Anne Ogborn wrote:
>
> Since it's likely that my incoming data will be 180 deg phase inverted,
> I'm using a different scheme to transmit data.
>
> I transmit 500 uSec of '+1' followed by 500uSec of '-1' for
> a 1. I transmit 250uSec of '+1' followed by 250uSec of '-1' for
> a 0.
>
> Then I just send bits - I check some 380 uSec after seeing a positive
> transition and if the pin is 1, it's a 1.
>
> I transmit about 1K of data, which ends with a CRC. After recording the unit r
esets.
>
> Sooo... my question is, anybody have any opinions on the advisability
> of making some sort of sync pulse every 8 or 16 bits?
>
> It's a 1 way channel, so I really
> can't do a retransmit request, hence complex error detection is
> irrelevant. If the CRC (16 bit) doesn't match, I flash an LED and
> the user tries again.

Well, it looks like you already have the clock included into the data
itself, why not use it? You have a 0 to 1 transition for every bit
sent...  if instead of 250us of -1 for a 0 you hold it for 750us, you
will ahve a bit for each 1ms, and the 0 to 1 transition will happens
always for each bit, 0 or 1, it could resync your receiver for every bit
received.

There is another easy way, the old NRZ/NRZI, just make a transition from
0 to 1 for bits 1 and a transition com 1 to 0 for bits 0, so your
receiver would work in edge sensitive mode, expecting it in a narrowed
window time (to avoid noise). Using this technique you transmit embedded
clock, and can increase speed and see what happens.

Wagner