Transceiver Architectures

Modulation techniques -> Multiple Access techniques -> Transceiver Architecture

• General Considerations

• Receiver
o Heterodyne
o Homodyne
o Image Reject
o Digital IF
o Sub-sampling receiver

• Transmitter

o Homodyne
o Heterodyne

Jayanta Mukherjee 1
LNAs and PAs for 5G
• Carrier Aggregation – LNA and PA should be able to cater to multiple bands
simultaneously.
• Flexible Numerology – Channel BW is flexible.
• More channels means higher concentrated RF power in and around the base
station, so the problem of isolation between channels without mutual
interference is exacerbated.
• LNA must have high dynamic range performance in order to remain robust in the
presence of high power signals.
• For Massive MIMO – large no of LNA(Rx) and PA’s(Tx) needed in a given volume.
Size has to be small.
• Thermal management must be addressed with the increased electronics’ and
transmitters’ power.
2
Jayanta Mukherjee
General Considerations
The primary criteria for selecting transceiver architectures have been:
• Performance
• Complexity
• Cost
• Power dissipation
• Number of external components (cost and need to drive 50 ohm
impedance)
However as IC technologies evolve (e.g., RF CMOS and SiGe BiCMOS) the relative
importance of each of these criteria changes and new previously impractical
approaches (e.g., Digital IF, Direct conversion) will become possible.

Jayanta Mukherjee 3
Wireless in urban areas is often called an hostile communication environment.
Impact of narrow bandwidth (e.g., 30 kHz in IS-54 and 200 kHz in GSM) in a
receiver:

• need for coding, compression, bandwidth efficient modulation

• avoid leakage to adjacent channels (Fig. 5.1a)
• reject strong neighbor interferers (Fig. 5.1b)

Jayanta Mukherjee 4
Jayanta Mukherjee 5
Example (Fig 5.2):
Consider a 30 kHz channel at 900 MHz. A second order LC Band Pass filter providing 63 dB
of attenuation at 45 KHz needs to have a Q  f c / 2f of 107 , where 2f  90 Hz is the 3 dB frequency.
 Difficult to implement even with surface acoustic wave filter
 Typically calls for a tradeoff between in-band loss (noise figure) and Q
 Equivalent to a tradeoff between out-band rejection and in-band loss.
 In band loss is translated as increased NF.

6
Jayanta Mukherjee
Example (Fig 5.2):
Now Consider a f  30 kHz channel at 10 MHz. A second order LC Band Pass
filter providing 63 dB of attenuation at 45 KHz needs to have a Q  f c / 2f of 1.1 105 .
 Much easier to implement.
 Both in band loss is lesser and out of band rejection can be higher.

10 MHz

Jayanta Mukherjee 7
Band and Channel
It is important to distinguish between the band and the channel. For example in GSM:
• band is 935-960 MHz (25 MHz)
• channel bandwidth is 200 kHz

Band selection and channel selection are usually implemented separately:

• The RF front takes care of the band selection (removes out-band interferers).
• The IF or baseband stage take care of the IF channel selection (removes in-
band interferers).

Example: Figure 5.3

Jayanta Mukherjee 8
Importance of Non-Linearities in Receiver
Large in-band interferers still accompany the received signal after the front-end band-pass filter.

The linearity (IP3) of the LNA and mixer must be high to reduce intermodulation.

See Figure 5.5

Jayanta Mukherjee 9
 Band select filter

Jayanta Mukherjee 10
Other Receiver Issues

Dynamic Range
• Typically 100 dB

Supply noise and voltage changes as the PA is turned on and off

• Need for noise immunity and supply rejection in all building blocks.
• Frequency synthesizer in transceiver are pushed by supply
fluctuation.

Jayanta Mukherjee 11
Front End Duplexer Transition Bandwidth and
Rejection
See example in Figure 5.4:
• Rejection of only 20/56 dB at 20/40 MHz offset between the transmit and receive band.
• Leakage leads to :
o desensitization: 1W (30 dBm) transmitted signal at the antenna leads to -26 dBm which is close to the -25 dBm, the
1dB compression point of LNA.
o frequency synthesizer of receiver pulled by transmitter (Fig 5.6) i.e the PLL in Rx can shift frequency based on Tx
leakage.
o TDD better for preventing leakage (preferred in 5G). Leakage power at
40 MHz offset
=30-56 = -26 dBm

P1,dB = -25 dBm  only 1 dB

allowance of signal power to pass
undistorted.

P=(20/2)2/(2x50))
= 1 watt
= 30 dBm

12
Jayanta Mukherjee
Transmit Path Issues
Impact of Loss on Transmit Path
• Consider a 1 W amplifier. A 2 dB loss in the front-end bandpass filter leads to a
370 mW loss of power. Typically more than the entire receive path.

Spectral Regrowth
• The front end bandpass filter cannot suppress the out-of-channel spectral
regrowth since BP filter is for all channels.
• PA needs to be linear by design or a proper modulation should be used (recall
GMSK vs QPSK). In band spectral regrowth
still present
BPF
PA

Band

Jayanta Mukherjee 13
Heterodyne Receiver
The Heterodyne Receiver is based on down conversion:

• Translates the signal to a lower frequency fIF : the intermediate frequency IF using a mixer and a local oscillator fLO. (See
Figure 5.7a)

ωIF = |ωLO – ωRF|

• This relaxes the Q requirement of the channel selection: Q = fIF / Δf

Requirements of Down-Conversion

• Need to use an LNA before mixer as real mixers are noisy (Noise Figure on the order of the conversion loss: F = Lc) (See
Figure 5.7b)
• Down-conversion is not demodulation (except for zero-IF down conversion)
• The local oscillator does not need to be synchronized to the transmitter RF carriers even for subsequent coherent
demodulation (except for zero-IF )
• The local oscillator needs to be very stable and low noise (-120 dBc/Hz at 60 kHz offset set) so as to not scramble the
channels and corrupt the phase (zero-crossing).
Jayanta Mukherjee 14
 Mixer

Down conversion
mixing
cos  1 t  cos  0 t

  fIF (channel)
or
 RF  LO

1 1 Down conversion
 cos 1  0  t  cos 1  0  t
2  2  fLO
     
filteredout downconver
ted product

2  1   0   IF

Mixers are noisy so a LNA is

needed to obtain overall lower noise figure

Jayanta Mukherjee 15
 mean
  
squared noise power

n(t) 2
p(  (t ))  p ( A(t ))
 4A 4 
PSD of phase PSD of amplitude
var iation
 var iation

Amplitude of signal
Jayanta Mukherjee 16
"Amplitude to phase noise conversion in electronic circuits" Edoardo Milotti, PHYSICAL REVIEW E, Jan 1998
How to Select the LO and IF?
Problem of Image (See Figure 5.9)
• If ωRF,desired = ωLO - ωIF then the image ωRF,image = ωLO + ωIF is also downconverted.
• The image power being usually out-of band can be much higher than the desired signal
• Mitigation of the image problem can be done using
- an image reject mixer (Hartley and Weaver discussed later on for homodyne architecture)
- an image reject filter between LNA and mixer with large ωIF .

ω1 = ωLO – ωIF
ωim = 2ωLO – ω1 = ωLO + ωIF

ωLO

17
Jayanta Mukherjee
How large can 2ωIF be?
See Figure 5.10
• A large ωIF is required to get good image rejection since ωRF,image = ωRF,desired + 2 ωIF .

• For large ωIF , image will also be reduced by preliminary BPF, but if ωIF is small, then it is a problem

• But with a large ωIF the Q of the IF filter becomes an issue.

• Large IF leads to a poor channel selection as the Q of the IF filter degrades.

• Calls for a trade off between image-rejection and channel selection.

• Referred as a trade off between sensitivity and selectivity.

• High-side (supradyne) versus

• low-side injection (infradyne)

- Infradyne: ωLO < ωRF leads to easier design of LO and mixers (since LO freq is lesser than RF freq)
- Supradyne: ωLO > ωRF if required to avoid noisy lower band

• In FDD the duplexer may also serve to reject the image if the IF is high.
18
Jayanta Mukherjee
 Filtering more difficult since IF higher, but
less in band interferers, so Pi,mds
requirement lower (higher sensitivity)

Or tolerate
more in band
loss

tolerate more in band loss

-> undesirable Easier filtering since IF lower, but
more in band interferers, so Pi,mds
requirement higher (lesser sensitivity).

Jayanta Mukherjee 19
Other factors influencing ωIF

• Availability and size of filters => Higher Q implies higher size

=> Need external components. 868 - 895 MHz SAW filter

• Form Factor (especially for MIMO)

 Smaller filters => lower Q => lower IFs more suitable.

• Loss of image reject filter, image power

Jayanta Mukherjee 20
Problem of Half IF 𝜔RF,desired + 𝜔 Image
Reject
2
Filter 𝜔 𝜔
LNA
𝜔RF,desired
BPF IF filter
Assume we have (Scenario 1): +𝜔
2𝜔 IF Amplifier
RF,desired   LO
 an interferer at LO
2
 the interferer undergoes second-order distortion in the LNA doubling its frequency.
 the LO contains significant amount of 2nd order harmonic 2LO the mixer output includes then an interferer component:
 RF,desired  LO   2LO  IF
𝜔RF,desired + 𝜔 Image
Reject 𝜔
2
Filter 2 𝜔
Alternately (Scenario 2): LNA
RF,desired   LO BPF IF filter
 an interferer at 𝜔RF,desired + 𝜔
𝜔 IF Amplifier
2 2

RF,desired  LO RF,desired  LO IF

 down converted to  LO   LO
2 2 2
 the interferer undergoes second - order distortion in the IF amplier doubling its frequency to IF .

Jayanta Mukherjee 21
Remediation Techniques for Half IF Problem
• Reduce second order distortion in RF and IF paths. (common in MOSFETS)
• Reduce 2nd order harmonics in LO by maintaining a 50% LO duty cycle (no second
harmonics).
 RF,desired   LO
• Image reject filter providing sufficient attenuation at
< 50 % duty cycle hence 2nd harmonic 2
50 % duty cycle hence no 2nd harmonic present

t t
t t

LO o/p LO o/p

Jayanta Mukherjee 22
 Image
Reject
Filter IF Amplifier

BPF IF filter
LNA

LO

in  LO
2 Using image reject filter to remove
interferer
Jayanta Mukherjee 23
Dual IF Topology
The trade-off between sensitivity and selectivity in a single IF heterodyne is usually too severe. The
solution is to introduce a 2nd IF (see Figure 5.12)

• A large 1st IF can then be selected to achieve a high image rejection

• A low 2nd IF can then be selected to achieve a high channel selectivity

Need For Progressive Filtering:

• Assume 40 dB gain from A to G (Figure 5.12). If the two IF filters provide no channel selection, the IP3
of the IF amplifier would typically need to be more than 40 dB above that of the LNA IP3(+ 30 dBm).
• Instead each IF filter gradually reduces the adjacent channel interferers relaxing the linearity
requirement.
• Loosely stated as: “Every dB of gain requires one dB of pre filtering."

Jayanta Mukherjee 24
 40 dB gain
If no filtering done In rx
chain then iip3 of IF amplifier
should be i/p power + 40
ω IF1 ω IF2 dBm
ω RF

BPF1
Out of band Interferers BPF2
In band Interferers Band
Image suppressed
But interferers remain
but Q requirement is lower

ω RF ω RF

BPF3

first down conversion

Could not eliminate interferers completely ω RF Image folding
But relaxes Q requirement. ω IF1
BPF4
second
ω IF1 ω IF2 down conversion
second IF complete
suppression of interferers
With lower Q filter

IIP3 requirement of IF amplifier

Is much relaxed ω IF2
ω IF2

Jayanta Mukherjee
25
Disadvantages
• More power consumption
• Noise also is folded and added.
• Multiple filtering needed (increase in NF) to avoid high IP3 requirement.
• Mixer spur – multiple mixing can generate additional frequency components.
• In modern receivers LO1 and LO2 are related. Say LO2= (1/2) LO1

IF 1  RF  LO1

LO1
IF 2  0  IF 1  LO 2  RF  LO1  LO 2   RF  LO1 
2
3 2
 RF   LO1   LO1  RF
2 3

Jayanta Mukherjee 26
What Happens after the 2nd Down Conversion?
In FM systems the demodulation is performed at the 2nd IF frequency ωIF,2 generating the
baseband signal.

In digital systems the second down-conversion generates the in-phase (I) and quadrature (Q)
components of the signal while translating the spectrum to zero frequency(See Figure 5.13).

The A/D converters sample then the I and Q channels. Their resolution (4-10 bits) depends on
the SNR required and the sampling rate on the frequency response required.

Digital non-zero IF or sampling IF i.e. when signals are directly sampled from the IF instead of
baseband are possible but have prohibitively high speed and high linearity requirements(for
mitigating intermodulation) and are used so far in base stations where many channels need to
be processed simultaneously.

Jayanta Mukherjee 27
 AGC to reduce
or increase signal

second IF

Jayanta Mukherjee 28
Drawbacks and Advantage of Heterodyne
Drawbacks:

• Complexity: need for careful frequency planning (spurious frequency components)

• Needs for external components if using image-reject filter
• Trade-off between gain, noise-figure, stability and power dissipation.

Advantages:

• Excellent selectivity and sensitivity

• Still considered the most reliable reception techniques

Jayanta Mukherjee 29
Example +21 dB to -38 dB 130 nm BiCMOS
2.7 V power supply, 50 mA current

900 MHz

(dc)

TDD
1.3 GHz 400 MHz
(external) (external) (dc)

GMSK
dc
900 MHz

GMSK
dc

30
Jayanta Mukherjee