Jee Labs Talk

This forum has moved to forum.jeelabs.net - this read-only archive is for reference only!

1. benny
   

   Can anyone tell me how to interpret negative temperature numbers returned from a DS18B20 in rooms.pde ?

   OK 1 146 3 (below zero)
   OK 1 236 3 (even more below zero)

   Above numbers came in from a FROZEN DS18B20, numbers above zero are ok.

   Posted 14/12/2009 18:02:05

2. benny
   

   problem solved (by reading datasheet ...)

   Posted 14/12/2009 18:51:27

3. jcw
   

   Ah, good one. It's passed back as 10-bit value in tenths of degrees, and it's little-endian. If the second byte is 2 or 3, it's negative. That means the first byte is then in 2's complement. So my guess would be that "146 3" is -11.0°C (256-146) and "236 3" is -2.0°C (256-236) - doesn't match your note that the 2nd result is even more below zero, but that's what I get...

   So to get the full range, do as follows: calculate the 10-bit value as "T = byte 1 + 256 byte2". If T is 512 or more, turn it into a negative value using "T = 1024 - T".

   Does that produce correct results for you?

   Posted 14/12/2009 19:35:01

4. benny
   

   Hi JC,

   Yes this is what I found out, thanks anyway !

   Posted 16/12/2009 20:04:01

5. mauzer
   

   Hi,

   I am using this:

       struct { 
         int temp0, temp1, temp2, temp3, temp4, temp5, temp6, temp7; 
         byte lobat :1;
       } payload;

       payload.temp0 =  23.50  100;
       payload.temp1 =  23.50  100;
       payload.temp2 =  23.50  100;
       payload.temp3 =  23.50  100;
       payload.temp4 =  23.50  100;
       payload.temp5 =  -5  100;
       payload.temp6 =  -5;
       payload.temp7 =  -11;
       payload.lobat = rf12lowbat();

   The result:

       OK 34 46 9 46 9 46 9 46 9 46 9 12 254 251 255 245 255 0

   How to calculate negative values?

   Posted 02/11/2010 12:32:24

6. jcw
   

   > How to calculate negative values?

   The geeky answer: flip the high bit, subtract 128, i.e.

       int v = lo + 256  ((hi ^ 0x80) - 128);

   or

       int v = ((lo + 256  hi) ^ 0x8000) - 32768;

   This sort of trickery is sometimes needed, to sign-extend 10-bit values etc.

   For 16-bit values as in your example, it's much simpler to use C itself:

       int v = (short) (lo + 256  hi);

   Since ATmega's are little-endian, the low byte comes first.

   Posted 02/11/2010 14:01:30

7. tankslappa
   

   Ooh, I've learnt a new C word (that sounds a bit wrong) :-)

   (short)

   Although my geeky mind will still prefer:

       int v=((lo+256hi)^0x8000)-0x8000;

   ..because, to me at least, it's more obvious what it does!

   Now having recently bought a couple of DS18B20+ (after a discussion about reading temperature in another post), I'm off to check my code because I suspect my temperature monitoring might be about to go a bit wrong when we get the first cold night!

   slaps self for not testing properly

   Posted 02/11/2010 16:41:00

8. mauzer
   

   It works! Thanks!

   Posted 02/11/2010 17:39:27