IR Record and Playback

william chan

Hi guys,

The 38Khz carrier using CTRA works perfectly.
But now the record and playback in SPIN only works for about 20% of equipments I tested.
For example, it works with Panasonic remotes, but not Philips or LG remotes.

Could it be SPIN is too slow to make accurate sampling at 80Mhz?
The record_ir() code I copied from the IR_Kit in the OBEX.
I wrote the playback_ir code(), but it seems difficult to accurately match the time per loop for record and playback.

CON
 _maxbuff = 500          'number of marks and spaces max 
 _quiet   = 10000         'ir completely quiet  

VAR
        long    buff[noparse][[/noparse]_maxbuff]
        long    samples


pub record_ir | iron, iroff

  dira[noparse][[/noparse]ird1_pin]~                          'make irdetector pin input 
  samples~
  repeat while ina[noparse][[/noparse]ird1_pin]==1     'wait any mark
  repeat
    iron~
    repeat while ina[noparse][[/noparse]ird1_pin]==0  'time mark
      iron++
      if iron > 10000                      'this just balances mark/space counts
         return
    buff[noparse][[/noparse]samples++]:=iron 
    iroff~
    repeat while ina[noparse][[/noparse]ird1_pin]==1  'time space
      iroff++
      if iroff>_quiet                        'end of train
        buff[noparse][[/noparse]samples++]:=iroff        'save last
        return
    buff[noparse][[/noparse]samples++]:=iroff
    if samples => _maxbuff-2
      samples -= 2                         'don't overfill the buffer


PUB playback_ir | i, count
    
  i~
  repeat
    if i => samples                  ' done?
       quit
    count := buff[noparse][[/noparse]i++]               
    dira[noparse][[/noparse]irled_pin]~~               ' Make pin output to send IR signal
    repeat while count>0         ' mark
       count--
       if count > 10000             ' this just balances time per loop   
    count := buff[noparse][[/noparse]i++]                  
    dira[noparse][[/noparse]irled_pin]~                 ' Make pin input to pause IR signal
    repeat while count>0         ' space 
       count--
       if count > 10000             ' this just balances time per loop 

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
www.fd.com.my
www.mercedes.com.my

Post Edited (william chan) : 12/10/2008 1:31:02 AM GMT

Chuck Rice

How do you know you are running at 80MHz? The code as posted would run at 12MHz. What happens if you insert

_clkmode = xtal1 + pll16x                             'Set to run at 80 MHz
_xinfreq = 5_000_000

At the top of the file?

william chan

I am running at 80Mhz, I am sure.
I didn't post the clock and freq code b'cos I don't want too much clutter on the example code.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
www.fd.com.my
www.mercedes.com.my

Mike Green

Since you want to check for a timeout, both in the time to wait for a pulse and the pulse width

noPulse := false
temp := cnt
repeat until ina[noparse][[/noparse]ird1_pin] <> 0
   if (waitTime := cnt - temp) > timeout
      noPulse := true
      quit
if noPulse
   ' no pulse was detected
else
   tooLong := false
   temp := cnt
   repeat until ina[noparse][[/noparse]ird1_pin] == 0
      if (pulseWidth := cnt - temp) > timeout
         tooLong := true
         quit
   if tooLong
      ' width of pulse > timeout
   else
      ' we have a valid pulse, save it and the off time prior to the pulse to the array

Note that you'd have two values for each pulse, the off time prior to each pulse and the on time of the pulse.
You'd obviously ignore the first off time since its unrelated to the pulse train. To play it back, you'd set the
DIRA bit to 1 and do a WAITCNT(value[noparse][[/noparse] x ] + cnt) where value[noparse][[/noparse] x ] is the first time value, then toggle the
DIRA bit, increment x, and do another WAITCNT repeating the whole sequence for each recorded time value.
At the end of the pulse train, you'd just set the DIRA bit to 0 turning off the IR pulses.

Post Edited (Mike Green) : 12/10/2008 6:00:54 AM GMT

william chan

Mike,

I don't quite understand why your code is totally different from mine.
Would your code's timing be more accurate?

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
www.fd.com.my
www.mercedes.com.my

william chan

Ok, I am beginning too understand your code better now. Thanks.

But there is a small problem.

Sometimes when we WAITCNT(buff[noparse][[/noparse] x ] + cnt) during playback, the value of buff[noparse][[/noparse]x] may be too small, causing the cog to wait forever.
If we set a minimum to the value of buff[noparse][[/noparse]x], it will cause the pulse train to become unfaithful.

So how now?

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
www.fd.com.my
www.mercedes.com.my

Mike Green

1) The code I wrote is essentially independent of the execution speed of the various statements and only dependent on the system clock.

2) You're correct about there being a minimum WAITCNT increment. It's roughly 400. You might set 1000 as a minimum. It's only about 12us, less than a half cycle of the 38KHz carrier. You shouldn't see delays this short.