Frank Describing the IMD of a mixer is more complicated, so I'll carry on with the discussion of an amplifier.
The graph in Wikipedia is correct, and shows the relationship between the levels of the input signals on the x-axis and the output signals on the y-axis, and the input and output intercept point. It also shows how the levels of the IMD products increase at 3 times the rate of the input signals. However, the graph is deliberately not scaled - there are no numbers on either axis, and there is no origin. If a unity-gain amplifier was used, then the graph would indeed go through the origin; the input signals (on the x-axis) would have the same levels as the output signals (on the y-axis). If an amplifier was used with a gain of 7db, then the graph would show a straight line with a gradient of 1 for the wanted signals and the y-values being higher by 7 dB. In this case the Output IP3 would be 7dB greater than the IIP3. If a lossy device was used instead, such as a PIN-diode attenuator, with a loss of (say) 12dB, then the graph would still show a straight line with gradient of 1 for the wanted signals and a gradient of 3 for the IMD3 signals, but the output level on the y-axis would be 12dB lower than the input levels on the x-axis. In this case, the IIP3 would be 12dB greater than the OIP3, and this fact is sometimes used by manufacturers - they usually quote the IP3 with the largest number! Buyer beware! The relationship between the levels of the input signals and the the 3rd order intermod products is always 3:1, as long as the device is not in compression. Note that all values are in dB, otherwise much more complicated maths is required. Leif's maths are correct; I find that often the maths doesn't help that much in understanding the problem, more often than not it either confuses people or switches them off, so I try and describe things in other ways. But we can talk maths if you want to. regards Grant > 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. So, the difference between the levels of the two, http://en.wikipedia.org/wiki/Image:Interceptpoint.png One more thing, is it always a 3:1 ratio for the slopes, just like the graph? Also is it correct to say that vertically, for a zero gain mixer, the 1:1 graph passes by the origin and when the mixer has a non-zero gain, the graph is is just shifted up or down by the dB value of the gain? Thanks, Frank
