At 05:48 AM 10/5/2003 -0500, michael brown wrote:

>David VanHorn wrote:
>> At 04:59 AM 10/5/2003 -0500, michael brown wrote:
>>
>>> Hello all,
>>>
>>> In my unending struggle to decode the 1200 baud FSK tones
>>> constituting Caller-ID, I've made progress but have some problems.
>>> The goal is to decode the FSK data by measuring the period of the
>>> tones being sent from the phone company.  The two tones being
>>> detected are 1200hz and 2200hz. Using the CCP module I can now
>>> successfully measure the period of a full cycle and determine the
>>> current tone.  The problem lies in the fact that at 1200 baud, I get
>>> exactly one full wave of the 1200hz MARK tone and 1.83 cycles of the
>>> 2200hz SPACE tone per bit time.
>>
>> So set your timer for 1.4 cycles of 2200, and if the int happens
>> early it's 2200, and late it's 1200.
>
>That's sorta what I'm doing now if I understand you correctly, 1600hz is
>the dividing line.  I have no trouble distinguishing 1200 and 2200, it's
>the timing of my output bit that's giving me fits.  I guess the problem
>is that I don't really know exactly what happens when a tone transition
>occurs.  I mean do they do anything to insure a zero crossing occurs
>near a bit time milepost?  Does it compress or stretch the waveform?
>Obviously when a transition occurs, some kind of FM or PM occurs.  I'm
>wondering how I can see this in software using the CCP module.

I used to know this.. :)
I think the transition starts on a zero crossing.
You may see half-cycles of a given tone, but I'm not sure. Check an encoder chip's spec.


>> Look at the DTMF tones. Not an integer relationship in the bunch.
>
>This is intentional, for the filtering reasons mentioned above, correct?

Yes.

--
http://www.piclist.com hint: The list server can filter out subtopics
(like ads or off topics) for you. See http://www.piclist.com/#topics