Hi Julian, > By channel coherence time do you mean time when the signal is readable? >
The channel choerence time is a property of a (fading) channel which gives an idea of the time interval over with the channel response is approximately *constant*. If you drive your car at 100 km/h and tune your car radio to a far and weak station in the 88/108 MHz FM broadcasting band you have probably noted that the station fades out quite fastly, say with an average rate of 10 Hz, you therefore might expect that the channel response is approximately constant for no more than a small fraction of 1/10Hz = 100 ms. This occurs because the signal you are receiving is the sum of (usually many) different scattered components, each of them coming from a random direction which is not necessary the direction you are driving through. Some of this components could come exactely from the direction you are driving through and are affected by a positive Doppler shift. Some other components could come from the direction you are coming from and they are affected by a negative Doppler shift. Other components could come from directions which form a right angle with yours and the would exhibit no doppler effect. The sum of all of these components can be treated by a stochastic ideal model which is called the (flat) Rayleigh channel model. This (ideal) channel model is essentially characterized by two parameters: 1) the maximum Doppler frequency shift (which is called the "channel Doppler spread") and 2) the average channel attenuation. The Doppler spread (Fd) depends upon a velocity v (the velocity of your car) and the signal carrier frequency Fc through the formula Fd = v/c*Fc (c = light speed). If you do the calculation with v = 27.8 m/s (100 km/h) and Fc = 100 MHz you will find that the Doppler spread is approximately 10 Hz (9.3 Hz, for the sake of precision). Interestingly, the autocorrelation function R(T) (how much two samples of the process are correlated given the time interval T they are separated by) of the flat Raileigh channel model it's quite easy to compute: it's the Bessel function Jo(k*T*Fd) (k is a constant, I don't remember its value, maybe PI or something like that). For T*Fd << 1, the autocorrelation function is not different from unity and this tell us that in a time interval T << 1/Fd the channel response is strongly correlated. This means that if the channel response assumed a value X at time t, the probability that its amplitude does not differ so much from X at time t+T is large. > I can see how this would work using widely separated frequencies. However we > have all observed that when a signal goes down in QSB, it does down right > across the passband. This is not always true and has to do with another parameter which is called the coherence bandwidth of the channel and which is inversely proportional to the "channel time spread". In the HFs it's not unfrequent that the channel time spread is some milliseconds and that the channel coherence bandwith is few hundreds Hertz. Multiple reflections from the F and E ionosphere layers are an example in which this happens. >So do you actually gain anything by spreading the transmission by only 2.2kHz, >other than the ability to annoy people who consider it a selfish waste of >bandwidth? Not always, sure, but you could figure out by yourself the amplitude of public crucifixions if some amateur transmission were designed to be spreaded by 10 kHz or more just because a 2.2 kHz spread is not sufficient to exploit frequency diversity as the ionosphere characteristics would require :-) In any case it is more easy to design a new communication system which copes with a given ionosphere rather than to alter the ionosphere itself. For now we just managed to alter the atmosphere with massive CO2 emissions (and it took one hundred years). For the ionosphere we need more time... 73s Nico, IV3NWV
