Power-delay product: Visual walkthrough — Hardware deep dive

Before line one, let's agree on the picture we'll draw over and over.

Step 1 — Meet the gate as a switch and a bucket

WHAT: We replace the scary transistor gate with two switches and a bucket. Top switch (PMOS) connects the bucket to the supply; bottom switch (NMOS) connects the bucket to ground (0 V).

WHY: A CMOS Inverter never has both switches closed at once. To output logic-1 it closes the top switch and fills the bucket to $V_{D D}$ . To output logic-0 it closes the bottom switch and empties the bucket to ground. Everything about energy comes from watching this bucket fill and empty — so we strip away all the silicon and keep the bucket.

PICTURE: Look at the figure. The lavender box is $C_{L}$ (our bucket). The top coral switch reaches up to the supply rail at height $V_{D D}$ ; the bottom mint switch reaches down to ground at height $0$ .

Step 2 — Fill the bucket: how much charge goes in?

WHAT: Close the top switch. Charge flows from the supply into the bucket until the bucket's voltage equals $V_{D D}$ . The amount that fits is $Q = C_{L} V_{D D}$ .

WHY this formula and not another? A capacitor's defining property is linear: the charge it holds is directly proportional to the voltage across it. Double the height you fill to, and double the charge fits. The constant of proportionality is exactly $C_{L}$ — that's what "capacitance" means. So we don't guess; the definition hands us $Q = C_{L} V_{D D}$ .

PICTURE: The rising mint curve shows voltage on the bucket climbing from $0$ up to $V_{D D}$ over time. The shaded lavender area is the charge $Q$ that ended up inside.

Step 3 — Count the energy the supply pays out

WHAT: The supply sits at a fixed height $V_{D D}$ and pushes charge $Q$ up. Because the supply never sags, every coulomb it delivers is pushed through the full height $V_{D D}$ . So the energy the supply spends is $E_{s u ppl y} = V_{D D} Q = C_{L} V_{D D}^{2}$ .

WHY multiply voltage by charge? Energy = (push) × (amount pushed). In electricity, "push" is voltage and "amount pushed" is charge, so energy $= V \times Q$ . Since the supply's push is constant at $V_{D D}$ , we simply multiply — no integral needed, the height never changes. (Formally $E_{s u ppl y} = \int V_{D D} i d t = V_{D D} \int i d t = V_{D D} Q$ ; the integral just re-collects all the little charge packets $i d t$ into the total $Q$ .)

PICTURE: The tall coral rectangle is the supply's energy: a full-height box of area $V_{D D} \times Q$ . Notice it's a rectangle — full height for every bit of charge, because the supply is stubborn.

Step 4 — The famous factor of one-half: where half the energy hides

WHAT: Of that $C_{L} V_{D D}^{2}$ the supply paid, only $\frac{1}{2} C_{L} V_{D D}^{2}$ actually ends up stored in the bucket. The other $\frac{1}{2} C_{L} V_{D D}^{2}$ was burned as heat in the top switch while charging.

WHY only half is stored — the triangle vs. rectangle: The supply pushes every charge through the full height $V_{D D}$ (a rectangle). But the bucket fills gradually — the first charge drops into an empty bucket (voltage ≈ 0, tiny stored energy), the last charge lands on an almost-full bucket (voltage ≈ $V_{D D}$ ). Averaging, the bucket only stored charge through half the height on average. Half of a rectangle is a triangle — and $\frac{1}{2} C_{L} V_{D D}^{2}$ is literally the area of that triangle.

PICTURE: Overlay the coral rectangle (supply energy) and the mint triangle underneath (stored energy). The leftover region between them — the other triangle, shaded coral — is the heat burned in the switch. Two equal halves.

Step 5 — Empty the bucket: the fall burns the stored half

WHAT: Now open the top switch, close the bottom one. The bucket, holding $\frac{1}{2} C_{L} V_{D D}^{2}$ , drains to ground. Every joule that was stored is dumped through the bottom switch and turned into heat.

WHY nothing is "saved": On the way down there is no supply to catch the energy — the charge just falls to $0$ V. All $\frac{1}{2} C_{L} V_{D D}^{2}$ of stored energy is dissipated. This is why a capacitor's stored energy is not "free": you paid to fill it, and you burn it all when you empty it.

PICTURE: The bucket voltage curve now falls from $V_{D D}$ to $0$ ; the mint triangle that was "stored" is re-colored coral to show it becoming heat in the bottom switch.

Step 6 — The degenerate cases: why speed and resistance drop out

WHAT: We test the two extreme "what ifs" that usually break formulas: what if the switch is nearly a perfect wire ( $R \to 0$ )? and what if it's a terrible switch ( $R$ huge)? The answer: the burned energy $\frac{1}{2} C_{L} V_{D D}^{2}$ is identical in both.

WHY it survives every case: Fast charge = big current for a short time; slow charge = small current for a long time. The instantaneous heat is $i^{2} R$ , but the total heat integrates to the same $\frac{1}{2} C_{L} V_{D D}^{2}$ every time — the $R$ inside cancels against the time it takes. This is precisely why PDP is honest: it is blind to how fast (and thus blind to Propagation delay and frequency), so you can't cheat it by slowing down. (Contrast with Dynamic power dissipation $P_{a v g} = α C_{L} V_{D D}^{2} f$ , which does carry frequency $f$ — see Switching activity factor.)

Degenerate voltage case: If $V_{D D} = 0$ , then PDP $= 0$ — no push, no energy, no flip. Sensible. And halving $V_{D D}$ quarters the PDP (the $V_{D D}^{2}$ ), the core lever of Voltage scaling / DVFS.

PICTURE: Two heat curves — a tall narrow coral spike (low $R$ , fast) and a low wide lavender bump (high $R$ , slow). Different shapes, but the figure labels their areas as equal.

The one-picture summary

This final figure compresses all six steps: the supply rectangle (paid), split by the diagonal into a stored triangle and a heat triangle, with the fall dumping the stored triangle back to heat — total per flip = one triangle = $\frac{1}{2} C_{L} V_{D D}^{2}$ , immune to speed.

Recall Feynman: the whole walkthrough in plain words

Picture a bucket you fill from a water tower and then tip over, again and again. The tower pushes every drop up the full height, so it spends a full "box" of effort — but the bucket only holds half of that as real stored water; the other half splashes and warms things up as it slams in. When you tip the bucket, that stored half spills out and warms things too. So each single action — one fill, or one tip — costs exactly half a box: $\frac{1}{2} C_{L} V_{D D}^{2}$ . And here's the magic: whether you fill it in a gulp or a slow trickle, the same amount of warmth is made — the splashing loss doesn't care about speed. That's why the "energy per flip" is an honest number: you can't fake it by going slow. Fill higher (more voltage) and each flip costs more, growing with the square of the height. That's the Power-Delay Product, seen with your eyes.

Active Recall

What is the energy the supply delivers to fill

C_{L}

V_{D D}

C_{L} V_{D D}^{2}

(a full rectangle: full height × total charge).

Why does the bucket store only half of what the supply pays?

The supply pushes through the full height (rectangle) but the bucket fills gradually, so charge is stored through only half the height on average (triangle).

What is PDP per single transition?

\frac{1}{2} C_{L} V_{D D}^{2}

Why does switch resistance

R

not appear in the burned energy?

A slower charge spreads the same total

\frac{1}{2} C_{L} V_{D D}^{2}

over more time;

R

only sets the timing, not the total heat.

V_{D D} = 0

, what is the PDP?

Zero — no push means no charge moved and no energy per flip.

Halving

V_{D D}

changes PDP by what factor?

One quarter, since PDP

\propto V_{D D}^{2}