try this

# Temperature ( oC) = Vout * 100 oC/V

http://www.facstaff.bucknell.e...sonshtml/Sensors/TempLM35.html

We note that the LM35 puts out a voltage proportional (with no offset) to the temperature T in degrees Celsius. Since your micro's A/D can't convert negative values, this means your circuit will work only for temperatures above freezing.

As mentioned, the scale is 10 mV/degree. So

Vout = 0.010*T

Then for a 10-bit A/D converter with a 5 V reference,

code = 1023*V/5 = 1023*0.010*T/5 = 2.046*T

so

T = code/2.046

If we want to evaluate this using only integer math, then we might write T = (code*22)/45. That doesn't overflow with 16-bit math, since 22*1023 = 22507 < 32768. I chose that approximation by truncating the continued fraction, so there's a sense in which it's optimal. This doesn't really matter, as long as it's much better than other sources of error (for example, the tolerance on the voltage reference, often around 1%) in the system.