Digital system design

Table of contents

1. Introduction
2. Circuits
3. Logic gates
   1. Inverters and buffers
   2. OR gates
   3. AND gates
   4. NOR gates
   5. NAND gates
   6. XOR gates
4. Multiplexers
5. Decoders
6. Clock
7. Flip flops
8. Field Programmable Devices
9. References

Introduction

Digital systems represent information using discrete signals, typically bits. This is opposed to analogue systems, which represent information using continuous signals.

The main reason for using binary is that it’s simple to represent 0 and 1 as high/low voltage (which also makes binary systems tolerant to noise).

As of 2020, microprocessors are some of the most complicated digital systems and so this section focusses on microprocessors.

A microprocessor is made up of multiple subcomponents (e.g. MMU, an instruction sequencer, caches). These subcomponents are made of the same building blocks: wires and transistors. Transistors are joined together to make larger abstractions (e.g. logic gates and flip-flops) that are used to build increasingly complex circuits [1, Pp. 2-3].

An integrated circuit (chip) is a set of electronic circuits on a single piece of semiconductor material (normally silicone). A chip contains a number of connections to the outside world, many of which are connected to a low-voltage DC power supply (usually between 1-5 volts). The energy provided by the power is what drives the processor as it moves electric charge from one place to another. In the process energy is dissipated as heat [1, P. 3].

A clock signal that drives the processor is generated on the motherboard and then sent to the chip [1, P. 4].

Circuits

Circuits can be broken into two types:

• Combinational logic (CL) circuits
• Sequential logic (SL) circuits

CL circuits take inputs and combine them to produce a single output after a small propagation delay. The output is only a function of the inputs (they are pure functions) [1, P. 8].

SL circuits have state—the output depends on the current input and memory values. Registers are an example of SL circuits.

Most digital systems boil down to registers separating combinational logic. Registers hold inputs stable for the clock cycle, once the CL circuit evaluates a new value, it’s captured in a register. After the clock edge, the register sends out the new value on an output wire or bus.

Logic gates

A logic gate is a circuit with one or more input signals and only one output signal [2, P. 19].

Logic gates are normally restricted to a maximum of four input signals for performance reasons [3, P. 4].

Logic gates are the building blocks for more complex digital logic circuits. Normally only a subset of logic gates are used to create a processor’s circuits [3, P. 5].

NOR and NAND gates are known as universal gates because they can be used to build any other logic function without another gate type [4, Pp. A-8].

Inverters and buffers

An inverter (NOT gate) is a gate with one input signal that outputs the opposite signal [2, P. 19].

A small circle (bubble) on the output side signals an inverter [2, Pp. 19-20].

A buffer is a circuit that takes an input signal and outputs the same signal. A buffer can be constructed by cascading two inverters. Buffers can be used for isolating two different circuits [2, P. 20].

OR gates

OR gates take two or more inputs and return high if any of the inputs is high, or low if none are high [2, P. 20].

AND gates

AND gates take two or more inputs and return high if all of the inputs is high, or low otherwise [2, P. 22].

NOR gates

NOR gates have two or more inputs. The output is high if all inputs are low, otherwise the output is low [2, P. 32].

NAND gates

NAND gates have two or more input signals. The output is low if all input signals are high, otherwise the output is high [2, P. 34].

XOR gates

XOR gates (exclusive-or) have two or more inputs. For the binary case, the output is high if only one input is high otherwise it is low [2, P. 37].

In the general case, an XOR gate output is high if an odd number of inputs are high and low otherwise [2, P. 37].

Multiplexers

A multiplexer (mux) is a circuit with multiple inputs and only one output [2, P. 58].

When a multiplexor has only two data inputs, the selector is a single bit that selects one of the inputs to be the output if it is true and the other if it is false [4, Pp. A-10].

With $n$ data inputs there must be $\vert log_2 n \vert$ selector inputs [4, Pp. A-10].

Read ports are often implemented using muxes [4, Pp. A-53].

Decoders

A decoder is a logic block that has an n-bit input and $2^n$ outputs. Only one output is asserted for each input [4, Pp. A-9].

A logic component called an encoder performs the opposite action, converting $2^n$ input lines into an n-bit output [4, Pp. A-9].

Write ports are often implemented using decoders [4, Pp. A-53].

Clock

Processors are implemented as synchronous digital circuits, where a clock signal is used to regulate the signals throughout the system by controlling when state elements should be updated (as opposed to asynchronous circuits which are clockless or self-timed) [1, P. 5].

A clock signal is a square wave signal [2, P. 93].

A clock cycle (tick) is the time from any place on the waveform to the same place on the next instance of the waveform. Normally it’s measured from the rising-edge of the signal to the next rising-edge (rising meaning rising from low to high) [1, P. 4].

The rate at which the clock runs at is measured in hertz (Hz). Modern processors often have clock speeds measured in gigahertz (GHz).

The clock must have a long-enough period so that all signals in combinational logic blocks stabilize before the clock edge samples those values for storage in the state elements [4, Pp. A-48].

Systems suffer from clock skew where two different state elements receive the clock signal at slightly different times (due to longer paths for one than the other). Designers must minimize clock skew by routing the clock in a way that minimizes difference in arrival time [4, Pp. A-73].

A clocking methodology defines “when signals can be read and when they can be written”. A clocking methodology makes hardware predictable by avoiding the situation where a value could be written to at the same time as it’s read [4, P. 261].

A simple example of a clocking methodology is positive edge-triggered clocking. In positive edge-triggered clocking, values stored in sequential logic elements are only updated on a rising clock edge (a transition from low to high) [4, P. 261].

The setup time is the minimum amount of time that an input must be valid before the clock edge [4, Pp. A-52].

The hold time is the minimum time that an input must be valid after the clock edge [4, Pp. A-53].

Clock-to-Q is the time it takes for a signal to propagate through a flip-flop so that the flip-flop reaches a stable state [4, Pp. A-72].

Asynchronous devices (like I/O devices which often have their own clock) can communicate with a CPU through a series of handshaking steps [4, Pp. A-75].

Flip flops

A flip-flop is a circuit that has two stable states. It remains in one of these states until triggered into the other [2, P. 90].

Computers use flip flops to regulate the flow of electricity through the system [2, P. 93].

Registers can be built from multiple flip-flops.

The clock signal is sent to each flip-flop in a computer to control when a flip-flop changes state [2, P. 93].

Field Programmable Devices

Field programmable devices are integrated circuits containing combinational logic (and possibly sequential logic) that can be configured by a user [4, Pp. A-77].

There are two types of FPDs:

• Programmable logic devices (which only contain combinational logic)
• Field programmable gate arrays (which contain combinational logic and flip-flops)

[4, Pp. A-77]

In FPDs, gate and register locations are static, but the connections can be configured [4, Pp. A-77].

One way to implement FPDs is by using an SRAM, which is downloaded at power-on, to control the settings of switches (meaning the FPD can be reconfigured multiple times) [4, Pp. A-77].

References

1. [1] J. Wawrzynek, “CS61c: Introduction to Synchronous Digital Systems.” 2007.
2. [2] A. Malvino and J. Brown, Digital Computer Electronics, 3rd ed. 1993.
3. [3] J. Wawrzynek, “CS61c: Representations of Combinational Logic Circuits.” 2007.
4. [4] L. Hennessy J and A. Patterson D, Computer Organization and Design: The Hardware / Software Interface: ARM Edition. 2017.