compared to reasonable phase-noise specs power cannot be decreased without killing the oscillation Sjöland, TCAS-II May of 59

49 Very-Wide-TR VCOs III o Switchable L Ultra-wide TR possible Difficult to obtain low PN at high FoM Additional issue: switchable 8-shaped inductor Sadhu et al., CICC of 59

50 Very-Wide-TR VCOs IV o Transformer-based VCOs Two resonances with overlapping TRs TR > 1 octave Difficult to design an 8-shaped transformer Bevilacqua et al., TCAS-II Apr. 2007; Li et al., JSSC June of 59

51 Very-Wide-TR VCOs V o Mode-switching VCO 4 inductors, two oscillation modes Rejects external magnetic fields TR > 1 octave Excellent PM and FoM Large area Taghivand et al., ISSCC of 59

52 Very-Wide-TR VCOs VI o Double-core VCO Two concentric 8-shaped coils do not interfere (much) with each other TR > 1 octave; saves inductor area, sub-optimal Q Fanori et al., ISSCC of 59

53 Very-Wide-TR VCOs VII o Reconfigurable active core Standard LC tank design (i.e., with very large capacitance) Negative resistance: either single-switch (nmos) pair SS mode or, double (complementary nmos-pmos) switch pair DS mode DS mode avoids power waste at lower frequencies Liscidini et al., ISSCC 2012, JSSC Mar of 59

54 SS pair vs. DS pair, again DS oscillation limit SS oscillation limit DS SS DS max FoM SS max FoM I B 4I B 54 of 59

55 Very-Wide-TR reconfigurable VCO Phase noise DS-mode, I DC = 4 ma Phase noise SS-mode, I DC = 8 ma DS SS SS SS DS SS o STM 28nm UTBB FD-SOI CMOS o GHz o -154<PN o 186 < FoM (dbc/hz) < 189 o 300kHz < 1/f 3 corner < 3MHz Fanori et al., RFIC of 59

56 Conclusions o Rigorous phase noise results For transconductor-based oscillators o Class-B VCOs simple, robust, ubiquitous Tail filter improves phase noise, even largely Recent proposals: common-mode tank resonance at 2w 0 o Class-C VCOs better efficiency than standard class-b, but more complicated Class-C must be enforced for all working conditions Excellent for BJT VCOs o Several techniques for very wide frequency tuning range None is a clear winner 56 of 59

57 References I 1. A. Hajimiri and T. H. Lee, A general theory of phase noise in electrical oscillators, IEEE J. Solid-State Circuits, vol. 33, no. 2, pp , Feb F. Pepe and P. Andreani, A General Theory of Phase Noise in Transconductor-Based Harmonic Oscillators, IEEE Trans. Circuits Syst. I, vol. 64, no. 2, pp , Feb P. Andreani et al., A study of phase noise in Colpitts and LC-tank CMOS oscillators, IEEE J. Solid-State Circuits, vol. 40, no. 5, pp , May P. Andreani and A. Fard, More on the 1/f 2 phase noise performance of CMOS differentialpair LC-tank oscillators, IEEE J. Solid-State Circuits, vol. 41, no. 12, pp , Dec M. Garampazzi et al., Analysis and Design of a dbc/hz Peak FoM P-N Class-B Oscillator With Transformer-Based Tail Filtering, IEEE J. Solid-State Circuits, vol.50,no. 7, pp , Jul D. Murphy et al., A VCO with Implicit Common-Mode Resonance, in Proc. of the IEEE ISSCC 2015, pp , A. Liscidini et al., A 36mW/9mW Power-Scalable DCO in 55nm CMOS for GSM/WCDMA Frequency Synthesizers, in Proc. of the IEEE ISSCC 2012, pp , A. Liscidini et al., A 36mW/9mW Power-Scalable DCO in 55nm CMOS for GSM/WCDMA Frequency Synthesizers, IEEE J. Solid-State Circuits, vol. 49, no. 3, pp , Mar A. Ismail and A. Abidi, CMOS differential LC oscillator with suppressed up-converted flicker noise, in Proc. of the IEEE ISSCC 2003, pp , E. Hegazi et al., A filtering technique to lower LC oscillator phase noise, IEEE J. Solid- State Circuits, vol. 36, no. 12, pp , Dec of 59

58 References II 11. M. Babaie et al., Ultra-low phase noise GHz clip-and-restore oscillator with 191 dbc/hz FoM, in Proc. of the IEEE RFIC 2013, pp , M. Babaie and R. B. Staszewski, An Ultra-Low Phase Noise Class-F 2 CMOS Oscillator With 191 dbc/hz FoM and Long-Term Reliability, IEEE J. Solid-State Circuits, vol.50,no.3,pp , Mar M. Shahmohammadi et al., A 1/f Noise Upconversion Reduction Technique Applied to Class-D and Class-F Oscillators, in Proc. of the IEEE ISSCC 2015, pp , D. Murphy et al., A Complementary VCO for IoE that Achieves a 195dBc/Hz FOM and Flicker Noise Corner of 200kHz, in Proc. of the IEEE ISSCC 2016, pp , F. Pepe et al., Suppression of Flicker Noise Up-Conversion in a 65-nm CMOS VCO in the 3.0-to-3.6 GHz Band, IEEE JSSC, vol. 48, no. 10, pp , Oct A. Mazzanti and P. Andreani, Class-C harmonic CMOS VCOs, with a general result on phase noise, IEEE J. Solid-State Circuits, vol. 43, no. 12, pp , Dec L. Fanori and P. Andreani, Highly Efficient Class-C CMOS VCOs, Including a Comparison With Class-B VCOs, IEEE J. Solid-State Circuits, vol. 48, no. 7, pp , July F. Padovan et al., Design of Low-Noise E-Band SiGe Bipolar VCOs: Theory and Implementation, IEEE Trans. Circuits Syst. I, vol. 62, no. 2, pp , Feb F. Boscolo et al., A 21GHz 20.5%-Tuning Range Colpitts VCO with -119dBc/Hz Phase Noise at 1MHz Offest, to be presented at IEEE ESSCIRC L. Fanori and P. Andreani, Class-D CMOS Oscillators, IEEE J. of Solid-State Circuits, vol. 48, no. 12, pp , Dec M. Babaie and R. B. Staszewski, A Class-F CMOS Oscillator, IEEE J. Solid-State Circuits, vol. 48, no. 12, pp , Dec of 59

59 References III 22. M. Garampazzi et al., An Intuitive Analysis of Phase Noise Fundamental Limits Suitable for Benchmarking LC Oscillators, IEEE J. Solid-State Circuits, vol. 49, no. 3, pp , Mar M. Nilsson et al., A 9-band WCDMA/EDGE transceiver supporting HSPA evolution, in Proc. of the IEEE ISSCC 2011, pp , A. Hadjichristos et al., Single-chip RF CMOS UMTS/EGSM transceiver with integrated receive diversity and GPS, in Proc. of the IEEE ISSCC 2009, pp , H. Sjöland, Improved Switched Tuning of Differential CMOS VCOs, IEEE Trans. Circuits Syst. II, vol. 49, no. 5, pp , May B. Sadhu et al., A CMOS GHz Wide Tuning Range, Low Phase Noise LC VCO, in Proc. of the IEEE CICC 2009, pp , A. Bevilacqua et al., Transformer-based dual-mode voltage-controlled oscillators, IEEE Trans. Circuits Syst. II, vol. 54, no. 4, pp , Apr G. Li et al., A Low-Phase-Noise Wide-Tuning-Range Oscillator Based on Resonant Mode Switching IEEE J. Solid-State Circuits, vol. 47, no. 6, pp , June M. Taghivand et al., A 3.24-to-8.45GHz Low-Phase-Noise Mode-Switching Oscillator, in Proc. of the IEEE ISSCC 2014, pp , L. Fanori et al., A 2.4-to-5.3GHz Dual-Core CMOS VCO with Concentric 8-Shaped Coils, in Proc. of the IEEE ISSCC 2014, pp , L. Fanori et al., A 2.8-to-5.8 GHz Harmonic VCO in a 28 nm UTBB FD-SOI CMOS Process, in proc IEEE RFIC, pp , of 59