Hades logoHades applet banner
Gate vs. wire delay demonstration

applet icon

The image above shows a thumbnail of the interactive Java applet embedded into this page. Unfortunately, your browser is not Java-aware or Java is disabled in the browser preferences. To start the applet, please enable Java and reload this page. (You might have to restart the browser.)

Circuit Description

One of the most important aspects of electronics circuits modelling concerns if and how to describe circuit timing. On the lower abstraction levels, the circuit behavior can be modelled very accurately via systems of differential equations for the voltages and currents in the devices. Those equations can include temperature and other environmental effects, and even circuit aging or electromigration. Unfortunately, solving the differential equations is computationally very expensive and is impractical on large circuits. Also, the level of detail created by such simulations is extremely high. For example, a digital designer usually is not interested in the actual voltages at the gate outputs, as long as the voltage is within the bounds for the logical low or high values at the required time.

Therefore, other modelling techniques have to be used on the higher abstraction levels. The simplest scheme for clocked circuits is simply to count clock cycles and to ignore all timing details within a clock cycle. This implies that timing information, e.g. to determine the maximum clock rate, will have to be created with other tools. However, such cycle-based (in German: periodenorientiert) simulations can be made to run very fast on standard workstations and use (comparatively) little memory. Therefore, cycle-based simulation has recently become the preferred method to simulate large circuits like modern microprocessor systems.

Naturally, each logical gate needs some time before its output stabilizes after input changes. This motivates the gate-delay model, where each component in a circuit is assigned a set of delay values. Those delays can be taken from actual device measurements or from device simulations. For example, a two-input NAND gate in a 1.0 micron CMOS process might be assigend a single delay value of 5 nanoseconds.

However, the new gate output value also needs some time to travel across the output wire before it reaches the next gates. This wire propagation delay can be computed accurately once the actual wire layout of a chip (or circuit board) is known - which is not true during the early design stages. Therefore, the wire delays are often approximated by the number of components connected to the wire and a coarse estimate for wire length between components.

The example applet demonstrates those delay models:

  • The circuit in the middle uses just the default gate-delay model. Open the property sheet of the AND gate to change its delay.

  • The circuit on top just connects a switch to a LED component, but there are a few 'delay-nodes' inserted into the wire to demonstrate a wire-delay. You can toggle the input switches before the previous value has reached the target gate, an effect that is similar to the real traveling waves in the physical modelling of a transmission line.

    (Note: due to its simulation algorithm, Hades cannot really model wire-delays. The small delay-nodes are actually just standard simulation components, each of which uses the gate-delay model. Still, you should get the idea of electrical waves traveling along the signal wire as the source of delay).

  • Finally, the circuit at the bottom combines both delay-models. This allows for the most accurate modelling, but it is also obviously the most complicated.

Until very recently, slow devices were the dominant source of delays in most technologies (discrete, TTL, and early MOS-circuits). Therefore, the gate-delay model was and is very popular, and it is also the delay-model used in Hades. However, the miniaturization of CMOS-devices at increasing chip sizes has reduced the ratio of gate- to wire-delays. In sub-micron CMOS-technology the wire-delays are often the dominant source of delay, and elaborate heuristics are used to estimate wire delays at early design stages, long before actual circuit layout data is available.

Print version | Run this demo in the Hades editor (via Java WebStart)
Usage | FAQ | About | License | Feedback | Tutorial (PDF) | Referenzkarte (PDF, in German)
Impressum http://tams-www.informatik.uni-hamburg.de/applets/hades/webdemos/12-gatedelay/10-delaydemo/gate-vs-wire-delay.html