6.976 High Speed Communication Circuits and Systems Lecture 19 Basics of Wireless Communication

Transcription

1 6.976 High Speed Communicaion Circuis and Sysems Lecure 9 Basics o Wireless Communicaion Michael Perro Massachuses Insiue o Technology Copyrigh 23 by Michael H. Perro

2 Ampliude Modulaion (Transmier) Transmier Oupu x() y() 2cos(2π o ) Vary he ampliude o a sine wave a carrier requency o according o a baseband modulaion signal DC componen o baseband modulaion signal inluences ransmi signal and receiver possibiliies - DC value greaer han signal ampliude shown above Allows simple envelope deecor or receiver Creaes spurious one a carrier requency (wased power)

3 Impac o Zero DC Value Transmier Oupu x() y() 2cos(2π o ) Envelope o modulaed sine wave no longer corresponds direcly o he baseband signal - Envelope insead ollows he absolue value o he baseband waveorm - Envelope deecor can no longer be used or receiver The good news: less ransmi power required or same ransmier SNR (compared o nonzero DC value)

4 Accompanying Receiver (Coheren Deecion) Transmier Oupu x() y() 2cos(2π o ) Receiver Oupu y() z() Lowpass r() 2cos(2π o ) Works regardless o DC value o baseband signal Requires receiver local oscillaor o be accuraely aligned in phase and requency o carrier sine wave

5 Impac o Phase Misalignmen in Receiver Local Oscillaor Transmier Oupu x() y() 2cos(2π o ) Receiver Oupu y() z() Lowpass r() 2sin(2π o ) Wors case is when receiver LO and carrier requency are phase shied 9 degrees wih respec o each oher - Desired baseband signal is no recovered

6 Frequency Domain View o AM Transmier avg(x()) X() Transmier Oupu Y() x() y() - o o 2cos(2π o ) - o o Baseband signal is assumed o have a nonzero DC componen in above diagram - Causes impulse o appear a DC in baseband signal - Transmier oupu has an impulse a he carrier requency For coheren deecion, does no provide key inormaion abou inormaion in baseband signal, and hereore is a wase o power

7 Impac o Having Zero DC Value or Baseband Signal avg(x()) = X() Transmier Oupu Y() x() y() - o o 2cos(2π o ) - o o Impulse in DC porion o baseband signal is now gone - Transmier oupu now is now ree rom having an impulse a he carrier requency (or ideal implemenaion)

8 Frequency Domain View o AM Receiver (Coheren) X() Transmier Oupu Y() x() y() - o o 2cos(2π o ) Lowpass Z() Receiver Oupu - o o -2 o - o o 2 o y() z() Lowpass r() 2cos(2π o ) - o o

9 Impac o 9 Degree Phase Misalignmen X() Transmier Oupu Y() x() y() - o o 2cos(2π o ) - o o -2 o Lowpass Z() - o o Receiver Oupu = 2 o y() z() Lowpass r() 2sin(2π o ) j - -j

10 Quadraure Modulaion I () Q () i () - o o 2cos(2π 2 ) 2sin(2π 2 ) q () y() Transmier Oupu Y i () j - o o Y q () j - -j o - o -j Takes advanage o coheren receiver s sensiiviy o phase alignmen wih ransmier local oscillaor - We essenially have wo orhogonal ransmission channels (I and Q) available o us - Transmi wo independen baseband signals (I and Q) ono wo sine waves in quadraure a ransmier

11 Accompanying Receiver I () Q () Demodulae using wo sine waves in quadraure a receiver - Mus align receiver LO signals in requency and phase o ransmier LO signals i () j - o o 2cos(2π 2 ) 2sin(2π 2 ) q () - -j y() Transmier Oupu Y i () j - o o Y q () o - o -j 2cos(2π ) 2sin(2π ) Receiver Oupu Proper alignmen allows I and Q signals o be recovered as shown y() - o o Lowpass j - -j Lowpass ir () qr () I r () 2 Q r () 2

12 Impac o 9 Degree Phase Misalignmen I () Q () i () - o o 2cos(2π 2 ) 2sin(2π 2 ) q () y() Transmier Oupu Y i () j - o o Y q () j y() - -j 2sin(2π ) 2cos(2π ) Receiver Oupu I r () i r () 2 Q r () q r () 2 j - -j o - o -j - o o I and Q channels are swapped a receiver i is LO signal is 9 degrees ou o phase wih ransmier - However, no inormaion is los! - Can use baseband signal processing o exrac I/Q signals despie phase ose beween ransmier and receiver

13 Simpliied View Baseband Inpu I () Q () For discussion o ollow, assume ha - Transmier and receiver phases are aligned - Lowpass ilers in receiver are ideal - Transmi and receive I/Q signals are he same excep or scale acor In realiy - RF channel adds disorion, causes ading - Signal processing in baseband DSP used o correc problems i () q () 2cos(2π ) 2sin(2π ) Lowpass ir () 2cos(2π ) 2sin(2π ) Lowpass qr () Receiver Oupu I r () 2 Q r () 2

14 Analog Modulaion Baseband Inpu Receiver Oupu i () q () 2cos(2π ) 2sin(2π ) Lowpass ir () 2cos(2π ) 2sin(2π ) Lowpass qr () I/Q signals ake on a coninuous range o values (as viewed in he ime domain) Used or AM/FM radios, elevision (non-hdtv), and he irs cell phones Newer sysems ypically employ digial modulaion insead

15 Digial Modulaion Baseband Inpu Receiver Oupu i () Lowpass ir () 2cos(2π ) 2sin(2π ) 2cos(2π ) 2sin(2π ) q () Lowpass qr () Decision Boundaries Sample Times I/Q signals ake on discree values a discree ime insans corresponding o digial daa - Receiver samples I/Q channels Uses decision boundaries o evaluae value o daa a each ime insan I/Q signals may be binary or muli-bi - Muli-bi shown above

16 Advanages o Digial Modulaion Allows inormaion o be packeized - Can compress inormaion in ime and eicienly send as packes hrough nework - In conras, analog modulaion requires circuiswiched connecions ha are coninuously available Ineicien use o radio channel i here is dead ime in inormaion low Allows error correcion o be achieved - Less sensiiviy o radio channel imperecions Enables compression o inormaion - More eicien use o channel Suppors a wide variey o inormaion conen - Voice, ex and messages, video can all be represened as digial bi sreams

17 Consellaion Diagram Q Decision Boundaries I Decision Boundaries We can view I/Q values a sample insans on a wodimensional coordinae sysem Decision boundaries mark up regions corresponding o dieren daa values Gray coding used o minimize number o bi errors ha occur i wrong decision is made due o noise

18 Impac o Noise on Consellaion Diagram Q Decision Boundaries I Decision Boundaries Sampled daa values no longer land in exac same locaion across all sample insans Decision boundaries remain ixed Signiican noise causes bi errors o be made

19 Transiion Behavior Beween Consellaion Poins Q Decision Boundaries I Decision Boundaries Consellaion diagrams provide us wih a snapsho o I/Q signals a sample insans Transiion behavior beween sample poins depends on modulaion scheme and ransmi iler

20 Choosing an Appropriae Transmi Filer daa() p() x() * S daa () P() 2 S x () Transmi iler, p(), convolved wih daa symbols ha are viewed as impulses - Example so ar: p() is a square pulse Oupu specrum o ransmier corresponds o square o ransmi iler (assuming daa has whie specrum) - Wan good specral eiciency (i.e. narrow specrum) / /

21 Highes Specral Eiciency wih Brick-wall Lowpass daa() p() x() * - S daa () P() 2 S x () /(2 ) /(2 ) Use a sinc uncion or ransmi iler - Corresponds o ideal lowpass in requency domain Issues - Nonrealizable in pracice - Sampling ose causes signiican inersymbol inererence

22 Requiremen or Transmi Filer o Avoid ISI daa() p() x() * Time samples o ransmi iler (spaced apar) mus be nonzero a only one sample ime insan - Sinc uncion saisies his crierion i we have no ose in he sample imes Inersymbol inererence (ISI) occurs oherwise Example: look a resul o convolving p() wih 4 impulses - Wih zero sampling ose, x(k ) correspond o associaed impulse areas

23 Derive Nyquis Condiion or Avoiding ISI (Sep ) daa() p() x() * impulse rain p() p(k ) Consider muliplying p() by impulse rain wih period - Resuling signal mus be a single impulse in order o avoid ISI (same argumen as in previous slide)

24 Derive Nyquis Condiion or Avoiding ISI (Sep 2) impulse rain P() / / * F{p(k )} / impulse rain p() p(k ) In requency domain, he Fourier ransorm o sampled p() mus be la o avoid ISI - We see his in wo ways or above example Fourier ransorm o an impulse is la Convoluion o P() wih impulse rain in requency is la

25 A More Pracical Transmi Filer Raised-cosine iler is quie popular in many applicaions p() P() -α 2 +α 2 - -/ / Transiion band in requency se by rollo acor, α Rollo acor = : P() becomes a brick-wall iler Rollo acor = : P() looks nearly like a riangle Rollo acor =.5: shown above

26 Raised-Cosine Filer Saisies Nyquis Condiion Nyquis Condiion Observed in Time p() Nyquis Condiion Observed in Frequency P() - -/ / In ime - p(k ) = or all k no equal o In requency - Fourier ransorm o p(k ) is la - Alernaively: Addiion o shied P() cenered abou k/ leads o la Fourier ransorm (as shown above)

27 Specral Eiciency Wih Raised-Cosine Filer daa() p() x() * - S daa () P() 2 S x () More eicien han when p() is a square pulse Less eicien han brick-wall lowpass - Bu implemenaion is much more pracical Noe: Raised-cosine P() oen spli beween ransmier and receiver /(2 ) /(2 )

28 Receiver Filer: ISI Versus Noise Perormance Baseband Inpu Raised-Cosine Filer I () Q () i () q () 2cos(2π ) 2sin(2π ) Lowpass i r () 2cos(2π ) 2sin(2π ) Lowpass q r () I r () Receiver Oupu Q r () Conlicing requiremens or receiver lowpass - Low bandwidh desirable o remove receiver noise and o rejec high requency componens o mixer oupu - High bandwidh desirable o minimize ISI a receiver oupu

29 Spli Raised-Cosine Filer Beween Transmier/Receiver Baseband Inpu Raised-Cosine Filer I () Q () i () q () 2cos(2π ) 2sin(2π ) i r () 2cos(2π ) 2sin(2π ) q r () Raised-Cosine Filer Addiional Lowpass Filering I r () Receiver Oupu Q r () We know ha passing daa hrough raised-cosine iler does no cause addiional ISI o be produced - Implemen P() as cascade o wo ilers corresponding o square roo o P() Place one in ransmier, he oher in receiver Use addiional lowpass ilering in receiver o urher reduce high requency noise and mixer producs

30 Muliple Access Techniques

31 The Issue o Muliple Access Wan o allow communicaion beween many dieren users Freespace specrum is a shared resource - Mus be pariioned beween users Can pariion in eiher ime, requency, or hrough orhogonal coding (or nearly orhogonal coding) o daa signals

32 Frequency-Division Muliple Access (FDMA) Channel Channel 2 Channel N Place users ino dieren requency channels Two dieren mehods o dealing wih ransmi/receive o a given user - Frequency-division duplexing - Time-division duplexing

33 Frequency-Division Duplexing Duplexer Anenna Transmier TX TX RX Receiver RX Transmi Band Receive Band Separae requency channels ino ransmi and receive bands Allows simulaneous ransmission and recepion - Isolaion o receiver rom ransmier achieved wih duplexer - Canno communicae direcly beween users, only beween handses and base saion Advanage: isolaes users Disadvanage: deplexer has high inserion loss (i.e. aenuaes signals passing hrough i)

34 Time-Division Duplexing Transmier TX Swich Anenna Receiver RX Use any desired requency channel or ransmier and receiver Send ransmi and receive signals a dieren imes Allows communicaion direcly beween users (no necessarily desirable) Advanage: swich has low inserion loss relaive o duplexer Disadvanage: receiver more sensiive o ransmied signals rom oher users swich conrol

35 Time-Division Muliple Access (TDMA) Time Slo N Time Slo Time Slo 2 Time Slo N Time Slo Time Frame Place users ino dieren ime slos - A given ime slo repeas according o ime rame period Oen combined wih FDMA - Allows many users o occupy he same requency channel

36 Channel Pariioning Using (Nearly) Orhogonal Coding Uncorrelaed Signals Correlaed Signals A B - - x() c() y() B B - - x() c() y() - - -A B - - x() c() y() -B B - - x() c() y() - - Consider wo correlaion cases - Two independen random Bernoulli sequences Resul is a random Bernoulli sequence - Same Bernoulli sequence Resul is or -, depending on relaive polariy

37 Code-Division Muliple Access (CDMA) Separae Transmiers x () y () Transmi Signals Combine in Freespace PN () y() x 2 () y 2 () PN 2 () T c Assign a unique code sequence o each ransmier Daa values are encoded in ransmier oupu sream by varying he polariy o he ransmier code sequence - Each pulse in daa sequence has period Individual pulses represen binary daa values - Each pulse in code sequence has period T c Individual pulses are called chips

38 Receiver Selecs Desired Transmier Through Is Code Separae Transmiers x () y () Transmi Signals Combine in Freespace PN () x 2 () y 2 () PN 2 () PN () Lowpass r() x() y() Receiver (Desired Channel = ) Receiver correlaes is inpu wih desired ransmier code - Daa rom desired ransmier resored - Daa rom oher ransmier(s) remains randomized

39 Frequency Domain View o Chip Vs Daa Sequences T c daa() p() x() - - * T c - daa() p() x() * - S daa () P() 2 S x () T c T c / /T c / /T c Daa and chip sequences operae on dieren ime scales - Associaed specra have dieren widh and heigh

40 Frequency Domain View o CDMA S x () S y () Transmier x () y () T c / PN () S x 2 () Sy2 () Transmier 2 /T c y() PN () S x () S x () Lowpass r() / x 2 () PN 2 () y 2 () CDMA ransmiers broaden daa specra by encoding i ono chip sequences CDMA receiver correlaes wih desired ransmier code - Specra o desired channel revers o is original widh - Specra o undesired channel remains broad T c /T c T c / /T c Can be mosly ilered ou by lowpass

41 Consan Envelope Modulaion

42 The Issue o Power Eiciency Baseband Inpu Baseband o RF Modulaion Power Amp Transmier Oupu Variable-Envelope Modulaion Consan-Envelope Modulaion Power amp dominaes power consumpion or many wireless sysems - Linear power amps more power consuming han nonlinear ones Consan-envelope modulaion allows nonlinear power amp - Lower power consumpion possible

43 Simpliied Implemenaion or Consan-Envelope Baseband Inpu Baseband o RF Modulaion Transmi Filer Power Amp Transmier Oupu Consan-Envelope Modulaion Consan-envelope modulaion limied o phase and requency modulaion mehods Can achieve boh phase and requency modulaion wih ideal VCO - Use as model or analysis purposes - Noe: phase modulaion nearly impossible wih pracical VCO

44 Example Consellaion Diagram or Phase Modulaion Q Decision Boundaries I Decision Boundaries I/Q signals mus always combine such ha ampliude remains consan - Limis consellaion poins o a circle in I/Q plane - Draw decision boundaries abou dieren phase regions

45 Transiioning Beween Consellaion Poins Q Decision Boundaries I Decision Boundaries Consan-envelope requiremen orces ransiions o allows occur along circle ha consellaion poins si on - I/Q ilering canno be done independenly! - Signiicanly impacs oupu specrum

46 Modeling The Impac o VCO Phase Modulaion Recall unmodulaed VCO model Phase/Frequency modulaion Signal S Φmod () Overall phase noise Φ n () S ou () o Phase Noise Spurious Noise Φ mod () Φ ou 2cos(2π o +Φ ou ()) Relaionship beween sine wave oupu and insananeous phase Impac o modulaion - Same as examined wih VCO/PLL modeling, bu now we consider Φ ou () as sum o modulaion and noise componens ou()

47 Relaionship Beween Sine Wave Oupu and is Phase Key relaionship (noe we have dropped he acor o 2) Using a amiliar rigonomeric ideniy Approximaion given Φ n () <<

48 Relaionship Beween Oupu and Phase Specra Approximaion rom previous slide Auocorrelaion (assume modulaion signal independen o noise) Oupu specral densiy (Fourier ransorm o auocorrelaion) - Where * represens convoluion and

49 Impac o Phase Modulaion on he Oupu Specrum Phase/Frequency modulaion Signal S Φmod () Overall phase noise Φ n () S ou () o Phase Noise Spurious Noise Φ mod () Φ ou 2cos(2π o +Φ ou ()) ou() Phase/Frequency modulaion Signal S Φmod () Overall phase noise Φ n () S ou () o Φ mod () Φ ou 2cos(2π o +Φ ou ()) ou() Specrum o oupu is disored compared o S Φmod () Spurs convered o phase noise

50 I/Q Model or Phase Modulaion Applying rigonomeric ideniy Can view as I/Q modulaion - I/Q componens are coupled and relaed nonlinearly o Φ mod () S a () S Φmod () Φ mod () cos(φ mod ()) sin(φ mod ()) i () q () S b () cos(2π 2 ) sin(2π 2 ) y() S y () - o o