Considerations When Conducting Noise Figure Measurements
By Sean Flavin, Anritsu Company
The signal-to-noise ratio (SNR) at the input of a radio receiver is a key parameter of communication systems. To achieve desired SNR, design engineers must accurately measure the noise figure (NF) performance of receivers in order to ensure their products meet specified performance. There are many factors that engineers should consider to conduct accurate NF measurements, including the performance of the test instrument and external factors.
NF measurements can be made with different instruments, including vector network analyzers (VNAs) and signal analyzers/spectrum analyzers. If the latter test solution is used, engineers need to evaluate a few key specifications.
In a spectrum analyzer, a low-noise amplifier (LNA) – called a preamp – is used before the first mixer to improve the Displayed Average Noise Level (DANL). The preamp is most effective when it is positioned at the front-most block. When measuring the low power levels of Noise Figure, it is best to use an internal preamp to be above the spectrum analyzer DANL. Attaching another external preamp to the RF input of the spectrum analyzer can improve the DANL even further.
An example of this preamp design can be found in the Anritsu MS269xA and MS2830A signal analyzers (Figure 1). They have DANL of -155 dBm/Hz (30 MHz to 2.4 GHz) and -153 dBm/Hz (30 MHz to 1 GHz), respectively, that allows an accurate Noise Figure measurement.
This kind of performance is required if the direct method is used to measure absolute noise power directly when calculating Noise Figure. The advantage of the direct method is a simple system configuration, but the test instrument used must have high performance. The uncertainty of the calculated NF value can be very different depending on the measured noise level, so the spectrum analyzer must be able to measure small NF values.
Engineers can also measure NF using the Y factor method. In this method, a noise source generates two signals with different power levels, which are input to the DUT. The NF of the DUT is calculated by comparing the SNR of the inputs and outputs for the two signals. The first step is to calibrate the measurement system by connecting the noise source directly to the spectrum analyzer input. The noise source is pulsed by a 28 Volt DC power supply from the spectrum analyzer to provide "hot" on and "cold" off power levels. Additionally, DC voltage is sometimes output, depending on the noise source. When using this type of noise source, a DC block must be used.
Measurement conditions must be established by setting the noise source parameters, followed by measurement frequency range, number of measurement points, measurement bandwidth, analysis time and Storage On/Off. Lengthening the analysis time and setting the averaging processing with the Storage On/Off setting improves the measurement accuracy but increases measurement time.
Figure 2 illustrates the measurement uncertainty during calibration of the test system and when the DUT is being tested. When determining the NF of the DUT using the Y factor method, the DUT gain must be measured. When measuring the relative values of the two levels at the spectrum analyzer (at calibration and with DUT connected), the uncertainty is standardized as the linearity error. The values of these uncertainties vary, depending on the DUT and the Excess Noise Ratio (ENR) of the noise source. As a result, an Uncertainty Calculator tool can be used to calculate the uncertainty by inputting parameters.
Since NF measurements involve measurement of extremely small noise powers, it is necessary to consider the status of the DUT. Extraneous wireless signal interference can make it difficult to obtain accurate measurement results, but errors can be prevented by shielding the DUT from this external RF energy by using a shielded enclosure or screen room. Measurement errors may be prevented by protecting the DUT from external factors, such as RF energy, by making NF measurements in a screen room or using a shield case.
For NF measurements, it is necessary to consider the total power injected into the mixer of the spectrum analyzer. Consider the case when using a noise source with an ENR of 24 dB. When this noise source is off, wideband noise of about –174 dBm/Hz is output; when it is on, wideband noise of about –150 dBm/Hz is output. These noise components are band controlled by the stage before the mixer input due to the spectrum analyzer internal blocks and are input to the first mixer of the spectrum analyzer. As a result, when the noise source is connected directly to a spectrum analyzer with 6 GHz bandwidth, and the noise source is on, the mixer input level is – 52 dBm. (–150 dBm/Hz + 10*Log [6 GHz] )
Spectrum analyzer linearity error performance must be evaluated because of the effect on the spectrum analyzer measurement. If a higher input level is input to the spectrum analyzer, the gain cannot be measured accurately because linearity cannot be assured due to distortion at the internal semiconductor parts. Always choose a noise source with an ENR matching the gain and bandwidth of the DUT to be measured, and set the attenuator at measurement.
The wideband noise of the diode is created by applying a high voltage reverse DC bias to create an electron avalanche but this can lead to a slight DC offset on the RF output of the noise source. Some noise sources have an internal filter to block this DC offset from reaching the spectrum analyzer input but a DC blocking filter might have to be inserted at the input to the spectrum analyzer to make accurate measurements.
Accurate Noise Figure measurements are critical when evaluating the performance of a communication receiver system. The designer must be aware of equipment measurement limitations, in addition to environmental factors.
this article to a friend!