You are on page 1of 26

Architectures and RF System Design Issues

for Integrated Receivers and Transmitters in


3rd Generation Wireless Handsets

Walid Y. Ali-Ahmad
Senior Member of Technical Staff
Wireless Communications Group
Maxim Integrated Products
Sunnyvale, CA

10/26/01 Walid Y. Ali-Ahmad 1


TALK OUTLINE

 Introduction: Example of present level of integration in an RF chipset


for CDMA cellular radio
 Receiver Architectures
 Heterodyne Receiver
 Image-Reject Receiver
 Direct-Conversion Receiver
 Low-IF Receiver
 Transmitter Architectures
 IF-Modulation / Up-conversion Transmitter
 Direct-Modulation Transmitter
 Offset-PLL Transmitter
 Summary

10/26/01 Walid Y. Ali-Ahmad 2


Example of Present Level of Integration in an RF chipset
for CDMA Cellular Radio

10/26/01 Walid Y. Ali-Ahmad 3


Heterodyne Receiver

L
Advantages:
 Down-conversion to baseband I&Q is done at an “Intermediate-Frequency” (IF) lower than RF.
This results in superior I & Q matching.
 Its selectivity, which measures the receiver’s capability to process a desired small channel in
the presence of close-in strong interferes, is done partly at IF using a highly selective SAW filter
and at baseband I&Q using low-pass baseband filters.
 The use of the IF SAW filter relaxes the linearity requirements (IIP2, IIP3) of the succeeding IF
and baseband stages.
 DC offsets at baseband I&Q do not limit its sensitivity because they are minimized by the fact
that the first LO frequency is not equal to the input RF carrier frequency.

10/26/01 Walid Y. Ali-Ahmad 4


Heterodyne Receiver (cont’d)
Design Issues:
 The need to use an an off-chip passive BPF for image rejection adds cost and board space
requirements.
 Currently, due to technology limitations, this “image-reject” BPF and the IF SAW filter can not
be integrated on-chip.
 The trade-off between image rejection and channel selection is key in determining the IF
frequency.
 Low frequency IF SAW filters (40-150MHz) have high Q and provide high adjacent channel
selectivity. However, These filters tend to be large.
 High frequency IF SAW filters (150-400MHz) have a relatively smaller physical size, but
provide a lower adjacent channel selectivity.
 Good frequency planning is essential in order to minimize spurious responses generated in the
receiver’s front-end (Fs = ±m⋅ FRF ±n⋅ FLO1 ±p⋅ FLO2) .
 The “half-IF” spurious response at (FRF+ FLO)/2 can be a serious problem in the case of low IF
frequency. The front-end mixer after LNA should have a low 2nd-order distortion and a high
suppression of the (2FLO× 2FRF) product.
H a lf-IF

W a n te d
S i g n al 2 x 2 p ro d u ct
R F IF
fL O f R F - fI F / 2 fR F IF W a n te d
S i g n al
LO

10/26/01 Walid Y. Ali-Ahmad 5


Importance of IF Selectivity for Suppression of 3rd -order IM products

R F
 Typical Two-stages cascaded IIP3 equation, in linear format::
1 ( g il) ;“g1” is gain of stage 1; “il” is insertion loss of IF filter
= 1 + 1
iip3 iip31 iip3 2

 Equivalent IIP3 of IF block when including selectivity “S” (dB) ahead of IF stage:

P1
IIM 3 2 = 3 ⋅ (Po1 − S ) − 2 ⋅ IIP 3 2 = 3 ⋅ Po1 − 2 ⋅ (IIP 3 2 +
3
2
⋅ S) ⇒ IIP 3 e2 = IIP 3 2 +
3
2
⋅ S; (dBm)

 Generalized equation for overall IIP3 of a receiver chain with M cascaded stages:

I I M1 3
1 1 g1 g1 ⋅ g 2 g1 ⋅ g 2 ⋅  ⋅ g M −1
= + + ++

P
iip3 iip31 iip3 2 ⋅ s13 / 2 iip3 3 ⋅ (s1 ⋅ s2 )3 / 2 iip3 M ⋅ (s1 ⋅ s2 ⋅  ⋅ sM-1 )3 / 2

10/26/01 I Walid Y. Ali-Ahmad 6


Importance of RF Selectivity for Suppression of 2rd -order IM products

R F
 Two-stages cascaded IIP2 equation, in linear format:
1 1 ( g il) ; “g1” is gain of stage 1; “il” is insertion loss of image-reject filter
1
= +
iip2 iip21 iip22

 Equivalent IIP2 of mixer’s block, including selectivity “S” (dB) ahead of mixer stage:

P1
IIM 22 = 2 ⋅ (Po1 − S ) − IIP 2 2 = 2 ⋅ Po1 − (IIP 2 2 + 2 ⋅ S ) ⇒ IIP 2e2 = IIP 22 + 2 ⋅ S; (dBm)

 Generalized equation for overall IIP2 of a receiver chain with M cascaded stages:
g1 ⋅ g 2 g1 ⋅ g 2 ⋅  ⋅ g M −1

I I M1 2 O
1 1 g1
= + + ++
iip22 ⋅ s12 iip23 ⋅ (s1 ⋅ s 2 )2 iip2M ⋅ (s1 ⋅ s 2 ⋅  ⋅ sM-1)2

P
iip2 iip21

10/26/01 I Walid Y. Ali-Ahmad 7


n
Image-Reject Receiver

M
Desired

Advantages:
 It facilitates the integration of the heterodyne receiver’s front-end by eliminating the use of the
Image
off-chip RF image-reject filter and accomplishing the image rejection on-chip through phasing.
 It works nicely in receiver systems which do not suffer from strong out-of-band blockers and do
not require interstage filter between front-end LNA and MIXER blocks.
 It is suitable for receiver systems using a very low IF frequency (e.g. 10.7MHz, 45MHz), since it
eliminates the need for a very high Q bandpass RF filter in order to reject the image .

10/26/01 Walid Y. Ali-Ahmad 8


Image-Reject Receiver (Cont’d)
Design Issues:
1– Hartley Architecture:
 Image rejection is limited by amplitude and quadrature phase mismatches. Amplitude
mismatches are minimum when using ω IF= 1/RC. Phase mismatches are due mainly to errors in
the LO quadrature generation circuit.
Image Rejection Ratio IRR (dB) = 10∗LOG([(∆ A/A)2 + θ 2 ]/4); for (∆ A/A)<<1, θ <<1rad.
 For typical matching in integrated circuits, image suppression falls in the range of 30 to 40dB
(0.2-0.6dB gain mismatch & 1°-5° quadrature phase mismatch). In most RF systems, 60-70dB of
image rejection is required. The front-end filter (duplexer) normally makes up for the remaining
required image rejection (~30dB).
 For the RC-CR 90° phase-shift network, (∆ A/A) = (∆ R/R) + (∆ C/C), at ω RC≈ 1.

2– Weaver Architecture:
 It is also sensitive to mismatches, but it is free from gain imbalances due to the RC-CR phase
shift network, thereby achieving greater image rejection despite process and T° variations.
 The Weaver architecture suffers from the “secondary” image problem because of the use of a
second mixing operation. The LPF (or HPF) in between 1st and 2nd mixing stages is used to
suppress the secondary image.
 The problem of secondary image can be eliminated if we choose a Zero IF frequency at the
output (FLO1± FLO2 = FRF). To its advantage also, 2nd-order distortion in the signal path can be
removed by the bandpass filters following the first mixing operation.

10/26/01 Walid Y. Ali-Ahmad 9


Image-Reject Receiver (Cont’d)
 A 1.9GHz Wide-Band IF Double Conversion Receiver (J.C. Rudell, et al., UC Berkeley, IEEE
JSSC, December 1997):

10/26/01 Walid Y. Ali-Ahmad 10


Direct-Conversion Receiver

 In current cellular systems where signals are either frequency- or phase-modulated, direct
downconversion must provide quadrature outputs so as to avoid loss of information, since the
two sidebands of the RF spectrum contain different phase information.
Advantages:
 The problems of image frequency and “half-IF” spurious response are eliminated since FIF = 0.
 RF BPF after LNA is optional; it is only needed for additional rejection of out-of-band interferers
and TX power leakage.

B P F L N# 1
 The bulky off-chip IF SAW filter is eliminated. All channel selectivity is done at baseband with
low-pass filters and baseband amplifiers.
 One VCO and one PLL are needed for the whole receiver.

10/26/01 Walid Y. Ali-Ahmad 11


Direct-Conversion Receiver (Cont’d)

Design Issues:
1– DC offsets:
 Static offsets are caused by process mismatch and drift of analog circuitry that vary slowly vs. T°,
aging, and current gain setting.
 Time-variant offsets are caused mainly by parasitic LO coupling to mixer RF port, LNA input port,
and antenna port. LO “self-mixing” occurs in mixer and it produces a dc component at the mixers
I & Q baseband outputs.
 Time-variant offsets can also be caused by a large interferer which can leak from LNA or mixer
input to LO input port and self-mix with itself to produce a dc offset at mixers outputs.
 The time-variance is due to reflection of LO leakage against moving objects back to receiver and
due to receiver movements.

ν max: maximum moving object or car speed.

10/26/01 Walid Y. Ali-Ahmad


B
 Maximum frequency content of time-variant DC offset due to Doppler shift = 2∗ν max

P
/λ ; where

12
F
Direct-Conversion Receiver (Cont’d)

2– DC offsets Cancellation Techniques:


DC offsets can easily saturate the receiver’s final baseband output stages. Hence, DC offset
removal or cancellation is required in direct-conversion receivers:
− DC blocking or High pass filtering:
− it is feasible in non-burst mode systems which are receiving continuously (FDD).
− In order to minimize distortion of signal, the high-pass corner should be < 0.1% of the data
rate for random binary M−ary data.
− The baseband signal in the transmitter can be encoded to result in “dc-free” modulation
scheme, such as FSK with β m > 1 or wideband Direct-Sequence Spread-Spectrum signals.
− DC calibration loop: In TDD systems, periodic offset cancellation can be performed during idle
times where the DC offset is stored on a capacitor and then subtracted from the received signal
during actual reception.
− Adaptive DSP techniques have been used for DC offset-cancellation in TDD systems.

10/26/01 Walid Y. Ali-Ahmad 13


Direct-Conversion Receiver (Cont’d)

Eye Diagram
Distortion with
HPF:
a) No filtering;

b) Fc = 1% of
Rb;

c) Fc = 0.1% of
Rb

DC offset cancellation in Pager system


using HP filtering

10/26/01 Walid Y. Ali-Ahmad 14


Direct-Conversion Receiver (Cont’d)
3– LO leakage:
 LO coupling to the antenna will be radiated out and will create interference in the receive band of
other users equipment using the same wireless standard.
 In order to minimize this problem of LO leakage and re-radiation, it is important to use differential
LO and RF inputs to the receiver IC to cancel out common mode signals. In addition, LO leakage
is further reduced by fully integrating the RF VCO tank on chip.
4– Flicker Noise:
 The 1/f noise of devices in the baseband section of a Zero-IF receiver can substantially corrupt
the down-converted signal after the mixers I/Q outputs, especially in MOS implementations (1/f
corner ~ 200kHz).
 The effect of flicker noise can be reduced by the use of active mixers with bipolar transistors in
switching pairs and by the use of large MOS devices for baseband filters and amplifiers.
 High pass filtering at baseband, when used as part of the DC offset cancellation, can reduce the
integrated 1/f noise at baseband.
 Integrated total noise at baseband including 1/f noise can be expressed as following:
f
2 1/ f
V =S ⋅f ⋅ ln( ) + ( f − f )⋅S ;
n th 1 / f f H c th
c
S : Thermal noise power spectral density, f :1/f corner, f : low frequency corner, f : High frequency corner
th 1/ f c H

10/26/01 Walid Y. Ali-Ahmad 15


Direct-Conversion Receiver (Cont’d)

Interference due to down-converted 1/f Mixer Topology that minimizes 1/f


noise and DC offset noise at its baseband output

10/26/01 Walid Y. Ali-Ahmad 16


Direct-Conversion Receiver (Cont’d)

5– I/Q mismatch:
 Assuming that the received signal can be written as vin(t)= I(t)∗sin(ω Ct)+Q(t)∗cos(ω Ct), and the
amplitude and quadrature phase imbalances in the I/Q down-converter are “ε ” and “θ ”,
respectively, we can write the baseband I/Q outputs, after demodulation, as:
vBB,I (t)= I(t).(1+ε /2).cos(θ /2) − Q(t).(1+ε /2).sin(θ /2);
vBB,Q (t)= −I(t).(1−ε /2).sin(θ /2) + Q(t).(1−ε /2).cos(θ /2);
 As we see from equations above, Gain error “ε ” appears as a non-unity scale factor in the
amplitude, while phase imbalance “θ ” results in cross-talk between demodulated I and Q
waveforms degrading the SNR (I & Q data streams are usually uncorrelated).
− In practice, ε < 1dB and θ < 5° for SNR degradation less than 1dB (for QPSK signals).
 The full on-chip integration of the Zero-IF receiver and the minimization of devices mismatch
reduce drastically the amplitude and quadrature phase imbalances in the I/Q down-converter.
6– Channel Filtering:
 Baseband channel low pass filters in a Zero-IF receiver need to have a high dynamic range:
− Receiver sensitivity can’t be compromised.
− Close-in interferes should be rejected without causing in-band distortion.
 External capacitors at mixers I&Q outputs can be used to provide additional selectivity to
blocking and out-of-band signals (pole @ 1/RCext)

10/26/01 Walid Y. Ali-Ahmad 17


Direct-Conversion Receiver (Cont’d)

7– Even-Order Distortion:
 Assume vRF(t) = A1∗cos(ω 1t)+A2∗cos(ω 2t), the LNA output will contain an IM term at frequency f1-f2,
resulting from 2nd order non-linearity in the LNA. This IM2 product at LNA output will leak to
mixer’s output because of finite feedthrough from RF input to IF output (−30…−40dBc).

I n t e
 Special attention to the mixer’s design is required since IM2 product can also be generated in the
mixer’s RF port by two tones interferes after being amplified in LNA.
 The 2nd order non-linearity in the LNA and in the mixer will also demodulate any AM component
on the received signal due to fading during propagation or Nyquist filtering.
 Based on input two-tone interferers level and the resultant low-frequency IM2 level at baseband
output, a receiver’s 2nd order intercept point (IP2) can be derived.
 Using differential LNA output and differential mixer’s input will suppress the generated common-
mode 2nd-order IM products. As a result, receiver’s IIP2 can be improved.

10/26/01 Walid Y. Ali-Ahmad 18


Low-IF Receiver

Advantages:
The received signal is down converted to a low-IF frequency, which is normally one to two times
the signal BW.
 It has the same advantage of Zero-IF receiver in terms of the integration of channel filters.
 It is less susceptible to 1/f noise.
 It is less susceptible to DC offsets since the bulk of signal energy is not centered around DC.

B P F L N# 1
 DC offsets cancellation scheme can be simplified.
 Very low frequency IM2 products can be easily blocked.

10/26/01 Walid Y. Ali-Ahmad 19


Low-IF Receiver (cont’d)

Design Issues:
 The A/D converters at baseband output require to have a higher sampling rate than in the case of
a Zero-IF receiver.
 Image suppression is an issue; Very good amplitude and phase matching between I & Q
baseband channels is required to obtain > 35dB image suppression.
 Careful choice of the IF can place the image signal in the adjacent channel.
− In order to discriminate between these two signals, it is essential to process I & Q outputs as
a complex pair.
− Complex Polyphase filters is essential to obtain the necessary reject in the adjacent channel
 2nd -Order Distortion can still result in in-band channel intereference.

10/26/01 Walid Y. Ali-Ahmad 20


IF-Modulation / Up-Conversion Transmitter

M i x
I
 Baseband I&Q signals undergo quadrature modulation at an intermediate IF frequency (ω IF).
The following IF filter (BPF1) rejects the harmonics of the IF signal. The IF modulated signal is
then up-converted to (FIF ± FLO2).
 The unwanted sideband imposes tough rejection requirements on BPF2, typically 50-60dB, in

c o ω I Fs t
order to meet transmitter’s spurious emissions levels imposed by standards.
 This topology does not allow full transmitter’s integration because of use of off-chip passive
devices such BPF2 and BPF1.
 On-chip I and Q matching is superior since modulation is done at IF and not at RF. This will
lead to better EVMs and lower cross-talk between I & Q channels.
 IF filtering reduces transmitted noise in RX band.
 Wide power control dynamic range because control it is distributed between RF and IF sections.

10/26/01 Walid Y. Ali-Ahmad 21


Direct-Modulation Transmitter

M i x e r
I
 In direct-conversion transmitters, the baseband signal is directly modulated unto the RF carrier.
The output carrier frequency is equal to the LO frequency at mixers inputs.
 This topology is attractive for full transmitter’s integration since it does not use an intermediate IF

c o ω L s tO ()
stage with upconversion and interstage IF filter.
 Its main disadvantage is the corruption through “injection pulling” of the VCO spectrum by the
high level PA output. Isolation required is normally > 60dB.
 The isolation can be highly improved by “offsetting” the LO frequency by using 2xLO off-chip and
dividing by 2 on-chip or by adding or subtracting another oscillator.
 The power control dynamic range is limited by the carrier feedthrough. A fully integrated

0
differential transmitter architecture will minimize carrier feedthrough because of higher of
common mode rejection (differential LO inputs and modulator output).

10/26/01 Walid Y. Ali-Ahmad 22


Direct-Modulation Transmitter (cont’d)

I ( t ) A I
O I & OQ are DC offsets at baseband I & Q inputs, respectively; AI & AQ are amplitude errors of I & Q paths, respectively.
Φ e is modulator quadrature phase error; I(t) and Q(t) are input baseband signals.

1 - The Error Vector magnitude (EVM) and phase (EVP) components can be calculated as following :

O
∴ EVM(t) = (( AI − 1) I (t ) + OI AI + (Q(t ) + OQ ) AQ sin(Φ e )) 2 + ( AQ cos(Φ e ) − 1)Q (t ) + OQ AQ cos(Φ e )) 2

c
[(
∴ EVP(t) = tan −1 AQ cos(Φ e ) − 1)Q(t ) + OQ AQ cos(Φ e ) ) ( ( AI − 1) I (t ) + OI AI + (Q(t ) + OQ ) AQ sin(Φ e ) )]
2 - The carrier and sideband suppression can be calculated as following : I
(assuming single tone modulating signal ωm ; I(t) = cos (ωm t); Q(t) = sin (ωm t))
  AQ   AQ 
2  1  2  
  2
∴ Carrier Suppression (dB) = 10 ⋅ LOG O I + 2 ⋅   ⋅ O I ⋅ OQ ⋅ sin (Φe ) +   ⋅ OQ2   ⋅ 1 + 2 ⋅  AQ  ⋅ cos(Φ ) +  AQ   
 AI  AI   4  AI  AI   
e
         
  AQ   AQ  
2  2 

∴Sideband Suppression (dB) = 10 ⋅ LOG  1 − 2 ⋅  ⋅ cos(Φ e ) +  1 + 2 ⋅  AQ  ⋅ cos(Φ ) +  AQ  
  AI   AI     AI  e
 AI  
  

10/26/01 Walid Y. Ali-Ahmad 23


Direct-Modulation Transmitter (cont’d)

Modulation error or EVM in Transmitters

10/26/01 Walid Y. Ali-Ahmad 24


Offset-PLL Transmitter

M i x
I
 Baseband I&Q signals undergo quadrature modulation at an intermediate IF frequency (ω IF).
Instead of upconverting the IF signal, the phase modulation in the IF signal is transferred
faithfully to the TX VCO via the offset PLL topology, with condition that the loop BW of PLL is
chosen properly.

c o ω I Fs t ) (
 This phase translation scheme to RF is only valid for constant envelop modulation signals such
in GSM (GMSK).
 The PLL LPF helps tremendously at suppressing the out-of-band noise generated by the
modulator and, hence, meeting the stringent GSM requirements for the thermal noise in the
receive band. This eliminates the need for the off-chip bulky Duplexer.
 This topology is quite attractive for low-cost high performance integrated transmitters using
constant-envelop modulation. However, enough isolation is required between the TX VCO and

0
the PA in order to suppress VCO pulling by PA output noise.

10/26/01 Walid Y. Ali-Ahmad 25


SUMMARY

 Need for Fully Integrated VCOs.


 Need for Mixers with low level of 2nd-order Distortion (High IIP2).
 DC offset cancellation schemes have to be implemented without distorting signal.
 High Dynamic Range Baseband filters are key for direct-conversion receivers.
 Variable Gain PAs improve power control dynamic range in Direct-Modulation Transmitters.
 Low Noise Direct-Modulation Transmitters and Direct-Conversion Receivers enable low-cost radios for 3G applications.

10/26/01 Walid Y. Ali-Ahmad 26

You might also like