Digital Doherty linear analysis
Plots of two important parameters (efficiency at a fixed, interpolated, output power, and saturated output power) are presented at two different frequencies.
While efficiency and saturated power are often the two most important performance metrics, The plots of measured data show that the best performance does not coincide at constant amplitude and phase difference values.
Optimum values for amplitude/phase split are pretty much unique, at different frequencies and for the different parameters.
First step
This simple first step already provides information and insight that was hitherto unavailable. At the two test frequencies, it is clear that both saturated output power and energy efficiency, cannot simultaneously be maximized using the conventional constant phase-amplitude input split. Even worse, for the amplitude and phase input differentials at which those maxima occur, the other three parameters are located on steep rate-of-change slopes. Luckily, alternative implementations are possible.
Parameter sensitivity
At this point, the engineer has visibility of at least three different solutions – the fixed input split (often with phase offset), the programmable input split, and the dispersive input split. Then, in another post-processing step, we can simulate industrial efficancy; that is to say, how sensitive the performance parameters are to variations in parameters.This simulation was performed by creating normally distributed random values for the amplitude and phase. This population of values was stored to enable replication in other experiments
Value extraction
Using the measurement database as a look-up-table (LUT), values for efficiency and saturated output power were extracted. These can now be parametrically plotted, to demonstrate the spread in those parameters at the low- and high- frequencies.
Worst case
By further taking the worst case value for each instance, a statistical spread of performance can be extracted and an exemplary specification, supported by analysis can be written.
Amplitude and phase variation
Next consider the case whereby different amplitude and phase values are used for the two frequencies. This replicates the scenario whereby either dispersive split, or programmable split is implemented.
Maximum efficiency operation
Again, as Doherty is intended to be an efficiency enhancement technique, nominal values were chosen that gave maximum efficiency operation (but also constrained by a limit on the amplitude offset). It is interesting to note that the optimum phase required to extract that maximum efficiency, is some 30-40 degrees apart across the frequency band, and is not a constant value.
Improved worst case
Repeating the exercise to extract the predicted industrial spread of efficiency and saturated power yields a much better result for the worst case efficiency.
Thus, the designer now has illumination of their design capability – even better, is armed with at least three different cost-performance paradigms.
