LM35 temp sensor reading (10°C off)

STIX · Joined: 17 Jun 2006 Posts: 12

Hi:

I connected a LM35 temperature sensor to my PIC16F877A MCU to take temperature readings. The following is my code:

SherpaDoug · Joined: 07 Sep 2003 Posts: 1640 Location: Cape Cod Mass USA

Put a voltmeter on the LM35 to see if it is a LM35 problem or a PIC problem. Also try putting the sensor in a glass of ice water to give a larger range of test data.
_________________
The search for better is endless. Instead simply find very good and get the job done.

Steve H · Guest

LM35's also like to oscillate - many times this oscillation looks like a reading that is way off and/or not particularly stable. Check the output with a scope or put the voltmeter that Sherpa suggested on AC millivolts - there should be no AC noise on the output of the LM35.

The data sheet shows how to prevent oscillation by the addition of a large C and a resistor.

HTH - Steve H.

caduhitec · Joined: 06 Feb 2007 Posts: 3

use this

{
delay_ms (500);
value = read_adc ();
value1 = (value*49+1)/100;
printf ("Temp:%5.1fC\n\r", (float)value1);
}

STIX · Joined: 17 Jun 2006 Posts: 12

jma_1 · Joined: 08 Feb 2005 Posts: 147 Location: Wisconsin

Greetings,

Try explicitly casting your result as an int32. Your multiplication of a 10 bit adc by 5 and 100 -> 1023 * 5 * 100 = 511500. The maximum value of an int16 is 65536. Alternatively, declare value1 as an int32.

When debugging code, examine the basics. Print out the adc value of your channel only (without the formula applied to convert v to c). Is the value correct? Apply a known voltage. Does the voltage match the adc reading? This well help determine if your channel is setup correctly (correct reference voltage, channel, etc).

Do you need to try and convert the integer into a float for displaying it on the lcd? You will never have a fraction portion of the number (integers only).

Please post your results.

Cheers

BrunoCF · Joined: 14 Aug 2007 Posts: 2 Location: SP - Brasil

I had a the same issue. Guess what was the problem: the power suply of my PIC was 4V instead of 5V... :P All my calculations were made for 5V, so they were all wrong.

E_KARVELAs · Joined: 14 Feb 2007 Posts: 46 Location: Greece & Cyprus

HI
I was looking for same issue with u with lm35 and pic16f877a , i have find the answer.
take a look of my code for me have been worked wery well.
I think if u try the command i have pointed u with arrow it will work !!!

Ttelmah · Guest

I'm sorry, but it is rare to see so many bad bits in one section of code....

First, *=16, is needed, if he is intending to write a larger main program, using more memory locations latter.
Second, using 'ADC_CLOCK_INTERNAL, is a bad decision. It will make the results _worse_, since it introduces extra randomness into the noise. Read the chip's data sheet. Note what it says under table 11-1, with regards to using the internal RC clock. If this made your code work, are you sure that you were using the right divider dfrom the processor clock...
Third, all you are doing, with your RTC code, is adding a long pause between the readings. If this fixed your problem, it suggests that you were not allowing long enough between the readings (the capacitor inside the ADC in the PIC, takes _time_ to charge, and is only charging when you are not reading). The existing code being tried, has a good delay present.

Now, the original poster, says he has put a large resistor and capacitor on the output. This is _not_ what is needed. The manufacturers recommended damping, when significant capacitance is present, is 75R, _in series_ with 1uF to the ground rail. Try this first. The PIC's ADC input, presents a lot of capacitance when it connects and switches during the conversion. Unless your voltmeter is able to see signals lasting for only a few uSec, it wont detect these spikes, and this could easily give the odd behaviour.
I the reading is then still consistently off, you really sould look doubly carefully at the supply rail. Again, it might look 'good' on a DVM, but if noise spikes are present on this, this would give spurious results. What damping is on this rail, how close to the PIC?.

Best Wishes

E_KARVELAs · Joined: 14 Feb 2007 Posts: 46 Location: Greece & Cyprus

metalm · Posted: Wed Jan 02, 2008 5:20 am

The LM35 don't output few volts for connecting it directly to the AN pin of the pic? Don't need an operational amplifier??

E_KARVELAs · Joined: 14 Feb 2007 Posts: 46 Location: Greece & Cyprus

soulraven · Joined: 08 Feb 2009 Posts: 72 Location: campulung muscel

but for negative temperature?how to use LM35?

Ttelmah · Guest

You have to add a -ve voltage source, and an external op-amp as a level shifter. Look at figure 2 in the data sheet, and then a simple op-amp 'adder'. Use this to add sufficient offset voltage so that the lowest reading you want to use is going to be above 0v.
Probably cheaper and simpler, to consider a different IC. The LM135, gives the same 10mV/degree, but with this starting at 0v=-273.15C. You get 2.73v, at 0C, and the IC itself happily works down to -40C (lower in some versions).

Best Wishes

asmboy · Joined: 20 Nov 2007 Posts: 2128 Location: albany ny

i'ved used the LM-34 with the F886/887 - and rely on using an external 2.5 v V prec. ref -- to read temps mostly in the area of 150F to 212F.

looking at the schemo i drew - back when- - all i do is run the sensor on the same modestly regulated 5V as the PIC - but with a 4.7 uf tant cap on the 5V supply line to the sensor - the sensor cable is a flexible single pair audio cable - where the shield is the NEG return and the +5 and ret signal are the "pair" inside.

typical displacement is about 3 to 4 ft from the controller board.

i just tried one with 30 feet of cable and it still works FINE.

the receiver components going TO the pic are nothing more than a series 100 ohm resistor - shunted to ground by .1uF at the PIC pin input pin.

i get excellent matching at the 212F cal point ( +/- 1.5 deg F ) between LM34's and have never observed anything you could call instability.