How Electronic Components Are Manufactured

Most product founders think about electronics at the board level: you pick parts, someone lays out a PCB, a factory assembles it. How the parts themselves come into existence stays invisible until something goes wrong — and then it matters a great deal, because the reason a capacitor lost 60% of its capacitance under DC bias, or a connector failed after 200 mating cycles, is buried in how that component was made.

This article walks through the four families that make up almost every bill of materials — semiconductors, passives, connectors, and cable assemblies — from the factory floor up. You do not need to become a component engineer, but you do need to understand why the datasheet says what it says.

Semiconductors: the fab and the package

Every chip in your product starts as a slice of purified single-crystal silicon. Modern fabs work on 200 mm and 300 mm wafers, which are polished flat to a degree that is hard to describe and then run through several hundred process steps.

The front end

Front-end processing builds the transistors through repeated cycles of deposition, photolithography, and etch. A layer goes down, photoresist is spun on, light through a patterned mask exposes it, the unexposed resist washes away, and an etch removes material everywhere the resist is not. Add doping steps that implant impurities, and repeat — dozens of times, each layer aligned to the one below within nanometers.

Two consequences matter to you. First, a fab cycle takes roughly two to three months of continuous processing, which is why semiconductor lead times cannot be compressed by shouting. Second, the enormous fixed cost of a mask set means custom silicon only makes sense at very large volumes; everyone else buys standard parts, which is why choosing the right microcontroller is a sourcing decision as much as a technical one.

Test, dice, and package

Finished wafers are electrically probed die by die, and failures are inked or mapped out. Yield at this step — the fraction of good die per wafer — is the single biggest driver of what a chip costs. The wafer is then thinned and diced into individual die.

Packaging is a separate industry, usually in a separate country. The die is attached to a leadframe or substrate, connected by fine gold or copper wire bonds or by solder bumps in flip-chip designs, encapsulated in molding compound, plated, marked, and tested again. Package choice is not cosmetic: it sets thermal resistance, board area, assembly difficulty, and whether a rework station can ever touch it — which is part of why DFM for electronics exists.

Passives: the parts nobody thinks about until they fail

A typical connected product has three to five hundred passive components and maybe a dozen active ones. Passives are made in staggering volume by a small number of manufacturers, and their production method explains almost everything about their behavior.

Multilayer ceramic capacitors (MLCC)

An MLCC is built by screen-printing metal electrode patterns onto thin ceramic dielectric sheets, stacking hundreds of those sheets in alternating orientation, pressing the stack, cutting it into individual chips, and firing the whole thing at high temperature so the ceramic sinters into a monolithic block. Terminations are dipped on and plated.

That construction produces three properties that surprise people:

• DC bias derating. Class II dielectrics like X5R and X7R lose capacitance as voltage is applied across them — often 30% to 70% of nominal at rated voltage in small case sizes. Your 10 µF decoupling capacitor may be delivering 3 µF in circuit. Class I dielectrics such as C0G/NP0 do not do this, but store far less charge per unit volume.
• Mechanical fragility. The fired ceramic is brittle and rigidly bonded to a flexing board. Board bend during depaneling, connector insertion, or a drop cracks capacitors, and cracked MLCCs fail as intermittent shorts — one of the most miserable field failures to diagnose.
• Microphonics. Class II ceramics are piezoelectric, so a capacitor in a switching supply can literally sing.

Resistors

Thick-film chip resistors — the overwhelming majority — are made by screen-printing a resistive paste onto an alumina substrate, firing it, then laser-trimming a notch into the element until the measured resistance hits target. That trim is why 1% tolerance costs barely more than 5%, and why the temperature coefficient is a separate specification from tolerance. Thick film typically runs 100 to 250 ppm per degree Celsius; thin-film resistors, sputtered rather than printed, reach 10 to 25 ppm and cost several times as much. If your product measures something precisely, the tempco of a divider resistor may dominate your error budget.

Inductors and magnetics

Wound inductors and transformers are made by winding wire on a ferrite or powdered-iron core. They are labor-intensive, geometry-sensitive, and the part of the BOM most likely to vary between suppliers claiming equivalence. Saturation current, not rated current, is usually the number that matters.

Connectors and cable assemblies

Connectors are precision assemblies of stamped metal contacts and molded plastic housings. The contacts are progressively stamped from a strip of beryllium copper or phosphor bronze, plated — commonly with nickel underplate and a gold flash on the mating surface — and inserted into a housing molded to tight dimensional tolerance. The stamping and plating steps are the entire quality story: too thin a gold layer and the contact resistance climbs after a few hundred cycles; the wrong base alloy and the spring force relaxes at temperature.

Cable assemblies are the least automated part of most products. Wires are cut, stripped, crimped, and loaded into housings, and crimp quality is verified by pull-force testing and cross-sectioning. This is manual work, which means it is the assembly most likely to arrive wrong, and why harness and connector design that survives assembly pays for itself immediately.

What the grades actually mean

The AEC-Q family is worth understanding even if you are not building a car. AEC-Q100 covers integrated circuits, Q101 discretes, Q200 passives. Qualifying a part to those documents means the manufacturer subjected sample lots to a defined battery of stress tests — temperature cycling, humidity bias, high-temperature operating life, mechanical shock — and committed to strict process change notification. That last item is the real value: an automotive-grade part cannot have its internal construction quietly changed on you. Products that plug into vehicles inherit these expectations, as covered in automotive electronics development.

Why this affects your product

Three practical consequences follow from all of the above.

Availability is a manufacturing constraint, not a purchasing one. When a fab or a passive plant is allocated, no amount of ordering earlier helps, which is why planning a product around component shortages belongs in the design phase.

Parts have lifecycles. Semiconductor lines get retired, and a part you designed in three years ago goes end-of-life with a last-time-buy notice. Having a plan for a component going end-of-life before it happens is the difference between a redesign and a crisis.

Grey-market parts are physically different objects. Because component manufacture is concentrated and demand spikes are sharp, brokered supply attracts remarked, recycled, and outright fake parts. The countermeasures in keeping counterfeit components out of your supply chain are worth reading before you ever need them.

Projects House selects, sources, and qualifies components for connected products, including derating analysis, second-source planning, and lifecycle review. If you want a second opinion on a BOM before it goes to a factory, send it through our contact form.