ECE 265A – Winter 2019

Lecture 13
Communications Concepts (3)
QAM, constellations
Demodulations
Intro. to system specs

Vincent Leung

© Leung, ECE265A Winter 2019 1

 Outline

 QAM
 Spectral regrowth (ACPR, or ACLR)
 Demodulation: matched filter and coherent detection
 TDD and FDD (Time/ Freq division duplexing)
 System Example (GSM)
o Sensitivity, P1dB, IMD3
 Effects of LO Phase noise
o Tx LO PN: Rx desensitization
o Rx LO PN: Reciprocal mixing

 References:
o Razavi textbook. Chapter 3, page 114-119, 130-136.
o Won’t cover page 119-130

© Leung, ECE265A Winter 2019 2

 Quadrature Amplitude Modulation (QAM)

 QPSK halves the bandwidth of PSK (while transmitting at the same data
rate)* – a “16QAM” could reduce it by another factor of two by allowing
4 possible amplitudes for the quadrature carriers :
o ,
o where and – a total of 16 combinations
 For a given transmitted power, the constellation becomes denser, thus
making the detection more sensitive to noise (detectability issue).
 Higher QAM also exhibits a larger envelope variations, requiring a
highly-linear (less efficient) PA.
 There is an example of tradeoff 3
between (1) bandwidth efficiency,
(2) detectability and (3) power efficiency.
1
(±1, ±3)

2-bit, 4 levels -3 -1 1 3
“Sending -1
4-bit
at a time”
(per symbol)
-3

© Leung, ECE265A Winter 2019 3

 64QAM

 QAM is rising*.
o For example, if 6 consecutive bits in the binary baseband stream
are S/P into 3-bit (8 levels) amplitudes for the quadrature carriers,
we get 64QAM
o WIFI (802.11a/g) employs 64QAM and achieve max. 54Mbps**
o Bandwidth will be reduced by 6 (to that of BPSK), but the detection
and power amplifier design becomes more difficult.
Q
*All pulse-shaping and
amplitude selection are D/A*
performed in the digital
domain (before the DAC) (±1, ±2, ±3, ±4)

3-bit, 8 levels
“Sending I
6-bit
at a time
(per symbol)

* “QAM is rising: 1024QAM and beyond”, Microwaves & RF, May 2014
** It employs 48 sub-carriers in an “OFDM” fashion. Each carrier can be BPSK,
© Leung, ECE265A Winter 2019 QPSK, 16QAM or 64QAM modulated (accomplished in the digital domain). 4
 Spectral Regrowth

 PA nonlinearities can cause “spectral regrowth” on variable-envelope

signals, such as QPSK* or QAM.
o “Spectral regrowth” corrupts the adjacent channels
 Consider a “constant envelope” signal:
o , where constant
o The 3rd-order nonlinearity will give:
•

o The spectrum in the vicinity of is un-affected.

 Consider a general, “variable envelope” signal (*see lecture 12, page 24)
o
•
( )

o The spectra of and tend to be “broadened” – and they

centered around

© Leung, ECE265A Winter 2019 * QPSK is considered variable envelope signal due to its baseband 5
pulse-shaping, but no significant information is carried in its amplitude.
 Spectral Regrowth (illustration)

 Nonlinear PA on “constant envelope” signal: shape of spectrum remains

unchanged:
o
 Nonlinear PA on “variable envelope” signal, spectrum grows:
( )
o

Spectral regrowth
is characterized in by
ACPR (adjacent channel power
ratio) or ACLR (leakage), in dB
(or dBc). See the Table Constant envelope
(transmitter session) in Lecture
11, page 9.

Variable envelope
Spectral
regrowth
© Leung, ECE265A Winter 2019 6
 Optimal Detection

 The baseband signal is digital, , where

o
o the detector must sample it every bit period,
, to determine its value. It should be time-
synchronized for optimal detection.
 Assuming the input is corrupted with additive
white noise, is there a filter that can help?
 It can be proven that a matched filter:
, would maximize the SNR at
the sampling instance.
o 𝒑(𝒕)
o 𝒙(𝒕) 𝒚(𝒕) 𝒚(𝑻𝒃 )
o
correlator
o
𝒙(𝒕)
 So, one can re-draw the optimum
detector using a correlator (as shown) 𝒚(𝒕)
𝒑(𝒕)

© Leung, ECE265A Winter 2019 ** If 𝑝(t) is zero outside (0, 𝑇 ) 7

 Matched filter & RRC filter

 As an intuition, assume is a rectangular pulse of duration ,

o The matched filter output will be a triangular pulse, which peaks at
, thus more favorable for detection.

𝒑(𝒕)

 It was discussed that “raised-cosine filter” could provide a tight spectrum

and ISI-free operation.
o In practice, the raised-cosine filter is decomposed into two sections,
called “Root-raised cosine filter” (RRC) -- one placed in the
transmitter and the other in the receiver (as in WCDMA*)
o The two sections together allow Nyquist signaling while the section
used in the receiver also operates as a matched filter.
Transmitter Receiver
channel

© Leung, ECE265A Winter 2019 * See ACLR screenshot two pages ago 8
 Coherent Detection

 Take the BPSK signal as an example.

o , where
 To demodulate it, one should first multiply it with ,
o
 Using the LPF to reject the 2nd-harmonic term: 𝑦 (𝑡)
LPF 𝑦(𝑡)
o 𝑥

o We can easily find the original binary

sequence . cos(𝜔 𝑡)

 We inherently assume perfect phase coherence between the modulated

signal (at the Tx) to the demodulation LO (at the Rx)*. Otherwise,
o

o
 As approach , the receiver will fail to recover at all!
* Sophisticated carrier recovery loop (in the digital domain) will phase-lock the
radio receiver to transmitted carrier frequency. This minimizes phase errors in
the data detection process. [Ref. CDMA Mobile Radio Design, Artech House,
© Leung, ECE265A Winter 2019 pg. 103-106] 9
 Coherent QPSK Demodulator

 QPSK demodulator can be thought of two BPSK signal streams

orthogonal to each other. They are each demodulated by “phase-
coherent” quadrature LOs.
 “Carrier recovery” ensures the LO precisely aligned in phase to the
transmitted carrier.
 “Symbol timing recovery” extracts the data clock, and ensures optimal
time-sampling within the bit period.
 On the in-phase (I) path:
o
o

 Similarly,
𝑦 𝑡
LPF ADC RRC 𝑏
A high-level view of a
coherent QPSK demodulator cos(𝜔 𝑡)
SYN/ Carrier Timing
𝑥 𝑇
LO recovery recovery
sin(𝜔 𝑡)
𝑦 𝑡
LPF ADC RRC 𝑏
© Leung, ECE265A Winter 2019 10
 Terminologies

 We won’t spend time on these topics. If they interest you, please read
the textbook and further references.
 Mobile RF communications
o Cellular system – frequency reuse
o Co-channel interference
o Hand-off
o Path loss ( ) and multi-path fading (Rayleigh distribution)
o Diversity (space or antenna)
o Delay spread
o Interleaving
 Multiple Access Techniques
o TDD, FDD, … (how Tx/ Rx of a mobile unit co-exist)
o TDMA, FDMA, CDMA … (how multiple users co-exist)
• CDMA (Direct sequence spread spectrum, DS-SS)
• Spreading, near/far effect (power control)

© Leung, ECE265A Winter 2019 11

 Time and Freq Division Duplexing

TDD FDD

Tx, Rx Simultaneous
takes turn Tx and Rx

 TDD  FDD (*most cellular standards)

o Same frequency band ( ) is o Employs two different frequency
utilized for both Tx and Rx paths. bands for the Tx and Rx paths
o System transmits half (some) of the o The frequency separation is called
time, and receives for the other half duplex frequency:
o Key component: T/R switch o Key component: Duplexer (filter)
o Tx/ Rx do not interfere, because Tx is o Duplexer reduces interferences from
turned off during Rx to save power Tx of other mobile units (Good)
(Good) o Own Tx signal (and phase noise)
o Strong signals by nearby mobile Tx fall leaks into Rx band (only ~50dB
in the Rx band, thus desensitizing the isolation) (Bad)
Rx (Very bad) o Duplexer is more complicated and
© Leung, ECE265A Winter 2019 higher loss than T/R switch (Bad) 12
 T/R Switch and Duplexer

FDD – Duplexer filter

TDD – T/R Switch

Rx Rx

Tx Tx

 Properties
o Low Insertion loss  Properties
o High isolation o Sharp cutoff
(High attenuation,
o High linearity
low loss)
o High power handling
o Low passband ripple
o (ref: www.skyworksinc.com)
o (ref: Qorvo TQQ7101)
© Leung, ECE265A Winter 2019 13
 GSM Air Interface & Sensitivity

 GSM (Global System for Mobile Communications) – 2G standard

worldwide (starting in Europe).
o TDMA / FDD System – 8 time-multiplexed users/ channel
o GMSK modulation
o 200kHz-wide channel, 25MHz-wide  ~1000 users

-102 dBm (sensitivity)

SNR=9dB
-111 dBm (noise floor)

 GSM sensitivity (MDS) requires* -102dBm

o Detection of GMSK with acceptable bit error rate (BER=10-3)
requires SNR of 9dB:
o NF = 10dB* * Phone manufacturers will outperform this.
** Using the equation from Lecture 6:
© Leung, ECE265A Winter 2019 𝑃 𝑑𝐵𝑚 = −174 + 10log(𝐵) + 𝑁𝐹 𝑑𝐵 + 𝑆𝑁𝑅 𝑑𝐵 14
 GSM – Blocking requirements
 GSM Rx blocking
requirements:

Same blocker
spec on this side
(mirror image)

 Desired Rx signal in the blocking test is 3dB above the sensitivity level.
 Blocker: (unmodulated) tone at frequency offset
o As frequency offset increases, so is the blocker power!
o OOB (beyond band edge+20MHz) blocker power jumps to 0dBm,
but it is expected to be heavily filtered by the duplexer
 1dB Compression Requirement [Assignment 3, Q4]:
o For example: worst-case blocker* = -23dBm; ,
gain compresses by 0.1dB.
* Other smaller blockers dictate
© Leung, ECE265A Winter 2019 VCO phase noise requirements 15
 GSM – IMD3 Test
 GSM Intermodulation Test:
-49dBm (interferences)

IMR3
-99 dBm (signal)
3dB
-102 dBm (sensitivity)
SNR=9dB
-108 dBm (noise + IMD3)
3dB
-111 dBm (IMD3input-referred
and the noise floor)
 IMD3 test conditions:
o Desired signal test is 3dB above the sensitivity level.
o A tone and a modulated interferences (at 800kHz and 1.6MHz
offsets) at -49dBm are applied
 Goal: satisfy BER (=10-3) requirement  SNR (or SNDR) = 9dB
 The noise floor is at -111 dBm (found previously for sensitivity test)
 IIP3 requirements*:
* Using the equation from Lecture 9:
( )
o dBm • 𝐴 𝑑𝐵 = 𝐴 𝑑𝐵 +

 So, the blocker test of the last page is more stringent

© Leung, ECE265A Winter 2019 16
 GSM – Transmitter
 Tx power is > 2W (+33dBm) in 900-MHz (or 1W for 1800MHz band)
 Need power control from +5dBm to max, at 2dB step
 GSM spectrum must satisfy “Emission Mask” – GMSK modulation/
pulse-shaping is done accurately.
 The maximum noise the Tx can emit in the Rx band (RxBN) must be
less than -129dBm/Hz.
o +33dBm – ( –129dBm/Hz) = –162dBc/Hz [note the unit]
o This calls for Tx LO phase noise spec of -162dBc/Hz at duplexer
offset (of 45MHz) – a tough spec.

-162dBc/Hz

RxBN

© Leung, ECE265A Winter 2019 17

 Tx LO Phase noise – Rx Desense
 Consider the effects of Tx LO phase noise with this numerical example*
 NF with Tx desense:
/ ( 𝟏𝟔𝟐 )/
o
o
 We see that with the transmitter on (FDD), the receiver is “desensitized by
0.33 dB”
Mechanism:
Tx noise at Rx
band adds to the
TXLO with
Rx Noise
phase noise -162dBc/Hz TXLO
Tx = +24dBm

freq
X
𝒇𝑻𝑿 𝒇𝑹𝑿

duplexer
iso = -45dB
* These numbers are
more pertinent to WCDMA system. X
But the concept still applies
RX (small signal NF = 2dB)
© Leung, ECE265A Winter 2019 18
 Rx LO Phase noise – Reciprocal Mixing

 Consider the effects of Rx LO phase noise with this numerical example*:

 NF with reciprocal mixing:

o / ( 𝟏𝟔𝟎)/

o
Tx = +24dBm
Mechanism:

duplexer
Phase noise of

iso = -45dB
RXLO down- RX (small signal NF = 2dB)
converts
the leaked Tx
X

RXLO Pj = -21dBm
* These numbers are
more pertinent to RXLO with
-160dBc/Hz
WCDMA system. phase noise
But the concept still
applies 𝒇𝑻𝑿 𝒇𝑹𝑿 freq
𝒇𝑻𝑿 𝒇𝑹𝑿 freq
© Leung, ECE265A Winter 2019 19