Impact of S22 on Performance of an RF Power Amplifier

Blog Post created by KeysightEEsofEDA Employee on Jul 12, 2016

We were asked a very interesting question on our YouTube channel regarding our video, How to Design an RF Power Amplifier: The Basics.

"Hey Matt, that was an awesome tutorial. However, I'd like to have your opinion on the impact of S22 on the performance of an RF Power Amplifier. And how it might affect the various Figures of Merit. Thanks. :)"

Since it's quite a long discussion, so we've posted the answer here.

It can be confusing. The load line match needed for the optimal PA operation is not at all the conjugate match scenario which is described in standard microwave engineering courses where one tries to present S22 conjugate back to the transistor through an output matching network. You may recall that a conjugate match is set up to obtain maximum power transfer between two small signal blocks. If you think about it, the assumption that there is that the power available from the generator is constant, and in that case the conjugate match results in maximum output power (and by extension maximum power gain) because as much power as could be transferred to the load was transferred to the load. The difference is that power from the generator in the large signal (PA) case is not constant at all. In fact for large signal cases, the power generated depends heavily on the load presented to the device because this load effectively limits the possible excursions of the voltage and current waveforms.

In the classical theory, there is only a load line match; a conjugate match is not relevant. Practically, maximum power generation described by the load line approximation only occurs for one single condition of bias and input drive.

For the classical modes the power predicted using the loadline approximation is the output power achieved somewhere around the 1 dB gain compression point, give or take a few tenths of dBs. However, most communication systems have complex modulation formats whereby the input signal is not fixed power at all; instead, the useful information is modulated onto the drive signal, resulting in variation in both amplitude and phase. The power amplifier is likely to see some drive signals which are low power and others which are high power all contained within the same modulated envelope. Some of these signals will have drive levels that will not push the PA very hard, others will push the PA to its limits (maybe even past the gain compression point). To a first order, the transition of the gain (and phase) characteristic from the small signal to large signal regions is what determines the level of spectral regrowth in the frequency domain. High levels of spectral regrowth can interfere with nearby adjacent channels. So then perhaps there is some need to consider PA performance for small signal drive levels after all …[sigh].

It turns out that there can actually be many different loads presented to the device which can all result in basically the same power and efficiency in a large signal sense. For an example of this, check out this video on continuous modes of operation.

If there are in fact multiple large signal loads that are all pretty much the same performance wise, what’s the best one to choose?

To make a long story short, it may be possible to pick a load (and bias level) which both delivers the desired large signal output power plus efficiency at compression while at the same time providing a small signal match which results in a favorable backed off power gain characteristic. In that way, the transition between the small signal region and large signal compressed region can be managed so that it is not overly expansive or compressive but occurs in a smooth, idealistic manner with only minimal compression at the maximum output power required, thereby giving minimal spectral regrowth.

All else being equal, that type of amplifier which has a good mix of large signal and small signal performance will perform best when put into a real life communication system. It’s likely that the load chosen is neither a perfect load line match nor a perfect conjugate match, but a tradeoff between the two.