433MHZ=433 waves in a 1 us or 216.5 waves in 0.5 us. if i measure the time period of those 216 waves then frequency can easily be calculated. In FSK we need at least 2 frequencies for 0 and 1 .lets say 433,000,005 hz=1 & 433,000,006 for 0. This never happens in real world why?

 

To have a circuit capable of distinguishing between two frequencies so close together would require an impossible value of Q. And if you COULD achieve such selectivity, the circuits would "ring" so long that the symbols would totally overlap. You can't fool mother nature....or the laws of bandwidth!

Eric

 

p.s. good question though!

Eric

 

Grammar is not clear to me in post #1

 

Because shfting the frequency takes some time?

 

Here would be an interesting simulation to perform:

Set up a system to use FSK about some carrier of +/- 0.5 Fshift so that the symbols for a 0 and a 1 are Fshift apart. Say Fshift = 100 kHz with a 100 MHz carrier.

Now modulate a very slow data signal that is just going back and forth between 0 and 1 at, say, 1kHz. Do an FFT and see how wide the signal is about each frequency. Now do it again but have a very low duty cycle data stream of, say, one 1 followed by ninety-nine 0s. What does the spectrum look like now? Now repeat for random data.

Now repeat this with a data rate of 10 kHz and 100 kHz. Can anything be inferred about how far the shift frequencies have to be in comparison to the data bit rate?