RF System Design Outline - Columbia EE [PDF]

RF System Design Peter Kinget Bell Laboratories Lucent Technologies Murray Hill, NJ, USA

Outline • Circuits for Wireless • Wireless Communications – duplex, access, and cellular communication systems – standards

• Receivers: – heterodyne – homodyne – image reject

• Transmitters – modulation – up-conversion

• Transceivers – frequency synthesis – examples © Peter KINGET

03/99 Page2

RF IC design Market Requirements Communication Theory Microwave techniques

TRANSCEIVER Receiver Freq.Synth. Transmitter

Discretes

Modulation

Standards Architectures

IC design RF, mixed-mode, digital © Peter KINGET

Circuits for Wireless

03/99 Page3

Circuits for Wireless - Overview • Noise limits the smallest signal • noise figure • cascade of stages

• Distortion limits the largest signal – large (interfering) signals: • compression, blocking, and desensitization • inter-modulation • cascade of stages

• Dynamic Range

© Peter KINGET

03/99 Page5

Noise Figure • Max. thermal noise power from linear passive network e.g. antenna: Nmax = kT ⋅ BW added

N (S / N)in • Noise Factor: F = = 1 + eq@ input (S / N)out Nin • Noise Figure: NF = 10 log10 (F ) ≥ 0dB NF F [dB] 0 1 1 1.25 2 1.6 3 2

(S/N)out=1/2 (S/N) in © Peter KINGET

03/99 Page6

Cascade of Stages: Friis Equation Avail. Power Gain: Ap1 Noise factor: F1 Rs Eg

Vi1 Ri1

Ap2 F2 Ro1

Ro2

Vi2 Ri2

Av1Vi1

F = 1 + (F1 − 1) +

Ap3 F3

Av2Vi2

Ro3

Vi3 Ri3

Av2Vi2

(F2 − 1) (F3 − 1) + A P1 AP 1 A P 2

later blocks contribute less to the noise figure if they are preceded with gain © Peter KINGET

03/99 Page7

Noise of lossy passive circuit • Lossy passive circuit (e.g. filter): • E.g. Band-select filter & LNA:

F = Loss

LNA

F = Ffilt + F=L+

(FLNA − 1) AP

(FLNA − 1) = L ⋅ FLNA 1 /L

Loss adds immediately to noise figure ! © Peter KINGET

03/99 Page8

Sensitivity • sensitivity = minimal signal level that receiver can detect for a given (S/N) at the output: F=

P (S / N) in 1 = signal _ in ⋅ (S / N) out Pnoise _ in (S / N) out

Psignal _ in = F ⋅ (S / N) out ⋅ Pnoise _ in = F ⋅ (S / N) out ⋅ kT ⋅ BW

• E.g. GSM (BW=200kHz, (S/N)out > 9dB): Psignal _ in = NF + (S / N) out − 174dBm / Hz + 10 log10 (BW) = 6 + 10 − 174 + 53 = − 105dBm

for a receiver with a noise figure of 6dB © Peter KINGET

03/99 Page9

Distortion: xin

yout

• Circuits have non-linearities – hard: e.g. supply clipping 2 – weak: y out = G1 ⋅ x in + G 2 ⋅ x in

• Effects:

3

+ G 3 ⋅ x in + L

G1 >> G 2 & G1 >> G3

– Gain compression – Blocking & Desensitization – Inter-modulation: IP2 & IP3

• Cascade of stages © Peter KINGET

03/99 Page10

Gain Compression Pout [dBm]

Pin

1dB

Pout

1

Gain [dB] Pin [dBm]

P-1dB

Pin ä è Gain æ è desensitization è blocking ©

Peter KINGET 03/99 Page11

Inter-modulation: 2nd order

ω2+ω1

DC

ω2-ω1

ω1 ω2

2ω1

2ω2

P out [dBm]

1

2

P IIP2

P in [dBm]

© Peter KINGET

03/99 Page12

Inter-modulation: 3rd order

2ω2+ω1

2ω1+ω2 ω1 ω2 3ω1 2ω1-ω2 2ω2-ω1

3ω2

P out [dBm]

1

3

P IIP3

P in [dBm]

© Peter KINGET

03/99 Page13

IIP3 for a cascade of stages xin

1 A

2 IIP 3

≤

A GA1

y1

1 A 2A IIP 3

B GB1

y2

C GC1

y3

G2A1 G 2A1 ⋅ GB21 + 2 + A B IIP 3 A 2C IIP 3

• worst-case approximation for narrow band systems ! • voltage/current levels and gains • effect of non-linearities more important at later stages ! © Peter KINGET

03/99 Page14

Spurious Free Dynamic Range dynamic range =

max. input level min. input level

• under certain conditions: – min. level such that (S/N) out is sufficient – max. level such that: effects of non-linearities are ≤ noise i.e. IM3 products ≤ noise

• other applications use different conditions

© Peter KINGET

03/99 Page15

© Peter KINGET

03/99 Page16

Spurious Free Dynamic Range

Pout [dBm] SFDR 1

3

SNRout G+NFL Pinmin SFDR

Pinmax Pin [dBm]

Wireless Communication Systems

Wireless Communications - Overview • ‘ether’ is one medium shared by all • 1st problem: Duplexing – how to arrange for a two way communication link

• 2nd problem: Multiple Access – how to arrange for multiple users

© Peter KINGET

03/99 Page18

Duplexing - Overview • Establish two way communications: – Time division duplex: • same rcv and xmt frequency channel • alternating in time between rcv & xmt – Frequency division duplex • different frequency channel for rcv and xmt • full duplex possible

© Peter KINGET

03/99 Page19

© Peter KINGET

03/99 Page20

Time Division Duplex (TDD)

• peer to peer communications • antenna switch

TDD design issues + mobile units can communicate + Switch low loss ( 9dB – signal range: -102dBm to -15dBm for signal of -99dBm: – blocking: in band: -43 up to -23dBm out of band: 0dBm – inter-modulation: -49dBm @800kHz & @1600kHz for signal of -82dBm: – co-channel test: 9dB smaller interferer in same channel – adjacent channel (@200kHz): 9dB larger – alternate channel (@400kHz): 41dB larger

© Peter KINGET

03/99 Page33

GSM Type approval (summary) • Transmitter

Power

Frequency

– close-in: modulation spectrum (spectral mask) – wide-band: noise spectrum e.g. • noise@3MHz < -115dBc/Hz • noise@6MHz < -130dBc/Hz • noise@25MHz < -130/-136dBc/Hz – average phase error < 5 deg.RMS – output power • up to 2-3 Watt: 33-35dBm • power control: 28dB – carrier leakage < 40dBc © Peter KINGET

03/99 Page34

Receivers

Radio Receiver Problem (e.g. GSM) • small signal: down to -102dBm • narrow band signal: 200kHz on ~900MHz • very hostile environment è interference – e.g. blocking signals ~100dB larger than signal !!

© Peter KINGET

03/99 Page36

Filter as RCV • e.g. GSM

fo=900MHz BW=200kHz • Quality factor: ~4500 – high Q è high loss è high NF

• High rejection & sharp filter • Tunable filter – center frequency accuracy

No Filter Technology available © Peter KINGET

03/99 Page37

Heterodyne Receiver • down-convert signal to lower fixed intermediate frequency (IF) for filtering è Q lower è fixed frequency • Mixer – zout = K ⋅ x in ⋅ y in – frequency translation: • xin@ω1 & yin@ω2 è zout@|ω2 +/- ω1| – conversion gain: • CVG = zout / x in = K ⋅ y in © Peter KINGET

03/99 Page38

Heterodyne Receiver: IMAGES ….

0

fIF

fLO- fIF

fLO

fLO+ fIF

0

fIF

fLO- fIF fLO fLO+ fIF

• fo+fIF & fo-fIF mix with fo to same fIF • potential interference • add IMAGE REJECT FILTER before mixer

© Peter KINGET

03/99 Page39

Heterodyne: choice of IF

fLO= fRF+/- fIF1

fIF= fIF1+/- f IF2

• high IF

+ more relaxed image filter + smaller IF filter - higher Q è higher loss • multiple IFs: distribute channel filtering • filter-amplify-filter-amplify • gain at different frequencies: no oscillation risk © Peter KINGET

03/99 Page40

Mixer Spurious Responses • image frequency • feed-through to IF: (LO è IF and RF è IF) • mixer: never only second but also higher order – e.g. spurious response table for double balanced mixer 6 5 4 3 2 1

fRF fRF fRF fRF fRF fRF

fLO 2fLO -100 -92 -97 -90 -84 -86 -90 -84 -97 -75 -63 -66 -70 -72 -72 -60 0 -35 -60 -60

• frequency planning

3f LO -95 -72 -86 -72 -70 -15 -70

4fLO -100 -92 -97 -72 -82 -37 -72

5fLO -100 -70 -90 -58 -62 -37 -72

6f LO -95 -95 -100 -86 -75 -45 -62 © Peter KINGET

03/99 Page41

Frequency Planning: spurious responses fIF 0

fRF fLO

• e.g. low side injection difference mixer – fIF=fLO-fRF – e.g. GSM RCV • RF in: 925-960MHz • IF: 71MHz • LO: 996-1031MHz

• find all spur frequencies fs

– |n fs +/- m fLO | = fIF – n: 0, 1, 2 ….; m: 0, 1, 2, ….

© Peter KINGET

03/99 Page42

(LO order, RF order)

Spur Frequency [MHz]

Spurious Responses

Channel frequency [MHz]© Peter KINGET 03/99 Page43

Spurious Responses (zoom) (LO order, RF order) Spur Frequency [MHz]

(2*996)-(2*960.5)=71 Image

Desired Channel frequency [MHz] © Peter KINGET

03/99 Page44

Level Diagram

IF2 AMP

IF2 Cer Filt

IF Mixer

IF AMP

IF SAW

Mixer

RF SAW

LNA

Duplex

Min. Signal Max. Signal (011) Max. Signal (001) Noise IM3_interferer IM3_product OoB Blocker

Antenna

50.00 30.00 10.00 -10.00 -30.00 -50.00 -70.00 -90.00 -110.00 -130.00

© Peter KINGET

03/99 Page45

Band-limited signal: Complex envelope r(t) ωo

r (t) = a(t) ⋅ cos(ωo ⋅ t + φ(t)) r (t) = I(t) ⋅ cos(ωo ⋅ t) − Q(t) ⋅ sin(ωo ⋅ t) 2

a(t) = I(t) + Q(t)  Q(t)  φ(t) = tan   I ( t )   −1

cos(ωo ⋅ t)

2

a(t) φ(t)

S P

I(t)

+

Q(t)

-

r(t)

sin( ωo ⋅ t ) © Peter KINGET

03/99 Page46

Homodyne Receiver cos(ωo ⋅ t) BS

PGA

LNA

PGA

a(t)

I(t)

Q(t)

S P

φ(t)

sin( ωo ⋅ t )

• • • •

fLO=fRF è fIF=0 image=signal quadrature down-converter lowpass filter does channel selection © Peter KINGET

03/99 Page47

Homodyne design issues (1) • Lowpass filters for channel selection – can be integrated on IC – high dynamic range required • preceded by limited gain or filtering – a lot of (programmable) gain at DC • parasitic feedback can cause stability problems – DC offset – 1/f noise

© Peter KINGET

03/99 Page48

Homodyne design issues (2) cos(ωo ⋅ t ) PGA

Q(t) BS

PGA

LNA

a(t)

I(t)

S P

φ(t)

sin(ωo ⋅ t )

• Time-varying DC offsets – self-mixing • LO leakage • RF leakage

• LO emission • I/Q mismatches © Peter KINGET

03/99 Page49

Homodyne design issues (3) • Even order distortion – – – – –

IM2@LNA -> LF signal -> mixer RF/IF feed-through IM2@Mixer -> LF signal & DC differential circuits but P/A single-ended -> antenna SE -> LNA SE single-ended to differential conversion at RF ….

© Peter KINGET

03/99 Page50

Why not for IF • • • •

Passive IF filters: high DR DC offset out of band: ac coupling IM2 out of band: ac coupling @IF 1/f noise low DC offset out of band • fLO=fRF +/- f IF : emission filtered • Modern IF: zero-IF back-end to go into DSP cos(ωIF1 ⋅ t)

fLO= fRF+/- fIF1

sin(ωIF1 ⋅ t ) © Peter KINGET

03/99 Page51

Image Reject Receiver: Hartley BS

LNA

cos

0 90

ωIF

sin

0 90

ωRF ± ωIF

• no IMR filter • image rejection depends on – quadrature accuracy – gain matching

• 90 degrees shift in signal path © Peter KINGET

03/99 Page52

Image Reject Receiver: Weaver ωIF = ωRF ± ω1 ± ω2 BS

LNA

cos

cos

sin

sin

0 90

0 90

ω1

ω2

• use 2 nd quadrature mixing stage instead of 90deg. shift • additional secondary image © Peter KINGET

Transmitters

03/99 Page53

Transmitters - Overview • Basic functions: – modulation: • encode the information on a waveform’s amplitude, phase or frequency – up-conversion: • move signal to desired RF carrier frequency – power amplification • amplify signal to deliver wanted power to antenna for emission

© Peter KINGET

03/99 Page55

Direct VCO modulation freq. modulation

dφ dt

Ref. Freq.

XMT-on VCO

P/A

PFD RF

RF

/N

• only constant envelope modulation • VCO in open loop during XMT – – – –

frequency drift pushing/pulling close-in VCO noise switch time XMT/RCV includes lock time

• compact

© Peter KINGET

03/99 Page56

Quadrature Modulator cos(ωo ⋅ t) a(t) φ(t)

S P

+ a(t) cos(ωo t + φ(t))

I(t)

-

Q(t)

sin( ωo ⋅ t )

• Any modulation format • see complex envelope

• But unwanted sideband when – non perfect quadrature – gain mismatches © Peter KINGET

03/99 Page57

Quadrature modulator: Side-band rejection (1 + ∆ 2) cos( ωLO t + ∆φ 2)

cos(ωIFt )

+

sin( ω IF t)

-

cos(( ωLO + ω IF ) t) + γ cos((ωLO − ω IF )t)

(1 − ∆ 2) sin( ωLO t − ∆φ 2)

ωLO- ωIF ωLO ωLO+ ωIF © Peter KINGET

03/99 Page58

Multi-step Up-conversion a(t)

cos(ωIF ⋅ t)

broadband noise reduction

S φ(t) P Q(t)

Image Reject Pre-amp

P/A

I(t)

IF

sin( ωIF ⋅ t)

RF

RF

RF

(ωRF ± ωIF )

• good image reject filter necessary • potential for other spurs • extra filter to reject broadband noise

© Peter KINGET

03/99 Page59

Direct Up-conversion a(t)

cos(ωRF t)

broadband noise reduction

Pre-amp

P/A

I(t)

φ(t)

S P Q(t) RF

RF

RF

sin( ωRF t )

• no IF and no spurs: relaxed filtering • extra filter to reject broadband noise • potential RF VCO re-modulation by P/A out – VCO shielding

• quadrature RF signal required © Peter KINGET

03/99 Page60

Indirect VCO modulation a(t)

cos( ωIF t )

broadband noise reduction

Loop Filter Tx-VCO

I(t)

S φ(t) P Q(t)

P/A

PFD RF

IF

sin( ωIF t)

IF

RF

(ωRF ± ωIF )

• • • •

only constant envelope modulation loop filter BW > signal BW low broadband noise ! Tx-VCO: high power & low noise (e.g. Pout 10dBm typ. in GSM) • potential for spurs © Peter KINGET

03/99 Page61

Power amplifier & output filters • TDD: P/A - switch - antenna • ~1dB loss in switch

• FDD: P/A - duplexer - antenna • ~2-3dB loss in switch • 30-50% of P/A power dissipated in duplexer

• average efficiency P/A