OT: Technical Question for Jeff Liebermann

On 10/13/2011 7:56 PM, GreenXenon wrote:
> On Oct 13, 7:44 pm, miso<m...@sushi.com> wrote:
>
>
>> First of all, I don't think you have a
>> handle on what AM means. For instance, do you mean OOK (on off keying)?
>> Are you taking bytes of data and PCMing? Just strong enough to be
>> clearly recognized by the DAC? What does that mean? You feed a DAC
>> digital signals. There is nothing ambiguous in the data.
>>
>> Intrinsic in any communications scheme is data framing and clock
>> recovery, not to mention having to whiten the data (scrambler).
>
> 1. Why does the data have be scrambled?
>
> 2. No, I'm talking about OOK.

Since we haven't nailed down the modulation, I really can't go into the
scrambler much, but basically data could be a bazillion zeroes or one.
That can lead to modulation that doesn't really "flog" the system (use
all possible symbols). For instance, if you had a simple 4 point
constellation, it would only use two out of the 4 locations. You run the
data through a scrambler to whiten it, i.e. make it look more random. It
is generally hard to do clock recovery on a signal that isn't scrambled.
Note this is not encryption. I'm having a hard time describing this
because whitening the data is generally job 1 so I have to think hard
about the bad things that happen if you don't whiten it.
> http://en.wikipedia.org/wiki/Scrambler
Actually this wiki looks good.

Let's take the case of simple FSK with no scrambler and worse yet, no
framing. (Note FSK doesn't really need clock recovery, though in
practice, you dejitter the recovered data.) If you sent a long string of
zeros, it would correspond to one frequency of the FSK pair. Your data
recovery would have a hard time locking on the data. Now if yo framed
it, say with start and stop bits, you could now lock on the data. But
for analysis purposes, say estimating a bit error rate, you rather have
both frequencies used. Better yet, most of the comm schemes have already
been analyzed, so you can plug and chug, but the analysis generally
assumes white noise and randomized data.

OOK doesn't need clock recovery either. You could look at how an
aircraft transponder works if you need an example. You do need apriori
knowledge of the data rate to recover the data, and usually it is
oversampled to aid in recovery.