Spreading resistance — the principle

A KTY sensor is a single-crystal silicon die with a circular metal contact deposited on one face and a much larger contact on the opposite face. Current injected through the small contact spreads into the bulk silicon — the geometry imposes a 1/r resistance profile that integrates to a finite total. Because the silicon is lightly doped, its bulk resistivity is dominated by phonon scattering, which rises with temperature. The result is an inherently positive temperature coefficient that is almost linear across the practical operating range.

Two consequences fall out of this geometry:

  1. The resistance is set by the doping and the contact diameter, both of which can be controlled to better than 1 % in modern silicon fabs. KTY parts therefore have very tight part-to-part repeatability without trimming.
  2. Two KTY sensors in series doubles the resistance — useful when you want to compensate direction-dependent thermal asymmetry without adding complexity to the read circuit.

The KTY81 / 83 / 84 family

The numbering is historical, originating with Philips Semiconductors (now NXP). The three live variants in current production are:

PartPackageR25Operating RangeTC at 25 °C
KTY81DO-34 axial glass1 kΩ or 2 kΩ−55 to +150 °C+0.79 %/K
KTY83SOD-68 axial cylindrical1 kΩ−55 to +175 °C+0.76 %/K
KTY84DO-34 high-temperature glass1 kΩ at 100 °C−40 to +300 °C+0.61 %/K (measured at 100 °C)

The KTY84 deserves a separate mention. Its reference resistance is specified at 100 °C rather than 25 °C because the part is intended for embedded motor-winding use where the rest is well above room temperature. The +300 °C operating limit makes it the only sensor in the family suitable for large industrial machines and Class H insulation.

Why the curve is near-linear

In a typical NTC the resistance follows eβ(1/T - 1/T0) — exponential, with a curve that is steep at low temperature and flat at high. In a KTY the spreading-resistance physics gives roughly:

R(T) = R₀ · (1 + α₁·ΔT + α₂·ΔT²)

with α1 ≈ 7.9 × 10−3 /K and α2 ≈ 1.8 × 10−5 /K² for KTY81/120. The quadratic term contributes only a few percent of the total response across most of the operating range, which is what makes the curve appear near-linear.

The practical effect: a simple two-point linear interpolation in firmware gets you within ±1 °C across the full operating range — far better than any single-β NTC fit, and good enough that you rarely need to implement a polynomial in code.

Sensitivity comparison

At 25 °C a KTY81-120 changes by about 7.9 Ω per °C (0.79 % of 1 kΩ). An NTC 10 kΩ / 3 950 K changes by about 450 Ω per °C at the same temperature — nearly 60× more sensitive. So why use KTY?

Because that NTC sensitivity rapidly falls off the cliff. At 150 °C the same NTC has about 0.2 % per degree of change; the KTY81 still has 0.4 % per degree at the same point. Across a wide operating range KTY is much more uniform — less ADC dynamic range needed, less curve-fit complexity, more usable resolution everywhere.

Where KTY wins

  • Motor winding temperature feedback to a VFD — the drive’s firmware can do per-tick compensation arithmetic but it cannot afford a 32-bit Steinhart-Hart calculation. KTY linearity lets it run a simple lookup with < 1 °C error across the whole winding range.
  • Automotive ECU temperature inputs — the AEC-Q qualification process is significantly faster for KTY than for NTC because of the tighter part-to-part repeatability and the simpler curve fit.
  • Replacement of an obsolete RTD where cost matters — KTY costs about 1/5 of a comparable Pt100 transmitter chain, gives 1-2 °C accuracy, and integrates with a microcontroller's ADC directly without a separate excitation source.

Where KTY does NOT win

  • Sub-degree precision — if you need 0.1 °C, use an RTD with Steinhart-Hart-equivalent calibration.
  • Above 300 °C — silicon stops being a useful resistor; use Pt100, Pt1000 or a Type-K thermocouple.
  • Very high voltage isolation — the silicon die and glass package have lower dielectric strength than a fluoropolymer-encapsulated RTD. Use the latter when AC 2.5 kV+ is required.

Wiring

KTY parts are typically two-wire only — the higher reference resistance means lead error is proportionally small (a 5 m cable adds about 350 mΩ, or 0.05 °C of error against a 1 kΩ element). For long cables in motor controllers the same 3-wire and 4-wire topology described for RTDs can be used, but it is rarely necessary.

The KTY family is polarised — the small contact and the large contact have different series resistance characteristics. Datasheets specify the preferred current direction; reverse it and the part still works but the R(T) curve shifts slightly. Production builds should use a colour-coded lead pair (or a non-symmetric package like SOD-68) to enforce polarity.

Calibration grades

Standard KTY parts come in tolerance grades named by the numerical suffix in the part number. For example, KTY81-110 / 120 / 121 / 122 / 150 are all 1 kΩ / +0.79 %/K parts with progressively looser tolerance:

PartR25 toleranceTypical use
KTY81-110±1 %High-precision motor feedback
KTY81-120±2 %General motor winding
KTY81-121±2 % (selected)Automotive ECU
KTY81-122±2 % (selected)Industrial control
KTY81-150±5 %Cost-driven appliance

Sourcing in 2026

The original Philips/NXP KTY81/83/84 family went end-of-life some years ago; the parts are still manufactured by a number of Asian and European suppliers who have replicated the silicon design. Jianlu’s KTY series silicon temperature sensors are produced as drop-in replacements across all standard tolerance grades, with optional custom trimming and lead-forming for OEM production lines.