Do logic families use different type of transistors?

Question

Do the logic families such as TTL and CMOS represent only different logical structure? Or do they also use different type of transistors? Latest microprocessor chips with millions of components use CMOS. Is that only because CMOS "operating logic" is superior or also because CMOS family use entirely different transistors?

So far I understood is there are transistors which are used in analog circuits and there are tiny ones which are used in IC chips. I guess analog use TTL level transistors?

The transistors in ICs use many different logic families such as TTL and CMOS. So there are different transistor manufacturers for all these?

Answer 1

Here is an awfully incomplete answer :

By doping, oxyding, metallising a slice of sillicon, you can create on the surface several types of components : wires, bipolar transistors (aka BJT), metal-oxyde transistors (aka MOS), resistors, inductors, capacitors... (but it is often easier to create a transistor than these passive components).

Components are created at once by applying patterns and exposing the die. Transistors are not 'pick and placed', even for analog integrated circuits.

Bipolar transistors used in TTL gates are 'current controlled' and work very differently than MOS transistors which are 'voltage controlled'.

TTL (transistor to transistor logic) is traditionally defined as logic levels and I/O characteristics, which were (in the '70s-'80s) optimal for bipolar transistors working with 5V power supply. It is now possible to build with MOS transistors chips compatible with the TTL logic levels.

Logic gates can be created with bipolar and MOS transistors, but, nowadays, MOS is used almost exclusively for creating logic circuits like microprocessors, memories...

Bipolar transistors (and j-fets) are currently mainly used in analog components (for example operational amplifiers).

It is possible to mix bipolar and MOS on a single die, but the additional number of steps and constraints makes that technology more expensive and reserved to specialty component (for history, the first Pentium used BiCMOS which mixes bipolar and CMOS, this technology is nowadays used for analog or mixed signal components like ADC/DAC)

Answer 2

TTL logic uses bipolar transistors and CMOS logic uses, well, CMOS transistors. The structure of gates is also quite different between the two since the two types of transistors are so different and therefore need to be used differently.

Answer 3

Each logic family has different circuits for the same function (e.g. an AND gate), since each was an advance on earlier technology.

The earliest logic families used bipolar junction transistors.  Some examples are:

RTL (1963) - resistor transistor logic (used in the Apollo Guidance Computer)
DTL (1962) - diode transistor logic (used in the Minuteman II Guidance Computer)
ECL (1962) - emitter coupled logic (faster, used in the IBM 7030 Stretch Computer)
TTL - transistor-transistor logic (the most popular logic family prior to CMOS)
      74xx (1964) - original TTL line (also 54xx for military)
      74Sxx (1969) - used Schottky transistors for speed
      74Lxx (1964) - low power
      74LSxx (1976) - low power Schottky
      74ALSxx (1976) - advanced low power Schottky
      74Fxx (1979) - fast than normal Schottky

The following families used CMOS transistors:

CMOS - complementary metal–oxide–semiconductor logic
      CMOS (1970) - CD4000 series
      CMOS HC (1982) - high speed CMOS, used same pinouts as 74LS family
      CMOS HCT (1982) - CMOS logic but as TTL logic levels (made combining them possible)

CMOS has the advantage over bipolar in that it uses no power except when it is switching from 0 to 1 and back. The use of CMOS transistors has made it possible to cram millions or even billions of transistors on a single chip without requiring cooling,

The most popular integrated circuits are made by at least two manufacturers (and in many case several); this is called second-sourcing and is required by large companies and the military to assure parts will be available. However sometimes chips are made by only one manufacturer, and it is somewhat risky to incorporate them into a design.

Analog ICs can use either bipolar or CMOS transistors, whichever is more suitable to the task. They are by no means limited to the voltage supplies used by various logic families, be it TTL or otherwise. In addition, it is not uncommon for them to have a negative supply, e.g. ±15v.

The main difference between analog and digital, is the latter operates on two levels only, 1 and 0. For this reason, the transistors are turned completely off or on, the later is called saturated. Whereas the transistors in analog circuits generally operate in what is called the "active" region between on and off.

Answer 4

There's more to transistors than just the basic type (MOSFET vs. JFET vs. BJT). CMOS, by definition, involves two types of MOSFETs (P-channel and N-channel). And not every transistor of the same type will be identical. For instance, the microcontrollers I work on use two voltages -- 3.3V for the IOs, and 1.2V for the core logic. The core transistors are smaller and have thinner gate oxides because they don't need to handle as much voltage. In analog circuits, it's also common to change the width to length ratio.

MOSFETs are not ideal for analog circuits because they have a high gate capacitance and low transconductance. But they're better in digital circuits because CMOS logic doesn't consume any DC power[1]. Plus, it's turned out to be very easy to shrink MOSFETs, and in logic more transistors always wins. In mixed-signal ICs like microcontrollers, the analog circuits might use CMOS transistors just because that's what available. (Adding BJT would mean a bunch of extra processing steps, which costs money.) Aside from logic, MOSFETs are also used for switching large currents because of their low on-resistance.

BJTs have higher transconductance, but require base current to operate. In integrated circuits you can do weird things like have a BJT with multiple emitters or collectors, but I don't think you can get discrete devices like that.

JFETs are a sort of middle ground, with medium transconductance but minimal base current. They're commonly used to make the input stages of op amps.

In all cases, the transistors can be made differently to achieve the required voltage and current handling, switching speed, etc.

[1] This is a historical advantage. Modern CMOS transistors with gate lengths shorter than 100nm have significant DC leakage currents. On the newest processes this can be something like half of the chip's total power consumption.