How to design "matching" networks for maximum IP3
However, I do not know how to match the circuit to get maximum IIP3.
I have simulated (in @DS) the different impedances in the circuit, so making conjugate matching for maximum ConversionGain works well, but it is possible to do something similar to "match" for maximum IIP3?
Best regards
Joakim
matching networks can help you to maximize the power transferring, I think IP3 is more related on linearity, by adjusting the tail current source, the source degraded inductor or just remove it to achieve better linearity, it can help you to get better IP3
To achieve better linearity, you can choose to use resistor degeneration or inductor degeneration. If you have voltage headroom problem, you should use inductor, because it gives lesser voltage drop. But it will take up more areas, as you know, inductor takes up a huge area.
Matching network can affect IP3 indirectly.
Better matching means higher gain.
Higher gain means lower IP3.
There should be a compromise between these two parameters.
I do not know the specific answer, but you should be able to do some things in the matching networks to improve IP3. IP3 is the measure of how well a device does in NOT generating unwanted spurious signals. Say you input 1 GHz and 1.1 GHz tones into a mixer. The 1 GHz will double to 2 GHz, mix with the 1.1 GHz, and generate an 0.9 MHz unwanted tone. Similarly there will also be a 1.2 GHz unwanted tone generated by doubling the input tone to 2.2 GHz. But what if the matching network provides the right set of impedance conditions so that the original tones do not generate the 2.0 and 2.2 Ghz tones in the first place? Well, there will be nothing to mix with to generate the unwanted 0.9 and 1.2 GHz spurs.
So if you have an analysis program to play with, play around with the impedances at the 2nd harmonic of the input tones (like trying to put a short circuit or an open circuit or a good match at 2 x Fin at the devices terminals, etc) and see if things get better. You will probably have to muck around a little with the harmonic load phase angles to get things best.
If you think of how one makes active or diode multipliers, you use "idler" circuits that control the impedances, and therefore the currents and voltages at the device terminals, to either enhance or reduce a specific harmonic. You should be able to do a similar thing with mixers.
Also, in the above example of 1 and 1.1 GHz input tones, what are you doing at the device terminals at the difference frequency of 100 MHz? What impedances you present there (bias line capacitance to ground, etc) will also affect IP3.
biff44 is correct.
Minor modification: You will always create the IM3 tones based on the fundamental non-linearity of the device. The 2X frequency and Delta frequency conversion also creates IM3 tones but these tones are in opposite phase to the IM3 generated by the fundamental non-linearity. Therefore, an input/output matching network can adjust these re-mixed tones to cancel the fundamental tones and create an IP3 nulling.
I've written a pseudo paper on that, which can be found here:
http://www.electriceng.com/Files/IP3Nulling.pdf
In real life there's not much you can do. If either the output or input impedances are designed by an external customer, then giving them restrictions on 2nd harmonic terminations is too difficult. Also small variations in the fundamental frequency can translate to huge differences in the harmonics, and thus creating a re-producable situation is very difficult. For most part, this type of analyses is used to identify a possible problem rather then using it to tune out IM3's. However, your situation may allow you to do these fancy things, and if so, have fun at it! ;)
Greg