Have you noticed? Electronic boards used to be dominated by Pt100/Pt1000; today it is almost always NTC.
And yet “on paper” NTC looks like the worse choice.
Pt100/Pt1000 sensors (the PTC family) are accurate and, above all, linear: a direct match between resistance and temperature. A Pt100 reads 100 Ω at 0 °C and rises by 0.385 Ω/°C … zero point three eight five … (the Pt1000 with coefficients ×10). Read 103.85 Ω → 10 °C. Simple.
NTC sensors are the opposite: resistance falls as temperature rises (hence the “N” for Negative), with a strongly NON-linear curve and complicated formulas. In practice you need conversion tables, which differ for each of the many models available. Typical value: ~10K at room temperature.
So: less linear, less “convenient”.
So why are they now used almost exclusively?
1. They cost less
2. The acquisition circuit is far simpler => it costs less
“But these characteristics were just as true 20 years ago, so what changed?”
The electronics changed. Today every board carries powerful microcontrollers, and that has flipped the trade-off.
Handling non-linearity with analogue circuits or limited MCUs used to be hard: better an inherently linear probe, even at the cost of more refined signal conditioning.
Today linearising in software (with calculations or lookup tables) is trivial: you may as well save on the probe and the circuit.
There is also a purely practical installation benefit: the very high working value (10K) makes cable length and cross-section practically irrelevant.
With a Pt1000 the wiring makes itself felt; with a Pt100, throwing the reading off by a few degrees (upwards, of course…) is the norm. Four-wire measurement exists, of course — but it costs even more.
That is why, in temperature control, NTC sensors are now the only ones left.
Enjoy your work!