Serial Comms and Crystals

[richard]

Which I think the test logic would translate to...
If the count is less than 19.2us or greater than 80us then it's a bad read.
Otherwise if the count is greater than 49.6us and less than 80us the bit is a 1.

that's it although a '0' is 26-28uS and a 1 is 70us according to the data sheet I have added some tolerance (this helps if interrupts are involved) if the count is outside this range then its not right and the process is abandoned , this test is not strictly necessary and could be omitted

the 1/0 test is basically centered in between the 1 and the 0 times ie 50 is halfway between 30 and 70 once again to add some tolerance . if less than halfway its a 0 otherwise a 1


there is some risk in this method in that hangs are possible if the dht responds in an unforseen manner . in my production version the tmrx int is enabled and if it is triggered then the process is also aborted to recover from such a sensor hang .

[Scampy]

Tabsoft / Richard,

I've incorporated the code below into my main program and whilst it runs fine, it does flash up a fair few "bad read" errors. Is there anything I could tweak to reduce these occurancies

read_dht:
 for bits=4 to 0 step-1 
 dht[bits]=0 
 next
 
 tmr2=0
 dht_dir=0
 dht_data=0
 pause 18 'start it up
 dht_data=1
 pauseus 30
 dht_dir=1
 pauseus 40 'wait till middle of response pulse window
 if ( dht_data)then goto badresponse 'no response then give up
 while (!dht_data)
 wend 
 t2con=6
 while ( dht_data) 
 wend 
 t2con=0
 if ((tmr2 <45)||(tmr2>55))then goto badpresence ' confirm presence ?
 for bits =39 to 0 step -1
 tmr2=0
 while (!dht_data) 
 wend 
 t2con=6
 while ( dht_data) 
 wend 
 t2con=0
 if ((tmr2>12)&&(tmr2<50)) then ' noise ?
 if (tmr2 >31 ) then
 dht.0[bits] = 1 

 endif 

 else
 goto badread ' noise ?
 endif
 next 
 crc=0 
 for bits=1 to 4 
 crc=crc+ dht[bits]
 next
 if crc != dht[0] then goto badcrc 'crc

 return

 badread:
 LCDOut $FE,$94 +8,"Bad read"
 pause 1000
 LCDOut $FE,$94 +8,"        "
 ' next
 return
 badcrc:
 for bits=4 to 0 step-1 
 LCDOut $FE,$94 +16,hex dht[bits],"," 
pause 1000
 next 
 return
badresponse:
 LCDOut $FE,$94+8,"No Response"
 return

badpresence:
 LCDOut $FE,$d4,"pr",#tmr2
 return

[richard]

does your program use interrupts ? if so how long does the isr take to execute ?
the tolerance here is 10uS + or - so an interrupt can easily cause the timing tolerance to be exceeded
possible solutions
1. disable interrupts during dht read
2. increase tolerance
3. ignore bad reading and retry
4. remove the noise test

t2con=0
 
if ((tmr2>12)&&(tmr2<50)) then ' noise ?
 if (tmr2 >31 ) then
 dht.0[bits] = 1 

 endif 

 else
 goto badread ' noise ?

 endif

5. use a faster chip

[Tabsoft]

To get a better understanding what is going on, why don't you make a small change for debugging.

Instead of this...

for bits =39 to 0 step -1
    tmr2=0
    while (!dht_data) 
    wend 
    t2con=6
    while ( dht_data) 
    wend 
    t2con=0
    if ((tmr2>12)&&(tmr2<50)) then ' noise ?
        if (tmr2 >31 ) then
            dht.0[bits] = 1 
        
        endif 
    
    else
        goto badread ' noise ?
    endif
next

Do this instead...

 
 for bits =39 to 0 step -1
     tmr2=0
     while (!dht_data) 
     wend 
     t2con=6
     while ( dht_data) 
     wend 
     t2con=0
     if tmr2 <= 12 then goto tooshort
     if tmr2 => 50 then goto toolong
     if (tmr2 >31 ) then
        dht.0[bits] = 1 
     endif 

 next

Then create the 2 subs, tooshort and toolong and have each print a different error message.

[falingtrea]

TI and Silicon Labs have RH/Temperature devices with an I2C interface in the $5.00 single part range: Si7013, Si7020, Si7021 2%-4% accuracy TI HDC1000 and HDC1008 similar accuracy specification. Both are available at Digikey. Honeywell also has a device, HIH6130, but it is a bit more expensive, at $14.

[Scampy]

Guys, thanks for the additional replies,

@Tim, I'll check out those suggestions... cheers