So as I thought it was completely backwards, I did as you said and decremented after each loop of displaying data, what you said makes sense that the pointer is already pointing at 52H but thats not the case in my circuit, I feel that its pointing at 53H for starters because when I send data and decrement it displays "||" "0" "7" . So that showed me that the left most bit is in 50H so I decremented R1 three times before displaying data and incremented inbetween loops and I end up with "6" "7" "0" which matched theoutput of the LM34 slighty. When I tried to heat up the sensor it jumps in 10's, so 67, 77, 87 for example, no accuracy at all and when watching the output of the sensor with a DVOM as it rises above 72 it switches to displayed temp from 67 to 77. Does this have to do with the step size or the conversion? For instance the conversion for BIN-DEC is divide by ten which only leaves 6,7,0(per count of 3 times correct?) for the DEC-ASCII conversion are those OR'd by 30H according to my code and when the 6 is OR'd it shows the number 53 which in ascii is the number 5 which is not good when the ouput of the LM34 is 670mV. I know that with my other kit that uses a ATmega1XX the steps for the adc were 1024 I think using the same sensor and it was way more accurate but used C and I understand C less that assembler, so basically I'm asking if I change the step size and my conversion math will I get more accuracy. EDIT: Just to be clear I meant the step size referring to the adc and the Vref input.

Thanks
James