Visual walkthrough — Dynamic vs static power consumption

This page rebuilds the headline formula of the parent note

$P_{\text{switching}}=\alpha C{V}_{dd}^{2}f$

from absolutely nothing — no calculus, no physics assumed. We will earn every letter with a picture before it appears. By the end you will not just remember the formula; you will see where each symbol comes from and why it sits where it sits.

If you have met the switch itself before, it lives in CMOS Inverter Design; the "bucket" lives in Capacitance in VLSI. You do not need either to follow this page — we build both from zero.

Step 1 — The switch and the bucket

WHAT. We draw the object we are studying: one CMOS gate. Ignore the transistor internals. What matters is that a gate drives a wire and the next gate's input, and all of that behaves like a bucket that holds electric charge.

WHY this picture. Power is energy per second. Energy in a digital gate goes almost entirely into filling and emptying this charge-bucket. So before any formula, we must agree on what fills and what it fills into.

PICTURE. In the figure, the tank on the right is the bucket. The tap at the top connects it to the supply rail $V_{dd}$ (a fixed "high water pressure"). The drain at the bottom connects it to ground (pressure zero). The switch decides which pipe is open.

The one relationship tying them together — the definition of the bucket — is:

$q=C\cdot V$

• $q$ — charge currently sitting in the bucket (coulombs)
• $C$ — the bucket's size (farads); a fixed property of the wire and gate
• $V$ — the pressure across the bucket right now (volts)

So to raise the bucket's pressure all the way up to the supply pressure $V_{dd}$, you must pour in a total charge:

$Q=C\cdot V_{dd}$

Hold on to that number $Q=C{V}_{dd}$. It is the "full bucket" amount.

Step 2 — What one flip costs the supply

WHAT. We flip the switch so the output goes from 0 to 1. The tap opens, charge flows from the supply into the bucket until the bucket pressure equals $V_{dd}$. We now ask: how much energy did the supply spend?

WHY this question. The supply is the thing draining your battery. It doesn't care where the energy ends up — it only knows how much charge it pushed and at what pressure. Energy from a constant-pressure source is simply pressure × charge pushed.

PICTURE. The left bar is the supply, always pushing at the full pressure $V_{dd}$ (it never sags). Every coulomb it hands over is handed over at that same full pressure — that is the shaded rectangle.

$E_{\text{supply}}=V_{dd}\cdot Q=V_{dd}\cdot (C{V}_{dd})=C{V}_{dd}^2$

• $V_{dd}$ — the supply pressure, constant the whole time, so it comes straight out of the product
• $Q=C{V}_{dd}$ — the full-bucket charge from Step 1
• $C{V}_{dd}^2$ — the two $V_{dd}$'s multiply: one is the pressure, one is hidden inside $Q$

That squared voltage is our first sighting of the famous $V_{dd}^2$. Remember its origin: one factor is pressure, one factor is how much charge that pressure had to move. Same knob turned twice.

Step 3 — Where half the energy vanishes

WHAT. The supply spent $C{V}_{dd}^2$. But how much energy is now actually stored in the bucket, ready to be reused? We compare "spent" with "kept."

WHY this matters. If everything the supply spent got stored, nothing would turn into heat and chips would never get warm. The gap between spent and stored is exactly the heat. We must find that gap.

PICTURE. Two rectangles overlaid. The tall one (area $C{V}_{dd}^2$) is what the supply paid — charge delivered at full pressure the entire time. The triangle underneath is what the bucket kept — because the bucket's own pressure started at 0 and only climbed to $V_{dd}$ as it filled. The bucket felt an average pressure of only $V_{dd}/2$.

$E_{\text{stored}}=\frac{1}{2}Q{V}_{dd}=\frac{1}{2}C{V}_{dd}^2$

• $Q$ — same full-bucket charge
• $\frac{1}{2}{V}_{dd}$ — the average pressure the bucket felt while filling (not the full $V_{dd}$)
• $\frac{1}{2}C{V}_{dd}^2$ — exactly half of what the supply paid

The missing half is burned as heat in the switch's resistance as charge scrapes through it:

$E_{\text{heat, charge}}=E_{\text{supply}}-E_{\text{stored}}=C{V}_{dd}^2-\frac{1}{2}C{V}_{dd}^2=\frac{1}{2}C{V}_{dd}^2$

Step 4 — The empty stroke costs just as much

WHAT. Now flip back: output 1 → 0. The tap closes, the drain opens, and the bucket dumps its stored $\frac{1}{2}C{V}_{dd}^2$ down to ground. We track where that energy goes.

WHY include this. A real gate doesn't just fill once — it fills and empties, over and over. If we stopped at Step 3 we'd be counting only half a cycle and would undercount the heat.

PICTURE. The bucket drains through the lower switch. None of the stored energy returns to the supply — it all rubs off as heat in the pull-down switch's resistance on the way to ground.

$E_{\text{cycle}}=\underset{\text{charging heat}}{\underbrace{\frac{1}{2}C{V}_{dd}^2}}+\underset{\text{discharging heat}}{\underbrace{\frac{1}{2}C{V}_{dd}^2}}=C{V}_{dd}^2$

• first term — heat in the pull-up switch during the fill
• second term — heat in the pull-down switch during the empty
• $C{V}_{dd}^2$ — the two halves add to a clean whole

So: one full flip of one gate costs $C{V}_{dd}^2$ joules of heat. That is the atom of dynamic power.

Step 5 — From "per flip" to "per second"

WHAT. Energy per flip is nice, but power is energy per second. We convert by asking: how many flips happen each second?

WHY. Your cooling system and your battery care about the rate of heat, watts (joules per second), not the cost of a single event.

PICTURE. A clock ticking $f$ times per second. Each tick is a chance to flip. Below it, the same clock but the gate only actually flips on some ticks — the busy ticks are marked.

If the gate flipped on every tick, power would be energy-per-flip times ticks-per-second:

$P=E_{\text{cycle}}\cdot f=C{V}_{dd}^{2}f$

• $E_{\text{cycle}}=C{V}_{dd}^2$ — joules per flip, from Step 4
• $f$ — flips per second (if it flipped every tick)
• product — joules per second = watts

Step 6 — The honesty knob: activity factor

WHAT. Real gates do not flip every tick. A gate holding a steady value sits idle even while the clock ticks. We shrink the count to only the ticks where a flip really happens.

WHY. Without this correction we would massively overestimate power. Most gates in a chip are quiet most of the time.

PICTURE. Ten clock ticks in a row; only two of them show a flip. The fraction lit up is the activity factor.

Scaling the flip-count by $\alpha$ gives the full switching-power law:

Notice the shapes of the dependence, straight from the derivation:

• $V_{dd}^2$ is quadratic — halve the voltage, quarter the power. This is why Dynamic Voltage Frequency Scaling (DVFS) leans on voltage first.
• $f$ is linear — double the clock, double this power.
• $\alpha$ is a discount — quiet chips run cooler for free.

Step 7 — Edge cases: pushing the knobs to their limits

WHAT. A formula you trust must survive its extremes. We drive each knob to zero or to its ceiling and check the picture still makes sense.

WHY. If the equation gave nonsense at $\alpha =0$ or $f=0$, we'd know a step was wrong. Passing these tests is how we earn belief in the result.

PICTURE. Four mini-panels, each a degenerate setting, with the surviving power value.

Step 8 — Two worked numbers

WHAT. We push real values through the law so the units become concrete.

WHY. Symbols only click once you watch pico and giga cancel into milli.

The one-picture summary

Every step folds into a single chain: bucket → fill cost → half lost → both strokes → per second → real flips.

Test yourself

Related vault stops: Thermal Design Power (TDP) (where these watts must be dumped as heat), Subthreshold Slope (the leakage half of the power story), and Amdahl's Law (why raw $f$ isn't the only speed lever worth its power).