M7 · A loss budget and a plan to measure it

Goal

• Read a converter’s losses by kind, across the load and the switching frequency, and say which one limits the efficiency where.
• Check one loss by hand against the tool.
• Plan the bench work that would confirm the budget: what to measure for each loss, with which instrument, and what limits the measurement.

Before you start

• Read the loss budget and the bench pages on probes, current sensing and thermal design.
• The buck example converts 24 V to 12 V at 24 W and 100 kHz.

Steps

1. The budget

Open the loss budget with the buck example:

▶ Open in the loss budget: Buck loss-budget example
Preset from the example loss-buck

It simulates the converter at each load with the duty ratio adjusted until the output is at its set value, and sorts the losses of the simulated period by kind.

2. One loss by hand

The gate drive does not depend on the load. With a gate charge of 10 nC at a drive of 10 V, work it out (gate-drive power) and compare it with the tool’s bucket. Enter it below. The driver and the gate resistances dissipate it, not the switch.

3. Full load and light load

Name the largest loss at full load and at a tenth of the load, and say why they differ: which losses follow the square of the current (conduction loss), and which are set by the switching frequency alone.

4. The switching frequency

Double $f_s$ and watch each bucket and the efficiency. Say which buckets doubled, and what the higher frequency buys in return (a smaller ripple for the same inductance, or a smaller inductance for the same ripple).

5. What the tool leaves out

The simulated transitions are instantaneous, so the overlap loss of a hard-switched transistor is not in the budget [Erickson & Maksimović 2020, Ch. 4], and the core loss is an estimate from coefficients measured with sinusoidal excitation [Erickson & Maksimović 2020, Ch. 10]. Decide how your plan would see both.

6. The bench plan

For each loss, write down the quantity you would measure, the instrument, and the limit that matters: the probe’s rise time against the switching edges (probe rise time), the burden and the pad resistance of a shunt, a meter that averages a pulsed current, the temperature rise of a part. End with how you would measure the efficiency itself.

Acceptance criteria

• The buck example's gate-drive loss, worked out by hand.
  Your value (W)Check
  Show the example's value

  10 mW (from loss-buck: Buck loss-budget example)

• The largest loss at full load, and why it is the largest there.
• The largest loss at a tenth of the load, and why it differs from the one at full load.
• What doubling $f_s$ does to each loss and to the efficiency, and what it buys.
• The losses the budget leaves out or only estimates, and how your plan sees them.
• A bench plan with, for each loss, the quantity, the instrument and the limit that matters, and how you would measure the efficiency.

Clear this mission

Your ticks stay in this browser only: no account, and nothing is sent.

Gotchas

• The efficiency at full load hides the light-load picture. Losses that do not fall with the load dominate there, so read the whole curve.
• A sense chain that clips on the peaks of a pulsed current reads a lower average than a meter across the same shunt (a current reading below the meter).
• A probe with a long ground lead rings with its own capacitance and adds its own edge to the one you measure (probes).

Go deeper

• Book chapter Steady-State Equivalent Circuit Modeling, Losses, and Efficiency (Ch. 3) — R. W. Erickson, D. Maksimović · retrieved 2026-09-23
  DC transformer models and how winding and semiconductor resistances limit the conversion ratio and efficiency.
• Application note AN-1144: Measuring Output Ripple and Switching Transients in Switching Regulators — Analog Devices · retrieved 2026-09-24
  How a probe's ground lead exaggerates switching transients, and the tip-and-barrel method for output ripple.

Quiz

1. The load current of a converter halves. Its conduction loss in a fixed resistance:

   1. Halves
   2. Falls to about a quarter
   3. Stays the same
   4. Doubles
   Answer and explanation

   Correct answer: B Conduction loss is the rms current squared times the resistance: half the current, a quarter of the loss (a little more, since the ripple does not shrink with the load).

2. Which losses stay the same when the load falls to a tenth?

   1. The conduction losses
   2. The diode's forward-drop loss
   3. The gate drive, and the capacitive loss at turn-on in CCM
   4. None
   Answer and explanation

   Correct answer: C The gate drive depends on the gate charge, the drive voltage and $f_s$; the capacitive loss on the node capacitance, the voltage at turn-on and $f_s$. Neither follows the load current, so at light load they dominate.

3. A gate charge of 20 nC is driven to 12 V at 250 kHz. What is the gate-drive loss?

   1. 60 mW
   2. 6 mW
   3. 0.24 mW
   4. 600 mW
   Answer and explanation

   Correct answer: A 20 nC times 12 V is 240 nJ per cycle, times 250 kHz: 60 mW, in the driver and the gate resistances.

4. The switching frequency doubles. Which losses double with it?

   1. Conduction
   2. Gate drive and capacitive turn-on
   3. Only the core loss
   4. All of them
   Answer and explanation

   Correct answer: B Both are an energy lost once per cycle, times the number of cycles per second. The core loss depends on the frequency and on the flux swing, which shrinks with the ripple, and the conduction loss follows the currents.

5. A measurement system has a bandwidth of 100 MHz. About what rise time does it show for an ideal step?

   1. 35 ns
   2. 3.5 ns
   3. 0.35 ns
   4. 10 ns
   Answer and explanation

   Correct answer: B For a single-pole response the rise time is about 0.35 over the bandwidth: 3.5 ns. An edge much faster than that shows at the system's own rise time.

6. Why can a loss budget from an ideal-transition simulation be too optimistic for a hard-switched converter?

   1. It ignores the overlap of voltage and current during each transition
   2. It ignores the conduction loss
   3. It counts the gate drive twice
   4. It is not: simulations always overestimate losses
   Answer and explanation

   Correct answer: A With instantaneous transitions, the switch never carries current while it blocks voltage, so the overlap loss is missing and must be estimated or measured separately.