PYTHON · NUMPY

Verify the instrument

Show that an energy monitor can be entirely dead while producing confident output, and that only a known load reveals it.

01Challenge

Here is a power trace from a real machine, taken while it was idle and again while ten CPU cores were pinned at full load. Decide whether the monitor built on this trace can be trusted.

Predict firstA laptop battery gauge reads 67.2 W at idle. Ten cores are then pinned at 100%. What does it read?

02Model

Before any energy claim, one question: does the instrument respond to a load you control. A battery gauge on a laptop read 67.2 watts at idle. Ten cores were then pinned at full tilt, and it read 67.2 watts. Not approximately — identically. Zero variance under a known load is the tell, and it is the only kind of failure that never announces itself, because a monitor built on that plane reports healthy forever and is never wrong in a way anyone notices. This is not a story about one bad gauge. It is the reason the first section of any energy result has to state the instrument, its noise floor against a known load, and the null control.

03Guided practice

Compute the variance under the known load, then decide the go/no-go. Write the check you would run before trusting any energy instrument.

04Feedback

PASS WHEN You rejected the instrument before it could produce a result.

If your run fails, check:

The gauge moved by exactly 0.000 W under a full-core load. It is not noisy — it is not measuring.
This failure is silent by construction: a monitor on a dead plane never raises an alarm, so it looks like a well-behaved system.
The rule this buys: no energy claim without the instrument, its noise floor against a known load, and a null control. All three, stated.

05Retrieve & space

What was the instrument's response to a full-core load, and why is that value damning rather than reassuring?
Why is a dead energy monitor harder to notice than a noisy one?
Name the three things every energy claim must state.

06Mastery & project

Run a known-load check on any energy instrument before using it, and report its response as part of your result rather than as a footnote.

Pick an energy instrument you have — a smart plug, a laptop gauge, a bench supply readout. Apply a load you can compute independently. Report whether it responded and by how much.

GO FURTHER · THE LIVE PLAYGROUND

PAI-280 · Where you put the meter ↗ · The plane-dependence result this discipline sits on.

← The leak you find with an IMU Who gets to define the method →

PYTHON · NUMPY

Verify the instrument

One instrument, two states. Decide whether it is fit for purpose before reading anything else from it.

You read

the arrays and values already in scope

You change

the code you write in each cell

Fixed

the dataset and the checks that grade you