High-Frequency Amplifiers:RFLNAs

RFLNAs

This section reviews the important performance criteria demanded of the front-end amplifier in a wireless communication receiver. The design of CMOS LNAs for front-end wireless communication receiver applications is then addressed. The following section considers the related topic of LNAs for optical receiver front-ends.

Specifications

The front-end amplifier in a wireless receiver must satisfy demanding requirements in terms of noise, gain, impedance matching, and linearity.

Noise

Since the incoming signal is usually weak, the front-end circuits of the receiver must possess very low noise characteristics so that the original signal can be recovered. Provided that the gain of the front- end amplifier is sufficient so as to suppress noise from the subsequent stages, the receiver noise performance is determined predominantly by the front-end amplifier. Hence the front-end amplifier should be an LNA.

Gain

The voltage gain of the LNA must be high enough to ensure that noise contributions from the following stages can be safely neglected. As an example, Figure 24.15 shows the first three stages in a generic front-end receiver, where the gain and output-referred noise of each stage are represented by Gi and Ni (i = 1, 2, 3), respectively. The total noise at the third-stage output is given by

According to Eq. (24.10), the gain of the first stage should be high to reduce noise contributions from subsequent stages. However, if the gain is too high, a large input signal may saturate the subsequent stages yielding intermodulation products which corrupt the desired signal. Thus, optimization is inevitable.

Input Impedance Matching

The input impedance of the LNA must be matched to the antenna impedance over the frequency range of interest to transfer the maximum available power to the receiver.

Linearity

Unwanted signals at frequencies fairly near the frequency band of interest may reach the LNA with signal strengths many times higher than that of the desired signal. The LNA must be sufficiently linear to prevent these out-of-band signals from generating intermodulation products within the desired frequency band, and thus degrading the reception of the desired signal. Since third-order mixing products are usually dominant, the linearity of the LNA is related to the ‘‘third-order intercept point’’ (IP3), which is defined as the input power level that results in equal power levels for the output fundamental frequency com- ponent and the third-order intermodulation components. The dynamic range of a wireless receiver is limited by noise at the lower and nonlinearity at the upper band.

For maximum power transfer, the input impedance of the LNA must be matched to the source resistance which is normally 50 W. Impedance matching circuits consist of reactive components and therefore are (ideally) lossless and noiseless. Figure 24.16 shows the small-signal equivalent circuit of a CS LNA input stage with impedance matching circuit, where the gate-drain capacitance Cgd is assumed to have negligible effect and is thus neglected [16,17]. The input impedance of this CS input stage is given by

CMOS CS LNA: Effect of Cgd

In the analysis so far, the gate-drain capacitance (Cgd) has been assumed to be negligible. However, at very high frequencies, this component cannot be neglected. Figure 24.18 shows the modified input stage of a CS LNA including Cgd , and where an input AC-coupling capacitance Cin has also been included. Small-signal analysis shows that the input impedance is now given by

then the input impedance can be easily matched to the source resistor Rs. Substituting Eq. (24.21) for ZL, the bracketed term in the denominator of Eq. (24.19) becomes

Cascode CS LNA

Input Matching

As outlined in the above section, the gate-drain capacitance (Cgd) degrades the input impedance matching and therefore reduces the power transfer efficiency. In order to reduce the effect of Cgd, a cascoded structure can be used [18–20]. Figure 24.19 shows a cascode CS LNA. Since the voltage gain from the gate to the drain of M1 is unity, the gate-drain capacitance (Cgd1) no longer sees the full input–output voltage swing which greatly improves the input–output isolation. The input impedance can be approx- imated by Eq. (24.11), thus allowing a simple matching circuit to be employed [18].

Voltage Gain

Figure 24.20 shows the small-signal equivalent circuit of the cascode CS LNA. Assuming that the input is fully matched to the source, the voltage gain of the amplifier is given by

From Eq. (24.26), the voltage gain is dependent on the ratio of the load and source inductance values. Therefore, high gain accuracy can be achieved since this ratio is largely process independent.

Noise Figure

Figure 24.21 shows an equivalent circuit of the cascode CS LNA for noise calculations. Three main noise sources can be identified; the thermal noise voltage from Rs, and the channel thermal noise currents from

M1 and M2. Assuming that the input impedance is matched to the sources, the output noise current due to vRS can be derived as

To improve the noise figure, the transconductance values (gm) of M1 and M2 should be increased. Since the gate-source capacitance (Cgs2) of M2 is directly proportional to the gate width, the gate-width of M2 cannot be enlarged to increase the transconductance. Instead, this increase should be realized by increas- ing the gate-bias voltage.