LM335 temperature sensors on 10m of speak wire (Help!)

[mearsy]

Hello,

I'm really not very good at the electronics side of things and could use some help understanding where my problem lies

I have built a remote temperature sensing circuit using an LM335. It seemed to work fine on the breadboard but when I have taken it apart on soldered it onto 10m of speaker wire it seems to behave oddly. I say it worked fine on the breadboard but my problem only seems to emerge when I try to take temperature readings above 45 degrees.

This is the speaker wire I have used

http://www.ebay.co.uk/itm/290585436641

I think this is 21Awg

According to this site

http://www.cirris.com/testing/resistance/wire.html

10m  (32.808399 feet) of 21 awg wire should have a wire resistance of 0.42 ohms

This is the circuit I have built



It has the 2.2kohn resistor at the Arduino end so basically it goes

5v --> resistor --> One short wire to analog pin
                    L--> 10m wire --> to the middle pin of the LM335.

Then the ground directly out --> to 10m of wire --> right pin of the LM335

I say it worked fine on the breadboard but in truth I didn't fully test it as I couldn't find a way to get the test temperatures up much above 38 easily without squashing the sensor so this is the first time I have really tested it.

The problem is when I place a hot teapot on top of the sensor, the temperature climbs to 45 degrees then it shows -40. I'm really not sure if I have an electrical problem or if I have done something stupid in my code. My code is at the bottom of this post

So, questions for the clever forum members please. I am happy to supply further readings

1.) I am assuming I have wired it correctly as I get a sensible sounding reading. As this is a diode if I wired it back to front I don't think I would get anything

2.) Similarly I am sure I have only used the two rightmost pins (I have clipped off the adjustment pin on the left of the image when the flat part is facing me)

3.) Is it my electronics or was it something in my code all along which I didnt spot up until now

Code: [Select]

void GetAverageTemps()                                                                                  // Function to get average temperatures on Pool & Panel in oC. Takes 1 second
{
        PanelTemp=0;                                                                                    // Reset the Panel Temp variable
        PoolTemp=0;                                                                                     // Reset the Pool Temp variable

        int PanelADCAvg=0;                                                                              // variable to hold Panel ADC Averages
        int PoolADCAvg=0;                                                                               // variable to hold Pool ADC Averages
        for (int i=1; i <=50; i++)                                                                      // Loop to get 50 values
        {
                delay(10);                                                                              // Delay 10ms to allow ADC to recover between reads
                PanelADCAvg=PanelADCAvg + analogRead(0);                                                // Read in 50 Panel ADC Readings
        }
        for (int i=1; i <=50; i++)                                                                      // Loop to get 50 values
        {
                delay(10);                                                                              // Delay 10ms to allow ADC to recover between reads
                PoolADCAvg=PoolADCAvg + analogRead(1);                                                  // Read in 50 Pool ADC Readings
        }
        PanelADCAvg = (PanelADCAvg/50);                                                                 // Work out the average Panel ADC reading
        PoolADCAvg = (PoolADCAvg/50);

        PanelTemp=(((PanelADCAvg * 4.965 * 100.0)/1024)-273.15)+config.PanelTempAdj;                    // Convert the PanelADCAvg to a temperature in oC and add any adjustment
        PoolTemp=(((PoolADCAvg * 4.965 * 100.0)/1024)-273.15)+config.PoolTempAdj;                       // Convert the PoollADCAvg to a temperature in oC and add any adjustment

        PanelTemp=(PanelTemp*4);                                                                        // Round the output to the nearest .25
        PanelTemp=round(PanelTemp);
        PanelTemp=PanelTemp/4;       

        PoolTemp=(PoolTemp*4);                                                                          // Round the output to the nearest .25
        PoolTemp=round(PoolTemp);
        PoolTemp=PoolTemp/4;
 
        PoolTemp=(constrain(PoolTemp,-40.00,100.00));                                                    // Limit the range of values to the LM335Z range
        PanelTemp=(constrain(PanelTemp,-40.00,100.00));                                                  // Limit the range of values to the LM335Z range
}


Notes
A couple of notes on the above. The reason I use 4.965 in the calculation is because I measure the arduino putting out between 4.96 and 4.97v on the 5v pin.

Ignore config.PoolTempAdj &config.PanelTempAdj. These are values used to trim the sensors (due to natural differences in manufacturing) For this code you can assume they are set to zero

ADDENDUM

I suspect I see my problem but I can't test it right now as I am at work so I would still appreciate some input. I think the problem may be in code.

A typical output for the LM335 at 25 degrees should be 2.98v. From memory I think this equates to an ADV value of 611 so I will use this for the example.

I take 50 readings and store them in an int called PanelADCAvg. 50 * 611 = 30550

I then divide this by 50 to get to my average. In this example its 611

I then run this calculation:
PanelTemp=(((PanelADCAvg * 4.965 * 100.0)/1024)-273.15)+config.PanelTempAdj
Which is ((611 * 4.965 *100.0)/1024)-273.15) =23.101 degrees c

The above is slightly lower than I stated. Perhaps I was wrong on the 611 = 25v when using 4.965 but no matter

An ADC read of 656 is around 45 degrees (Where the problem occurs)

So I take my 50 readings and store them in an int called PanelADCAvg. 50 * 656 = 32800

Obviously an int variable will roll over to negative when it gets above 32767
Whilst I am expecting PanelADCAvg to be 32800/50=656 in reality its actually -33/50=-0.66
Because an int can't hold a floating point -0.6 becomes -1
So whereas I think my calculation will be:
((656 * 4.965 *100.0)/1024)-273.15) =44.92 degrees c

it is actually

((-1 * 4.965 *100.0)/1024)-273.15) =-273.635 degrees c

I then do a bit of rounding to take this to -273.50 degrees c

And then finally I constrain it to the range of the LM335 (-40 to +100)

So when I go above 45 degrees I get a figure of -40 all because I used int instead of unsigned int

Do you agree? Or is there something else I am missing.

[mearsy]

As an addition to the above I just realised that when int is asked to store the number 32800 it rolls over and becomes -32734. Not -1 as I suggested

which / 50 yields a result of -655 for the ADC read.

Which makes the result of the calculation -590.735 (rounded to -590.75)

Which is then constrained to -40

Same net result. Just I got my understanding of how ints roll over a little wrong.

Still, please feel free to comment on anything above. Especially the electronics which is my weakness.

[DVDdoug]

So when I go above 45 degrees I get a figure of -40 all because I used int instead of unsigned int...

... just realised that when int is asked to store the number 32800 it rolls over and becomes -32734.

Congratulations for finding that!  It's not easy, because there's nothing special about the number 45.   There's also nothing special about the reading from the ADC, and it will just max-out/clip... The ADC never rolls-over or puts-out any "funny" numbers.

Once you've seen a problem like that, you'll never forget it!  As soon as I read your post, I  was thinking you might have a rollover problem.   Once a million years ago, the company I worked for had a customer with a chart recorder hooked-up to our equipment, and at a particular pressure reading, the chart recorder would go crazy.  As I remember, the digital display was OK but the analog chart recorder output was erratic.   And, I guess there were some customer-configurable gain/range settings because we couldnt duplicate the problem at the factory.   We had no idea what was going-on, but I think the programmer finally figured-out there was a binary number rolling-over somewhere in the code.  Then everything made sense, and I never forgot it!

[dc42]

I run a company that produces tools for developing and mathematically-verifying safety-critical embedded software. Occasionally, we get asked to verify software that has already passed testing and is considered production-ready. When we do this, the most common problem we find is that for certain combinations of inputs, there is an arithmetic overflow in some calculation. So you are not alone.

[mearsy]

Thank you to both commenters for their kind responses and for not making me feel too bad for my error.

I would still gladly welcome any comments on the way I've hooked up the sensors and the fact I have set the voltage in code to the actual Arduino output voltage of 4.965 volts.

I am due to ship this from the UK to Australia so would like to be confidenetI have done it right before I sent it.

[luxxtek]

the problem is the capacitive load of the long wires.
I used the LM335 before and had the same problem, switching to an LM35 solved the problem.

In your layout you show the LM35 so I'm surprised you don't use it.