Play

← fab24hr Blog

Chip basics8 min read

Why do chips use power? Leakage and standby current, explained

A chip uses power because it is full of tiny switches run by electricity. The power goes two ways. One is while it's working: every time a switch flips, it uses a little electricity. The other is while it's doing nothing: a switch is off, yet a tiny current still slips through. That's called leakage, or leakage current. The small current a chip draws from its battery while it's on but not doing anything is called standby current. A "power-saving" or "low-power" chip is one built to keep both of these down.

Where does the power go when a chip is working?

The most basic part of a chip is the transistor, a tiny switch run by electricity: on is 1, off is 0 (for what it looks like, see What is a transistor?).

Every time a switch flips, a little electricity is charged in and then let go, like filling a small cup and emptying it. So the more a chip does and the faster its switches flip, the more power it uses.

Voltage matters a lot: this part of the power is roughly proportional to the square of the voltage. Cut the voltage to 80 percent, and it drops to a little over 60 percent. That's why the first trick for saving power is often to lower the voltage.

Why does a chip use power when it's doing nothing?

Because its switches don't shut tightly.

Transistors are tiny. When the gate says "off", a little current still slips from the entrance to the exit, and some electrons go straight through the very thin insulating layer under the gate. That's leakage.

Think of a tap that doesn't quite close. While you're running water, a drip here and there doesn't matter. But let it drip all night, and by morning the bucket is full.

One switch leaks so little you could ignore it, but a chip has a huge number of them, and a phone chip has more than ten billion, so it adds up. And the power for switching is only used while the chip is working; leakage goes on every second the power is on.

The smaller the switches, the bigger the problem. From around the 2000s, leakage grew so large that the voltage couldn't come down much further, which is one reason Moore's law slowed down (see What is Moore's law?). With each new generation, like the ones the news calls "2 nm" (see The news says "2 nm"), engineers have had to find new ways to make the switches shut tightly.

Why do faster chips usually leak more?

How much voltage it takes to turn a switch on is decided when the chip is made. This is called the threshold voltage.

A low threshold means the switch opens at a nudge and runs fast, but it doesn't shut as tightly, so it leaks more. A high threshold shuts tightly but opens more slowly. So "fast" and "frugal" are often a trade-off.

Chips made from the same design don't come out exactly alike, either: the ones that came out a bit fast usually leak more. And the hotter a chip is, the more it leaks.

What is standby current?

It's the current a whole chip draws from its power supply while it's on but not doing anything. At that moment the switches are hardly flipping, so what's left is mostly leakage.

Power used while workingLeakage while doing nothing
When it happensWhen switches flipEvery second the power is on
What it depends onHow much work, how fast the switching, how high the voltageHow tightly the switches shut, how many there are, how hot it is

Why do hearing aids, watches and sensors care so much about standby current?

Because they have two things in common: small batteries, and a chip that is often waiting, in part or as a whole.

Many sensors are asleep most of the time; now and then they wake up, take a reading, send it off and go back to sleep. An electronic watch, too, spends most of its time just keeping time. Waiting takes far longer than working, so standby is often a big share of all the power they use.

A hearing aid is a small device that helps people hear more clearly. While it's worn, it is always picking up and processing sound, but not every part of the chip is busy all the time, and the parts that aren't busy keep leaking. It often runs on one small zinc-air battery, which lasts from a few days to a week or two, depending on the battery size, the model and how many hours a day it's worn.

How long a battery lasts is roughly how much charge it holds, divided by how much is used on average at any moment. Use 10 percent more on average, and it runs out about 10 percent sooner. Every bit more a chip leaks, the sooner the battery is gone.

How do designers make a chip use less power?

In plain terms, these are the common tricks:

  • Pause the parts not in use: a chip has a signal like a metronome, called the clock, and its switches flip in time with it. Stop the beat for a block that isn't needed, and that block saves its switching power; but it still has power, so it still leaks.
  • Cut the power to whole blocks not in use: switch the power off entirely, and the leakage is saved too. The cost: turning a block back on takes a little time, and anything it needs to remember has to be saved before it goes off.
  • Use fast switches only where they count: where speed is a must, use low-threshold switches; where there's no hurry, use high-threshold ones that shut tightly.
  • Keep the voltage as low as possible, and let the chip sleep most of the time: when full speed isn't needed, lower the voltage and run a little slower; wake the chip only when there's work, and put it straight back to sleep after.

How do fabs and test houses keep leakage in check?

How tightly a switch shuts is also a matter of how it's made. How thick the insulating layer under the gate is grown (some are grown in a furnace) and how many atoms are added to the silicon (often by ion implantation) both affect the threshold and the leakage. If a step comes out a little off, chips can come out a little faster and leak a little more. So before wafers leave the fab, the fab runs WAT, measuring current and leakage on test keys to see whether the batch came out off target.

At the test-and-assembly house, every single chip takes an exam (how it works: How are chips tested?). The exam paper is the test program, and one of its questions is standby current: put the chip in a state where it does nothing, measure how much current it draws, and stop it if it's over the limit, sorting it into a separate bin.

Where to draw the limit is a trade-off. Too tight, and good chips are stopped too, so yield falls. Too loose, and chips that leak too much reach the people using them, and the batteries don't last.

Testing asks how much a chip leaks now; whether it will fail after some use is a different question (see What is burn-in?).

In fab24hr

fab24hr's main story, The life of a chip, follows Glowmere, a fictional hearing-aid chip. It runs on a single button battery, so what everyone fears most is leakage in standby.

In chapter 10, "Standby Current", you're the test engineer at the test-and-assembly house. The first production wafers have just arrived, and the most common failure in the bins is standby current too high, gathered in a ring around the wafer's edge, right around the limit. Leo, the customer's product engineer, says their ads carry one line: "One battery, several more days." Whether to tighten the limit, loosen it, or set it from the R&D line's FF corner numbers (if you didn't have the FF corner run in chapter 1, that option isn't there) is your call; then you write it into the test program.

The finale is chapter 13, "Into His Hands". You're Mr. Shen. Your old hearing aid uses a lot of power: the battery you changed two days ago beeped twice again this evening. Your daughter brings home a new one: "It uses very little power: the same battery lasts several days longer." At half past ten at night, when the old one would have gone quiet long ago, your wife speaks in the kitchen, and you hear her from the living room. If you loosened the limit in chapter 10, three months later some hearing aids get sent back because their batteries last several days less than promised.

Tap "Clock in", choose "A day in the fab", tap "Test-and-assembly house" on the "Your role today" card and swipe to the test engineer: the morning shift is the chapter 10 shift. In the afternoon shift, you also decide whether to screen out three chips whose standby current is within the limit but stands well above their neighbors'.

The companies, chips, people and numbers in the game are fictional; "several more days" is a line from the story, not a figure for any real product. The parts about leakage and standby current come from public, general sources. For other jargon, see Chip fab jargon in plain English.

Where you'll meet it in the gameIn chapter 10 of the main story, The life of a chip, "Standby Current", you're the test engineer at the test-and-assembly house: the bins show the most common failure is standby current too high, and you decide where to draw the limit and write it into the test program. In chapter 13, Mr. Shen puts on a new hearing aid, and his daughter says the same battery lasts several days longer.

In fab24hr, a tighter standby current limit costs yield; a looser one means some hearing aids' batteries don't last. Where the limit goes is your call.

Play chapter 10, Standby Current, and decide where the limit goes

Teachers: this article can go with "fab24hr in the classroom" in class.

More articles

More on this site