Skip to content
Tutorials Center
My learning

Topics

Lesson 14 of 57 ยท Logic Gates

CMOS Gate Circuits, Logic Families and IC Scales

See how CMOS NAND, NOR, AOI and OAI gates are built from series and parallel transistors, compare RTL, DTL, TTL, ECL and CMOS logic families, and learn SSI to ULSI.

Logic Families and Timing explained noise margins, fan-out and delay. This lesson opens the box. You will see how a family's gates are actually wired, why CMOS gates have the shape they do, how the older families compared, and how chips are labelled by size, from a handful of gates to billions of transistors.

A short history of logic families

A logic family is a set of gate designs built with the same technology, so that its members share voltage levels and can be wired directly together. Early families were named after what was inside:

Family Meaning Notes
RTL resistor-transistor logic Very early. Resistors and bipolar transistors. Slow, poor noise margin.
DTL diode-transistor logic Diodes form the AND/OR function, a transistor inverts. Better than RTL, still slow.
TTL transistor-transistor logic Multi-emitter input transistors. Dominant from the 1960s. Fast, rugged, power-hungry.
Schottky TTL TTL with Schottky diodes The diode stops the transistor from saturating deeply, so it switches faster.
ECL emitter-coupled logic Transistors never saturate. The fastest bipolar family, but high power and small noise margin.
NMOS / PMOS single-polarity MOS Dense and simple, but they burn power when the output is low (NMOS) or high (PMOS).
CMOS complementary MOS Pairs of P and N transistors. Almost no static power. Today's choice.

Two ways to classify them are worth remembering for exams. Saturated bipolar families (RTL, DTL, TTL) let the transistor go deep into saturation, which costs storage time and so slows switching. Non-saturated families (Schottky TTL, ECL) avoid saturation and are faster. The other split is bipolar (RTL, DTL, TTL, ECL) versus unipolar or MOS (PMOS, NMOS, CMOS).

Rough comparison (typical figures; exact values depend on the part):

Property TTL ECL CMOS
Static power medium to high high tiny
Speed fast fastest fast, improving with each process
Noise margin good small very good (about 30-40% of the supply)
Packing density medium low highest
Weak spot power power, negative supply electrostatic discharge (ESD) damage, process variation

CMOS wins because most of the time a chip is not switching, and CMOS uses power essentially only when it changes state.

How a CMOS gate is wired

Every static CMOS gate has two networks that meet at the output:

  • A pull-down network (PDN) of N-type transistors connects the output to ground. An N transistor conducts when its gate input is 1.
  • A pull-up network (PUN) of P-type transistors connects the output to the supply. A P transistor conducts when its input is 0.

For every input combination exactly one network must conduct. The rule that guarantees it: the PUN is the dual of the PDN. Wherever the PDN has transistors in series, the PUN has them in parallel, and the other way round.

  • NAND: PDN has the N transistors in series (output goes low only if A and B are both 1). PUN has the P transistors in parallel (output goes high if either input is 0). 4 transistors.
  • NOR: PDN in parallel (either input being 1 pulls the output low). PUN in series. 4 transistors.
  • Inverter: one transistor of each kind. 2 transistors.

Series means AND, parallel means OR, when you read the PDN as "when does the output go low?". That is why a CMOS gate is naturally inverting: the PDN describes F', and the output is the complement of the PDN's condition.

This extends to compound gates. The AOI (AND-OR-INVERT) gate computes F = (AB + C)'. Its PDN is "A in series with B, that pair in parallel with C", and its PUN is the dual: "A in parallel with B, that pair in series with C". The OAI (OR-AND-INVERT) gate computes F = ((A + B)C)'. Each takes only 6 transistors, whereas building the same function from separate AND, OR and NOT gates would cost 8 to 12. That saving is why standard-cell libraries are full of AOI and OAI cells.

The simulator below describes each PDN as nested series and parallel groups, derives the PUN by swapping them, and checks every input combination.

Python A switch-level CMOS simulator

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

The assertion would fail if a PUN were not the dual of its PDN: you would get either a path from supply to ground (a short circuit) or an output connected to nothing (a floating output).

Chip size labels: SSI to ULSI

Integrated circuits are grouped by how much logic they hold. The labels date from the 1960s and 70s, and textbooks draw the borders a little differently, so treat the numbers as orders of magnitude.

Scale Name Typical content Example
SSI small-scale integration up to about 100 transistors (a few gates) a 7400 chip with four NAND gates
MSI medium-scale hundreds to a few thousand transistors decoder, multiplexer, 4-bit counter
LSI large-scale thousands to tens of thousands early microprocessors, small memories
VLSI very-large-scale hundreds of thousands to millions modern microcontrollers
ULSI ultra-large-scale tens of millions and beyond today's CPUs, GPUs, big memories

You also meet two other splits when reading about ICs. By technology, an IC is monolithic (everything fabricated on one piece of silicon, the normal case) or hybrid (separate chips and components on one substrate). By function, it is digital (gates, memories, processors), analog (amplifiers, regulators) or mixed-signal (converters).

The benefits of integration are the reason Moore's law mattered: small size, low cost per function, low power, speed (short wires), and reliability (fewer solder joints). The main drawbacks are that most ICs cannot be repaired, high-power parts need heat dissipation, and very small processes are expensive to design and sensitive to defects.

Python Label a chip by its transistor count

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

Note how the labels fall behind: today even a small microcontroller has more transistors than the "VLSI" label was invented to describe. The names survive mostly in textbooks and datasheets.

Check your understanding

Quick check ยท 5 questions

of 5 answered You scored / 5 ยท perfect!

  1. 1. In a static CMOS NAND gate, how are the N transistors of the pull-down network connected?

    The output must go low only when both inputs are 1, so the two N transistors are in series. The P transistors are in parallel.
  2. 2. Which statement about a CMOS gate's pull-up and pull-down networks is correct?

    Series and parallel are swapped between the two networks, which guarantees exactly one of them conducts for every input combination.
  3. 3. Which family is the fastest bipolar family because its transistors never saturate?

    ECL steers current instead of saturating transistors, making it fastest, at the price of high power and a small noise margin.
  4. 4. How many transistors does an AOI21 gate F = (AB + C)' need in static CMOS?

    Three inputs need three N and three P transistors: six in all.
  5. 5. A chip with about 20 million transistors would traditionally be called:

    Tens of millions of transistors and beyond is ULSI in the traditional naming.

Summary

  • Families evolved from RTL and DTL through TTL (and Schottky TTL, ECL) to CMOS, which dominates because it uses almost no static power.
  • Saturated families (RTL, DTL, TTL) are slower than non-saturated ones (Schottky TTL, ECL).
  • A static CMOS gate has a pull-down network of N transistors and a pull-up network that is its dual; series means AND and parallel means OR in the pull-down.
  • NAND and NOR use 4 transistors; AOI and OAI compound gates implement AND-OR-invert functions in far fewer transistors than separate gates.
  • ICs are labelled SSI, MSI, LSI, VLSI and ULSI by size; the boundaries vary and are mostly historical.

Next: Boolean Algebra Laws.