Grant, Thanks for your input. >Lets assume that the signals are quite >strong, and are at 27 and 28MHz, i.e. 1MHz spacing. >The closest IMD products will be at 26 and 29MHz, and will be at equal >amplitude. These are the third order IMD products, because they are >procuced by the 2nd harmonic of each of the two input signals mixing >with the fundamental of the other. We will ignore the other IMD >products for this discussion. >
Ok, let me do some number crunching with the numerical values you gave. Second harmonic of 27 and 28MHz are 54MHz and 56MHz. So a 54MHz signal is added two a 28 MHz signal, with some non linearity function after that. So I get the sum and products: 26 and 70MHz. Doing the same with 56 and 27, I get 29 and 83MHZ. With filtering, I will only keep the contents near the IF bannd: 26, 27, 28 and 28Mhz. Ok, I it matches so far. >Now, it gets more tricky - if the level of BOTH of the input signals >is increased by 1dB, then the levels of the 3rd order IMD products >will increase by 3DB. I am still struggling here. I would assume that a model of y(t)= a x(t)+ b x(t)^2 + c z(t) + d x(t) z(t) +e z(t)^2 x(t)=input signals = x1(t) + x2(t) z(t)=local oscillator I am trying to map this idea to the explanation you provided. I just don't see how come in the below graph http://en.wikipedia.org/wiki/Image:Interceptpoint.png that the two lines don't start at the origin. The graph seems to give me the below idea: y(t)= a x(t)+ b (x(t)- f)^2 + c z(t) + d x(t) z(t) +e z(t)^2 which means that the IMD power is subtracted by a constant, next scales up (slope b). Physically, how do you get on the mixer a sudden little piece of power cut off from the IMD total power? Frank > I hope that helps. > > regards > > Grant G8UBN >
