I remember making a thermostat control using just a normal silicon diode as the sensor, it was to replace a thermostat in a fridge, so i used a relay controlled by this little circuit which became so sensitive that as soon as you open the fridge door the relay would cut in within about a second, at first I suspected that the fridge light bulb coming on generated a spike that tripped the circuit on, so I tried suppressing for any spikes on my low voltage regulated power supply that supplied 12v dc to my circuit through a small transformer and regulator, but it still tripped each time you opened the fridge door, so then I tried another method to rule out that it wasn't definitely the bulb tripping the circuit on, so I took the bulb out, and it still tripped the relay, it was hyper sensitive, it did not have any hysteresis, but then I thought could the door switch also supply power to something else, so i eliminated this possibility by taping over the door contact switch, and again each time you open the door and more or less instantly the relay tripped in, the temperature control was hyper sensitive, so I had to enclose the diode sensor inside a plastic sleeve (heat shrink wrapping) so as to delay its thermal response, this stopped the relay coming on instantly and delayed it by about 10 seconds, but it still sensed an increase in temperature by a minute degree, so I then had to implement some hysteresis such that it won't trip on again until temperature rose by a full degree, so in other words 5a very closed loop feedback can also become nuisance and hence why TRVs do a reasonable job though their response could be shortened a little, so that occupants don't feel too cold for too long and nor too hot for too long.