Thanks for the reply! Ahh so it is a case of 'splitting' the differential balanced line into two single ended ones?
Yes, the second match on the smith chart is indeed a single ended one, using just one of the TX pins (12+j0) to 50 Ohms. This actually came about by accident, as I was just playing about with the chart to get from 25 to 12, but changed the load to '50' (wasn't sure if the balun was correct) and ended up with and inductor value identical to the reference design (left picture)... but with a cap double the value. This is handy, since it means I can 'do away' with the balun, just use single ended matching, sacrificing Tx power output from the transciever (halving its power) - but if I include a PA, I'm only losing 1-2dB for the convenience of NOT having to hand solder a 0603 size balun.
About the harmonic supression. This reference design was from 2003
And it has several low pass, and band pass 2.45Ghz ceramic filters after the TX chain, I guess that is how they suppressed the higher frequencies there. But I will look into a LP match, as it might prove better to lower band interference with my other transceivers I'm testing.
Can you explain the first picture? As I said, I tried matching 12 to 25 (25 ohm load) with the smith chart, but the values I get are (for a high pass match) 1.6nH shunt, with a series 5pF, and a shunt 22nH for the Tx bias. (actual input impedance is 12.3-j0.0). Now these are very different from the reference design, but the smith chart says its spot on. When converting these two single ended matches to differential, would I have to halve/double the values ? I'm asking because this document refers to it:
https://www.skyworksinc.com/uploads/documents/201156A.pdf
I have checked the PCB gerbers for the ref design, and I can't see any stubs or matching traces, its all very close together, so I can only assume that its the discrete LC parts which are doing all the matching. Also, thinking about it, given how close the parts are together, I'm not even sure a mismatch would be much of a problem, the balun is <5mm away from the chip via the caps - so traces between these parts are very short indeed. Output power is 3dB max for differential.
My main curiosity (remember, this is mainly academic, for learning) was the fact that my single ended match, 12->50 ohm, was the same as half the reference designs differential match, the only difference being, double the series capacitance.... that can't just be a co-incidence can it? Or have they made a mistake?
Thanks once again, sorry for asking lots of questions. I guess for my project is isn't 'critical', but I like to know whats going on