PYTHON · NUMPY

The one-volt race

Show that a comparator tolerance, not the control law, can decide where a machine's energy goes.

01Challenge

Two drives share a bus. Each has its own chopper with its own voltage threshold. Set the mismatch to zero and to one volt out of three hundred and eighty, and report where the braking energy lands.

02Model

Because the choppers watch a shared bus, the first one to fire pulls the bus voltage down and prevents every other chopper from ever reaching its threshold. A one-volt manufacturing tolerance therefore sends the entire machine's regeneration into a single resistor. Yaskawa's fix is not a supervisory optimiser — it is one wire: export each drive's regen-transistor state, tie them with a pull-up, and if any fires, all fire.

03Guided practice

Run the sweep. Note that the split is binary, not proportional — there is no partial sharing.

04Feedback

PASS WHEN You showed a tolerance, not a controller, deciding the outcome.

If your run fails, check:

The split is winner-take-all because the shared bus couples the comparators.
This is invisible at the mains and invisible in simulation unless the bus is modelled.
The industrial fix is a wired-OR — cheaper and faster than any supervisory scheme.

05Retrieve & space

Why does the lower threshold never fire?
What would a mains-side meter show as the mismatch changes?
Describe the one-wire fix and why it beats a supervisor.

06Mastery & project

Explain why component tolerance can dominate a control result, and name one place to look for it.

Find another shared-resource race in a system you know, where the first actor to act suppresses the others.

GO FURTHER · THE LIVE PLAYGROUND

Open the instrument ↗ — Two axes on a shared DC link. Move the meter, slide the phase, watch the number change.

← Three ways to lose the same joule 1/T³ and 1/T →

PYTHON · NUMPY

The one-volt race

Sweep the threshold mismatch and report the split.

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