BeginnerQuestion 21 of 50

What is an isolation cell and why is it needed?

From PDVerse Low-Power Physical Design Mentor Guide, part of the pdVerse Mentor Guide

By Tabish Iqbal, who has 9+ years of VLSI Physical Design experience

Short Answer

An isolation cell is a small gate on a signal leaving a domain that can switch off, and it forces that signal to a known value while the domain is off. It exists because an unpowered driver leaves the net floating, and the always-on receiver would otherwise read garbage and leak current. The cell is powered from a supply that stays on and is controlled by an enable from always-on logic.

Schematic: a driver in the hatched off domain PD_COP feeds an AND clamp gate powered from VDD1p0, whose output goes to a receiver in the tinted always-on PD_MYCHIP; ISE comes from the always-on controller U_PC. — Technical Reference DiagramWhat is an isolation cell and why is it needed?

Technical Reference DiagramWhat is an isolation cell and why is it needed?

Technical Explanation

An isolation cell is usually an AND or OR gate: one input is the data, the other is the isolation enable.
With isolation off, data passes straight through; with it on, the output is clamped to 0 or 1.
The cell must run on an always-on supply; if it shared the dead rail, the clamp would die too.
Its enable comes from the always-on power controller and is asserted before the switch opens.
In IEEE 1801-2015, set_isolation (UPF) names the domain, ports, -clamp_value, -isolation_supply, -isolation_signal and -isolation_sense.
Legacy form (still accepted by ICC2/PT): -isolation_power_net on set_isolation (UPF), with the enable in set_isolation_control (UPF).
A missing cell leaves a floating input in live logic: crowbar current and random behaviour during shutdown.

Common Mistake

The Trap: Thinking isolation "blocks" the signal, so any gate will do.

The job is to substitute a specific safe value, and the wrong clamp value can fire a request or hold a reset in the live domain.
Pick the clamp per signal, from what the receiver treats as inactive.

Follow-up Question & Model Response

"Where does the isolation cell physically sit: inside the off domain or in the always-on domain?"

Candidate Model Response: Either can work, and -location self or -location parent on set_isolation (UPF) picks it. Inside the switchable domain the cell needs an always-on secondary supply, so you need a dual-rail isolation cell there. In the always-on parent it can use that domain's normal rail. The rule that never changes is that its supply must stay on while the source is off.

Practical Example

Design Scenario: (illustrative; ISO_AND_X1 is an illustrative cell name) PD_COP drives irq into PD_MYCHIP. An ISO_AND_X1 cell sits on irq, powered by VDD1p0, with its enable driven by U_PC/ISE through an inverter so the AND input is low during isolation. Before PD_COP shuts down, U_PC asserts ISE and irq is held at 0. The interrupt controller in PD_MYCHIP sees "no interrupt" for the entire off period instead of a floating line. On wake-up, U_PC releases ISE only after VDD1p0_SW is back and irq is valid again.

Low-Power & UPF Handbook