Upload
simenli
View
230
Download
48
Embed Size (px)
Citation preview
RF Transceiver Module DesignChapter 4
RF Transceiver Architectures李健榮助理教授
Department of Electronic EngineeringNational Taipei University of Technology
Outline
• General Considerations
• Frequency Conversion
• Receiver Architectures Heterodyne Receiver
Direct-Conversion Receiver (DCR)
Image-Reject and Low-IF Receiver
• Transmitter Architectures Direct-Conversion Transmitter (DCT)
Heterodyne and Sliding-IF Transmitter
Open-loop and Closed-loop PLL-based Transmitter
Envelope Tracking and Envelope Following Transmitter
Polar Transmitter
Department of Electronic Engineering, NTUT2/110
Front-end General Considerations
• TX: Adjacent channel leakage
• RX: Rejection of inband and out-of-band interference
BPF
Power Amplifier (PA)Transmitted Channel
Adjacent Channels
ω
BPF
Low Noise Amplifier (LNA) Bandpass Filter
Response
AdjacentChannel
Alternate AdjacentChannel
1ff
Department of Electronic Engineering, NTUT3/110
Interferer Suppression
• High linearity to accommodate interferes without experiencingcompression or significant intermodulation. Filtering theinterferer can relax RXlinearity requirements.
• BPF high selectivity is required for near channel rejection.
• Variable BPF is required for different carrier frequencies, andit is difficult.
900 900.4( ) MHzf
20 dB35 dB
BPF Response
Hypothetical filter to suppress an interference
Department of Electronic Engineering, NTUT4/110
Channel-Selection Filter
• All of the stages in the RXchain that precede channel-selection filtering must be sufficiently linear to avoidcompression or excessive intermodulation
• Since channel-selection filtering is extremely difficult at theinput carrier frequency, it must be deferred to some other pointalong the chain where the center frequency of the desiredchannel is substantially lower and hence the required filterQ’sare more reasonable.
Department of Electronic Engineering, NTUT5/110
Band-Select Filter
• A band-select filter selects entire RXband and reject out-of-band interferers, thereby suppressing components that may begenerated by users that do not belong to the standard ofinterest.
• Trade-off between selectivity and in-band loss (higher-orderfiltering sections and arise NF).
BPF
LNADesiredChannel
Receive Band
f
f
Band-selection filtering
Department of Electronic Engineering, NTUT6/110
TX-RX Feedthrough
• TX leakage in a CDMAtransceiver (full duplex). The RXmust meet difficult linearity requirements.
• A BPF following the LNAcan alleviate the leakage.
Du
ple
xer
−20 dBmLNA
PA
1 W (+30 dBm)
−50 dB
Du
ple
xer
LNA
PA
−50 dB
f
f
TX Leakage
f
BPF Response
BPF
10
dB
/div
.
20 MHz/div.
1f2f
TX Band RX Band
50 dB30 dB
Department of Electronic Engineering, NTUT7/110
Frequency Conversion (I)
• Recall Chapter 1 (double sideband amplitude modulation)
( ) ( )cos2m cs t A t f tπ=t( ) ( )BBs t A t=
ff
cf0 Hzcf−0 Hz
USBLSBUSBLSBLSBUSB
cos2 cf tπ“real signal”
Real signal
f0 Hz
Complex conjugate
USBLSB
1f1f−cos2 cf tπ
0 Hzcfcf−
USBLSBLSBUSB
IF cf f+c IFf f−IF cf f−c IFf f− −
Double sideband (DSB)
Double sideband (DSB)
Department of Electronic Engineering, NTUT8/110
Frequency Conversion (II)
• Recall Chapter 1 (linear modulation)
• Yes, a modulated signal sm(t) is a real signal.
( ) ( ) ( ) ( )1 12 2
2 2j t j tj f t j f tA t A t
e e e eφ φπ π− −= +
( ) ( ) ( )( )1cos 2ms t A t f t tπ φ= +
( ) ( ) 12Re j t j f tA t e eφ π= ⋅
f1f0 Hz1f−
“complex”“complex” “real”
Complex conjugate
( )I t
1cos tω1sin tω−
( )Q t
( )ms t
Real signal
Complex envelope
Department of Electronic Engineering, NTUT9/110
Frequency Conversion (III)
0 Hz2f2f−
0 Hz2f2f−
USBLSBLSBUSB
Real signal
f0 Hz
Complex conjugate
USBLSB
1f1f−1 2f f+2 1f f−1 2f f−2 1f f− −2cos2 f tπ
RFIF
( )I t
cos IF tωsin IF tω−
( )Q t
( )IFs t
Modulated signal (real signal)
f0 Hz
USBLSB
IFfIFf−
cos 2 cf tπ
RF 0 Hzcfcf−
USBLSBLSBUSB
IF cf f+c IFf f−IF cf f−c IFf f− −
Double sideband (DSB) mixing
upconversion
upconversion
IF
LO
LO By filtering, you can choose only USB or LSBtransmission, which is call the single-sideband(SSB) transmission.
Department of Electronic Engineering, NTUT10/110
Frequency Conversion (IV)
0 Hz2f2f−
0 Hz2f2f−
Real signal
f0 Hz
Complex conjugate
1f1f−1 2f f+1 2f f−2 1f f−2 1f f− −2cos2 f tπ
IFRF
downconversion
2 1f f<
2f2f−
0 Hz2f2f−
0 Hz2f2f−
Real signal
f0 Hz 1f1f− 1 2f f+2 1f f−1 2f f−2 1f f− −2cos2 f tπ
IFRF
downconversion
2 1f f>
2f2f−
High-side injection
Low-side injection
LO
LO
Department of Electronic Engineering, NTUT11/110
Receiver Architecture
• Basic Heterodyne Receiver
• Modern Heterodyne Receiver
Hetero-dyne
Different-freq. Mixing
Department of Electronic Engineering, NTUT12/110
Basic Heterodyne Receivers (I)
• Translating the desired channel to a much lower centerfrequency to permit a channel-selection filtering with areasonableQ.
ωinωinω− 0
ωLOωLOω− 0
Downconversion by mixingRF input
inω ω
Mixer
0 cos LOA tω
vLPF IF Output
in LOω ω− − in LOω ω− + 0 in LOω ω+in LOω ω−
Filtered-outFiltered-out
LO
( ) ( )1 1cos cos cos cos
2 2in LO in LO in LOt t tω ω ω ω ω ω⋅ = + + −
Low freq.High freq.
Two IF frequencies:
Department of Electronic Engineering, NTUT13/110
Basic Heterodyne Receivers (II)
• Use of LNA to reduce noise
• Variable IF:
• Constant IF:
Mixer
0 cos LOA tω
vLPF IF OutputRF input
LNA
IFj RFj LOf f f= − (Constant LO freq. and variable IF freq.)
IF RFj LOjf f f= − (Variable LO freq. and constant IF freq.)
Precise LO freq. and steps provided by a “frequency synthesizer”
Constant IF approach is more common to simplify the design ofIF path; e.g., it does notrequire a variable-frequency channel selection filter.
LO
Department of Electronic Engineering, NTUT14/110
Basic Heterodyne Receivers (III)
• Constant-LO downconversion mixing
• Constant-IF downconversion mixing
1RFff f
f1LOf
IFf0
1RFff f
fLOf
1IFf0 2RFff f
f
2IFf0
LOf
2RFff f
f
IFf0
2LOf
Department of Electronic Engineering, NTUT15/110
• While each wireless standard impose constrains upon theemissions by its own users, it may have no control over thesignals in other bands. The image power can therefore bemuch higher than that of the desired signal, requiring proper“image rejection.”
Image Problem in Heterodyne RX
cos LOtω
vLPF
Desired signal
Image
inω imω ωIFω ω
IFω IFω
LOωω
High-side injection
( ) ( )cosd d dA t t tω φ + ( ) ( )cosim im imA t t tω φ +
( ) ( ) ( ) ( ) ( ) ( ) ( )1 1cos cos
2 2IF d LO d LO d d LO d LO dx t A t A t t A t A t tω ω φ ω ω φ = + + − − +
( ) ( ) ( ) ( ) ( ) ( )1 1cos cos
2 2im LO im LO im im LO im LO imA t A t t A t A t tω ω φ ω ω φ + + + − − +
Department of Electronic Engineering, NTUT16/110
DownconvertedSpectrum (I)
1ω−2ω− 1ω+ 2ω+0ω
ω
ω
0 LOω+LOω−
0
1ω−2ω− 1ω+ 2ω+0ω
ω
ω
0 LOω+LOω−
0
Downconversion for 1LOω ω< Downconversion for 2 1LOω ω ω> >
Department of Electronic Engineering, NTUT17/110
DownconvertedSpectrum (II)
1ω−2ω− 1ω+ 2ω+0ω
ω
ω
0 LOω+LOω−
0
1ω−2ω− 1ω+ 2ω+0ω
ω
ω
0 LOω+LOω−
0
Downconversion for 2 1LOω ω ω> > Downconversion for 2LOω ω>
1 2
2LO
ω ωω +=
Department of Electronic Engineering, NTUT18/110
• The most common “image rejection “ approach is to precedethe mixer with an “image-rejection filter.”
• The filter exhibits a relatively small loss in the desired bandand a large attenuation in the image band, two requirementsthat can be simultaneously met if 2ωIF is sufficiently large.
• A filter with high image rejection typically appears betweenthe LNA and the mixer to lower the noise contribution to theRX NF (The NF increases while the filter precedes the LNA).
Image Rejection
Image Reject Filter
inω imωω
2 IFωcos LOtω
v
ImageRejectFilter
LNA
Department of Electronic Engineering, NTUT19/110
Image Rejection v.s. Channel Selection
Image Reject Filter
inω imωω
2 IFω
cos LOtω
v
ImageRejectFilter
LNA
v
ChannelSelectFilter
Desired channel
Interference
ImageChannel Select Filter
(high-Q needed)
IFωω
0
0ω
IFωinω imω
2 IFω
ω
High IF
Low IF
• If the IF is high, the image can besuppressed but complete channelselection is difficult, and vice versa.
Department of Electronic Engineering, NTUT20/110
Image Noise Increases Noise Figure
• Even in the absence of interferes, the thermal noise producedby the antenna and the LNAin the image band arrives at theinput of the mixer. The thermal noises in the desired channeland image band are downconverted to IF (unless the LNAhasa limited bandwidth to suppresses the noise in the image band).
LOω
LNA
inω
Thermal Noise
LOω inωω
2 in LOω ω−
ω
Department of Electronic Engineering, NTUT21/110
Dual IF Receiver (I)
• The concept of heterodyning is extended to multipledownconversions, each followed by filtering and amplification,to resolve the trade-off between “image rejection” and“channel selection.”
• This technique performs partial channel at progressively lowercenter frequencies, thereby relaxing theQ required of eachfilter.
1LOω
vBPF2
LNA
vBPF1 vBPF3
2LOω
vBPF4
Band Select Filter
ImageReject Filter
RF MixerMX1
ChannelSelect Filter
IF MixerMX2
ChannelSelect Filter IF Amp.
A C E GB D F H
Department of Electronic Engineering, NTUT22/110
Dual-IF Receiver (II)
1LOω
vBPF2
LNA
vBPF1 vBPF3
2LOω
vBPF4
Band Select Filter
ImageReject Filter
RF MixerMX1
ChannelSelect Filter
IF MixerMX2
ChannelSelect Filter IF Amp.
A C E GB D F H
Department of Electronic Engineering, NTUT
Desired Channel Image
fA
C
E
G
B
D
F
H
BPF1BPF2
Image
ImageBPF3
BPF4
f
f
f f
f
f
f
23/110
Mixing Spurs (I)
• In practice, the mixing is the multiplication of the RF input byall harmonics of the LO. Thus the RF mixer producescomponents at and IF mixer, wherem andn are integers.
• For the desired signal, only is of interest. But ifan interferer, , is downconverted to the same IF, it corruptsthe signal; this occurs if .
1in LOmω ω± 1 2in LO LOm nω ω ω± ±
1 2in LO LOω ω ω− −
intωint 1 2 1 2LO LO in LO LOm nω ω ω ω ω ω± ± = − −
Department of Electronic Engineering, NTUT24/110
Mixing Spurs (II)
• An example of a 2.4 GHz dual conversion RX:
1 1.98 GHzLOω =
LNA
vBPF
2 400 MHzLOω =
420 MHz2.4 GHz
20 MHz
2.7
6 G
Hz2.4 GHz
2.8
GH
z 4.38 GHzf
Received Spectrum
820 MHz780 MHz
2800 MHz2760 MHz
22 800 MHzLOω =12 3.96 GHzLOω =
4380 MHz
20 MHz420 MHz
2 400 MHzLOω =20 MHz
Department of Electronic Engineering, NTUT25/110
Modern Heterodyne Receivers
ω
ω0
0
1IFω− 1IFω+
2LOω+2LOω−
ω
ω
Desired ChannelInterferer
0
0
• Zero Second IF: Avoid secondary image(assume no interferers aredownconverted as an image to a zero center frequency).
• Interferer appears in the adjacent channel
2LOω ω
ω1IFω ω
02 1LO IFω ω=
2 1LO IFω ω=
Department of Electronic Engineering, NTUT26/110
Signal Becomes its Own Image
• For symmetrically-modulated signal:
• For asymmetrically-modulated signal:LOf
f
f0 f
LOf
( )S f ( )LOS f f−
vVCO
( )BBx t
t fcf
1IFω+1IFω− 0ω
ω0
same information on both side of the carrier
downconversion
Corruption occurs if the signal spectrum is asymmetric
Department of Electronic Engineering, NTUT27/110
Avoid Self-corruption of Asymmetric Signals (I)
• One can downconvert signal to an IF equal to half of the signalbandwidth to avoid self-corruption of a signal with asymmetricspectrum.
1IFω+1IFω− 0ω
ω0
BWω
2BWω+
2BWω−
1 2BW
IF
ωω ≥
Department of Electronic Engineering, NTUT28/110
Avoid Self-corruption of Asymmetric Signals (II)
• Zero second IF with quadrature downconversion.
( )IFx t
( ),BB Ix t
( ),BB Qx t
2cos LO tω
2sin LO tω
2 1LO IFω ω=
Quadrature baseband signal
Though xBB,I(t) and xBB,Q(t) exhibit identical spectra, they are separated in phaseand together can reconstruct the original information
Department of Electronic Engineering, NTUT29/110
Zero Second IF Heterodyne RX
• Zero second IF heterodyne RX with quadrature downconverison
( )IFx t
( ),BB Ix t
( ),BB Qx t
2cos LO tω
2sin LO tω
1LOω
vBPF
LNA
No image rejection filter LNA/mixer interface can be optimized (need not 50 Ohms) for gain, noise, and
linearity with little concern for the interface impedance values. The lack of image rejection filter requires careful attention to the interferers in
the image band, and dictates a narrow-band LNA design (suppress imagenoise).
No channel-selection filter is shown at the first IF, but some “mild” on-chipBPF is usually inserted to suppress out-of-band interferers.
Department of Electronic Engineering, NTUT30/110
Sliding-IF Heterodyne RX (I)
( )IFx t
( ),BB Ix t
( ),BB Qx t
2,ILO
1LOω
vBPF
LNA
vLO1 v2÷
2,QLO
t
2,ILO
2,QLO
1LO
90
RF Input
1st LO
1st IF
2nd IF
inff
f
f
f
1LOf
1in LOf f−
11 2
LOin LO
ff f− −
For an input band [f1, f2], the LOmust cover a range of [(2/3)f1, (2/3)f2].
11 2
LOLO in
ff f+ = 1
2
3LO inf f=
The 1st IF is not constant, because
1
1
3IF in LO inf f f f= − =
Department of Electronic Engineering, NTUT31/110
Sliding-IF Heterodyne RX (II)
• As fin varies fromf1 to f1, fIF1 goes fromf1 /3 to f2 /3 (slide IF).
RF Range
LO Range
1st IF Range
f1f 2f
f
f
1
2
3f 2
2
3f
1
1
3f 2
1
3f
( )2 1
2 1
1 13 3% *100%
1 1 12 3 3
IF
f fBW
f f
−=
+
( )( )
2 1
1 2
% *100%12
RF
f fBW
f f
−=+
Department of Electronic Engineering, NTUT32/110
Sliding-IF Heterodyne RX (III)
• Image band of the sliding-IF heterodyne RX
Image Band RF Band
LO Band1
1
3f 2
1
3f 1f 2f
f
f
1
2
3f 2
2
3f
Department of Electronic Engineering, NTUT33/110
Direct-Conversion Receivers
• Direct-conversion receiver (DCR) is also called the “zero-IF”,or the “homodyne” receiver.
• As mentioned previously, downconversion of an asymmetric-modulated signal to a zero IF leads to self-corruption unlessthe baseband signals are separated by their phases.
I
Q
cos LOtωsin LOtωvBPF
LNAvLPF
vLPF
inω
LO inω ω=
Department of Electronic Engineering, NTUT34/110
DCR Advantages
• The absence of an image greatly simplifies the design process.
• Channel selection is performed by low-pass filters, which canbe realized on-chip as active circuit topologies with relativelysharp cut-off characteristics.
• Mixing spurs are considerably reduced in number and hencesimpler to handle.
• The LNA/mixer interface can be optimized for gain, noise, andlinearity without requiring a 50-Ω impedance.
Department of Electronic Engineering, NTUT35/110
DCR Issues − LO Leakage
• A DCR emits a fraction of its LOpower fromits antenna, andthe LO emission is undesirable because it may desensitizeother receivers operating in the same band. Typical acceptablevalues range from−50 to−70 dBm(measured at the antenna).
• In heterodyne receivers, since the LOfrequency falls outsidethe band, it is suppressed by the front-end band-select filters inboth the emitting receiver and the victimreceiver.
PadLNA
LOSubstrate
LO
LNA
LO leakage Cancellation of LO leakage by symmetry
Department of Electronic Engineering, NTUT36/110
DCR Issues − DC Offset
• The leakage causes “LOself-mixing” at the mixer to produce adc component in the baseband (because multiplying a sinusoidby itself results in a dc term). The zero second IF architecturealso suffers fromthis issue.
• LO leakage yields a very large dc offset due to the high gain ofthe receiving chain, and this saturates the baseband circuits(degrades the dynamic range), prohibiting signal detection.
• Time-varying dc offset
RF LOV kV+
PadLNA
IF DCV V+
Department of Electronic Engineering, NTUT37/110
Effect of DC Offset in Baseband Chain
• Since and , thus . Amplifiedby another 40 dB, this offset reaches 1-V at the basebandoutput.
1 31.6vA = ( )632 2 VleakV µ= 10 mVdcV =
cos LOtω
LPF
sin LOtω
LNA
1 30 dBvA = 2 40 dBvA =
0 cos inV tω ( )cosbb in LOV tω ω−
1 0bb vV A V=cosleak LOV tω
1dc v leakV A V=
Department of Electronic Engineering, NTUT38/110
Leakage of Quadrature Phases of LO
• The dc offset measured in the basebandI and Q outputs areoften unequal.
LOLNA
( )cosleak LO leakV tω φ+
( ), cosdc I leak LO leak circuitV V Vα φ φ= +
( ), sindc Q leak LO leak circuitV V Vα φ φ= − +
Department of Electronic Engineering, NTUT39/110
Cancelling DC Offset
LNA
cos LOtω
LPF1C
1R
bv
1ABaseband SignalHPF Response
1f− 1f+0f
• Using a HPF (ac coupling) removes dc offset but also removesa fraction of the signal’s spectrumnear zero frequency, therebyintroducing distortion.
• As a rule of thumb, the corner frequency of the HPF must beless than 1/1000 of the symbol rate for negligible distortion.This may require very large capacitance and thus difficult tointegrate on chip (especially for lowsymbol rate). For the slowresponse to transient inputs (LOswitch, LNA gain change), accoupling is rarely used in today’s receivers.
Department of Electronic Engineering, NTUT40/110
DCR Issues − Even-order Distortion
• DCRs are additionally sensitive to even-order nonlinearity inthe RF path (IM2 falls around DC to corrupt the desired signaland mixer feedthrough), and so are the heterodynearchitectures having a second zero IF .
Beat Component
cos LOtω
1 2ω ω−Feedthrough
LNA
DesiredChannel
Interferers0
ω
0ω
1ω ω2ω
Department of Electronic Engineering, NTUT41/110
Mixer Feedthrough– Simple Mixer
inVoutV
LO
1R
( ) ( ) ( ) ( ) ( ) ( )1 1
2 2out in in inV t V t S t V t S t V t = ⋅ = − + ⋅
where is the RF input and is the Ideal LO toggling between 0 and1 with 50%duty cycle, and
( ) 1
2S t −
DC-free square wave
( ) 1
2inV t ⋅ is the RF feedthrough to the output
( )inV t ( )S t
represents a
Department of Electronic Engineering, NTUT42/110
Mixer Feedthrough– Differential Mixer
• If the output is sensed differentially, the RF feedthroughVout1(t)andVout2(t) are cancelled while the signal components add.
• This cancellation is sensitive to asymmetrics, e.g., if theswitches exhibits a mismatch between their on-resistance, thena net RF feedthrough arises in the differential output.
( ) ( ) ( )1out inV t V t S t= ⋅
LO
LO1R
1R
1outV
2outV
inV( )S t
( )1 S t−
( ) ( ) ( )2 1out inV t V t S t= ⋅ −
Department of Electronic Engineering, NTUT43/110
Even-order Distortion (I)
• The 2nd order nonlinear effect cause the beat amplitude whichgrows with the square of the amplitude of the input tones.
(log scale)
(log scale)inA
IIP2A
IIP2A1Aα
22Aα
( ) ( ) ( )21 2out in inV t V t V tα α= +
( ) ( ) ( )2 21 1 2 2 1 2 2 1 2cos cos cos cosA t t A t A tα ω ω α ω ω α ω ω= + + + + + +⋯
( ) 1 2cos cosinV t A t A tω ω= +
Department of Electronic Engineering, NTUT
• Since the net feedthrough of thebeat depends on the mixer andLO asymmetries, the beatamplitude measured in thebaseband depends on the derivedimensions and the layout and istherefore difficult to formulate.
44/110
Even-order Distortion (II)
• Even-order distortion may manifest itself even in the absenceof interferers. Suppose in addition to frequency and phasemodulation, the received signal also exhibits amplitudemodulation.
• Both of the terms and are low-pass signals and,like the neat component, pass through the mixer with finiteattenuation, corrupting the downconverted signal.
( ) ( ) ( )0 cosin cx t A a t t tω φ= + +
( ) ( ) ( )21 2out in inV t V t V tα α= +
( ) ( ) ( ) ( )2 2 22 2 0 0
1 cos 2 22
2c
in
t tx t A A a t a t
ω φα α
+ + = + +
( )2 0A a tα ( )22 2a tα
Department of Electronic Engineering, NTUT45/110
DCR Issues − Flicker Noise (I)
• Linearity requirements limit the cascaded LNA/mixer gain, thedownconverted signal in a DCR is still relatively small andhence susceptible to noise in the BB circuits. Since the signalis centered around zero frequency, it can be substantiallycorrupted by the flicker noise.
• The mixers themselves may also generate flicker noise at theiroutput.
(log
sca
le)
fBWfCf
1000BWf
( )1 fS f
thS
,
1
2BW RF BWf f=1 fS
f
α= thc
Sf
α = th cS fα = ⋅
Assume noise components below fBW/1000 are unimportant
( )1 0.0015.9 ln
BW
BW
fc
n BW c th c th BW thfBW
fP df f f S f S f S
f f
α = + − = + +
∫
If no flicker noise 2n BW thP f S≈
Flicker noise penalty 1
2
n
n
P
P
Department of Electronic Engineering, NTUT46/110
DCR Issues − Flicker Noise (II)
• In a good design, the thermal noise at the end of the basebandchain arises mostly fromthe noise of the antenna, the LNA,and the mixer. Thus, a higher front-end gain directly raisesSth,thereby loweringfc and hence the flicker noise penalty.
• Flicker noise penalty:
An 802.11g RX with fc of 200 kHz:
10 MHzBWf =
1
2
1.04n
n
P
P=
A GSM RX with fc of 200 kHz:
1
2
16.4n
n
P
P=
(log
sca
le)
BWf200100
( )1 fS f
( ) kHzfthS
DownconvertedGSM Channel
Effect of flicker noise on a GSM channel
Flicker noise makes it difficult to employ DCR for a narrow channel bandwidth.In such cases, the “low-IF” architecture proves a more viable choice.
Department of Electronic Engineering, NTUT47/110
DCR Issues − I/Q Mismatch
• DCR require 90o shift of the RF signal and this generallyentails severe noise-power-gain trade-offs.
I
Q
LOV
vLPF
vLPF
90
RFV
I
Q
LOV
vLPF
vLPF
RFV90
Shift of RF signal or LO waveform by 90o
Department of Electronic Engineering, NTUT48/110
I/Q Mismatch (I)
• Errors in the 90o phase shift circuit and mismatches betweenthe quadrature mixers result in imbalance in the amplitudesand phases of the basebandI andQ outputs.
• The BB stages themselves may also contribute mismatches.
I
Q
LOV
vLPF
vLPF
RFV
90
Phase and Gain Error
Phase and Gain Error
Phase and Gain Error
Phase and Gain Error
Department of Electronic Engineering, NTUT49/110
I/Q Mismatch (II)
• I/Q mismatches tend to be larger in DCRs than in heterodynetopologies, because:
Propagation of a higher frequency experiences greater mismatches
LO quadrature phases suffer from greater mismatches at higher frequencies
cos LOtωsin LOtω
LNA
5 GHz
10 ps 18 @5 GHzT∆ = ⇒
( )IFx t
( ),BB Ix t
( ),BB Qx t
LNA
vLO1 4÷
5 GHz
4 GHz
1 GHz
10 ps 3.6 @1 GHzT∆ = ⇒
DCR Heterodyne RX
Department of Electronic Engineering, NTUT50/110
I/Q Mismatch – QPSK Example (I)
( ),BB Ix tvLPF
vLPF
( )inx t 90 ( )LOx t2
θ+
2
θ−
12
ε+
12
ε−
( ),BB Qx t
( ) cos sinin c cx t a t b tω ω= +
, 1a b = ±
( ), 2 1 cos2 2LO I cx t tε θω = + +
( ), 2 1 cos2 2LO Q cx t tε θω = − −
( ), 1 cos 1 sin2 2 2 2BB Ix t a bε θ ε θ = + − +
( ), 1 sin 1 cos2 2 2 2BB Qx t a bε θ ε θ = − − + −
The mismatch causes crosstalk between I and Q BB signals.
Department of Electronic Engineering, NTUT51/110
I/Q Mismatch – QPSK Example (II)
I
Q
t
tI
Q
Ideal
I
Q
t
t
I
Q
Ideal
Only amplitude mismatch : 0, 0ε θ≠ =
Only phase mismatch : 0, 0ε θ= ≠
( ), 12BB Ix t aε = +
( ), 1 cos2 2BB Qx t bε θ = −
( ), cos sin2 2BB Ix t a bθ θ= −
( ), sin cos2 2BB Qx t a bθ θ= − +
Department of Electronic Engineering, NTUT52/110
Correction of I/Q Mismatch
cos LOtωsin LOtω
LPF
LPF
LNAI
Q
IQ
PhaseMismatch
AmplitudeMismatch
t
LPF
LPF
ADC
ADC
Logic
φ
φ
cos LOtω
sin LOtω
LNA
• Calibration of quadrature phase and gain either at power up orcontinuously is usually needed for high performance system.
Test signal
Analog adaption
Digital adaption is more popularin nowadays systems
Department of Electronic Engineering, NTUT53/110
Mixing Spurs
• Unlike heterodyne systems, DCRs rarely encounter corruptionby mixing spurs. This is because, for an input frequencyf1 tofall in the baseband after experiencing mixing withnfLO, wemust havef1 ≈ nfLO.
• Since fLO is equal to the desired channel frequency,f1 lies farfrom the band of interest and is greatly suppressed by theselectivity of the antenna, the band-select filter, and the LNA.
• The issue of LOharmonics does manifest itself if the receiveris designed for a wide frequency band (greater then twooctaves). Examples include TVtuners, “software-definedradios,” and “cognitive radios.”
Department of Electronic Engineering, NTUT54/110
Image-Reject Receivers
• “Image-reject” architectures are another class of receivers thatsuppress the imagewithout filtering, thereby avoiding thetrade-off between image rejection and channel selection.
• Benefits froma 90o shifter (Hilbert transform, −j for f > 0, +j for f <0)
Re
Im
Re
Im
Re
Im
2
Aj+
2
A
2
A
2
Aj−
cω−
cω+ω
( )cos2
c cj t j tc
AA t e eω ωω −= +
Illustration of 90o phase shift for a cosine
( ) ( )9090cos 902
ccj tj t
c
AA t e e
ωωω − −− − = +
2 2c cj t j tA A
je jeω ω−= − +
sin cA tω=
Department of Electronic Engineering, NTUT55/110
Hilbert Transform(I)
• Hilbert transformation pair:
• The transform means a 90 degree phase shift in time domain, the impulse response of the Hilbert transformation .
( ) ( )( )
ˆx
x t dt
ττ
π τ∞
−∞=
−∫ ( ) ( )( )x t
x dt
τ τπ τ
∞
−∞= −
−∫
( ) 1h t tπ=
( ) 1h t
tπ=( )x t ( )x t
90( )x t ( )x t
Department of Electronic Engineering, NTUT56/110
Hilbert Transform(II)
• Simple relation between sine and cosine functions:
• It simply shows that if we want to make a transformbetweencosine and sine waveforms, a 90 degrees phase shift isrequired.
• FromEuler’s formula:
( ) ( )cos 90 sint tω θ ω θ+ − = + ( ) ( )sin 90 cost tω θ ω θ+ − = − +and
0 0
0cos2
j t j te et
ω ω
ω−+=
0 0
0sin2
j t j te et
j
ω ω
ω−−=
( ) ( )0 0
12
δ ω ω δ ω ω− + +
( ) ( )0 0
1
2 jδ ω ω δ ω ω− − +
F.T.
F.T.
Department of Electronic Engineering, NTUT57/110
Hilbert Transform(III)
• We like to find a transfer function, which is able transfer thecosine to since and since to cosine function.
( ) ( ) ( ) ( ) ( )0 0 0 0
1 12 2
H j jδ ω ω δ ω ω ω δ ω ω δ ω ω− + + ⋅ = − − − +
( ) ( ) ( ) ( ) ( )0 0 0 0
1 1
2 2H j
jδ ω ω δ ω ω ω δ ω ω δ ω ω − − + ⋅ = − − + +
( ) ( )sgnH j jω ω= − ⋅
( )1 , 0
sgn 0 , 0
1, 0
ωω ω
ω
>= =− <
( ) ( ) 0
0
1 1sgn
2 2 2j t j t j tj j
h t j e d e d e dt
ω ω ωω ω ω ωπ π π π
∞ ∞
−∞ −∞= − ⋅ ⋅ = − =∫ ∫ ∫
cosine
sine negative cosine
0 0ω ω= > ( ) ( ) ( )0
1 10 0
2 2H j jδ ω δ⋅ = −
0 0ω ω= − < ( ) ( ) ( )0
1 10 0
2 2H j jδ ω δ⋅ − =
0 0ω ω= = ( ) ( ) ( ) ( ) ( )1 10 0 0 0 0
2 2H jδ δ δ δ+ ⋅ = − − ( )0 0H =
( )H j jω = −
( )H j jω =
:
:
:
sine
phase
f90+
90−
Department of Electronic Engineering, NTUT58/110
90o Phase Shift (I)
• 90o phase shift for a modulated signal
Re
Im
( )X ω
ωcω+
cω−
( )X ω( )jX ω
( )jX ω−
( ) ( ) ( )cos cx t A t t tω φ= +
( ) ( ) ( ) ( )( ) ( )( ) ( ) ( )( ) ( )( )90 90cos 90
2 2c c c c
j t t j t t j t t j t t
c
A t A tA t t t e e je je
ω φ ω φ ω φ ω φω φ + − − + − + − + + − = + = − +
( ) ( )sin cA t t tω φ= +
( ) ( ) ( )90sgnX X jω ω ω= −
Department of Electronic Engineering, NTUT59/110
90o Phase Shift (II)
Re
Im
2
Aj+
2
A
2
Aj−
cω−
cω+ω
sin2
c cj t j t
c
e eA t A
j
ω ω
ω−−=
2
A−
sin cjA tω
Re
Im
2
A
2
Acω−
cω+ω
cos ctω
Re
Im
Acω−
cω+ω
cos sinc ct jA tω ω+
• Plot the spectrum of cos sinc cA t jA tω ω+
(SSB spectra)
Department of Electronic Engineering, NTUT
sine is the Hilbert transform of the cosine
To get a SSB spectra: (1) real signal (2) its Hilbert transform (3) = (1)+j(2)
60/110
90o Phase Shift (III)
Re
Im
( )S ω
ωcω+
cω−
( )S ω( )jS ω
( )jS ω−
Re
Im
( )S ω
ωcω+
cω−( )S ω−
( )jS ω
( )jS ω−
Re
Im
( ) ( )ˆS jSω ω+
ωcω+
• A narrowband signal with a real spectrumis shifted by 90o
to produce . Plot the spectrumof which is calledthe “analytic signal,” or the “pre-envelope” of .
( )s t
( )s t ( ) ( )ˆs t js t+( )s t
( )S ω ( )ˆjS ω
Department of Electronic Engineering, NTUT61/110
90o Phase Shift (IV)
• Use RC-CRnetwork to implement the 90o phase shifter
1C
1R
1R
1C
1outV
2outVinV
HPFH
LPFH
11
2
1 1
1
R C
ω
11 1tan
2HPFH R Cπ ω−∠ = −
11 1tanLPFH R Cω−∠ = −
ω2
π
2
π−
02
π
( ) 1 1 1
1 1 1out
HPFin
V R C sH s
V R C s= =
+
( ) 2
1 1
1
1out
LPFin
VH s
V R C s= =
+
We can consider Vout2 asthe Hilbert transform ofVout1 at frequencies closeto (R1C1)−1
Department of Electronic Engineering, NTUT
Amplitude response
Phase response
62/110
Quadrature Downconversion (I)
• Quadrature downconversion translate the spectrumto anonzero IF as a 90o phase shifter.
RFV
IFI
cos LOtωsin LOtω
IFQ
0cω− cω+ ω
0LOω− LOω+ ω
1
2+1
2+
1
2+ 1
2+
0IFω− IFω+ω
0cω− cω+ ω
2
j+
2
j−
LOω+0LOω− ω
0IFω−IFω+
ω
2
j−
2
j+
High-side injection
Department of Electronic Engineering, NTUT63/110
Quadrature Downconversion (II)
• Quadrature downconversion translate the spectrumto anonzero IF as a 90o phase shifter.
2
j+
2
j−
LOω+0LOω− ω
0
IFω−
IFω+ ω
2
j−
2
j+
1
2+ 1
2+
0IFω− IFω+ω
Low-side injection
RFV
IFI
cos LOtωsin LOtω
IFQ
0cω− cω+ ω
0cω− cω+ ω
0LOω− LOω+ ω
1
2+1
2+
Department of Electronic Engineering, NTUT64/110
Distinguish Desired Signal and Its Image
( )sig im IFI I+
cos LOtωsin LOtω
( )sig im IFQ Q+
Re
Im
sigI
IFω−sigQ
sigQ
sigI
IFω+ ω
Re
Im
imI
imQ
imQ
imI
IFω+ ω
Re
Im
cω+ ωimω+imω−
cω−
Signal Components
Image Components
IFω−
Department of Electronic Engineering, NTUT
Signal: low-side injection
Image: high-side injection
LOω+
LOω−
65/110
Image Reject RX – Hartley Architecture (I)
• Negate image
Department of Electronic Engineering, NTUT
C
A
LPF
LPF
B
cos LOtωsin LOtω
90
sigQ
imQ,90im
Q
,90sigQ
sigI
imI
IF Outputω
imω+ cω+cω− imω− 0
sigI
sigQ
imI
imQ
Re
Im
IFω−
IFω+ ω
Re
Im
IFω−IFω+
ω
Re
Im
IFω−
IFω+ ω
Re
Im
IFω−
IFω+ ω
66/110
Image Reject RX – Hartley Architecture (II)
C
A
LPF
LPF
B
cos LOtωsin LOtω
90
sigQ
imQ,90im
Q
,90sigQ
sigI
imI
IF Output
Department of Electronic Engineering, NTUT
sigI imI
Re
Im
IFω−
IFω+ ω
Re
Im
IFω−
IFω+ ω
,90sigQ
,90imQ
Re
Im
IFω−
IFω+ ω
Re
Im
IFω−IFω+
ω
Re
Im
IFω−
IFω+ ω
67/110
• Negate image
Image Reject RX – Hartley Architecture (III)
• Realization of 90o phase shift in Hartley receiver
1cos tω1sin tω
RF Input
LPF
LPF
IF Output
1R
1R
1C
1C
Department of Electronic Engineering, NTUT68/110
Image Reject RX – Hartley Architecture (IV)
• Downconversion of Hartley receiver output to baseband
LPF
LPF
1cos LO tω1sin LO tω
90
2sin LO tω2cos LO tω
I
Q
RF Input
Department of Electronic Engineering, NTUT69/110
Image Reject RX – Weaver Architecture (I)
A
1cos tω1sin tω
RF InputLPF
LPF
2cos tω2sin tω
LPF
LPFB
C
D
E
F
−
+
IF Input
Department of Electronic Engineering, NTUT
ωimω+ cω+cω− imω− 0
sigI imI
Re
Im
1IFω−
1IFω+ ω
Re
Im
1IFω−
1IFω+ ω
sigQ imQ
Re
Im
1IFω−
1IFω+ ω
Re
Im
1IFω−1IFω+
ω
70/110
• Negate image
Image Reject RX – Weaver Architecture (II)
A
1cos tω1sin tω
LPF
LPF
2cos tω2sin tω
LPF
LPFB
C
D
E
F
−
+
IF
Department of Electronic Engineering, NTUT
sigI imI
Re
Im
2IFω−
2IFω+ ω
Re
Im
2IFω−
2IFω+ ω
sigQ imQ
Re
Im
2IFω−
2IFω+ ω
Re
Im
2IFω−
2IFω+ ω
Re
Im
IFω−
IFω+ ω
Low-side inj.
71/110
• Negate image
Image Reject RX – Weaver Architecture (III)
1inω ω−
ω2ω
2 12 inω ω ω− +0
FirstIF
ω
1 2inω ω ω− −0
SecondIF
ω
2 12 2inω ω ω− +1ω inω0
RF Input
SecondaryImage
DesiredChannel
Department of Electronic Engineering, NTUT
• Problem of secondary image
72/110
Image Reject RX – Weaver Architecture (IV)
Department of Electronic Engineering, NTUT
1cos tω
1sin tω
RF Input
LPF
LPF
2÷ ( ),BB Qx t
−
+−
+
( ),BB Ix t
• Double quadrature downconversion Weaver architecture toproduce BB outputs. The second downconverion produceszero IF to avoid secondary image.
73/110
Low-IF Receiver (I)
cf
GSM Adjacent Channel Spec.
9 dB
ff0 100 kHz
cff
LOff
200 kHz
• It is undesired to place image within the signal band becausethe overall NF would raise by approximately 3 dB.
• In “low-IF” RXs, the image indeed falls in the band but can besuppressed by image rejection techniques.
• For a GSMRX, signal would be corrupted by flicker noise in azero-IF architecture. The noise penalty can be lower by usinglow-IF architecture (attractive for narrow-channel standards).
Moderate IRR is ok.
Department of Electronic Engineering, NTUT74/110
Low-IF Receiver (II)
2sin tω
LPF
2cos tω
LPF
1R
1R
1C
1C
IF OutputRF Input
Quadrature Phases of Image and Signal
sin LOtω
LPF
cos LOtω IF OutputRF Input
LPF
ADC
ADC
90
Department of Electronic Engineering, NTUT
• Adopt image cancellation technique with low-IF architecture
75/110
Low-IF Receiver (III)
• Low-IF receiver with double quadrature downconverter
Department of Electronic Engineering, NTUT
sin ctωcos ctω
IF Output
RF Input
( ),IF Qx t
−
+ +
+
( ),IF Ix t90
76/110
Polar Receiver
• Using the oscillator injection locking technique to accomplishmagnitude and phase extraction of the complex envelope. Thistechnique was also published to performenvelope eliminationand restoration in the Kahn EER transmitter.
LPF
LPFILO1 ILO2
Magnitude
Phase
Department of Electronic Engineering, NTUT77/110
Transmitter Architectures
• Basic Direct Conversion Transmitter (DCT)
• Modern DCT
• Heterodyne Transmitters
• OOK Transceivers
• Open-loop Phase Modulation Techniques
• Closed-loop Phase Modulation Techniques
• Polar Transmitter
Department of Electronic Engineering, NTUT78/110
• Quadrature upconverter:Server as the modulator.
• Power amplifier:Amplify the signal.
• Matching network :Provide maximum power delivery to antenna and filter out-of-band componentsthat result from the PA nonlinearity.
• xBB,I(t) and xBB,Q(t) are generated by BB circuits and hence has a sufficientlylarge amplitude, the noise of the mixers is much less critical here than inreceivers.
• A predriver is typically interposed between the upconverter and the PA to serveas a buffer.
Direct-Conversion Transmitter (DCT)
cos ctω MatchingNetwork
PA
sin ctω−( ) ( ) ( )cos cx t A t t tω φ= +
( ) ( ) ( ) ( )cos cos sin sinc cA t t t A t t tφ ω φ ω= −
( ) ( ) ( ), cosBB Ix t A t tφ=
( ) ( ) ( ), sinBB Qx t A t tφ=
Department of Electronic Engineering, NTUT79/110
I/Q Mismatch
• I/Q mismatch in the DCT:
• Constellation:
( ) ( ) ( )1 2cos sinc c c c cx t A A t A tα ω θ α ω= + ∆ + ∆ +
( ) ( )1 2 1cos cos sin sinc c c c c c cA A t A A A tα θ ω α α θ ω = + ∆ ∆ + − + ∆ ∆
1 2, 1α α = ±
1 21 cos , 1 1 sinc c
c c
A A
A Aβ θ β θ
∆ ∆= + + ∆ = − + ∆
1 21 cos , 1 1 sinc c
c c
A A
A Aβ θ β θ
∆ ∆= + + ∆ = − − + ∆
1 21 cos , 1 1 sinc c
c c
A A
A Aβ θ β θ
∆ ∆= − + ∆ = + + ∆
1 21 cos , 1 1 sinc c
c c
A A
A Aβ θ β θ
∆ ∆= − + ∆ = − + + ∆
Department of Electronic Engineering, NTUT80/110
I/Q Calibration (I)
• Apply a single sinusoidal to both inputs of the upconverter toreveal phase mismatch.
It can be shown that the output contains two sidebands of equal amplitudes andcarries an average power equal to
We observe thatε is forced to zero as described above, then .
Thus, the calibration of phase mismatch proceeds to drive this quantity to zero.
cos ctωsin ctω0 cos inV tω 3outV
+
−
( ) ( ) ( )3 0 01 cos cos cos sinout in c in cV t V t V tε ω ω θ ω ω= + + ∆ −
( )0 cos 1 cos cosin cV tω ε θ ω= + ∆
( ) ( )2 23 0 1 1 sinoutV t V ε θ= + + ∆
2 23 1 sinout outV V θ− = ∆
( )0 cos 1 sin 1 sinin cV t tω ε θ ω− ⋅ + ∆ +
Department of Electronic Engineering, NTUT81/110
I/Q Calibration (II)
• Applying a sinusoidal to one BB input while the other is set tozero for gain mismatch calibration.
The gain mismatch can be adjusted so as to drive this difference to zero.
0 cos inV tω
cos ctωsin ctω 1outV
cos ctωsin ctω
0 cos inV tω
2outV
( ) ( ) ( )1 0 1 cos cosout in cV t V t tε ω ω θ= + ⋅ + ∆ ( )2
2 201 02out
VV t V ε= +
( )2 0 cos sinout in cV t V t tω ω= ⋅ ( )2
2 02 2out
VV t =
( ) ( )2 2 21 2 0out outV t V t V ε− =
Department of Electronic Engineering, NTUT82/110
Carrier Leakage (I)
• The analog BB circuitry producing the quadrature signalsexhibits dc offsets, and so does the baseband port of eachupconvertion mixer:
• The upconverter therefore contains a fraction of theunmodulated carrier, called “carrier leakage”:
( ) ( ) ( )1 2cos cos sin sinout OS c OS cV t A t V t A t V tφ ω φ ω= + − +
( ) ( ) ( ) 1 2cos cos sinout c OS c OS cV t A t t V t V tω φ ω ω= + + −
( )
2 21 2
2Relative Carrier Leakage OS OSV V
A t
+=
Department of Electronic Engineering, NTUT83/110
Carrier Leakage (II)
• RX BB outputs suffer from dc offsets.
• In the presence of carrier leakage, ifthe TX power is controlled byvarying BB signals, it is difficultforThe base station to measure theactual signal power.
0 2OSV V+ +
Q
I
0 1OSV V+ +0V− 0V+0V−
0V+
cos ctωsin ctω
PA
Receiver
Base Station
Bas
eban
dP
roce
sso
r
Carrier Leakage
cωω
Department of Electronic Engineering, NTUT84/110
Reduction of Carrier Leakage
Bas
eban
dP
roce
sso
r
DACQ
DACI
Register ADC
PowerDetector
cos ctωsin ctω
• The BB swing, A(t), must be chosen sufficiently larger toreduce carrier leakage. However, asA(t) increases, the inputport of mixers becomes more nonlinear. A compromise istherefore necessary.
( )
2 21 2
2Relative Carrier Leakage OS OSV V
A t
+=
• Use BB offset control to reducethe carrier leakage. Duringcarrier leakage cancellation, theBB processor produces a zerooutput so that the detectormeasures only the leakage. Thus,the loop can use the DACs todrive the leakage toward zero.
Department of Electronic Engineering, NTUT85/110
Transmitter Linearity (I)
• Upconversion mixers in TXsense no interferers, however,excessive nonlinearity in the mixer BB port can corrupt thesignal or raise the adjacent channel power.
• In most cases, as the BB signal swings increase, the PAoutputbegins to compress before the mixer nonlinearity manifestsitself.
• Power back-off is required for
variable envelope signal to avoid
spectrumregrowth at PAoutput.
1-dB Compression PointoutV
inV0V
t
Department of Electronic Engineering, NTUT86/110
Transmitter Linearity (II)
• In the TX chain, the signal may experience compression in anyof the stages. Since the largest voltage swing occurs at theoutput of the PA, this stage dominates the compression of theTX. In a good design, the preceding stages must remain wellbelow compression as the PAoutput approaches P1dB. Toensure this, we must maximize the PAgain and minimize theoutput swing of the predriver and the stage preceding it.
cos ctωsin ctω
,BB IV
,BB QV
PAPredriver
XV drV outVXV
drV
outV
BBV
Department of Electronic Engineering, NTUT87/110
Oscillator Pulling
• The PA output (very large swing) would couple to variousparts of the systemthrough the silicon substrate, packageparasitics, and traces on the printed-circuit board. Thus, it islikely that an appreciable fraction of PAoutput couples to theLO to pull the oscillator.
outφLO
ω∆
LOω ω
LOω ω
ω∆PA
LO
I
LOω
Q
Output Spectrum
injω ω
Department of Electronic Engineering, NTUT88/110
Avoid LO Pulling (I)
• Most of today’s DCTs avoid an oscillator frequency to the PAoutput frequency by using frequency division and mixing.
• Since the PAnonlinearity produces a finite amount of power atthe second harmonics of the carrier, the LOmay still be pulledby using the following architecture.
• Very high speed divider is needed, but even a substantial efforton divider design to enable this architecture is well justified.
I
2LO cω ω=
Q
LO 2÷PA
cω 2 cωω
Department of Electronic Engineering, NTUT89/110
Avoid LO Pulling (II)
• Use a frequency double is possible to avoid LOpulling, butthe doubler typically dose not provide quadrature phases,necessitating additional quadrature generation stages such asthe poly phase filter.
• Advantage: no harmonic can pull the LO.
• Disadvantage: the doubler and polyphase filter suffer fromahigh loss, requiring the use of power-hungry buffers.
I
2c
LO
ωω =
Q
LO 2X PolyphaseFilter
PA
cω ω
Department of Electronic Engineering, NTUT90/110
Avoid LO Pulling (III)
• Use of mixing to avoid LO pulling.
• It is difficult to remove the unwanted carrier by means offiltering because the two frequencies are only differ by only afactor of 3. Even the filter is applied, the unwanted sidebandwould corrupt other channels or bands.
1
2
ωLO
2÷
1ω
1
2
ω 13
2
ωω
PAQuadratureUpconverter
( ),BB Ix t
( ),BB Qx t
chI chQLO
1
2
ω 13
2
ωω
1
2
ω 13
2
ωω
Department of Electronic Engineering, NTUT91/110
Suppress Unwanted Sideband (I)
• Use the single-sideband (SSB) mixing technique to suppressthe unwanted sideband instead of filtering.
• The harmonics of the input frequencies also corrupt the outputof an SSB mixer.
2cos tω
2sin tω
1cosA tω
1sinA tω
outV1ω
2ω
outV
( )1 2 1 2 1 2cos cos sin sin cost t t t tω ω ω ω ω ω− = +
Symbol of a SSB mixer
2ω 1ω1 2ω ω−0 2 13ω ω− 1 2ω ω+ 1 23ω ω−ω
Department of Electronic Engineering, NTUT92/110
Suppress Unwanted Sideband (II)
• For use in a DCT, the SSB mixer must provide the quadraturephases of the carrier. This is accomplished by nothing that
2cos tω
2sin tω
1cos tω
1sin tω
2cos tω
2sin tω
+
−
+
+
( )1 2sin tω ω+
( )1 2cos tω ω+
( )1 2 1 2 1 2sin cos cos sin sint t t t tω ω ω ω ω ω− = +( )1 2 1 2 1 2cos cos sin sin cost t t t tω ω ω ω ω ω− = +
SSB mixer providing quadrature outputs
Department of Electronic Engineering, NTUT93/110
Suppress Unwanted Sideband (III)
• Carrier provided by SSB mixing for a DCT.
• While suppressing the carrier sideband atω1/2, thisarchitecture presents two drawbacks: (1) the spurs at 5ω1/2 andother harmonic-related frequencies prove troublesome, and (2)the LOmust provide quadrature phases, a difficult issue.
( ),BB Ix t
( ),BB Qx t
12
3
ω1
3
ωI/QI/QLO 2÷
PA
1ω
DCT using SSB mixing in LO path
Department of Electronic Engineering, NTUT94/110
Heterodyne Transmitter
• Another approach to avoiding injection pulling involvesperforming the signal upconversion in two steps so that theLO frequency remains far fromthe PAspectrum.
• Smaller I/Qmismatch
1sin tω
1cos tω
2cos tω
I
BPF
PA
Q
1ωω
2ωω
1 2ω ω+ω
Department of Electronic Engineering, NTUT95/110
Sliding-IF TX
• The carrier frequency is equal to 3ω1/2.
1ω
PA
BPF
LO2÷
I
Q
1
1
2ω
1
32
ω
RF Mixer
Department of Electronic Engineering, NTUT96/110
Carrier Leakage
• The dc offsets in BB yield a component atω1/2 at the output ofthe quadrature upconverter, and the dc offset at the input of theRF mixer produces another component atω1. The former canbe minimized, and the latter (lower sideband) atω1/2 must beremoved by filtering. The leakage atω1 is closer to the uppersideband than the lower sideband is, but it is also much smallerthan the lower sideband. Thus, the filter following the RFmixer must be designed to attenuate both to acceptably lowlevels.
1
2
ω
IF Output
1
2
ω 13
2
ω1ω
ω
RF Output
Carrier leakage in heterodyne TX
Department of Electronic Engineering, NTUT97/110
Mixing Spurs (I)
• Heterodyne TXdisplays various mixing spurs that must bemanaged properly. The spurs arise fromthe mechanismwith1st LO and 2nd LO.
1
2
ω+
IF Output
13
2
ω+ 15
2
ω+015
2
ω− 13
2
ω− 1
2
ω−ω
1
2
ω+ 13
2
ω+ 15
2
ω+0 17
2
ω+13
2
ω− 1
2
ω−ω
RF Output
1
2
ω+ 13
2
ω+17
2
ω− 015
2
ω− 13
2
ω− 1
2
ω−ω
LO mixed with 2LO, 5LO
IF mixed with 2nd LO
Department of Electronic Engineering, NTUT98/110
Mixing Spurs (II)
• Effect of harmonics of 2nd LO on TX output. Upon mixingwith +3ω1, the IF sideband at−3ω1/2 is translated to+3ω1/2,thereby corrupting the wanted sideband (if the modulation isasymmetric). Similarly, the IF sideband at−5ω1/2 is mixedwith +5ω1 and falls atop the desired signal.
13
2
ω+
ω13
2
ω− 015
2
ω−
13
2
ω+
ω0
Department of Electronic Engineering, NTUT99/110
Reduce Unwanted Components
• Use of BB quadrature SSB mixing and IF SSB mixing toreduce the unwanted component.
( ) ( ), cosBB Ix t A t θ=
( ) ( ), sinBB Qx t A t θ=
+
−
+
+
1ω LO2÷PA
RF SSB Mixer
I
Q
13
2
ω
Department of Electronic Engineering, NTUT100/110
OOK Transceivers
• On-off keying (OOK) modulation is a special case of ASKwhere the carrier amplitude is switched between zero andmaximum.
• Less bandwidth-efficient as unshaped binary pulses modulatedon one phase of the carrier occupy a wide spectrum.
LO
PA
LO
PA
LNAEnvelopeDetector
Direct LO switching PA switching
OOK RX
OOK TX
Department of Electronic Engineering, NTUT101/110
Open-loop Modulation
• Open-loop modulation based-on a frequency synthesizer (orphase-locked loop).
• Wideband (high data rate).
• Poor accuracy due to VCOfrequency drifting.
reff
DAC
VCO
PFD Loop Filter
Div-by-N
[ ]BBs n
( )BBs t
( )ms t
Department of Electronic Engineering, NTUT102/110
Closed-loop Modulation (I)
• Closed-loop modulation based-on a frequency synthesizer (or phase-locked loop).
• Narrowband (lowdata rate).
• Good frequency accuracy.
• No DACs required.
∆ −∑
VCO
PFD Loop Filter
[ ]BBs n
/ 1N N÷ +
( )ms t
reff
Modulator
Department of Electronic Engineering, NTUT103/110
Closed-loop Modulation (II)
• Use the compensated filtering to increase the data rate.
∆ −∑
VCO
PFD Loop Filter
[ ]BBs n
/ 1N N÷ +
( )ms t
reff
ModulatorCompensated
Filter
Department of Electronic Engineering, NTUT104/110
Closed-loop Modulation (III)
• Use the two-point ∆-Σmodulation to increase the data rate.
∆ −∑
Two-point VCO
PFD Loop Filter
[ ]BBs n / 1N N÷ +
( )ms t
reff
Modulator
DAC
Department of Electronic Engineering, NTUT105/110
Envelope Detector
Envelope Following/Tracking Transmitter
• Dynamically adjusting bias to improve efficiency.
( )BBA t′
( )ms t
Linear PA
Antenna
Matching
( )BBA t
I/Q Modulator
AmplitudeModulator/Regulator
Department of Electronic Engineering, NTUT
( )I t
cos ctωsin ctω−
( )Q t
106/110
Polar Transmitter (I)
• Envelope Elimination and Restoration Scheme (Kahn EER TX,1952):
Department of Electronic Engineering, NTUT
Envelope Detector ( )BBA t′
( )ms t
Switching-mode PA
Antenna
Matching
( )BBA t
I/Q Modulator
AmplitudeModulator/Regulator
( )I t
cos ctωsin ctω−
( )Q t
Limiter
107/110
Polar Transmitter (II)
• Polar Transmitter
Department of Electronic Engineering, NTUT
( )BBA t
cos ctω
( )ms t
Switching-mode PA
Antenna
PhaseModulator
Matching
( )BBA t
( )BB tφ
Bas
eban
d
Pro
cess
orAmplitudeModulator
( ) 2Re c BBj f t te
π φ+
• Linear modulator to generate PM signal• Frequency synthesizer or PLL-based PM modulator
108/110
Polar Transmitter (III)
• Hybrid Quadrature and Polar Modulation TX (HQPM-TX):
Department of Electronic Engineering, NTUT
Bas
eban
d
Pro
cess
or
( )BBA t′
( )ms t
Switching-mode PA
Antenna
Matching
( ),BB DSMA t
I/Q Modulator
AmplitudeModulator/
Class-S
( )I t
cos ctωsin ctω−
( )Q t
109/110
Summary
• In this chapter, many receiver and transmitter architectureswere introduced. For receiving or transmitting, there are twomain categories including heterodyne and direct conversionarchitectures.
• For these transceivers, the modulation and demodulation canbe classified as “I/Q” and “polar” schemes. I/Qmodulator isan universal modulator with high linearity and signal quality,and the polar modulator is adopted for improving powerefficiency. I/Q demodulator is the conventional scheme todemodulate signals, and the polar demodulator is proposed forlow-cost and low-power applications.
Department of Electronic Engineering, NTUT110/110