Sources and References

Primary references — B. Razavi, RF Microelectronics, 2nd ed., Prentice Hall, 2012; D. M. Pozar, Microwave and RF Wireless Systems, Wiley, 2001. Supplementary texts — Kai Chang, RF and Microwave Wireless Systems, Wiley, 2000; J. S. Seybold, Introduction to RF Propagation, Wiley, 2005; M. Steer, Microwave and RF Design: A System Approach, SciTech, 2010. Online resources — MIT OpenCourseWare 6.776 High Speed Communication Circuits; IEEE Communications Society tutorials.

---

Chapter 1: Introduction to Modulation and Multiple Access Techniques

1.1 The Role of Modulation in Wireless Systems

Wireless communication requires that baseband information — voice, data, video — be translated to a radio-frequency carrier so that it can propagate through space efficiently. Modulation is the process of embedding an information-bearing signal onto a carrier waveform by varying one or more of its parameters: amplitude, frequency, or phase. The choice of modulation scheme determines the spectral efficiency, power efficiency, robustness to noise, and the hardware complexity of both the transmitter and receiver.

A generic passband signal can be written as

\[ s(t) = A(t)\cos\!\bigl[2\pi f_c t + \phi(t)\bigr] \]

where \(f_c\) is the carrier frequency, \(A(t)\) is the time-varying envelope, and \(\phi(t)\) is the time-varying phase. Amplitude modulation fixes \(\phi\) and varies \(A\); frequency/phase modulation fixes \(A\) and varies \(\phi\); modern digital schemes vary both.

1.2 Analog Modulation

1.2.1 Amplitude Modulation (AM)

In conventional AM the envelope carries the information:

\[ s_{AM}(t) = A_c\bigl[1 + m\,x(t)\bigr]\cos(2\pi f_c t) \]

where \(x(t)\) is the normalised message signal and \(m\) is the modulation index (\(0 < m \le 1\) for no overmodulation). The bandwidth is \(B_{AM} = 2W\) where \(W\) is the message bandwidth. Because the carrier is always present, AM is wasteful of power: at \(m = 1\) with a sinusoidal message, only one-third of the total power resides in the sidebands. Double-sideband suppressed-carrier (DSB-SC) and single-sideband (SSB) variants improve power efficiency.

1.2.2 Frequency Modulation (FM)

FM embeds information in the instantaneous frequency deviation:

\[ s_{FM}(t) = A_c\cos\!\Bigl(2\pi f_c t + 2\pi k_f \int_{-\infty}^{t} x(\tau)\,d\tau\Bigr) \]

The peak frequency deviation is \(\Delta f = k_f \max|x(t)|\), and the modulation index is \(\beta = \Delta f / W\). By Carson’s rule, the RF bandwidth is \(B_{FM} \approx 2(\Delta f + W) = 2(\beta + 1)W\). FM trades bandwidth for improved noise performance: the output SNR improvement over AM is approximately \(3\beta^2(\beta+1)\) for large \(\beta\), making wideband FM attractive for high-fidelity audio broadcasting.

1.3 Digital Modulation

1.3.1 Binary Phase-Shift Keying (BPSK)

BPSK maps binary data to two antipodal constellation points at phases \(0\) and \(\pi\):

\[ s_i(t) = \sqrt{\frac{2E_b}{T_b}}\cos(2\pi f_c t + (i-1)\pi),\quad i=1,2 \]

Spectral efficiency is 1 bit/s/Hz (ideal); the bit error probability over an additive white Gaussian noise (AWGN) channel is \(P_b = Q\!\bigl(\sqrt{2E_b/N_0}\bigr)\). BPSK has a constant envelope, which is favourable for nonlinear PA operation.

1.3.2 Quadrature Phase-Shift Keying (QPSK)

QPSK uses four phases (\(45°, 135°, 225°, 315°\)), transmitting 2 bits per symbol. At the same symbol rate as BPSK, QPSK doubles spectral efficiency to 2 bits/s/Hz while maintaining the same \(E_b/N_0\) BER performance. It is widely used in satellite links and CDMA downlinks.

1.3.3 Quadrature Amplitude Modulation (QAM)

\(M\)-QAM places \(M\) constellation points on a rectangular grid, achieving \(\log_2 M\) bits per symbol. 16-QAM delivers 4 bits/symbol; 64-QAM delivers 6 bits/symbol. The spectral efficiency gain comes at the cost of reduced noise margin: the minimum Euclidean distance decreases as \(M\) increases, requiring higher SNR.

1.3.4 Orthogonal Frequency Division Multiplexing (OFDM)

OFDM divides a wideband channel into \(N\) narrowband orthogonal subcarriers, each modulated independently (typically with QAM). The composite signal is

\[ s(t) = \text{Re}\!\left\{\sum_{k=0}^{N-1} X_k\, e^{j2\pi k \Delta f\, t}\right\},\quad \Delta f = \frac{1}{T_s} \]

A cyclic prefix (CP) of length \(T_{CP} \ge \tau_{max}\) (maximum multipath delay spread) eliminates inter-symbol interference, converting the frequency-selective fading channel into a set of flat-fading sub-channels equalisable by a single complex tap. The IFFT/FFT pair provides an efficient implementation. OFDM is the waveform of choice in 4G LTE, 5G NR, Wi-Fi (802.11a/g/n/ac/ax), and DAB/DVB.

1.4 Multiple Access Techniques

Multiple access allows many users to share a common spectrum resource simultaneously.

1.4.1 Frequency Division Multiple Access (FDMA)

FDMA allocates a distinct frequency band to each user for the duration of their call. Guard bands prevent inter-channel interference. FDMA was used in first-generation cellular (AMPS, TACS). Its drawbacks are rigid spectrum allocation and relatively low efficiency when traffic is bursty.

1.4.2 Time Division Multiple Access (TDMA)

TDMA assigns each user a distinct time slot within a repeating frame, all sharing the same carrier frequency. GSM uses TDMA with 8 slots per 200 kHz carrier. TDMA allows statistical multiplexing but requires accurate synchronisation and burst-mode transmission.

1.4.3 Code Division Multiple Access (CDMA)

In CDMA all users transmit simultaneously over the same bandwidth, distinguished by unique pseudo-noise (PN) spreading codes that are mutually orthogonal (or near-orthogonal). The processing gain is \(G_p = B_{SS}/R_b\) where \(B_{SS}\) is the spread bandwidth and \(R_b\) is the data rate. CDMA inherently provides soft capacity, multipath resistance (RAKE receiver), and soft handoff. cdmaOne (IS-95) and WCDMA/CDMA2000 (3G) are commercial implementations.

1.4.4 Orthogonal Frequency Division Multiple Access (OFDMA)

OFDMA extends OFDM to multiple access by assigning different subcarrier subsets to different users in the frequency and time domains. Schedulers can exploit multiuser diversity (assigning subcarriers on which each user has good channel gain). OFDMA is used in 4G LTE downlink and 5G NR.

---

Chapter 2: System-Level Key Characteristics — Receiver

2.1 Thermal Noise and Noise Figure

2.1.1 Thermal Noise Power

Any resistive element at physical temperature \(T\) (Kelvin) produces thermal (Johnson–Nyquist) noise with available power spectral density

\[ S_n = kT \quad [\text{W/Hz}] \]

where \(k = 1.38 \times 10^{-23}\) J/K is Boltzmann’s constant. In bandwidth \(B\), the available noise power is \(N = kTB\). At room temperature (\(T_0 = 290\) K) this is \(-174\) dBm/Hz.

2.1.2 Noise Figure Definition

The noise figure (NF) of a two-port is the ratio of the total output noise power spectral density to the portion attributable to the source alone:

\[ F = \frac{S/N_{\text{in}}}{S/N_{\text{out}}} = \frac{N_{out}}{G\,kT_0 B} \]

NF in decibels is \(\text{NF} = 10\log_{10} F\). An ideal noiseless amplifier has \(F = 1\) (NF = 0 dB). Practical LNAs achieve NF of 0.5–3 dB; mixers typically contribute 6–12 dB.

2.1.3 Friis Noise Figure for Cascaded Stages

For a cascade of \(N\) two-port stages with individual noise factors \(F_i\) and available power gains \(G_i\), the overall system noise factor is

\[ F_{sys} = F_1 + \frac{F_2 - 1}{G_1} + \frac{F_3 - 1}{G_1 G_2} + \cdots + \frac{F_N - 1}{\prod_{i=1}^{N-1} G_i} \]

2.2 Receiver Sensitivity

Receiver sensitivity \(P_{min}\) is the minimum detectable signal power for a specified output signal-to-noise ratio \(\text{SNR}_{min}\) (determined by the demodulation scheme and target BER):

\[ P_{min} = kT_0 B \cdot F \cdot \text{SNR}_{min} \]

In dBm:

\[ P_{min}[\text{dBm}] = -174 + \text{NF}[\text{dB}] + 10\log_{10}(B) + \text{SNR}_{min}[\text{dB}] \]

2.3 Dynamic Range and Blocking

2.3.1 Linearity Metrics: 1-dB Compression and IIP3

Real amplifiers deviate from linear behaviour at large signal amplitudes. A standard behavioural model retains third-order nonlinearity:

\[ y(t) = \alpha_1 x(t) + \alpha_2 x^2(t) + \alpha_3 x^3(t) \]

For a single-tone input \(x = A\cos\omega t\), the fundamental output amplitude is \(\alpha_1 A + \frac{3}{4}\alpha_3 A^3\). The 1-dB compression point (P1dB) is the input power at which the gain has dropped by 1 dB from its small-signal value:

\[ A_{1\text{dB}} = \sqrt{0.145\,\left|\frac{\alpha_1}{\alpha_3}\right|} \]

The third-order intercept point (IIP3) is found from a two-tone test (\(x = A\cos\omega_1 t + A\cos\omega_2 t\)). The intermodulation product at \(2\omega_1 - \omega_2\) has amplitude \(\frac{3}{4}|\alpha_3|A^3\). Extrapolating the fundamental and IM3 slopes to their intersection:

\[ A_{IIP3} = \sqrt{\frac{4}{3}\left|\frac{\alpha_1}{\alpha_3}\right|} \]

The relationship between P1dB and IIP3 (in dBm) is approximately:

\[ \text{IIP3} \approx \text{P1dB} + 9.6\;\text{dB} \]

2.3.2 Cascaded IIP3

For a cascade of two stages:

\[ \frac{1}{A_{IIP3,\,sys}^2} = \frac{1}{A_{IIP3,1}^2} + \frac{G_1^2}{A_{IIP3,2}^2} + \cdots \]

Or equivalently in voltage terms, the dominant contribution shifts to later high-gain stages. In contrast to noise figure, IIP3 is dominated by the stages with the highest gain (typically the mixer and IF chain, not the LNA).

2.3.3 Spurious-Free Dynamic Range (SFDR)

SFDR is the input power range over which the receiver operates without the IM3 products exceeding the noise floor. Given noise power spectral density \(N_0 = kTF\):

\[ \text{SFDR}[\text{dB}] = \frac{2}{3}\bigl(\text{IIP3}[\text{dBm}] - P_{noise}[\text{dBm}]\bigr) \]

SFDR is a key figure of merit for receivers that must handle large interferers while maintaining sensitivity to weak desired signals.

2.3.4 Blocking and Desensitisation

A large out-of-band interferer (blocker) at power \(P_b\) causes gain compression of the desired signal path, effectively increasing the apparent noise figure. The desensitisation is:

\[ \Delta\text{NF} \approx 20\log_{10}\!\left(1 + \frac{P_b}{P_{1\text{dB}}}\right) \]

Standards specify blocker levels at specific frequency offsets; 3GPP LTE, for instance, specifies a −25 dBm blocker 10 MHz from the band edge while maintaining sensitivity.

2.4 Transmitter System Parameters

2.4.1 Linearity Metrics: IMD, ACPR, EVM

For transmitters, linearity determines spectral purity. Adjacent channel power ratio (ACPR) measures the power leaked into adjacent channels relative to in-channel power:

\[ \text{ACPR} = \frac{P_{adj}}{P_{in\text{-}ch}} \]

Error vector magnitude (EVM) measures the deviation of actual constellation points from ideal positions, expressed as a percentage of the ideal signal magnitude. 3GPP NR requires EVM < 8% for 64-QAM and < 3.5% for 256-QAM on the downlink.

2.4.2 Power Amplifier Efficiency

Drain (collector) efficiency is \(\eta_D = P_{out}/P_{DC}\). Power-added efficiency accounts for input drive power:

\[ \text{PAE} = \frac{P_{out} - P_{in}}{P_{DC}} = \eta_D\!\left(1 - \frac{1}{G}\right) \]

2.4.3 Noise/Linearity Budget

A link budget analysis distributes NF and IIP3 allowances across all receiver stages so that the system meets its sensitivity and dynamic range targets. Each component’s contribution is calculated using the Friis formula and cascaded IIP3 formula, iterated to find an optimum allocation that minimises cost while satisfying specifications.

---

Chapter 3: Key Building Blocks

3.1 Low Noise Amplifiers (LNAs)

3.1.1 Objectives and Specifications

The LNA is the first active element in a receiver chain. Its primary objectives are:

• Minimum noise figure to set the system sensitivity floor
• Adequate gain to suppress downstream noise contributions (Friis)
• Sufficient IIP3 to handle large blockers without compression
• Input impedance matching to 50 Ω (or antenna impedance) for maximum power transfer and defined system reference

These objectives are in tension: increasing bias current reduces noise but increases power consumption; increasing gain improves sensitivity but may worsen IIP3 of downstream stages.

3.1.2 Noise in MOSFETs

The dominant noise source in a MOSFET at RF frequencies is channel thermal noise, modelled as a drain current noise with power spectral density:

\[ \overline{i_{nd}^2} = 4kT\gamma g_{d0}\,\Delta f \]

where \(g_{d0}\) is the output conductance at zero drain-source voltage and \(\gamma\) is a technology-dependent factor (typically 2/3 in long-channel devices, larger in short-channel). Gate resistance and induced gate noise are secondary sources at moderate frequencies.

3.1.3 Inductive Source Degeneration for Noise Matching

A widely used LNA topology employs inductive source degeneration (\(L_s\)) with a series gate inductor (\(L_g\)) to simultaneously achieve input impedance matching and near-minimum noise figure without a lossy resistive match.

The input impedance of the inductively degenerated common-source amplifier is:

\[ Z_{in} = j\omega(L_g + L_s) + \frac{1}{j\omega C_{gs}} + \frac{g_m L_s}{C_{gs}} \]

The last term is a real resistive contribution — controlled by \(L_s\) — that can be set to 50 Ω without adding thermal noise. Resonating \(L_g + L_s\) with \(C_{gs}\) at the operating frequency gives a purely real input impedance.

The minimum achievable noise figure in this topology is:

\[ F_{min} \approx 1 + 2\omega_0\sqrt{\frac{\gamma}{5}}\,\frac{C_{gs}}{g_m}\,\omega_0 \]

which decreases with increasing \(f_T = g_m/(2\pi C_{gs})\) of the transistor, motivating the use of advanced process nodes for RF LNA design.

3.1.4 Cascode LNA

The cascode topology (common-source plus common-gate) improves reverse isolation and reduces Miller capacitance, making the input matching stable across frequency. The cascode device contributes negligible noise (its noise is referred to the input divided by the gain of the lower device), while substantially improving IIP3 through the increased output impedance.

3.2 Mixers and Modulators

3.2.1 Mixer Function and Key Parameters

A mixer multiplies (in the ideal case) two signals to produce sum and difference frequencies. Given RF input \(v_{RF} = A_{RF}\cos(2\pi f_{RF} t)\) and LO \(v_{LO} = A_{LO}\cos(2\pi f_{LO} t)\):

\[ v_{IF} = \frac{1}{2}A_{RF}A_{LO}\bigl[\cos(2\pi(f_{RF}-f_{LO})t) + \cos(2\pi(f_{RF}+f_{LO})t)\bigr] \]

The wanted output is the difference (IF) frequency \(f_{IF} = f_{RF} - f_{LO}\). Key mixer parameters are:

• Conversion gain (CG): ratio of IF output amplitude to RF input amplitude, in dB
• Noise figure: computed at the IF output referred to the RF input; mixers typically contribute 6–12 dB
• IIP3: third-order intercept referred to the RF input port
• LO-to-RF isolation: attenuation of LO leakage at the RF port (problematic in direct-conversion)
• Image rejection: attenuation of the image frequency \(f_{image} = f_{LO} + f_{IF}\) (superheterodyne)

3.2.2 Passive vs. Active Mixers

Passive mixers (e.g., diode rings, FET switches) have conversion loss (typically −6 dB for a doubly balanced ring) but excellent linearity and low flicker noise contribution. Active mixers (transconductance + commutating pair) provide conversion gain at the cost of added noise and reduced IIP3.

3.2.3 Gilbert Cell Architecture

The Gilbert cell is the canonical active doubly-balanced mixer in CMOS and BiCMOS. It consists of:

1. A transconductance stage converting the RF voltage to a differential current
2. A commutating quad (four transistors switched by the LO) that multiplies the current by the LO square wave
3. A load (resistive or inductive) converting the multiplied current back to voltage

Because the LO switches the current differentially, the even-order intermodulation products and LO feed-through cancel at the differential output. The conversion gain of a Gilbert cell is:

\[ CG = \frac{2}{\pi}g_m R_L \]

where \(g_m\) is the transconductance of the input pair and \(R_L\) is the load resistance. The factor \(2/\pi\) arises from the Fourier coefficient of the square-wave switching function.

Noise in the Gilbert cell comes predominantly from the commutating transistors’ flicker noise during their partially-on state and from the tail current source. Techniques to reduce mixer noise include current bleeding (injecting extra DC current through cascode devices to reduce the fraction flowing through the commutating pair) and dynamic biasing.

3.2.4 IQ Modulator/Demodulator

The IQ (in-phase/quadrature) modulator generates a DSB-SC signal:

\[ s(t) = I(t)\cos(2\pi f_c t) - Q(t)\sin(2\pi f_c t) \]

Two mixers share the same LO, with one driven in quadrature (90° phase shift). IQ imbalances — amplitude mismatch \(\Delta A\) and phase mismatch \(\Delta\phi\) — degrade EVM and image rejection. The image rejection ratio (IRR) due to IQ imbalance is:

\[ \text{IRR} = \frac{1 + 2\epsilon\cos\phi + \epsilon^2}{1 - 2\epsilon\cos\phi + \epsilon^2} \]

where \(\epsilon = 1 + \Delta A/A\) and \(\phi = \Delta\phi\). For 40 dB IRR, the phase error must be below approximately 0.6° and the amplitude mismatch below 0.01 dB — motivating careful layout and digital calibration.

3.3 Oscillators

3.3.1 Barkhausen Criteria

An oscillator sustains steady-state oscillation when the loop gain satisfies the Barkhausen criteria: at the oscillation frequency \(\omega_0\),

\[ \lvert H(\omega_0)\rvert = 1 \quad\text{and}\quad \angle H(\omega_0) = 0° \text{ (or } 360°\text{)} \]

In practice the loop gain is designed slightly greater than unity at startup, with nonlinearity limiting the amplitude.

3.3.2 LC Oscillator Topologies

LC oscillators use a resonant tank (inductor \(L\), capacitor \(C\)) to define the oscillation frequency \(\omega_0 = 1/\sqrt{LC}\). Common CMOS topologies include the complementary cross-coupled pair (NMOS + PMOS), which presents a negative resistance of \(-2/g_m\) to compensate tank losses (characterised by parallel resistance \(R_p = Q_L \omega_0 L\)).

The tank quality factor \(Q = R_p / (\omega_0 L)\) largely determines phase noise performance. On-chip spiral inductors typically achieve \(Q = 5\)–20 at microwave frequencies; off-chip inductors or cavity resonators offer higher Q.

3.3.3 Phase Noise and Leeson’s Model

Phase noise is the spectral density of phase fluctuations of an oscillator output, expressed in dBc/Hz at offset \(\Delta\omega\) from the carrier. Leeson’s empirical model gives:

\[ \mathcal{L}(\Delta\omega) = 10\log\!\left[\frac{2FkT}{P_s}\left(1 + \left(\frac{\omega_0}{2Q\,\Delta\omega}\right)^2\right)\left(1 + \frac{\Delta\omega_{1/f^3}}{\lvert\Delta\omega\rvert}\right)\right] \]

where \(F\) is an excess noise factor, \(P_s\) is the carrier power at the oscillator output, \(Q\) is the loaded tank Q, and \(\Delta\omega_{1/f^3}\) is the corner frequency of the \(1/f^3\) phase noise region (related to transistor flicker noise).

Key design insights from Leeson’s model:

• Phase noise improves as \(Q^2\) — maximise tank Q
• Increasing oscillator power \(P_s\) improves phase noise (at the cost of power consumption)
• Flicker noise up-conversion (the \(1/f^3\) region) is reduced by symmetric waveforms and tail filtering

3.3.4 Voltage-Controlled Oscillators (VCOs)

VCOs are LC oscillators in which the resonant frequency is controlled by a tuning voltage applied to a variable capacitor (varactor). In CMOS, accumulation-mode MOS varactors provide a monotonic C-V characteristic. The VCO gain is \(K_{VCO} = d\omega_0/dV_{tune}\) (rad/s/V). The VCO is the core element of a phase-locked loop (PLL) used for frequency synthesis in transceivers.

3.3.5 Crystal Oscillators

Crystal oscillators use the piezoelectric resonance of quartz to achieve extremely high Q factors (\(10^4\)–\(10^6\)), resulting in very low phase noise and excellent frequency stability. They are used as reference oscillators, with PLLs multiplying the reference frequency up to the RF carrier.

3.4 Power Amplifiers

3.4.1 Classification by Conduction Angle

PA classes are defined by the transistor conduction angle \(\theta\) (fraction of each RF cycle during which the transistor conducts):

Class	Conduction Angle	Max Drain Efficiency	Linearity
A	\(2\pi\) (360°)	50%	High
B	\(\pi\) (180°)	78.5%	Moderate
AB	\(\pi\) to \(2\pi\)	50%–78.5%	Moderate–High
C	\(< \pi\)	> 78.5% (→100%)	Low
D	Switching	~100% (ideal)	Low
E	Switching	~100% (ideal)	Low
F	Harmonic tuned	~100% (ideal)	Low

3.4.2 Switching-Mode PAs (Classes D, E, F)

Switching-mode PAs treat the transistor as a lossless switch. In the ideal switch, the device is either fully on (zero voltage drop) or fully off (zero current), so no power is dissipated in the device — theoretically 100% drain efficiency.

Class E achieves zero-voltage switching (ZVS) and zero-derivative switching (ZdVS) by shaping the drain waveform with a series resonant network, eliminating capacitive discharge losses. Practical Class E PAs achieve 80–95% PAE at GHz frequencies.

Class F uses harmonic traps (short circuits at even harmonics, open circuits at odd harmonics) to shape the drain voltage toward a square wave and the current toward a half-sinusoid, maximising output power per unit of DC supply.

3.4.3 Efficiency Enhancement Techniques

Because most digital modulation standards have high peak-to-average power ratio (PAPR) — 8–12 dB for OFDM — PAs spend most of their time operating at backed-off power, severely reducing average efficiency.

• Doherty PA: combines a carrier (Class B) and peaking (Class C) amplifier through a quarter-wave transformer. The peaking amplifier turns on only near peak power, maintaining high efficiency over a 6–10 dB back-off range. Widely used in base stations.
• Envelope tracking (ET): dynamically adjusts the PA supply voltage to track the signal envelope, keeping the PA near saturation across the dynamic range.
• Digital pre-distortion (DPD): applies an inverse nonlinearity to the baseband signal to compensate PA nonlinearity, allowing the PA to operate near saturation while meeting spectral mask and EVM requirements.

---

Chapter 4: RF Radio Transceiver Architectures

4.1 Receiver Architectures

4.1.1 Superheterodyne Receiver

The superheterodyne (superhet) receiver, invented by Edwin Armstrong in 1918, remains the dominant architecture for demanding applications. It translates the RF signal to a fixed intermediate frequency (IF) using a first mixer+LO stage, then applies channel filtering at IF (where high-Q filters are feasible), before a second downconversion to baseband.

Signal path: Antenna → BPF (preselect) → LNA → BPF (image reject) → Mixer₁ → IF BPF (channel filter) → IF Amp → Mixer₂ → Baseband → ADC

Image frequency: The first mixer responds to two RF frequencies that both produce the same IF. If \(f_{IF} = f_{RF} - f_{LO}\), then the image at \(f_{im} = f_{LO} - f_{IF} = f_{RF} - 2f_{IF}\) also produces a signal at \(f_{IF}\). The image-reject filter (placed between the LNA and mixer) must attenuate the image to acceptable levels. A higher IF places the image further away, making filtering easier but potentially increasing VCO phase noise requirements.

Advantages of superheterodyne:

• Excellent selectivity via high-Q IF SAW/BAW filters
• High image rejection (image separated by \(2f_{IF}\))
• Stable, narrowband IF amplification

Disadvantages:

• Multiple off-chip components (image-reject and IF filters) increase bill of materials and board area
• Difficult to integrate fully in CMOS SoC
• Dual-conversion adds complexity

4.1.2 Direct-Conversion (Zero-IF) Receiver

The direct-conversion receiver (DCR) sets \(f_{LO} = f_{RF}\), translating the desired channel directly to baseband:

\[ f_{IF} = f_{RF} - f_{LO} = 0 \]

The image is the mirror image of the desired signal itself; IQ downconversion (separate I and Q mixers with 90° LO phase difference) allows full signal reconstruction from both sidebands.

Signal path: Antenna → BPF → LNA → IQ Mixer → LPF (channel select) → ADC → DSP

Advantages:

• No image frequency problem (or rather, image = desired signal’s other sideband)
• Channel filtering at baseband (LPF, implementable on-chip)
• Amenable to full CMOS integration
• Widely used in modern wireless SoCs (smartphones, IoT)

Disadvantages and challenges:

1. DC offset: LO leakage from the mixer self-mixes to produce a large DC component at the output, saturating subsequent amplifiers. Mitigated by AC coupling (sacrifices low-frequency data) or digital DC cancellation.
2. IQ mismatch: Amplitude and phase errors between I and Q paths cause image leakage and EVM degradation (see Section 3.2.4).
3. Flicker (1/f) noise: At zero-IF, the signal occupies the frequency range from DC to \(B/2\). MOSFET flicker noise, which is largest near DC, degrades the SNR of low-frequency signal components.
4. Even-order distortion: Unlike superheterodyne where IM2 products fall at low frequencies and are filtered at RF, in DCR the IM2 of large blockers falls in-band. Requires high IIP2 (>+40 dBm) through careful differential design and calibration.

4.1.3 Low-IF Receiver

The low-IF architecture is a compromise: the LO is offset from the RF by 1–2 channel bandwidths, placing the IF at a low but non-zero frequency. The image is an adjacent channel and can be rejected through complex (IQ) signal processing in the digital domain (Hartley or Weaver image reject architecture). Avoids DC offset and 1/f noise while remaining nearly fully integrable. Used in Bluetooth and GSM/EDGE receivers.

4.1.4 Wideband Receiver Architectures

Modern software-defined radios (SDRs) and 5G receivers use wideband sampling — digitising a wide frequency range and performing channel selection entirely in the digital domain. This requires high-speed, high-dynamic-range ADCs (16-bit at several GS/s) and powerful digital signal processors. The RF front-end provides only broad bandpass filtering and LNA gain.

4.2 Transmitter Architectures

4.2.1 Direct-Launch (Direct-Conversion) Transmitter

In the direct-launch (direct up-conversion) transmitter, the baseband I and Q signals are directly upconverted to the RF carrier using an IQ modulator:

\[ s_{RF}(t) = I(t)\cos(2\pi f_c t) - Q(t)\sin(2\pi f_c t) \]

Signal path: DSP → DAC → LPF → IQ Modulator → VGA → BPF → PA → Antenna

Advantages:

• Fewest components, highest integration potential
• No IF hardware

Challenges:

• LO pulling: The PA output at \(f_c\) can couple back into the VCO operating at the same frequency, pulling (perturbing) the VCO frequency — a major problem in CMOS SoCs. Mitigated by using an offset LO or injection locking techniques.
• IQ mismatch degrades modulation accuracy (EVM)
• Wide instantaneous bandwidth required of the DAC

4.2.2 Two-Step (Heterodyne) Transmitter

The two-step transmitter first upconverts the baseband IQ signal to an intermediate frequency \(f_{IF}\), applies IF channel filtering to remove DAC images and spurious products, then upconverts to the final RF carrier:

Signal path: DSP → DAC → LPF → IQ Mod (IF₁) → IF BPF → Mixer (LO₂) → BPF → PA → Antenna

Advantages:

• IF filtering removes spurious emissions before final upconversion
• LO at RF is offset from PA output (at RF + IF or RF − IF), reducing pulling
• Better spectral purity

Disadvantages:

• More components: two mixers, two LOs (or a PLL+divider chain), IF filter
• Increased power consumption and board area

4.2.3 Polar Transmitter

For constant-envelope or near-constant-envelope modulation, a polar transmitter separates the signal into amplitude and phase components:

\[ s(t) = A(t)\cos\!\bigl[2\pi f_c t + \phi(t)\bigr] \]

The phase-modulated carrier \(\cos[2\pi f_c t + \phi(t)]\) is generated by a phase-locked loop; the amplitude modulation \(A(t)\) is applied via the PA supply voltage (envelope tracking) or through the PA drive. Polar architectures enable highly efficient PAs (operating near saturation) while supporting amplitude-modulated signals, but require wide-bandwidth envelope paths and careful alignment of amplitude and phase.

---

Chapter 5: Antenna Fundamentals

5.1 Basic Antenna Operation

An antenna is a transducer that converts guided electromagnetic wave energy (in a transmission line or waveguide) into free-space radiation, and vice versa. The fundamental mechanism is the acceleration of charge — a time-varying current distribution on the antenna structure radiates electromagnetic fields.

The Friis transmission equation relates the received power \(P_r\) to the transmitted power \(P_t\) in free space:

\[ P_r = P_t\, G_t\, G_r \left(\frac{\lambda}{4\pi d}\right)^2 \]

where \(G_t\) and \(G_r\) are the gains of the transmit and receive antennas, \(\lambda = c/f\) is the wavelength, and \(d\) is the link distance. The term \((\lambda/4\pi d)^2\) is the free-space path loss factor (also written as \(1/L_{FS}\)). In dB:

\[ P_r[\text{dBm}] = P_t[\text{dBm}] + G_t[\text{dBi}] + G_r[\text{dBi}] - 20\log_{10}(d) - 20\log_{10}(f) + 147.6 \]

5.2 Antenna Parameters

5.2.1 Radiation Pattern

The radiation pattern is the spatial distribution of radiated power density (or field strength) as a function of angle \((\theta, \phi)\) in the far field. It is typically plotted in polar or rectangular coordinates, normalised to its maximum value.

Key features of a radiation pattern:

• Main lobe: the region of maximum radiation
• Side lobes: subsidiary maxima outside the main beam
• Back lobe: radiation directed toward the rear of the antenna
• Nulls: directions of zero radiation

5.2.2 Directivity and Gain

Directivity \(D\) is the ratio of the radiation intensity in a given direction to the average radiation intensity (i.e., compared to an isotropic radiator):

\[ D(\theta,\phi) = \frac{U(\theta,\phi)}{U_0} = \frac{4\pi U(\theta,\phi)}{P_{rad}} \]

Antenna gain \(G\) accounts for ohmic losses:

\[ G = \eta_{rad}\cdot D \]

where \(\eta_{rad} = P_{rad}/P_{in}\) is the radiation efficiency. Gain is expressed in dBi (decibels relative to isotropic) or dBd (decibels relative to a half-wave dipole; 1 dBd = 2.15 dBi).

5.2.3 Half-Power Beamwidth (HPBW)

HPBW is the angular width of the main lobe between the two directions at which the radiated power falls to half (−3 dB) of its maximum. For a uniformly illuminated aperture of width \(D_a\):

\[ \text{HPBW} \approx \frac{0.886\,\lambda}{D_a} \quad\text{(radians)} \]

Narrower HPBW (higher directivity) requires a larger aperture or a more complex array.

5.2.4 Effective Isotropic Radiated Power (EIRP)

EIRP is the product of the transmit power and the antenna gain, representing the power that an isotropic radiator would need to produce the same power density in the direction of maximum radiation:

\[ \text{EIRP} = P_t \cdot G_t \]

In dBm: \(\text{EIRP}[\text{dBm}] = P_t[\text{dBm}] + G_t[\text{dBi}]\). EIRP is regulated by spectrum authorities to limit interference.

5.2.5 Input Impedance and Matching

The input impedance of an antenna \(Z_A = R_A + jX_A\) includes radiation resistance \(R_{rad}\) (representing power radiated) and loss resistance \(R_{loss}\). For a half-wave dipole in free space, \(Z_A \approx 73 + j42.5\;\Omega\) at resonance (with slight shortening to cancel the reactive part). A matching network transforms this to 50 Ω for connection to the transmission line.

5.2.6 Polarisation and Bandwidth

Antenna polarisation describes the orientation of the electric field vector of the radiated wave. Common polarisations are linear (horizontal or vertical), circular (RHCP or LHCP), and elliptical. Polarisation mismatch between transmit and receive antennas causes a polarisation loss factor \(\text{PLF} = \cos^2\psi\) where \(\psi\) is the angular difference.

Antenna bandwidth is the frequency range over which the antenna maintains acceptable performance (e.g., VSWR < 2:1, or gain within 3 dB of maximum). Wideband antennas (log-periodic, Vivaldi) are used in multi-band systems.

5.3 Common Antenna Types

Half-wave dipole: The fundamental radiating element, resonant at \(\ell = \lambda/2\), with omnidirectional pattern in the plane perpendicular to its axis and gain of 2.15 dBi. The building block for many antenna arrays and practical designs.

Monopole over ground plane: Equivalent to half a dipole with an image in the ground plane; resonant at \(\ell = \lambda/4\), gain of 5.15 dBi (3 dB over a dipole) for an infinite perfect ground.

Patch (microstrip) antenna: A printed rectangular conductor over a ground plane, resonant when its length is approximately \(\lambda/2\) in the substrate. Low profile, low cost, conformable. Gain 5–9 dBi. Used in GPS, 5G handsets, and vehicle antennas. Narrow bandwidth (~2–5%) is a limitation.

Horn antenna: A waveguide flared into a horn shape. High gain (10–25 dBi), low side lobes, known aperture efficiency. Used as a feed for reflector antennas and as a gain standard for calibration.

Parabolic reflector: A dish antenna with gain \(G = \eta_A ({\pi D_a}/{\lambda})^2\) where \(\eta_A\) is the aperture efficiency (typically 0.55–0.75) and \(D_a\) is the dish diameter. Used for satellite communication and radar, achieving gains of 40–60 dBi.

---

Chapter 6: Antenna Arrays, Phased Arrays, and Beamforming

6.1 Array Factor and Uniform Linear Arrays

An antenna array consists of multiple identical antenna elements whose individual radiation patterns combine to produce a composite pattern with higher gain and directional control. The total field is the product of the element pattern and the array factor (AF).

For a Uniform Linear Array (ULA) of \(N\) elements spaced \(d\) apart along the z-axis, each element excited with amplitude \(a_n = 1\) and progressive phase shift \(\delta\), the array factor is:

\[ AF(\theta) = \sum_{n=0}^{N-1} e^{jn(kd\cos\theta + \delta)} \]

where \(k = 2\pi/\lambda\) is the free-space wavenumber and \(\theta\) is measured from the array axis. Defining \(\psi = kd\cos\theta + \delta\):

\[ AF(\psi) = \frac{\sin(N\psi/2)}{\sin(\psi/2)} \]

The maximum occurs at \(\psi = 0\), i.e., at angle \(\theta_0 = \cos^{-1}(-\delta/kd)\). By varying the progressive phase shift \(\delta\) electronically, the main beam can be steered without mechanical movement — the principle of electronic beamsteering.

The maximum gain of an \(N\)-element array over a single element is \(N\) (in power), or \(10\log_{10}N\) in dBi relative to one element.

6.2 Beam Steering in Phased Arrays

A phased array electronically steers the beam by applying a programmable phase shift \(\delta_n\) to each element’s excitation. To steer the beam to angle \(\theta_0\) from broadside:

\[ \delta_n = -nkd\cos\theta_0 \]

Phase shifters can be implemented using:

• Analog phase shifters: switched LC networks, loaded-line, reflection-type (RTPS) — fast, low power, but limited resolution
• Digital phase shifters: switched delay lines quantised to \(b\) bits, providing \(2^b\) phase states
• Time-delay units (TDUs): provide true time delay (TTD) rather than phase shift, enabling squint-free wideband beamsteering (important for 5G mmWave with wide instantaneous bandwidths)

6.3 Beamforming Architectures

6.3.1 Analog Beamforming

A single RF chain drives all antenna elements through a corporate feed network with phase shifters. Only a single beam is formed, but the hardware is minimal. Used in point-to-point mmWave links (60 GHz WiGig, E-band backhaul).

6.3.2 Digital Beamforming

Each antenna element has its own dedicated RF chain (LNA, mixer, ADC/DAC). Beamforming is performed in the digital baseband, enabling multiple simultaneous independent beams (multi-user MIMO) and full spatial processing flexibility. Requires \(N\) RF chains — expensive in power and cost, but the preferred architecture for sub-6 GHz massive MIMO.

6.3.3 Hybrid Beamforming

A compromise architecture for mmWave massive MIMO (e.g., 5G NR FR2): a small number of RF chains (\(N_{RF} \ll N\)) each drive a sub-array of elements through analog phase shifters. The digital precoder operates across \(N_{RF}\) streams while the analog network maps them to the \(N\) antenna ports. Achieves most of the gain of full digital beamforming at a fraction of the hardware cost.

6.4 MIMO Systems

Multiple-input multiple-output (MIMO) systems use multiple antennas at both transmitter and receiver. With \(N_T\) transmit and \(N_R\) receive antennas, the channel matrix is \(\mathbf{H} \in \mathbb{C}^{N_R \times N_T}\). The ergodic capacity is:

\[ C = \mathbb{E}\!\left[\log_2\det\!\left(\mathbf{I} + \frac{\rho}{N_T}\mathbf{H}\mathbf{H}^H\right)\right] \quad \text{[bits/s/Hz]} \]

In a rich scattering environment, the capacity scales approximately linearly with \(\min(N_T, N_R)\) — the spatial multiplexing gain. Spatial multiplexing (SM) transmits independent data streams on orthogonal spatial channels (singular value decomposition of \(\mathbf{H}\)), while transmit diversity (Alamouti scheme) uses redundancy across antennas to improve reliability without requiring CSI at the transmitter.

---

Chapter 7: Special Topics in Radio and Wireless Systems

7.1 Link Budget Analysis

A link budget is an accounting of all gains and losses in a radio link from transmitter to receiver, used to verify that the received signal power exceeds the receiver sensitivity by a sufficient margin (link margin).

A typical downlink budget:

1. Transmitter output power \(P_t\) (dBm)
2. \(+\) Transmit antenna gain \(G_t\) (dBi)
3. \(=\) EIRP
4. \(-\) Free-space path loss \(L_{FS} = 20\log_{10}(4\pi d/\lambda)\)
5. \(-\) Additional losses (body loss, cable/connector loss, rain fade margin)
6. \(+\) Receive antenna gain \(G_r\) (dBi)
7. \(=\) Received power \(P_r\) (dBm)
8. \(-\) Receiver sensitivity \(P_{min}\) (dBm)
9. \(=\) Link margin (dB) — must be positive, typically 10–20 dB to account for fading

7.2 Propagation Effects

7.2.1 Path Loss Models

Beyond free space, practical RF propagation is affected by terrain, buildings, and atmospheric phenomena. The two-ray model (direct path + ground reflection) gives path loss that scales as \(d^4\) at large distances (rather than \(d^2\)):

\[ P_r \approx P_t G_t G_r \frac{h_t^2 h_r^2}{d^4} \]

where \(h_t\) and \(h_r\) are the transmit and receive antenna heights. Log-distance path loss models for cellular environments are:

\[ PL(d)[\text{dB}] = PL(d_0) + 10n\log_{10}(d/d_0) + X_\sigma \]

where \(n\) is the path loss exponent (2 in free space, 3–4 in urban areas, 2–8 in indoor environments), \(d_0\) is a reference distance, and \(X_\sigma\) is a zero-mean Gaussian random variable (in dB) representing log-normal shadowing with standard deviation \(\sigma\) (4–12 dB typically).

7.2.2 Multipath Fading

When multiple reflected and scattered copies of the signal arrive at the receiver with different delays and phases, they interfere constructively or destructively, causing rapid amplitude variations known as multipath (Rayleigh) fading. The coherence bandwidth \(B_c \approx 1/(5\tau_{rms})\) (where \(\tau_{rms}\) is the rms delay spread) determines the frequency correlation of the channel. The Doppler spread \(f_d = v/\lambda\) (for mobile velocity \(v\)) determines the coherence time \(T_c \approx 1/(4 f_d)\).

OFDM addresses multipath by dividing the bandwidth into subcarriers narrower than \(B_c\), ensuring flat fading per subcarrier. The cyclic prefix ensures inter-symbol orthogonality as long as \(T_{CP} \ge \tau_{max}\).

7.3 Frequency Planning and Spectrum Regulations

7.3.1 ITU Spectrum Allocation

The International Telecommunication Union (ITU) divides the radio spectrum into allocations for different services (mobile, fixed, broadcasting, satellite, radar). Key bands include:

• HF (3–30 MHz): shortwave broadcasting, maritime
• VHF (30–300 MHz): FM radio (88–108 MHz), terrestrial TV, aviation
• UHF (300 MHz–3 GHz): cellular (700 MHz, 850 MHz, 1.9 GHz, 2.1 GHz, 2.6 GHz), Wi-Fi (2.4 GHz, 5 GHz), GPS (1.575 GHz)
• SHF (3–30 GHz): 5G NR FR2 (24–52 GHz), satellite, radar
• EHF (30–300 GHz): 60 GHz WiGig (802.11ad/ay), automotive radar (77 GHz)

7.3.2 Receiver Spurious Products and Image Management

In superheterodyne receivers, multiple mixing products can fall in-band:

• Image frequency: \(f_{im} = 2f_{LO} - f_{RF}\) (for low-side LO injection)
• Half-IF spurs: signals at \(f_{RF} \pm f_{IF}/2\) that pass imperfectly through the image-reject filter and produce IF products through second-order distortion
• LO harmonics: the \(n\)-th LO harmonic mixed with RF products at \(nf_{LO} \pm f_{IF}\)

These are mitigated through careful IF selection, band-select pre-filtering, and high-IP2 mixers. Frequency planning (choice of IF) seeks to place these spurious responses outside the receive band.

7.4 Noise in Digital Radio Systems

7.4.1 Phase Noise Effects

VCO phase noise causes reciprocal mixing: a large interferer located at offset \(\Delta f\) from the desired signal mixes with the phase noise skirt of the LO, transferring noise energy into the IF passband. The resulting degradation in SNR is:

\[ \text{SNR}_{degraded} = \text{SNR}_{0} - 10\log_{10}\!\bigl(P_{int}\cdot\mathcal{L}(\Delta\omega)\cdot B_{IF}\bigr) \]

where \(P_{int}\) is the interferer power, \(\mathcal{L}(\Delta\omega)\) is the phase noise at offset \(\Delta\omega\), and \(B_{IF}\) is the IF bandwidth.

Phase noise also causes in-band EVM degradation. For OFDM, the phase noise contribution to EVM is approximately:

\[ \text{EVM}^2 \approx \int_{-B/2}^{B/2} \mathcal{L}(f)\,df \]

(the integrated phase noise power over the signal bandwidth), motivating tight specifications on VCO phase noise in high-order modulation systems.

7.4.2 Quantisation Noise and ADC Requirements

The ADC in the receiver must have sufficient dynamic range to digitise both the desired signal (at the sensitivity level) and large blockers simultaneously. The spurious-free dynamic range of an ADC with \(b\) bits is approximately:

\[ \text{SFDR}_{ADC} \approx 6.02b + 1.76 \;\text{dB} \]

For a receiver with a 70 dB dynamic range (sensitivity to maximum input), a minimum of 12-bit ADC is required. 5G NR base station receivers typically use 16-bit ADCs.

7.5 5G NR Radio System Overview

5G New Radio (NR) operates in two frequency ranges: FR1 (sub-6 GHz, 410 MHz–7.125 GHz) and FR2 (mmWave, 24.25–52.6 GHz). Key technical features:

• Flexible numerology: subcarrier spacing \(\Delta f = 2^\mu \times 15\) kHz, with \(\mu = 0,1,2,3,4\) (15 kHz to 240 kHz), allowing adaptation from eMBB (enhanced mobile broadband) to URLLC (ultra-reliable low latency)
• Massive MIMO: up to 32 or 64 transmit/receive antenna ports in gNB, enabling multi-user MIMO and 3D beamforming
• Millimetre wave: 400 MHz instantaneous bandwidth, phased array with 64–256 elements, hybrid beamforming — path loss is severe (\(PL \propto f^2\)), requiring high-gain beams
• Non-orthogonal waveform support: Filtered-OFDM (f-OFDM) and window-OFDM reduce out-of-band emissions in flexible spectrum operation

7.6 System-Level Design Example: 4G LTE Receiver Noise Budget

Consider a 4G LTE Category 4 UE (user equipment) receiver operating in Band 4 (2110–2155 MHz downlink). The target is \(-97\) dBm sensitivity for 10 MHz bandwidth (QPSK, \(\text{SNR}_{min} = 1\) dB with FEC).

Budget allocation:

Stage	Gain (dB)	NF (dB)	IIP3 (dBm)
Antenna switch	−0.5	0.5	+60
BPF	−2.5	2.5	+50
LNA	+16	1.2	−5
Image filter	−1.0	1.0	+45
Mixer	+0	8.0	+10
IF filter	−3.0	3.0	+35
VGA	+20	5.0	−10

Cascaded NF (Friis):

\[ F_{sys} \approx F_{ant\text{-}sw} + \frac{F_{BPF}-1}{G_{ant\text{-}sw}} + \frac{F_{LNA}-1}{G_{ant\text{-}sw}G_{BPF}} + \cdots \]

Working through the cascade (all in linear ratios):

• After switch+BPF (combined loss 3 dB, \(G = 0.5\), \(F = 2.0\)): \(F = 2.0\)
• Adding LNA (\(G_1 = 40\), \(F_{LNA} = 1.32\)): \(F_{sys} = 2.0 + (1.32-1)/0.5 = 2.64\)
• Adding image filter (\(G = 0.79\), \(F = 1.26\)): \(F = 2.64 + (1.26-1)/(0.5\times 40) = 2.65\)
• Adding mixer (\(F = 6.3\), \(G = 1\)): \(F = 2.65 + (6.3-1)/(0.5\times 40\times 0.79) = 2.98\)

Total NF \(\approx 4.7\) dB. Then

\[ P_{min} = -174 + 4.7 + 70 + 1 = -98.3\;\text{dBm} \]

which meets the \(-97\) dBm target with 1.3 dB of margin.

---

Chapter 8: Advanced System Topics

8.1 Linearisation Techniques

8.1.1 Feedforward Linearisation

Feedforward linearisation measures the distortion at the PA output, amplifies and inverts it in a correction loop, then subtracts it from the output using a second (error) amplifier. It is wideband, fast, and can achieve 30–40 dB improvement in ACPR but requires a perfectly matched error path and the correction amplifier itself must be linear — increasing power consumption.

8.1.2 Feedback Linearisation

Cartesian (IQ) feedback linearises the transmitter by comparing the downconverted output with the input baseband signal and feeding the error back to correct the I and Q inputs. Stability requires the loop bandwidth to be well below the oscillation frequency of the closed loop. Suitable for narrowband signals.

8.1.3 Digital Pre-distortion (DPD)

DPD is the dominant linearisation technique in modern base stations. A memory polynomial or Volterra series model characterises the PA inverse nonlinearity:

\[ y_{DPD}(t) = \sum_{k=0}^{K}\sum_{m=0}^{M} a_{km}\,x(t-m)\,\lvert x(t-m)\rvert^k \]

The coefficients \(a_{km}\) are identified by least-squares regression on observed input-output data (using an observation receiver feeding the DSP). DPD compensates both memoryless nonlinearity and memory effects (frequency-dependent gain/phase variation). Commercial DPD systems achieve ACPR improvement of 15–25 dB, enabling PAs to operate 3–6 dB closer to saturation.

8.2 Frequency Synthesisers and PLLs

A phase-locked loop (PLL) is the standard method for frequency synthesis in RF transceivers. The PLL tracks the phase of a reference oscillator (crystal) and locks a VCO to \(N\) times the reference frequency:

\[ f_{out} = N \cdot f_{ref} \]

A charge-pump PLL consists of: phase-frequency detector (PFD), charge pump (CP), loop filter (LF), VCO, and frequency divider (\(\div N\)). The closed-loop bandwidth \(f_{BW}\) is set by the loop filter: inside \(f_{BW}\), the PLL suppresses VCO phase noise (the crystal reference dominates); outside \(f_{BW}\), VCO free-running phase noise passes through.

Fractional-N PLLs use a delta-sigma modulator to dither the divider ratio between integer values, achieving sub-Hz frequency resolution while maintaining loop stability. Third-order sigma-delta modulators with multi-bit output are standard in modern synthesisers for cellular and Wi-Fi.

8.3 Diversity Techniques

Diversity exploits multiple independent copies of the transmitted signal (from different paths, antennas, frequencies, or times) to combat fading.

• Space diversity: multiple antennas separated by \(\ge \lambda/2\) at the receiver (selection diversity, maximal ratio combining MRC, equal gain combining EGC)
• Frequency diversity: transmission over multiple frequency-separated bands (explicit) or OFDM exploiting coding across subcarriers (implicit)
• Time diversity: interleaving coded bits across multiple coherence times — effective but introduces latency
• Polarisation diversity: two antennas with orthogonal polarisation, typically yielding ~10 dB correlation reduction in rich scatter environments

MRC achieves the maximum output SNR: \(\text{SNR}_{MRC} = \sum_{i=1}^{N} \text{SNR}_i\), providing \(N\)-fold SNR improvement on average in Rayleigh fading. The outage probability with MRC over \(N\)-branch Rayleigh fading is:

\[ P_{out}(\gamma_0) = 1 - e^{-\gamma_0/\bar{\gamma}}\sum_{k=0}^{N-1}\frac{(\gamma_0/\bar{\gamma})^k}{k!} \]

which falls as \(\gamma_0^N\) for small \(\gamma_0\) — a diversity order of \(N\).

8.4 Transceiver Integration and RF SoC Design

Modern wireless SoCs integrate the entire RF transceiver — LNA, mixer, PLL, ADC/DAC, power management — on a single CMOS die. Achieving this requires careful attention to:

• Substrate coupling: digital switching noise couples into sensitive RF nodes through the common substrate. Techniques include deep N-well isolation, differential circuit topologies (which cancel common-mode substrate noise), and guard rings.
• Supply rejection: RF circuits must operate correctly despite supply voltage ripple from switching regulators. Regulated cascodes and on-chip decoupling capacitors help.
• Antenna tuner integration: impedance matching networks with tunable components (BST capacitors, MEMS switches) compensate for hand effects and antenna detuning.
• Process scaling: moving to 7/5/3 nm FinFET nodes improves \(f_T\) and reduces digital power but degrades passive quality factors (thinner metals, higher resistivity at mmWave frequencies), requiring architectural innovations.

---

Summary of Key Formulas

Formula	Description
\(F_{sys} = F_1 + \frac{F_2-1}{G_1} + \frac{F_3-1}{G_1 G_2} + \cdots\)	Friis noise figure for cascaded stages
\(P_{min} = kT_0 B \cdot F \cdot \text{SNR}_{min}\)	Receiver sensitivity
\(A_{IIP3} = \sqrt{\frac{4}{3}\lvert\frac{\alpha_1}{\alpha_3}\rvert}\)	Input-referred third-order intercept
\(\text{IIP3} \approx \text{P1dB} + 9.6\;\text{dB}\)	IIP3 to 1-dB compression relationship
\(\text{SFDR} = \frac{2}{3}(\text{IIP3} - P_n)\)	Spurious-free dynamic range
\(P_r = P_t G_t G_r (\lambda/4\pi d)^2\)	Friis transmission equation
\(\text{EIRP} = P_t G_t\)	Effective isotropic radiated power
\(AF(\theta) = \frac{\sin(N\psi/2)}{\sin(\psi/2)}\)	ULA array factor (\(\psi = kd\cos\theta + \delta\))
\(\mathcal{L}(\Delta\omega) \propto \frac{FkT}{P_s Q^2 \Delta\omega^2}\)	Leeson phase noise (simplified)
\(\text{PAE} = (P_{out}-P_{in})/P_{DC}\)	Power-added efficiency
\(C = \log_2\det(\mathbf{I} + \frac{\rho}{N_T}\mathbf{H}\mathbf{H}^H)\)	MIMO ergodic capacity