Staying inside the limits

Here's something that surprises most people when they start designing circuits: getting the thing to work is the quick part.

Wiring an LED, a sensor or a motor so that it does what you want on your desk takes an afternoon. What professional engineers spend most of their time on is the question that comes after: will it keep working? On a hot day, with a slightly different batch of parts, when the battery is flat, when someone plugs in the wrong adapter, and for years rather than minutes.

Nearly all of that comes down to one discipline: making sure every component stays inside its limits for voltage, current and heat, under the worst conditions it will ever see. This article covers what those limits are, why heat is the one that catches people out, and how to find the numbers in a datasheet.

If you haven't read Electricity, explained with water yet, start there. This article builds on it.

Three ways to break a part

Every component has three kinds of limit. Go past any one of them and it fails, sometimes instantly, sometimes months later.

Too much voltage. Every insulator has a pressure it can hold back before it breaks down, like a pipe wall that bursts. Inside a chip, the insulating layers are only a few atoms thick. Put too much voltage across them and a spark punches straight through, leaving permanent damage. This is how microcontroller pins die when they're connected to something higher than their own supply. The Vector S3's power docs warn that its battery pin reaches 4.2 V and must never go to a GPIO, for exactly this reason.

Too much current. Every conductor has some resistance, so current through it makes heat. Push enough through a thin wire, a narrow circuit board trace or the hair-thin bond wires inside a chip, and it melts. That's exactly how a fuse works, deliberately. It's also how an LED without a resistor dies.

Too much heat. This is the big one, and it's really the other two in disguise. Whenever a component has current flowing through it and voltage across it at the same time, it turns power into heat: P = V × I. Parts are rated for a maximum internal temperature. Exceed it and they drift out of spec, then shut down, then fail.

Heat is the one that gets you

Voltage and current limits are easy to check: you compare a number against a rating. Heat is harder, because the temperature a part reaches depends not just on how much power it burns, but on how easily that heat escapes.

Conveniently, heat flow works just like Ohm's law:

Thermal resistance, written θ (theta), tells you how many degrees hotter a part gets for each watt it burns. A small package on a small board might be 200 °C/W or more. A big package on a heatsink might be 20 °C/W. The rule is:

temperature rise = power × thermal resistance
T_junction = T_ambient + P × θJA

The junction is the silicon inside the package, which is always the hottest point. Ambient is the air around the part. θJA is the thermal resistance between the two.

The classic place this bites is the linear regulator, the part that turns a higher voltage into a steady 3.3 V for your microcontroller. It works by acting as a variable resistor, burning off everything above 3.3 V as heat. Feed it 12 V and draw 300 mA, and it has to throw away (12 − 3.3) × 0.3 ≈ 2.6 W. In a small package, that's a temperature rise of over 200 °C.

A few things to take from that:

• The voltage difference is what costs you. From 5 V to 3.3 V, a linear regulator is fine for most hobby loads. From 12 V, it struggles with even modest currents.
• The package matters as much as the part. The same chip in a bigger package, or on more copper, runs far cooler. The metal tab on many power parts is there to carry heat into the board, so it needs a generous area of copper soldered to it.
• Your surroundings eat into the margin. A project that's fine on the bench at 25 °C can fail inside a sealed box in the sun, because it starts 35 °C closer to its limit.
• When the drop is large, change the approach. A switching regulator (often called a buck converter) converts voltage by switching rapidly instead of burning off the excess. It's typically 85 to 95% efficient. For 12 V to 3.3 V at 300 mA, that's around 0.1 W of heat instead of 2.6 W.

Datasheets: where the limits live

Every component made has a datasheet: a document from the manufacturer describing exactly what the part does and what it can take. They're free, and you'll find them on the manufacturer's site or linked from any electronics distributor's product page.

Get the manufacturer's own datasheet, not a summary on a marketplace listing. Listings often copy the most flattering number from the front page and leave out the conditions that make it true.

A datasheet can run to a hundred pages, but for staying inside the limits you need only a few sections. They're in almost the same order in every datasheet, whoever makes the part.

Three habits cover most of what goes wrong.

Absolute maximum is not a target

The absolute maximum ratings table lists what the part survives, not what it works at. Treat those numbers as a cliff edge you never go near, and design to the recommended operating conditions instead. The gap between the two tables is the margin the manufacturer left for you. Don't spend it on purpose.

Design to min and max, never typical

Parts vary. Two resistors from the same reel differ slightly. Two regulators from the same batch give slightly different outputs. The typical column shows where most parts land at room temperature. The min and max columns are the guarantees. If your circuit only works with typical parts, it will fail on some of the boards you build.

Read the conditions

Every number in the electrical characteristics comes with test conditions: a temperature, a supply voltage, a current. Change the conditions and the number changes. A MOSFET's on-resistance quoted at 10 V on its gate can be much higher at the 3.3 V a microcontroller gives it. The graphs further into the datasheet show how each figure moves with temperature, voltage and current, and they're often more useful than the tables.

Worst case, not best case

Checking a part against its limits means checking it under the worst conditions it can meet, not the conditions on your desk. Go through the list:

• Supply variation. USB is allowed to be anywhere from 4.75 V to 5.25 V. A single LiPo cell runs from 4.2 V when full down to about 3.0 V when flat.
• Part tolerances. Common resistors are ±1% or ±5%. Capacitors are often ±10% or ±20%. Small ceramic capacitors also lose a large part of their capacitance as the voltage across them approaches their rating.
• Temperature. Inside an enclosure, near a motor, or in direct sun, everything starts hotter and has less margin left.
• Start-up. Motors draw several times their running current while they get going, and a stalled motor can draw five to ten times as much. Capacitors draw a surge of current when power is first connected.
• Transients. Plugging in a power adapter can cause a brief overshoot. Switching off a motor or relay produces the inductive spike described in Electricity, explained with water. Static from your fingers reaches thousands of volts.

Here's the LED from Electricity, explained with water, checked properly. It's a red LED with a 330 Ω ±5% resistor on USB power. The worst case for current is the highest supply, the lowest LED forward voltage (1.8 V), and the resistor at the bottom of its tolerance (313.5 Ω):

I = (5.25 − 1.8) ÷ 313.5 ≈ 11 mA
P = I² × R = 0.011² × 313.5 ≈ 0.04 W

11 mA is comfortably under the LED's 20 mA rating, and 0.04 W is a sixth of a ¼ W resistor's limit. It passes with margin to spare. That's what a real design check looks like: short, but done at the edges.

Derating: leave yourself margin

Even staying inside the recommended limits isn't the whole story. Parts that run close to their ratings wear out faster. Electrolytic capacitors dry out, and components that run hot age more quickly. So engineers derate: they choose parts whose ratings comfortably exceed what the circuit needs.

The exact margins vary by industry, but these rules of thumb are a good start:

Margin costs very little. A ½ W resistor instead of a ¼ W one, or a 25 V capacitor instead of a 16 V one, costs pennies. A field failure costs far more.

A checklist for every part

When you add a component to a design, run through these questions. The answers come from your circuit on one side and the datasheet on the other.

1. Voltage: What's the highest voltage it will ever see, including spikes and a worst-case supply? Is that inside the recommended range, with margin?
2. Current: What's the most current through it, including start-up and faults? Is that inside its rating, and inside the rating of the traces and connectors carrying it?
3. Power: How much power does it burn (P = V × I), and what temperature does that give with the package and cooling you actually have?
4. Conditions: Do the datasheet numbers you're relying on apply at your voltage, current and temperature, or only at the test conditions?
5. Tolerances: Does the design still work with every part at the wrong end of its min and max?

Get in the habit of asking these for every part, and the circuits you build will stop failing in ways that look like bad luck. For the power supply side of the same thinking, see Powering your project.