Bipolar Junction Transistor Amplifiers:Differences between Discrete and IC Amplifier Design

Differences between Discrete and IC Amplifier Design

Prior to the early 1960s, electronic circuits were constructed from discrete component circuits. Discrete resistors, capacitors, inductors, and transistors are packaged individually to be connected with wires or printed circuit board (PCB) conductors to create a functioning electronic circuit. The discrete component circuit is still required in certain applications in the 21st century, but the IC is now the dominant form

of implementation for an electronic system. There are major differences in the design principles used for the IC and the discrete circuit as the following section will explain [4].

Component Differences

Discrete resistors, capacitors, and inductors are available in a very large range of values. Although many amplifier circuits are designed with element values of 5–10% tolerances, very precise values can be obtained at a higher cost. The range of resistor values extends from a few W to many MW and capacitor values range from fF to hundreds of mF. Metal core or air-core inductors are available over a wide range of values as also are transformers with a wide choice of turns ratios.

For most standard fabrication processes, the IC chip becomes too large to be useful when the total resistance of resistors on the chip exceed some value such as 100 kW. The same applies if the total capacitance exceeds perhaps 50–100 pF. It is only in the last decade that inductors could be fabricated on a chip. Even now, integrated inductors with limited values in the nH range are possible and losses may limit Q values to the range 2 to 8 [5].

A major problem with IC components is the lack of precise control of values. Resistors or capacitors are typically fabricated with an absolute accuracy of around 20%. It is possible, however, to create resistive ratios or capacitive ratios that approach a 0.1% accuracy, even though absolute value control is very poor in standard processes.

The cost of a discrete circuit is determined by different variables than that of the IC. This results in a design philosophy for ICs that differs from that of discrete component circuits. Before ICs were available, a designer attempted to minimize the number of transistors in a circuit. In fact, some companies estimated the overall component cost of a circuit by multiplying the number of transistors in the circuit by some predetermined cost per stage. This cost per stage was determined mainly by the transistor cost plus a small amount added to account for the passive component cost. The least expensive item in the discrete circuit is generally the resistor which may cost a few cents.

In an IC, the greatest cost is often associated with the component that requires the most space. The BJT typically requires much less space than resistors or capacitors and is the least expensive component on the chip. An IC designer's attempt to minimize space or simplify the fabrication process for a given circuit generally results in much larger numbers of transistors and smaller numbers of resistors and capacitors than the discrete circuit version would contain. For example, a discrete circuit bistable flip- flop of 1960 vintage contained two transistors, six resistors, and three capacitors. A corresponding IC design had 18 transistors, two resistors, and no capacitors.

Matching of components in IC design can be considerably better than discrete design, but also presents unique problems as well. Components that are made from identical photographic masks can vary in size and value because of uneven etching rates influenced by nearby structures or by asymmetrical processes. Dummy components must often be included to achieve similar physical environments for all matched components. Certain processes must be modified to result in symmetrical results. Metal conductors deposited on the IC chip can introduce parasitics or modify performance of the circuit if placed in certain areas of the chip, thus care must be taken in placing the metal conductors. Of course, these kinds of considerations are unnecessary in discrete circuit design.

While discrete circuits can use matched devices such as differential stages, often single-ended stages can be designed to meet relatively demanding specifications. Discrete components can be produced with very tight tolerances to lead to circuit performance that falls within the accuracy of the specifications. Circuit design can be based on the absolute accuracy of key components such as resistors or capacitors. Although it may add to the price of a finished circuit, simple component-tuning methods can be incorporated into the production of critical discrete circuits.

Since the absolute values of resistors and capacitors created by standard processes for ICs cannot be determined with great accuracy, matching of components and devices is used to achieve acceptable performance. Differential stages are very common in IC design since matched transistors and components are easy to create even if absolute values cannot be controlled accurately.

Some analog circuits do not require high component densities while others may pack a great deal of circuitry into a small chip space. For low-density circuits, more resistors and small capacitors may be used, but for high-density circuits, these elements must be minimized. The next section considers circuits that replace resistors and capacitors with additional BJTs in IC amplifier building blocks. The BJT current mirror, which is quite popular in the biasing of IC amplifiers has been discussed in Chapter 10 of this handbook. This circuit is also used to replace the load resistance of a conventional amplifier to make it easier to integrate on a chip.

Parasitic Differences

One advantage of an IC is the smaller internal capacitance parasitics of each device. A proper layout of an IC leads to small parasitic device capacitances that can be much less than corresponding values for discrete devices. Although the discrete and IC device may be fabricated with similar processes, the collector and emitter of the IC device are typically much smaller and usually connect to fine pitch external leads/ connectors. The discrete device must make external connections to the base, collector, and emitter and these external connection can add a few picofarads of capacitance. An on-chip IC device generally connects to another device adding minimal parasitic capacitance. However, external connects from the chip to the package must be carefully designed to minimize parasitic capacitance as signals are brought off chip.

Another difference relates to the parasitic capacitance associated with IC resistors and inductors. These elements have capacitance as a result of the oxide dielectric between metal layers and the chip substrate. The frequency response of IC resistors and inductors must be considered in critical designs.

At higher frequencies (above a ~1 GHz) other considerations influence the IC design. To maintain the

advantage of low cost in IC circuits over discrete realizations, the devices on the chip are made much smaller than their discrete counterparts. Current and power levels must be limited in the smaller devices to avoid gain roll-off due to high-level injection effects or overheating. The lower currents necessitate higher impedance levels in most bipolar chip designs.

For many analog circuits, terminations of 50 W are used at I/O ports. This requires large current magnitudes to reach the specified voltage levels. For example, a 20 mA peak current would be required to develop a peak voltage of 1 V. This is not a difficult problem in discrete circuit design using PCBs. The limitation on current and power in the chip design leads to less flexibility with higher impedance interconnects to develop acceptable voltages. High-frequency chip designs must also deal with bond wire or other package inductance in the range 0.1–0.5 nH and the package capacitance that may reach 5 pF. A buffer is often needed to generate the off-chip signals with sufficient current to develop the specified voltages, but IC chip heating due to power dissipation of the buffer stage must be carefully considered. Common thermal centroid techniques are used to prevent unbalanced heating of matched devices.