A transition-edge sensor balances exactly on the knife-edge between superconducting and normal states, where a tiny photon energy causes a massive spike in resistance.
But if you drive it with a constant current, that resistance spike is fatal: power is I²R. More resistance means more heat, which means more resistance, a runaway spiral that destroys the edge.
If you drive it with a constant voltage, the math inverts: power is V²/R. The exact same resistance spike now chokes the power, cooling the sensor back down to equilibrium automatically.
The stability isn't a property of the sensor at all. It's a property of the bias. The choice between current and voltage is the choice between positive and negative feedback: whether the resistance sits in the numerator or the denominator.
