The capability of a Nonlinear Vector Network Analyzer (NVNA) repetitive sampling receiver, which is designed to perform pulsed RF S-parameter and load-pull measurements, is presented. Due to its unique sampler-based architecture, unlike a mixer based VNA, the dynamic range of the NVNA is independent of the duty cycle. This results in a good dynamic range even for tiny duty cycles. This NVNA is presented in combination with an open-loop active harmonic load-pull setup, thus enabling fast waveform engineering in pulsed mode.
Nowadays it is a big challenge for engineers dealing with high power RF/microwave transistors to perform an accurate and complete characterization of their devices and to build and validate a transistor model. The modeling process usually starts by extracting a simple linear model from S-parameter measurements. An initial nonlinear model is obtained using pulsed IV and pulsed S-parameter measurement systems. These methods are the ones typically used by the transistor modeling community. A second method, based on load-pull measurements, is used to check the validity of the model for large signal behavior.
Load-pull measurements are typically used by amplifier designers to find the optimum operating conditions of the transistor in order to meet specific amplifier requirements, but also to fully understand the operating behavior of their transistors, particularly when they are used in the nonlinear region. The RF designer can today simulate time domain slopes in RF software, and particularly load lines, to preview the behavior of devices. Therefore, it is imperative to validate simulations against measurements, which can be achieved by associating the NVNA measurement system with an active harmonic load-pull setup. This gives engineers the capability to perform time domain load-pull and S-parameters measurements in CW and pulsed mode.
The better way to understand nonlinear phenomena in microwave transistors is to work close to the saturation region with significant power compression. Moreover, most RF power amplifiers operate close to the saturation region, because the power added efficiency (PAE) can be maximized in this area. In the saturation region, nonlinear phenomena and parasitic effects have been observed and have direct consequences on the electrical performance of the components.
More high power devices are used in pulsed mode and in applications where high peak-to-average ratios are common. Consequently, the memory effects taking place in active devices (thermal effects and trapping effects) will be excited more, so they have to be accurately considered, in particular, a shift in the device characteristics. The RF designers can benefit from the transient envelope simulator tools, but, once again, such simulations need to be checked with measurements.
Figure 1 SWAP X402 sample-based NVNA.
The SWAP X402 (see Figure 1) is a sampler-based NVNA product from Verspecht-Teyssier-DeGroote (VTD). It contains four independent and synchronized RF sampler-based receivers, whose role is the conversion of the RF signals into intermediate signals (see Figure 2). The sampling clock, which is driving the four samplers, is generated by a precise and stable local oscillator (LO). This LO is phase-locked to a stable reference that is shared with the RF source and the analog to digital converter (ADC) clock.
Figure 2 Sampler-based NVNA architecture.
As shown the Figure 2, the RF incident and reflected waves a1, b1, a2, b2 at the ports of the DUT, which are separated by low loss couplers (wave-probe or directional coupler), are down-converted by the RF samplers using the high precision local frequency FLO=FRF-KFIF (K is an integer number) and then filtered by a 50 MHz low-pass filter, and acquired by four ADCs. This simple architecture is very efficient for high power measurements.
With this receiver configuration the full wave information is obtained at the calibration planes, as the wave data at the fundamental and harmonics frequencies are acquired simultaneously and coherently by using one-shot measurements. Using robust calibration algorithms, the time domain current and voltage slopes versus time or the load line cycles can be drawn at the reference planes. Moreover, all the information of the performance data, such as the output power, input power, gain, power added efficiency, can also be obtained easily.
Figure 3 Output power and gain vs. input power at 2 GHz.
By using one-shot measurements to get all the RF (fundamental and harmonic frequencies) information, this NVNA can measure 25 dynamic load lines per second, significantly faster than a mixer-based NVNA. It is also simple to integrate it into any load-pull environment, whether active or passive load-pull is being used.
Next is an example of classical time domain load-pull measurements. These measurements were performed on a power amplifier with a fundamental frequency of 2 GHz and whereby two harmonic frequencies are measured up to 6 GHz. The electrical performance is shown in Figure 3 and the time domain drain current and drain voltage waveforms in Figure 4. Figure 3 shows the output power and the gain versus the input power for a given set of impedances. An output power of 25 dBm has been reached at 10 dB of gain compression (circle point). At 10 dB of gain compression, we observe the time domain waveform in Figure 4. These figures convey the nonlinear phenomena and the distortion of the signal at the output of device due to the large compression of the gain.
Figure 4 Time domain wave form and load line for different input power at 2 GHz.
With the presence of a large signal in CW mode, we risk the destruction of the device. Alternatively, the excitation of the high power device with pulse signals allows the characterization of the components with signals that have large amplitudes sometimes up to the avalanche point while reducing the risk of damaging the device. It also highlights some nonlinear phenomena like the thermal effects and the trap effect in newer high power devices. So there is a real need for power amplifier designers to perform transistor characterization under pulsed operating conditions.
This sampling process of CW large signals can be explained in the time domain as a stroboscope. It allows the instrument, in a one-shot measurement, to acquire both the fundamental and harmonic frequencies data for all the incident and reflected waves. The PIV and pulsed RF measurement techniques have already proven their efficiency for the characterization of high power devices. The authors have developed a new approach for the pulsed measurements, one that operates like a second-level stroboscope. This is the so-called Time Domain Approach (TDA) for RF pulse measurements. It is based on a progressive acquisition of all the required data points before sending them to the Fast Fourier Transform (FFT) of the NVNA. Inside every pulse, a defined number of samples is stored, and put together in phase with the preceding ones. This principle, shown in Figure 5, is a stroboscope approach because the RF sampler shots are slightly shifted compared to the observed RF frequency. The exact computation of all the related frequencies of the system – RF signals, RF samplers, ADC acquisition – is a mandatory point, to be sure of the phase coherence between the samples to be put together before the FFT.
Figure 5 Synchronized samples in the IF band frequency.
This method affects the total acquisition time which depends directly on the measurement duty cycle, but does not have any effect on the dynamic range because there are no relative long-term jitter. The amount of detected energy with this approach is the same as that in CW, although the measurement time is extended. This approach was validated with a duty cycle as small as 1/10000 without any dynamic loss. To ensure good performance, the TDA needs a common reference for the triggers of the ADCs and for the RF receiver; all the frequencies must be very accurately synthesized in order to ensure exact ratios and phase shifts. Table 1 shows the acquisition measurement time for different duty cycles.
Active Load-Pull Setup with NVNA
The VTD NVNA measurement system lends itself well to use within pulsed load-pull measurements and as such, the development team at Mesuro has implemented it as the receiver in its current open-loop Active Harmonic Load-Pull solution. The instrument's ability to define the number of samples taken from each pulse and to adjust the time from the pulse trigger signal to the start of the sampling window proves very useful when measuring devices that are intended to be operated within pulsed environments. In simple terms, this means that one is able to gather as much data from each pulse as required while being able to avoid the transition areas. Good correlation has been established so far between CW performance and pulsed measurements on the same device.
Figure 6 Setup for pulse time domain load-pull measurements.
Figure 6 shows the Mesuro/VTD setup for the pulse time domain open-loop Active Harmonic Load-Pull measurement. Figure 7 shows contour plots of a device under test, both under CW and pulsed conditions, establishing the optimum output power performance in each case. Work has also been undertaken to confirm correlation of the waveforms at the output current generator reference plane, where both output voltage waveforms and dynamic load lines have been validated.
Figure 7 Output power contour plots, CW vs. pulsed.
Figure 8 S11 and S22 plots of the transistor from 1 to 18 GHz.
The greatly reduced thermal loading on the device under test, by operating in a pulsed signal environment, enables the designer to undertake measurements far beyond the safe CW operating region and under more realistic operating conditions. This significantly extends the capabilities of the Active Harmonic Load-Pull setups, allowing for a practical methodology for data gathering for use within measurement-based models. It also allows for waveform engineering to be undertaken on devices viewed as thermally vulnerable in order to achieve the high efficiency modes of operation.
The NVNA can also perform scattering parameter measurements in both mode CW and pulsed mode. Here, we present some results of S-parameters measurements in the pulsed mode. These measurements were performed on a GaN device and on-wafer configuration. The pulse configuration was fixed for a pulse width of 200 ns with 0.2 percent duty cycle and without any dynamic loss. The frequency range for these measurements is 1 to 18 GHz. The quiescent point was fixed on the gate and the drain with Vgs = -4.5 V; Vds = 20 V and Ids = 100 mA. Figure 8 shows the data for S11 and S22 of the transistor.
Figure 9 The magnitude and phase of S21 from 1 to 18 GHz.
The magnitude (dB) and phase (degrees) of the S21 are shown on Figure 9. These measurements are very accurate. It is observed that the transistor model is able to accurately represent the S-parameter variations versus the bias points values. The measurements in Figure 10 show the dynamic range up to 18 GHz for two different duty cycles 0.2 and 1 percent. This is the S21 for an open circuit between port 1 and port 2 in CW and pulse mode with a duty cycles up of 0.2 percent. The dedicated stroboscopic approach allows the NVNA to conserve a dynamic range of up to 65 dB.
Figure 10 Dynamic range vs. frequency up to 18 GHz for different duty cycles.
Pulse measurements allow engineers to study nonlinear behavior and thermal effects in devices that might otherwise be damaged under high power CW conditions. Harmonic Active Load-Pull solutions with the capability to handle very high power devices allow designers to perform waveform engineering. A cost-effective sampler-based NVNA with extended pulse capabilities that is compatible in terms of hardware and software with an active load-pull system has resulted in a very versatile and fast Time Domain Load-Pull setup running with the same level of performances in CW or pulsed mode.