hi accuracy with ad590

[evmav]

is it posible to obtain higher accuracy with ad590 and 16f877 (for example 0.05 degree C)
i'm using 10bit a/d , but i'm not sure what to do with +Vref and -Vref

[Acetronics2]

Hi

AD 590 is given for a +/- 0.5 °C precision ....

so, +/- 0.05°C do not mean aything here, also, do you really think you can make a current to voltage amplifier able to sense as small as 0.05 µA ...( 1µA/Deg )

so, it's simple : without special calibration +/- 0.5°C !!!
repetability is +/- 0.1°C : so do not try more than +/- 0.1°C, that with a fine calibration ...

Alain

[Warrier]

Hi:

Like Alain was saying the AD590 is trimmed for +/- 0.5 °C from actual value - so if you are trying to measure absolute temperature you cannot do any better. However,

if you are doing relative measurements (i.e. looking to measure changes with +/- 0.05 °C accuracy) you may be able to do that provided you understand ADC, Vref and amplifier limitations, component drift and PCB leakage paths. You would need a low input bias amp with as low a bias current drift as possible - an electrometer type. For the AD590 with a scale factor of 1uA/K, the resolution you are looking for is 50nA. Designing a current to voltage is not difficult - with a 100k feedback resistor 0.05 °C=5mV. - With Vref=5V your 10bit gives you an LSB=5mV and you can get about 50° max range. But keep in mind PIC doesn't really give you true 10bit digitization....
-warrier