2.3 Two-Tone Characterization Tests
2.3.3 Two-Tone Characterization Setups
Because mixing products involving a combination of both1and2have frequen- cies that are different from either 1 or 2 inputs, two-tone test measurements use a spectrum analyzer. The most commonly used arrangement for such a setup is shown in Figure 2.10.
Table 2.1 Out-of-Band Distortion Components Generated from a Two-Tone Excitation Mixing
Product dc 2−1 21 2+1 22 31 21+2 22+1 32
Order 2nd 2nd 2nd 2nd 2nd 3rd 3rd 3rd 3rd
Figure 2.9 Output fundamental power per tone and distortion power at the baseband for an equal amplitude two-tone excitation.
Although, in its simplest form, this setup would only involve two signal genera- tors of variable level, a power combiner and the spectrum analyzer, the implementa- tion depicted in Figure 2.10 requires many more laboratory components. They are intended to guarantee accurate measurements of very high signal-to-distortion ratios.
First of all, it is assumed that the combination of the two input signals is not made with a simple T-junction, but by using a true power combiner. This not only guarantees port matching, as it profits from adjacent port isolation. And that is of paramount importance as it prevents each signal to mix with the other in the nonlinear output stages of the generators. The measurement error that may be induced by this parasitic IMD is so dramatic that it may be found useful to artificially boost combiner isolation with the two isolators shown in Figure 2.10. Moreover, harmonics of the generated signals can also mix with the other fundamental to produce further residual distortion either in the DUT, the signal generators’ output stages, or both. Since the amplitude and phase of this residual IMD is unknown, but it may be amplified by the DUT’s linear gain, it will add to the wanted DUT’s IMD producing an unpredictable error. That is why two lowpass filters were included at the signal generator outputs to improve their signal spectral purity.
Beyond those sources of IMD measurement error, there is also the spectrum analyzer nonlinear input stage. Remember that if you are characterizing a very linear device, everything in the setup (including the spectrum analyzer) should be even more linear. Obviously, one way to get rid of that additional distortion caused by the DUT’s fundamentals in the spectrum analyzer front-end stage would be to use a large attenuator at its input. However, that attenuator adds a certain amount of noise, as it masks, to the same amount, the very small DUT’s IMD components.
The solution is to take advantage, as much as possible, of the available spectrum analyzer’s dynamic range.
Two-ToneCharacterizationTests41
Figure 2.10 Most commonly used two-tone test measurement setup.
This can be done by studying the dynamic range characteristics of that equip- ment, and then choosing the optimum input power level (adjusting the spectrum analyzer’s input attenuator value) and convenient sweep time (determined by the selected resolution bandwidth, and thus, correspondent noise floor) [8], as illus- trated in Figure 2.11.
An alternative way to accomplish the same objective consists of tuning the spectrum analyzer’s settings until the best results are met. For that, the first step should be to empirically choose the minimum required input attenuation. Starting with the highest end of available attenuation values, the desired attenuation value is the minimum one in which the carrier-to-distortion ratio actually read is still unchanged.
The second step should be to take advantage of the available spectrum analyzer’s sensitivity. This demands a reduction in resolution bandwidth, which must be inevitably accompanied by a correspondent increase in sweep time and probable reduction in frequency span. This, in turn, may obviate the simultaneous observa- tion of the output fundamentals and the distortion components, as it may call for low phase noise and highly stable frequency-synthesized generators.
In the next section we will show how an imaginative modification of the setup may circumvent the majority of these difficulties.
2.3.3.1 Bridge Setup
The above discussion, on the distortion measurement difficulties, has shown that the main source of error introduced by the spectrum analyzer is due to the DUT’s
Figure 2.11 Different spectrum analyzer error entities when measuring nonlinear distortion.ATT2
(ATT3) represents the optimum spectrum analyzer input attenuation value for max- imized signal to noise plus second (third)-order distortion ratio.
fundamentals. Therefore, eliminating those components, without perturbing the desired distortion, would constitute a foremost benefit. Nevertheless, except for very special cases where out-of-band distortion is sought, in which tone separation is so high that the fundamentals can be rejected without significantly perturbing the closely located distortion components, or, eventually, where the measurement band is previously fixed, filtering is out of the question. Therefore, the elimination of the DUT’s output fundamentals demands more ingenious solutions.
The bridge setup presented in Figure 2.12 is one of such possible methods [9].
It relies on the fact that, since the signal present at the DUT’s input is an exact replica of the output fundamentals, but includes no distortion components, it can be used to cancel out those fundamentals, while preserving the desired distortion.
In doing that, the amplitude and phase of a sample of the excitation are tuned in the auxiliary bridge arm and then combined with the DUT’s output to produce the sought cancellation. Naturally, it is herein also assumed that both the attenuator and phase shifter are linear components, and so they cannot introduce any distortion of their own.
Note that, except for the bridge network composed by a 3-dB power splitter, linear gain and phase control cells—a variable attenuator and phase shifter—and the final 3-dB power combiner, the bridge setup is essentially equal to the one of Figure 2.10.