The diode: a one-way valve for current
Plumbing has a part that lets water flow one way and slams shut the other: a check valve. Electronics has its twin, and it is the most fundamental switch in all of power conversion — the diode. Push current through it in the *forward* direction and it conducts, dropping only a small, almost-fixed voltage across itself. Try to push the other way and it becomes a wall, blocking with only a whisper of leakage. A diode has no control terminal; it decides for itself, purely from the polarity of the voltage you hand it. That makes it the dumbest switch in the box — and, for turning AC into DC, exactly the right one.
That "small forward drop" matters more than beginners expect. A silicon PN-junction diode drops about 0.7 V when conducting; a Schottky diode, built from a metal-semiconductor junction instead, drops only 0.3–0.4 V and switches off faster. In a 5 V supply, losing 0.7 V twice in a bridge is a real tax. So the first design choice in any rectifier is *which diode* — and that choice is a trade between forward drop, reverse blocking voltage, and how fast the device recovers when current reverses.
IDEAL DIODE REAL SILICON DIODE
I | I | /
| | | / (slope = 1/r, the
| | (on, V=0) | / on-resistance)
───┼──────┼──── V ───┼─────/──── V
| | /|
| (off, I=0) | / |
| | 0.7V ← turn-on knee
on OR off, no loss real: a knee + a slope + leakageThe bridge rectifier: folding AC into pulsating DC
A single diode only lets the positive half of an AC wave through and throws the negative half away — *half-wave* rectification, which wastes half your energy and leaves big gaps. The clever fix is to take the negative half and fold it up to be positive too. Four diodes wired in a diamond — the bridge rectifier — do exactly this. On the positive half-cycle, two diodes route current to the load one way; on the negative half-cycle, the *other* two diodes route the now-reversed source current through the load in the *same* direction. The load never sees a negative voltage. You get *full-wave* rectification: every hump of the original wave, all pointing up.
AC in Full-wave bridge
~ ┌──►|──┬──►|──┐
/\ D1 ►| │ D1 │ D3 │
/ \ ┌────────────────┤ ●──────┤────● +Vout
────/ \──┤ AC source │ │ │ │
\ / └────────────────┤ ●──────┤────● 0V (load here)
\ / D2 ◄| │ D2 │ D4 │
\/ └──|◄──┴──|◄──┘
v(load): /\ /\ /\ /\ ← every hump up,
/ \ / \ / \ / \ twice per AC cycleWorked example: smoothing capacitor and ripple
Those humps are still bumpy — useless to a chip that wants a flat rail. So we hang a fat capacitor across the output. Think of it as a water tank fed by a pump that only pulses: while each hump is rising the capacitor charges up to the peak; when the hump falls away, the capacitor takes over and feeds the load by itself, sagging gently until the next hump arrives to top it up. The leftover sag is the ripple voltage, and you can estimate it with one beautifully simple formula.
The capacitor's defining law is I = C·(dV/dt). During the gap between humps, the load pulls a roughly constant current I from the capacitor for a time Δt, so the voltage droops by ΔV ≈ I·Δt / C. For a full-wave bridge on 60 Hz mains, the humps arrive every 1/(2·60) = 8.3 ms, so Δt is at most that gap. Let's put numbers on it.
GOAL: 12 V rail, load draws I = 0.5 A, on 60 Hz mains.
Full-wave → ripple period T_r = 1/(2·60) = 8.3 ms
Pick a target ripple of ΔV = 0.5 V (peak-to-peak).
ΔV ≈ I · Δt / C (capacitor discharge)
I · Δt 0.5 A · 0.0083 s
⇒ C ≈ ──────── = ──────────────── ≈ 8300 µF
ΔV 0.5 V
Round UP to a standard 10000 µF / 25 V cap.
──────────────────────────────────────────────
Rule of thumb (full-wave, 60 Hz):
ΔV(volts) ≈ I(amps) / (120 · C(farads))The thyristor: a latching valve you can trigger
The plain diode conducts whenever physics says so — you don't get a vote. The thyristor, or SCR (silicon-controlled rectifier), adds one priceless thing: a *trigger*. It blocks in both directions until you send a brief pulse into its gate; the instant you do, it latches fully on and conducts like a diode. The catch — and it's a defining one — is that you can't gate it *off*. Once latched, it stays on until the current through it naturally falls to zero. On AC, that happens at every zero-crossing for free, which is exactly why thyristors and AC are made for each other.
Now combine that trigger with the bridge and something powerful appears: phase control. If you wait — delaying the gate pulse until partway through each half-cycle — the thyristor only conducts the *late* portion of each hump. Fire early and the load gets nearly the full wave; fire late and it gets just a sliver. You have built an adjustable DC supply with no moving parts, dialing the output voltage purely by *when* you choose to trigger. This is how megawatt motor drives, electroplating baths, and the front-ends of HVDC links set their DC level.
Phase-controlled hump (one half-cycle):
v | ╭───╮ fire EARLY → most of
| ╭─╯ ╰─╮ the area delivered
| gate╱ ╲
| pulse│ ╲___
└───────┼──────────────► ωt
α (firing angle)
small α → big DC output (nearly full-wave)
large α → small DC output (thin sliver)
Adjustable DC, set purely by WHEN you trigger.The power MOSFET: the controllable workhorse
The power MOSFET is the switch every modern converter is built around. Like a wall light switch you hold a finger on, it's *fully* commandable: put a voltage on the gate and it turns hard on; remove it and it turns hard off — no waiting for a zero-crossing, and it does it in tens of nanoseconds. The gate is insulated, so steering it costs almost no steady current. That combination — voltage-controlled, fast, and self-latching only when *you* say so — is why a power MOSFET can switch hundreds of thousands of times a second. Rungs 3–6 do nothing but flip these on and off in clever patterns to build buck and boost converters.
But a power MOSFET is not the ideal switch of rung 1 — and the gap between ideal and real is where the entire craft of power electronics lives. Two non-idealities dominate, and you will fight both for the rest of this track:
- On-resistance (Rds(on)). When fully on, the MOSFET isn't a perfect wire — it's a small resistor, maybe a few milliohms to an ohm. The conduction loss is I²·Rds(on), paid the whole time current flows. Carry 10 A through 10 mΩ and you burn 1 W as heat, continuously. Lower Rds(on) is the headline spec on every power MOSFET datasheet.
- Switching loss. A real MOSFET can't snap between on and off instantly. During each transition there's a brief window where it carries high current *and* drops high voltage at the same time — and their product is wasted power. That energy is paid *every single switching event*, so switching loss grows with frequency. Switch faster for a smaller converter, and you pay more here: the central tension of the whole field.
The toolbox, ranked
Step back and the three devices form a clean ladder of *control*. The diode decides for itself. The thyristor lets you say *go* but not *stop*. The MOSFET lets you say both, fast. Each step up in control buys you flexibility and costs you complexity — and you pick the cheapest device that does the job, never more.
device | turn ON | turn OFF | typical job
───────────┼──────────────┼──────────────┼────────────────────
diode | by itself | by itself | rectify AC→DC
thyristor | gate pulse | wait for I=0 | phase-controlled DC,
| | | mains-frequency only
MOSFET | gate voltage | gate voltage | high-freq switching
| (fast!) | (fast!) | (buck/boost, SMPS)
───────────┴──────────────┴──────────────┴────────────────────
control: none → trigger-only → full command, both edges