Chapter 2 Ultra-Wideband Communication System
2.5 IR-UWB Receiver
The main characteristic of UWB impulse radios is that very low emission power density can be achieved by spreading the energy of short-time pulses in wideband.
These radios present the advantage of not requiring up/down conversion of frequency,
which results in reduced complexity and low cost of manufacturing. According to the demodulated method, IR-UWB receivers can be divided into two architectures: coherent receiver and non-coherent receiver. Coherent receivers, as shown in Fig.2-13, rely on the correlation of the received pulses and a local template demand complex implementations [12,13]. Because of short pulse adoption and known position, it has some advantages of high data rate and high SNR. Unfortunately, it needs precision timing synchronization between transmitter and receiver ends. Generally the requirement is realized by delay lock loop (DLL) circuits to achieve synchronization.
On the other hand, the non-coherent receiver shown as Fig.2-14 requires neither pulse synchronization nor estimation of the shape of the incoming pulses. Instead, it recovers the energy of the pulses during a symbol time and compares it to the noise level in order to determine the presence or absence of a symbol [14]. The main drawback of using this kind of detector is that the UWB pulses cannot be detected when the signal-to noise ratio (SNR) is very low, and hence, it cannot make use of the processing gain that spread spectrum systems have. Accordingly, the non-coherent impulse receiver will only work properly when the SNR is above a threshold which is close to the noise level.
Due to the very limited power that is allowed in the UWB transmitter, this only occurs at very short distances.
As to sensitivity, the BER vs. SNR curves of a non-coherent receiver is simulated and compared to that of a coherent receiver using BPSK modulation [14]. The simulation results are presented in Fig.2-15. Both simulations are executed for 10, 25, 50, and 100 Mbps using a pulse rate of 100MHz. The monocycles pulses are shaped to use the low part of the UWB spectrum 3.1GHz-5GHz. As expected, the coherent receiver presents processing gain and hence requires less SNR than the non-coherent receiver for a fixed BER. In addition, the simulation shows that the non-coherent receiver can still detect the UWB pulses for SNR close 0 dB.
Figure 2-13 Architecture of the coherent receiver
Figure 2-14 Architecture of the non-coherent receiver
Figure 2-15 Performance curves of a non-coherent receiver and a coherent receiver using BPSK modulation
Chapter 3
Design of the Ultra-wideband LNA and Correlator
3.1 Overview
In this chapter we propose two circuits, which can apply in the front-end receiver of impulse-radio UWB system. The first one is a UWB LNA with transformer-feedback matching network. It employs the characteristic of the transformer to achieve good input matching and noise performance. Subsequently, the correlator with dynamic gain control has presented. In addition, the wide bandwidth is achieved by canceling the dominant pole at the internal node with the zero introduced by the shunt inductor at the loading stage. These circuits have been fabricated by TSMC 0.18μm 1P6M CMOS technology. We also exhibit the simulation and measurement results of each circuit.
3.2 A UWB LNA with Transformer Feedback Matching Network
In this section we propose another UWB LNA design method. It utilizes transformer feedback for wideband matching and noise degradation. First, the general monolithic transformer prototype is introduced. Then we analysis different kinds of transformer and list each advantages and disadvantages. Then we demonstrate that the design concepts in the proposed UWB LNA in detail. Subsequently, we show the simulation and measured results. Finally, the differences between the simulation and measured results are discussed.
3.2.1 Introduction of Monolithic Transformers
Transformers have been used in radio frequency (RF) circuits since the early days of telegraphy. Recent works have shown that it is possible to integrate passive
transformers in silicon IC technologies because of useful performance characteristics [15,16,17]. In general, the operation of a passive transformer is based upon the mutual inductance between two conductors. Basically, the transformers are used for the following three different functions.
1. Impedance matching: depending on the number of windings, the transformer has the property to change the impedance of the primary or secondary when measuring from the opposite port.
2. Balun: balanced to unbalanced conversion and vice versa
3. DC isolation: obtained with the magnetic (nonelectric) connection between the primary and secondary
Fig.3-1(a) summarizes the basis of an ideal transformer where the primary self-inductance and the secondary self-inductance are characterized with ideal inductors. The mutual inductance is represented by
L1 L2
M , the primary and secondary currents and voltages are , , , and , and the primary and secondary winding numbers are and , respectively. The coupling factor k is defined by
i1 i2 v1 v2
N1 N2 M , ,
and and represents the energy transmitted from the primary port to the secondary port [18,19].
L1
L2
The behavior of the ideal transformer in Fig.3-1(a) is ruled by its characteristic equations
When transformers are applied in silicon ICs, the ideal model cannot be easily establish
because of substrate loss and parasitic effects. Thus, the electrical model of an integrated transformer must be redefined, and the equivalent model of an integrated transformer is shown in Fig.3-1(b).
(a) (b)
Figure 3-1 (a) Electrical model for an ideal transformer (b) Equivalent circuit of an integrated transformer
where and represent the ohmic losses due to the resistivity of the inductor metal tracks; is the capacitive coupling caused by the voltage difference between the turns in the same metal, which form a spiral; represents the capacitive coupling caused by the voltage difference between the turns of the primary and secondary spirals;
is the capacitive coupling between the metal used for each inductor and ground;
R1 R2 Cp
Cm
Cox
C and si R represent the coupling and ohmic losses due to the conductive substrate. si
(a) (b)
(c) (d)
Figure 3-2 Monolithic transformer winding configurations (a) Parallel conductor winding (b) Interwound winding (c) Overlay winding (d) Concentric spiral winding
A monolithic transformer is constructed using conductors interwound in the same plane or overlaid as stacked metal layers, as shown in Fig.3-2(a). The type of parallel conductor winding is interwound to promote edge coupling of the magnetic field between windings. The primary and secondary windings lie in the same plane, as illustrated in the cross-section at the right of Fig.3-2(a). Because of the asymmetric characteristic, the ratio of transformer turns is not unity.
As shown in Fig.3-2(b), the type of interwound winding has feature of symmetry.
It ensures that electrical characteristics of primary and secondary are identical when they have the same number of turns. Another advantage of this design is that the transformer terminals are on the opposite sides of the physical layout, which facilitates connections to other circuits.
Integrated transformers with multiple conductor layers are illustrated in Fig.3-2(c).
The overlay winding utilizes both edge and broadside magnetic coupling to reduce the
overall area required in the physical layout. Flux linkages between the conductor layers can be improved as the intermetal dielectric is thinned. In addition, there is a large parallel-plate component to the capacitance between windings due to the overlapping of metal layers, which limits the frequency response.
Transformers can also be implemented using concentrically wound planer spirals as shown in Fig.3-2(d). The periphery between the two windings is limited to just a single turn. Therefore, mutual coupling between adjacent conductors contributes mainly to the self-inductance of each winding and not to mutual inductance between the windings. As a result, the concentric spiral transformer has less mutual inductance and more self-inductance than other types. This kind of low ratio of mutual inductance to self-inductance is useful in applications such as high-performance broadband amplifiers.
The four kinds of monolithic transformers above are summarized in Table 3-1. We can select an appropriate type for an specific design. For instance, when designing an amplifier for UWB, we can use the type of concentric spiral winding for achieving wideband response.
Table 3-1 Comparison of differential transformers Transformer type Coupling coefficient
LSelf fSR
Parallel conductor winding Middle Low High
Interwound winding Middle Low High
Overlay winding High High Low
Concentric spiral winding Low Middle High
3.2.2 Design Concepts
Figure 3-3 Schematic of the proposed UWB LNA.
Transformer-feedback matching network
The proposed LNA is shown in Fig.3-3. The transformer consists of two inductors.
The primary winding is shunt with the gate end of M1; the secondary winding is series to the source end of M1. There are no extra lumped elements placing at the input port excluding ac-coupling capacitance and bypass capacitance . In general, the common-source amplifier with the inductive degeneration has been popularly used because it can generate real impedance to match source resistance. The imaginary part can be cancelled by the reactive elements located at the gate of M1. But the method only suits in narrow band amplifier design because it satisfies above only at a specific frequency. Therefore, we modify the conventional matching skills for broadband matching purpose as explained in the following. The simplified input impedance diagram is shown in Fig.3-4.
1
CD Cb1
TLs
Figure 3-4 The simplified input impedance network
r is the parasitic resistance of s and the parasitic resistance of is neglected for simplifying the analysis. is the gate-source capacitance of M1. The input impedance can be represented in s-domain as
Ls Lp
The imaginary part forms a equation comprising three zero-point frequencies, one resonant frequency is zero, the others are both 1/ LSCgs1 . It means that if the operating frequency is equivalent to 1/ LSCgs1 , the impedance matching can accomplish due to the cancellation of the imaginary part. Generally is a fixed
value as long as the biasing is determined. Then, would need to be decreased as the operating frequency rises in order to meet the matching condition. This is the reason that we use transformer in the matching network. The self-inductance and mutual inductance of the transformer are appropriately selected and the practical inductance looked into the source of M1 decreases apparently. The broadband matching can be realized by the phenomenon of the transformer feedback.
1
Cgs
Ls
Noise Analysis and Current-reuse Technique
The noise performance of the proposed topology is determined by two main contributors: the losses of the input network and the noise of the first amplifying device (M1). The general expression for noise figure as given by a classical noise theory [20]:
2 is the real part of the source admittance. We can derive the optimal noise figure if the condition of
In order to have adequate power gain at the condition of the limited power consumption, we adopt the current-reuse technique at gain stages. The second stage ( ) is stacked on the top of the first stage ( ). A coupling capacitor ( ) and a bypass capacitor ( ) are required for this topology. Both and are metal-insulator-metal (MIM) capacitors. The choice of large capacitance of is preferred to perform better signal coupling. However, too large MIM capacitors may suffer from parasitic capacitance between the bottom plate of the capacitor and ground, which would degrade the circuit gain. The value of is chosen to be as large as possible to provide ideal ac ground. In addition, and are designed to have a peaking characteristic to compensate the low frequency roll-off of the device.
M2 M1 CD2
LB connects between the main stage and buffer in order to enhance wideband characteristic. The further bandwidth extension is achieved due to a series LC resonance with the gate capacitance of . Various values of can cause different performances of bandwidth enhancement. Fig.3-5 exhibits the gain performance with different value of . We select =2.75 nH for the trade-off between bandwidth and
M4
LB
LB
LB
gain flatness for our design purpose.
Figure 3-5 Gain performance with LB variation
A source follower is adopted as the output buffer for a wideband output matching purpose. The current source is different from the conventional one because it is driven by itself without extra bias voltage. The buffer is not needed if the LNA integrates directionally to mixer or other circuit block.
Layout consideration
The geometric layout of the transformer [19] will impact the input return loss, noise figure, and stability, etc. Rectangular-type and concentric spiral windings are adopted to reduce the Q factor for wideband applications. Turn numbers of separate inductor and the spacing between the two inductors inside the transformer are appropriately chosen to achieve feasible self-inductance and mutual inductance, respectively. The transformer of this proposed LNA is simulated by ADS Momentum and the geometric photo is shown in Fig.3-6. Metal 6 is used to layout the transformer because that thicker metal reduces the ohmic losses in the primary and secondary windings of a planar transformer. The size of M1 is 150×0.18 μm2 for noise optimum
source impedance. The 130×0.18 μm2 of size is selected as the trade-off between gain and nonlinearity. The value of is 4.8nH to make sure in inter-stage stability.
M2
Lint
Figure 3-6 Diagram of the proposed transformer applied in this design
3.2.3 Simulation and Measurement Results
The Ultra-wideband LNA with the transformer feedback matching network is fabricated using TSMC 0.18μm RF CMOS process. Fig.3-7 shows the microphotograph and the total area is about 0.6 mm2. The whole measurement is on-wafer test on the RF probestation. Two three-pin GSG RF probes are used for transmission of input and output signal, and one three-pin PGP DC probe provides the biasing and supply voltage.
The scattering parameters are measured by HP 8510C. Fig.3-8~3-11 illustrate the measured results of individual s-parameters. The simulation results are also put together to compare the difference. In Fig.3-8 the measured input return loss (S11) is better than -9.8 dB over the entire UWB band with two dips at 3.7GHz and 6.2GHz. The performance proves that transformer feedback is feasible in wideband matching. The output return loss (S22) is -10.4 dB below from Fig.3-11. Fig.3-10 shows that the
measured isolation form output to input is under -28 dB. Fig. 3-9 is the simulated and measured results of the power gain. The measured power gain has a peak value in 12.4dB and the average value is 11.2dB with 1.2dB ripple from 3.1GHz to 10.6GHz.
The result presents a good feature of flatness. For the noise figure measurement, we exploit noise figure analyzer (Agilent N8975A) and noise source (Agilent 346C). The results of simulated and measured noise figure are shown in Fig.3-12. It is obvious that the curves of and are mostly overlap in the UWB band. It indicates that transformer feedback has an optimal noise figure performance as well. The measured results of NF have the minimum value 3.2dB at 5GHz and the average is 4dB.
NFsim NFmin
As to in-band linearity, the 1-dB compression input point ( ) with 5GHz is -14dBm as shown in Fig.3-13. Two-tone signals of 5GHz and 5.01GHz are applied to the LNA to observe the input referred third-order intercept point (IIP3), and the spectral diagram has shown in Fig. 3-14. The measured value is in the range of -5 to -10dBm over 3 to 10-GHz system while is about -22dBm to -14dBm. From Fig.3-15 it can observed that IIP3 is around 10 dB less than over the entire band.
P1dB
P1dB
P1dB
Table 3-2 makes a comparison between simulation and measurement results. It demonstrates that they are only slightly different. Table 3-3 lists some previously-proposed references. Each input matching method and characteristic of LNA is summarized to make comparison. It can be observed that our UWB LNA with transformer feedback has competitive performances at I/O return loss, power gain, power consumption, and even chip size.
Figure 3-7 Microphotograph of the proposed LNA
0 2 4 6 8 10 12 14 16
Frequency (GHz) -30
-25 -20 -15 -10 -5 0
S11 (dB)
Sim.
Mea.
Figure 3-8 S11 simulation and measured results of UWB LNA
0 2 4 6 8 10 12 14 16
Figure 3-9 S21 simulation and measured results of UWB LNA
0 2 4 6 8 10 12 14 16
Figure 3-10 S12 simulation and measured results of UWB LNA
0 2 4 6 8 10 12 14 16
Figure 3-11 S22 simulation and measured results of UWB LNA
2 4 6 8 10 12 14
Figure 3-12 NF simulation and measured results of UWB LNA
Figure 3-13 Measured input 1dB compression point at 5GHz
Figure 3-14 Spectral diagram of the IIP3 measurement (one tone is 5GHz, and the other tone is 5.01GHz)
3 4 5 6 7 8 9 10
Frequency (GHz)
-22 -20 -18 -16 -14 -12 -10 -8 -6 -4
P1dB(dBm) IIP3(dBm)
Figure 3-15 Measured P1dB and IIP3 curves over the UWB band
Table 3-2 Comparison of simulation and measurement results of UWB LNA Simulation results Measurement results
S11 (dB) <-10.7 <-9.8
S21 (dB) 12 11.2±1.2
S12 (dB) <-36 <-24
S22 (dB) <-11.6 <-10.4
NF (dB) <4.6 3.2~5.5
P1dB (dBm) -14 ~ -23 -14 ~ -22
IIP3 (dBm) -3 ~ -13 -5 ~ -10
Power dissipation (mW) 11.4 11.1
Table 3-3 Performance of summary and comparison to other wideband LNAs
[23] 3~5 Resistive-shunt feedback
[16] 3.1~10.6 Dual feedback <-11.2 10.8~12 4.7~5.6 -12 ~-10.6 10.57*
@1.5V
* without adding consumption of buffer stage
3.2.4 Discussion
In section 3.2, we propose a novel UWB LNA and implement successfully by TSMC 0.18μm CMOS process. This circuit uses the transformer feedback topology to realize broadband matching and noise optimization. At the gain stage, it adopts current-reuse technique to have adequate gain performance under power consumption limits. A source follower with self-biasing can reduce extra bias voltage and achieve output matching. The measured results of S11, S21, and S22 are similar to the simulation results. The measured NF is close to simulation result at the low band and only higher than simulation result about 1dB at the high band. The reason may be that parasitic resistance affects apparently at the higher frequency band. As to linearity, the maximum IIP3 is -5dBm at 6GHz while is -14dBm. It can be observed that our UWB LNA with transformer feedback has competitive performances at I/O return loss, power gain, power consumption, and even chip area.
P1dB
3.3 Analog Correlator with Dynamic Gain Control
This section describes the circuit design principle of a correlator suitable for UWB systems. The technique of dynamic gain control is inserted for a VGA-like architecture in this correlator. The simulation and measurement results are exhibited and discussed in the final of this section.
3.3.1 Design Concepts
The function of correlator is detecting and demodulating the received signal for the following A/D converter or comparator. Normally a correlator incorporates a separate multiplier and a separate integrator. There are some main problems in designing the correlator of pulse-based UWB receiver [24]. For instance, the multiplier should have very wide-band input frequency response, even the lower band UWB pulse is very large depending on data and applications. Therefore, though the DC offset current or 1/f noise current must be much smaller than the signal current, it can integrate on the integral capacitor for a much longer time the multiplied signal. Another problem is that after the pulse correlation, the output voltage should maintain for a long time for the ADC to sample, but the output impedance of the integrator normally is not large enough, and results in large leakage from the integrator. These ultra-wideband characteristics of the input pulses make the digital domain correlation not suitable in this application.
According to above discussions, we adopt the analog correlator in UWB system receiver.
Generally multipliers are much more difficult to design than mixers. For mixers, the gain doesn’t need to vary linearly with the LO signal, and only the first harmonics of the LO is needed to multiply with the input signal. Working in linear region makes the multiplier very difficult to bias. In this design of multiplier we apply a Gilbert-cell with bandwidth extension and linear adjustment for UWB systems.