In the RF and microwave world, low phase noise is a major aim—alongside higher bandwidth, greater analysis of the environment, better detection of potential threats—and of course, all achieved wirelessly. However, the road to achieving this aim is not smooth, with one particular barrier being the design tools available.
Although phase noise analyzers are becoming increasingly important in engineering laboratories, they come in various types and sizes and use traditional analog techniques or digital signal processing. Researchers have tried to assess such instruments, usually by comparing their phase noise floor limits, but almost none have tried to compare their accuracies. A key element of a phase noise analyzer’s accuracy relates to calibration, and the most popular technique to calibrate them relies on the generation of calibrated spurious signals that can be swept in offset frequency and amplitude. This way, the accuracy of the instrument can be traced back to a national institute, such as NIST, LNE, NIM or NPL depending on the country where the instrument is used, as the spurious generators are themselves traceable.
Calibrating signal source analyzers, phase noise analyzers or spectrum analyzers for phase noise is usually a time consuming process, requiring that the user extract the instrument from its operating environment and ship it to an external calibration provider—with the customs risks and delays that might occur. Such issues could be reduced or eliminated if it were possible to verify the accuracy on-site with a simple, independent solution, which is why Noise XT has developed the PNG-A phase noise generator, which has the added advantage of including amplitude noise generation in the same package. This device is completely independent from any analyzer or any manufacturer’s current design, and is claimed to be very affordable.
The PNG-A, with its internal processor, computes the perfect sine wave in real-time and adds the exact amount of random and discrete phase and amplitude noise. Thanks to the graphical user interface (GUI) (see Figure 1), the user defines noise profiles that can be programmed to reflect the best application where the calibration has to be done. Those profiles are not just basic white noise from a diode or a resistor. The main issue with these techniques is that the reference level is difficult to tune, as the noise will be proportional to 4kBTB, where kB is the Boltzmann constant, B is the bandwidth in Hertz and T is the temperature in Kelvin. The level of this signal phase noise depends on external parameters such as temperature, aging, VSWR and nonlinearity in the chain, which might experience long term fluctuations. Accuracy in such analog solutions would be fully based on another calibration process involving another set of accuracies.
Numerical generation circumvents these issues. By applying deterministic contributions on the phase, amplitude and frequency of a sinusoidal modulation, the user can generate a signal with a well-defined phase noise profile. A CW signal, on top of which the user can add several tunable random or sinusoidal modulations (see Figure 2), is generated inside the Xilinx Zynq system on a chip. Later, in the FPGA fabric, the amplitude and phase noise are mathematically combined, such as shown in Figure 3, then converted to RF. Thus, the noise does not depend on measurement conditions. A purely designed noise profile matches the real world, as its noise density reduces as the offset increases, as in all frequency sources currently available. The PNG-A generates both the sine wave and its relative noise with the same digital-to-analog converter; this guarantees that the small index AM and PM noises will always be at the correct calibrated level. Experimental measurements made with the NXA-6 phase noise analyzer, shown in Figure 4, match expectations well.
One of the key advantages of this digital modulation is the flicker noise generation. This 1/f noise is difficult to produce with analog filters or analog modulation and is seen as a major issue for many applications. As a cherry on the pie, all these parameters can be tuned in “real-time,” making this device stand out and useful in a dynamic calibration process.
The PNG generates a 5 to 35 MHz sine wave which may, at first, seem limited. However, the purpose of this calibrated source is to validate the phase and amplitude noise accuracy, not the whole frequency coverage of an analyzer. Traditional calibrations will not verify operation at all frequencies, as the analyzer “core” is common and its inherent accuracy is basically the same across its complete input frequency range. Calibrating at 10 MHz from 1 Hz offset to 1 MHz offset will be sufficient in 95 percent of the applications, which lowers the cost. The PNG approach to phase noise calibration is particularly relevant in two situations: one, calibrating on-site and in-lab in a matter of minutes, where the hassle of going through an “old fashioned” solution is not really an option, and two, where having the capability for an independent low-cost solution to verify that the data is “real” will satisfy the customer’s expectations.
Technologies (Noise XT)