Chapter 1 Introduction
1.2 T HESIS O RGANIZATION
The chapter 2 of this thesis investigates the basic principle and the consideration of two types of RF PAs. One of them is the narrowband class-AB PA and the other is the ultra-wideband class-AB distributed PA. A thorough discussion on the design equations and performance matrices are presented.
Chapter 3 investigates several ESD protection strategies. Two protection schemes for RF application are proposed. One of them is to use an inductive device as the ESD clamp, and the other is to use SCR devices as the ESD clamp. The co-design methodology and flow is proposed and investigated. Circuit topologies and
narrowband PA and the ultra-wideband distributed PA, along with the corresponding ESD protection schemes, both fabricated in a standard 0.13-um CMOS process. The measurement results are compared and analyzed. The result reveals that both of the proposed ESD protection schemes can indeed provide excellent ESD robustness.
Chapter 5 is the conclusions of this thesis and the future works on this topic.
Chapter 2 RF PA Basics
2.1NARROWBAND CLASS-ABPA
2.1.1 Conventional Architecture
Fig. 2.1 is the conventional class-AB PA architecture. A class-AB PA is basically a transconductor (Gm-cell), providing a sufficient amount of the output current corresponding to the input driving power. Fig. 2.2 is a typical circuit implementation of this architecture. The cascode topology provides good voltage gain and good isolation; it also prevents drain overstress since the voltage swing may approach 2 times the VDD. The input and output matching networks provide suitable impedance matching for RF signal integrity, power transfer efficiency, and the narrowband selectivity [13].
2.1.2 Load-line Design
The loading impedance of the class-AB PA is designed for maximum output power capability, and therefore the load-line design aims to exploit the maximum usefulness of the circuit limit. To accomplish this goal, the load-line of the class-AB PA is designed as shown in Fig. 2.3. Fig. 2.3 is a typical I-V relationship of the conventional class-AB PA core output, along with the maximum output power
RL is the target impedance for the output matching network. VMAX and Vknee are the maximum and minimum output voltage operation limit, respectively. IMAX is the maximum output current capability of the device. All these parameters are set whenever the sizes and the bias of the active devices are set.
Once the class-AB PA is designed with a PA core loading of the optimal load-line impedance RL, the PA can theoretically operate with maximum output power efficiency. Therefore, the output matching of a PA is also named as the load-line match.
2.1.3 Matching Network Design
Lossless devices (inductors and capacitors) are used in order not to dissipate any of the signal power. Their behaviors and the effects of impedance transform are shown in Fig. 2.4 on the ZY Smith chart. There are three commonly used matching network configurations to transform impedance to a desired position on the Smith chart. Namely, they are L-section (Fig. 2.5), T-section (Fig. 2.6), and π-section (Fig.
2.7). Form these figures and the attached design equations, proper input and output matching networks can be designed accordingly.
In the typical circuit implementation of this architecture in Fig. 2.2, the output matching network is in the π-section configuration, and the input matching network is as the L-section configuration.
2.1.4 The Design Principle of the Narrowband Class-AB PA
The output power is defined in (2-2).
(2-2)
Once the output power specification is given, RL is set by (2-2), according to the maximum output voltage swing available from the given circuit topology. The required maximum output current (IMAX) is set by (2-1). Thus, the sizes and bias of the active devices are set.
Once the sizes and bias of the active devices are set, the input and output impedance of the active devices is defined. Then, the input and output matching networks can be designed to transform the input and output impedance of the active devices to a suitable impedance level accordingly. After the input and output matching networks are designed properly, the narrowband class-AB PA is ready to work.
2.2UWBCLASS-ABDISTRIBUTED AMPLIFIER
2.2.1 Conventional Architecture
The distributed amplifier is an elegant way to overcome the limitation of maximum gain-bandwidth product [14]. This architecture achieves a gain-delay trade-off without the penalty on bandwidth. Theoretically, this architecture can provide possibly infinite bandwidth with arbitrary gain. Therefore, ultra-wideband amplification is accomplished.
Fig. 2.8 is the conventional distributed amplifier architecture. Each Gm-cell acts as a transconductor to provide a certain amount of output current corresponding to the input driving voltage signal. While the input driving voltage signal propagates down the input line, each Gm-cell is being excited in succession, producing the output
other half of the output current signals ultimately sum in time coherence if the delays of the input and output lines are matched. Therefore, the output current waves sum up coherently in constructive superposition manner.
2.2.2 Load-line Design of Each Gm-cell
Fig. 2.9 is a typical circuit implementation of this architecture. The cascode topology provides good voltage gain and good isolation. The input and output lines can be synthesized by lumped passive devices, exhibiting a transmission line characteristic impedance of Zo.
Fig. 2.10 shows the loading condition of each Gm-cell. The active device output is loaded with a characteristic impedance of Zo in both directions. Equivalently, each Gm-cell is loaded with Zo/2. Therefore, it is easy to show that the voltage gain, Av, of the distributed amplifier is governed by equation (2-3).
(2-3)
The n, the number of stages, in (2-3) is the number of Gm-cell in the distributed amplifier.
The distributed amplifier architecture provides the capability to achieve simultaneously 50-ohm conjugate match and load-line match. Since each direction seen by the active device output is designed to be 50-ohm for minimum signal reflection, the total loading seen by the active device output is 25-ohm as the 50-ohm output matching condition. If each Gm-cell is also designed to be with a 25-ohm optimal load-line RL, as shown in Fig. 2.10, 50-ohm conjugate match and optimal load-line match is simultaneously achieved. Therefore, minimum output signal reflection and PA maximum output power efficiency can be accomplished at the same time. Note that this is impossible in the case of narrowband class-AB PA, which must
trade the output 50-ohm matching and the PA maximum output power efficiency, since the optimal load-line impedance is usually much smaller than 50-ohm.
Finally, the total output power can be derived and shown in (2-4).
(2-4) Pouteach is the output power of each Gm-cell; Pouttotal is the total output power appear at the output port.
2.2.3 Input and Output Line Design
Fig. 2.11 is the distributed amplifier architecture whose input and output lines are synthesized by lumped devices. In such manner, the input and output lines are named as the artificial lines.
From Fig. 2.11 the governing equation (2-5) of the characteristic impedance of Zo of the line is also straight forward [].
(2-5) To achieve 50-ohm matching, the characteristic impedance Zo of the line is designed to be 50-ohm. The terminal resistor Rt at the end of the line is also 50-ohm to ensure no signal reflection back to the input and output port.
There are three configurations for the artificial lines (Z and Y), namely low-pass line, high-pass line, and band-pass line, as shown in Fig. 2.12. From Fig. 2.11, it can be observed that the overcome of the bandwidth limitation of this architecture comes
into the input and output line, causing entirely no degradation on the circuit operation speed. Therefore, until the cutoff frequency of the line itself is approached, the input and output impedance remains constant and equal to Zo, and the overall operation bandwidth is controlled solely by the input and output lines. It is obvious that the band-pass line structure is the most convenient way to control the overall band-pass type bandwidth.
2.2.4 Design Principle of the UWB Distributed Amplifier
First of all, the optimal load-line RL condition of each Gm-cell is designed as the 25-ohm, as shown in Fig. 2.10. In this case, conjugate matching condition and maximum output power efficiency condition can be simultaneous achieved. Therefore, minimum signal reflection and excellent power efficiency can be guaranteed.
Once the optimal load-line RL is set, the size and bias of the active devices is set.
Also, the output power of each Gm-cell is defined, as shown in (2-4). Therefore, the input and output impedance of the active devices is defined. With the information of the input and output capacitances, along with the bandwidth specification, the input and output artificial line can be designed, according to Fig. 2.12.
Finally, from the output power specification, the number of stages can be defined, according to (2-4), and the ultra-wideband distributed amplifier is ready to work.
2.3BASIC RFPAFIGURES OF MERITS
2.3.1 Scattering Parameters, S-parameters
To characterize the behavior of a RF two port network, scattering parameters (S-parameters) are used [14]. Fig. 2.13 is the demonstration of a two-port network
characterized by S-parameters.
S-parameters are defined by power waves. For RF systems, the signals are actually in the form of power waves, and thus S-parameters are suitable for characterizing these systems which operate at high frequency. On the other hand, the traditional Z-parameters and Y parameters are not capable of defining such systems. It is because the definitions of Z-parameters and Y parameters require perfectly open and short conditions, which are difficult to obtain at high frequency. Also, active devices, such as diodes, BJTs, and MOSFETs, cannot function stably under open and short terminal conditions. Therefore, S-parameters are the most common way to define a microwave/RF system.
S-parameters are defined in (2-6).
(2-6) In (2-6), a1 and a2 represent the incident waves from the source and the load, respectively; b1 and b2 represent the reflected waves from Port 1 and Port 2 of the system to the source and the load, respectively. The physical meaning of the terms in the S-parameters matrix is in (2-7).
Port 2 to Port 1.
S-parameters of a PA are commonly used to demonstrate the small signal power handling capability of the PA over the frequency range of interest. S11 describes the input matching condition of the PA. S22 describes the output matching condition of the PA. S21 describe the small signal power gain of the PA. S12 describes the reverse isolation condition of the PA and stands for the unilateral property of the PA. Fig.2-14 shows a typical S-parameters behavior of a PA. Note that the S-parameters are indexed with frequency. Therefore, it can be somehow viewed as the frequency response of the PA small signal power handling capability.
From the S-parameters the stability of a RF amplifier can be extracted. The stability of a RF amplifier indicates the property that the RF amplifier can function properly without going into oscillation. This is crucial for RF amplifiers, especially for RF power amplifiers, since power amplifiers consume and output a large amount of power, which is prone to oscillation. Oscillation happens when the real part of the input or output impedances becomes negative. This condition results in the situation that input or output reflection coefficient, Γin or Γout, is greater than unity. Under this circumstance, the RF amplifier would no longer stably amplifying signals but behave as an oscillator.
The stability can be quantified by the S-parameters of the active devices, the matching conditions of the amplifier circuit, and the source and load impedances.
There are two factors to describe this quantity of stability, k-factor and μ-factor. []
(1) k-factor:
Given a S-parameters of a two-port network between the input source and the output loading, the necessary and sufficient conditions for unconditional stability are that k is greater than unity and |Δ| is less than unity, as shown in (2-8).
⇒ (2-8) This expression can be further manipulated into another equivalent condition, as shown in (2-9).
(2-9) The k factor is also named as stability factor and b factor is named as stability measure.
(2) μ-factor
The μ-factor gives the geometric distance from the center of the Smith chart to the nearest output (load) stability circle. This stability factor is given by (2-10).
(2-10)
The single necessary and sufficient condition for unconditional stability of the 2-port network is that μ is greater than unity. This condition actually describes the fact that the output reflection coefficient, Γout, is less than unity, and thus the output impedance is not negative resistive. Positive real part of the output or input impedances guarantees unconditional stability.
Alternatively, the single necessary and sufficient condition for unconditional stability of the 2-port network is that μ′ is greater than unity, as shown in (2-11).
(2-11)
that the input reflection coefficient, Γin, is less than unity, and thus the input impedance is not negative resistive, which also guarantees the unconditional stability.
2.3.2 Large Signal Transfer Characteristics
The plot of output power versus input power gives the power transfer characteristics (PTC) of a PA. PTC is commonly used to demonstrate the large signal power handling capability of a PA [13]. Fig.2-15 shows a typical PTC of a PA.
A commonly used unit for describing RF power is dBm, which is defined in (2-12).
(2-12) From the PTC, two properties of the PA can be read. The first property is the saturation power, which is the maximum output power capability of the PA.
The second property is the large signal power gain, which is defined in (2-13).
(2-13) PTC can be further manipulated to extract the linearity property of the PA. One of the most important linearity indicators is the 1-dB compression point (P-1dB).
From the PTC of a PA, it can be noticed that, for sufficient large input power, the output power would no longer be amplified by the same constant gain. The output power would eventually saturate, and there would be no further linear amplification relationship between the input and the output.
The reason for this phenomenon is that, for linear PAs, usually the active devices should remain in their active regions during the whole operation cycle. For MOSFETs the active region is its saturation region. For small output signal power this assumption stays true. However, when the output power increases, the signal swing at
the active device output node (for MOSFETs it means drain) would gradually increases, and eventually the devices would swing into the nonlinear region of the device I-V curve (for MOSFETs it means triode region). Thus, the output swing would be compressed, and so as the output power. No further output power can be delivered to the load since from now on the active devices have been driven to their maximal operation limit, as shown in Fig.2-16.
1-dB compression point, P1dB, quantifies this phenomenon. It is defined in (2-14).
(2-14) OP1dB is the output power at P1dB, and IP1dB is the corresponding input power.
The definition of P1dB is that at this power level the amplifier is unable to amplify the signal at the same constant rate as it can for much smaller signal level. At P1dB, the actual gain (OP1dB(dBm)-IP1dB(dBm)) is 1 dB smaller than the small signal linear gain.
It is then defined that this point is the starting point at which the output power begins to be noticeably compressed. Therefore, OP1dB can be treated as the maximum linear output power capability of a PA.
Another performance metric for PA large signal operation is efficiency.
Efficiency of a RF amplifier indicates the utilization of power participated in the process of transforming input power into output power. There is always inevitable loss in this transforming process; no one hundred percent perfect utilization of power.
Thus the efficiency of this transforming process needs to be quantified to describe the PA performance of utilizing power to satisfy its output power requirements.
There are three factors to describe the quantity of efficiency: drain efficiency,
(1) Drain efficiency, η
Drain efficiency, η, describes the utilization of the DC supplying power (PVDD) in the viewpoint that a PA’s duty is to output power. Therefore, this metric describes how efficiently the supplying power is used to satisfy the output power requirement.
(2) Power added efficiency, PAE
Power added efficiency, PAE, describes the utilization of the DC supplying power (PVDD) in the viewpoint that a PA needs to output and amplifies power. In such manner, PAE describes how efficiently the supplying power is used in the PA to satisfy its output power capability and the power amplification capability, and thus is more suitable for a power amplifier metric of efficiency.
(3) Total efficiency, ηT
Total efficiency, ηT, describes the utilization of the DC supplying power (PVDD) in the viewpoint that a PA output power by consuming supplying power and input power. In such manner, ηT describes how efficiently the supplying power and the input driving power are used in the PA to satisfy its output power capability, and thus is suitable for specifying overall system performance.
Fig. 2.1 Conventional class-AB PA architecture.
Fig. 2.2 Typical circuit implementation of the conventional class-AB PA architecture.
Fig. 2.3 Typical I-V relationship and the optimal load-line for maximum output power efficiency, of conventional class-AB PA core operation.
Fig. 2.4 Behaviors and effects of impedance transform on the ZY Smith chart.
Fig. 2.5 Four types of the L-section with the corresponding design equations.
Fig. 2.6 Four types of the T-section with the corresponding design equations.
Fig. 2.7 Four types of the π-section with the corresponding design equations.
Fig. 2.8 Conventional distributed amplifier architecture.
Fig. 2.9 Typical circuit implementation of the conventional distributed amplifier architecture.
Fig. 2.10 Loading condition of each Gm-cell.
Fig. 2.11 Distributed amplifier with artificial line.
(a) (b)
(c)
Fig. 2.12 Detailed artificial line structure and corresponding design equations of (a) low-pass line, (b) high-pass line, and (c) band-pass line.
Fig. 2.13 Demonstration of the system characterized by the S-parameters.
Fig. 2.14 Demonstration of the S-parameters of a typical PA.
Fig. 2.15 Power transfer characteristic (PTC) of a PA.
Chapter 3
RF ESD Protection Basics
3.1ESDTESTING BASICS
3.1.1 ESD Testing Pin Assignments
Fig. 3.1 shows a typical pad layout allocation of a RF system IC. The input and output (I/O) pads of a RF system are in GSG style, providing good shielding and wave-guiding capability for signal power waves. The VDD pad provides the DC power supply for the system. All the GND pads should be carefully connected all together internally or externally to provide a stable voltage reference level for RF signals.
The ESD energy enters an IC device most likely through its I/O VDD, and GND pads, in the form of a discharging current or voltage overstress. All these pads serve as the communication interface between the IC internal circuits and the external environment. Through these pads, however, ESD current can also find its way and discharge into the IC, causing damage to the IC internal circuits. Therefore, the first priority of the ESD protection circuits is to shunt away the ESD current at the very first moment whenever the ESD current appears at the I/O, VDD, or GND pads.
There are four configurations of ESD testing pin assignments for I/O pads.
Namely, they are PS-mode, NS-mode, PD-mode, and ND-mode, as shown in Fig. 3.2
Namely, they are PS-mode, NS-mode, PD-mode, and ND-mode, as shown in Fig. 3.2