In this chapter, you will learn about various kinds of low-pass filters, such as Butterworth filters and maximally flat envelope delay (MFED) filters. You will see the small-signal bandwidth these filters exhibit as well as how they respond to step response. You will learn that all high-frequency circuits are eventually limited in their bandwidth because of output resistance driving an input capacitance. And then you will observe that it is possible to increase bandwidth by the judicious use of inductance added in the right place. You will find that it is possible to obtain as much as 2.72 times the bandwidth of a simple RC (resistance/capacitance) low-pass filter while maintaining good transient response. These techniques are referred to as 'peaking' networks.

This chapter develops the concept of difference amplifiers as applied to high-frequency design. The author has used the models developed in the last chapter to analyze these circuits. The author found that by setting the time constant of the parallel combination of R_eC_e to the transistor T_t, the input impedance at the base would look like a pure capacitor that could be series-, shunt-, or T-coilpeaked. The authors developed the concept of a level-shift PNP transistor that allowed a complete gain block to be created. The gain block is designed so that it has zero DC volts at both the input and output along with the same impedance levels (usually 50-100 Ω). This allows gain blocks to be directly cascaded.

Up to now, all the circuits the author has talked about have used series feedback, where the local feedback element is in series with both the input and output currents. In this chapter, the author looks at shunt feedback, where some of the output voltages are sampled and shunted back to the input. This is the technique used by almost all op-amp circuits. However, since the same technique can be applied successfully to high-frequency circuits, the authors need to investigate this important concept. Moreover, there is another important circuit that does not fit well into the other categories. This is a circuit made up of three transistors that produces the equivalent of a single transistor. This composite circuit has a higher equivalent ft and twice the β of the devices making up the circuit. Because it can be used almost anywhere a single transistor can be used, it is an extremely useful circuit - especially handy when designing integrated circuits, because it is easy to add transistors to an integrated circuit. This circuit provides almost the same bandwidth-improvement factor as series peaking. But because series peaking requires inductors, the composite circuit is much easier to implement on an integrated circuit.

Another reason to change the actual Gummel-Poon parameters (while keeping f_t the same) is to see the effect on the circuit of a particular parameter (the usual one being C_jc) without upsetting the over f_t of the device. C_jc affects things outside the immediate transistor. For example, it is the main component in s₁₂; typically, if C_jc were zero, s₁₂ would be zero. It is also the principal contributor to preshoot. It supplies the capacitive sneak path that causes preshoot. To see if these effects are dominant in your circuit, it is sometimes necessary to absorb the effects of C_jc on f_t into one or both of the other parameters (C_je or τ_f) so that the transistor f_t does not change if C_jc is set to zero. The outside circuit effects will be changed with C_jc equal to zero. In this way, a valid comparison can be made and we can determine the importance of C_jc. The author used this technique when talking about preshoot in earlier chapters to illustrate certain points.

The following was contributed by Dr. Allen Podell (he teaches Advanced Wireless Microwave Techniques at Besser Associates and is an expert in microwave engineering). It is a discussion of how T-coils can be derived from a lattice network with two baluns and more importantly, how their 'ideality' (their frequency response) can be extended by including an extra shield wire.