FSK constellation and eye diagrams

Question

why do constellation diagram and eye diagram not show any useful output for FSK? I have tried to find that on Simulink Mathwork, but the progress was negetive.

Answer 1

An eye diagram shows amplitude against time but FSK has nominally constant amplitude and carries zero amplitude information. Apart from that, an eye diagram is used for the raw data and not a carrier modulated by data.

A constellation diagram pin points regularities in amplitude and phase but FSK uses two frequencies therefore the phase relationship is lost on one of the two FSK frequencies and amplitude is constant anyway.

I would expect FSK to look like a dot or a circle in a constellation diagram (and possibly a partial arc of a circle) because FSK doesn't necessarily link carrier phase angle to the data changes (as is necessary in PSK).

Answer 2

Eye diagrams and constellation diagrams are generated from voltage(s) that represent the "symbol space" of the baseband signal. Non-quadrature signals have a single voltage; quadrature signals have two. In the case of FSK, this baseband voltage would be the output of the frequency detector.

A constellation diagram is produced by sampling the voltage(s) using the recovered symbol clock, and plotting the resulting points on a scatter plot. The quality of the decoding can be judged from how well the points cluster and how clear the spaces between the clusters are. It is not possible to separate the effects of noise and timing jitter in a constellation diagram. For non-quadrature signals, the points all lie along a single line, so this type of diagram is normally only used for quadrature signals.

An eye diagram is normally used for non-quadrature signals. It is produced by plotting the single signal voltage versus time, again triggering on the recovered symbol clock, with the timebase adjusted so that one or two complete symbols appear in the center of the display. The signal quality is judged from how clear the "eye" openings are. Noise can be evaluated at the top and bottom of the eye, and timing jitter can be evaluated at the crossings.

So the answer to your question is that you probably want to make an eye diagram at the output of your frequency discriminator. You'll need a symbol clock to trigger the individual plots, and a much higher frequency sample clock to capture the sample points. You should be able to get a very useful diagram.