Types of RF Signal Analyzers
When capturing RF signals, engineers are typically interested in signal characteristics such as amplitude, frequency, and phase. Depending on the characteristics you need to analyze, you might choose either a spectrum analyzer or a vector signal analyzer. (Note that a third instrument, the vector network analyzer, also performs analysis, but we do not address it in this paper.) The spectrum analyzer is used to capture only the frequency and power information of an RF signal. The typical output of this instrument is a power versus frequency graph, which we observe later in this article.

A vector signal analyzer, on the other hand, is capable of the same measurements as a spectrum analyzer but with added capabilities. Because a vector signal analyzer captures the time-domain of the RF signal as well, you can acquire phase information to produce a constellation plot, shown in Figure 1.

Traditionally, spectrum analyzers and vector signal analyzers used different instrument architectures. The traditional spectrum analyzer consists of basic components such as a tunable local oscillator (LO), a mixer, a bandpass filter, and a power sensor. To make spectrum measurements, the traditional analyzer simply tunes the LO to each frequency bin and makes a power-in-band measurement on the resulting signal. By sweeping through each frequency bin, a traditional spectrum analyzer can provide power information through a broad range of frequencies. Some spectrum analyzers still operate in this mode, known as the "swept" mode. The architecture of a traditional spectrum analyzer is shown in Figure 2.

Many modern spectrum analyzers are designed in much the same way as a vector signal analyzer. This architecture, shown in Figure 3, uses a tunable LO mixed with the RF signal to produce a wideband intermediate frequency (IF). Rather than retuning the LO for each frequency bin, however, the analyzer simply performs a fast Fourier transform (FFT) on the IF signal. The FFT can provide power and frequency information that spans a broad frequency range with a single acquisition. As you might expect, the architecture of a more modern vector signal analyzer is actually quite similar to that of a vector signal generator. See Figure 3 for a simplified block diagram of a superheterodyne vector signal analyzer, which uses an IF signal.

As the diagram illustrates, an analog-to-digital converter (ADC) captures a broader spectrum of data. By acquiring a broader spectrum, you can capture the phase information of the RF signal as well. In addition, this enables a vector signal analyzer to perform spectrum measurements with a simple FFT calculation. Attenuation and Reference Level
RF signal analyzers are designed to measure many types of RF signals with the greatest dynamic range possible. One way to maximize the dynamic range over a broad range of signals is to use attenuation to adjust the signal level to the ideal amplitude for a given signal. Typically, RF analyzers are designed to have a broad range of reference or attenuation levels, specified in decibels (dB). From a user experience, the reference level is typically set to a power level that is slightly higher than the maximum expected power. The instrument then applies appropriate gain or attenuation to the signal. Attenuation or gain is typically applied as close to the RF front end as possible to maintain a constant signal level at the mixer and to achieve maximum dynamic range on the signal being analyzed.

Programmable attenuation or gain is important because it allows an RF instrument to measure signals at a variety of power levels. For example, if you were to attach a broadband antenna to a signal analyzer, you would notice that many of the over-the-air wireless communications signals operate at greatly varying power levels. Most FM radio stations can be observed at maximum amplitudes of around -50 dBm. By contrast, unless you are close to a base station, it is hard to find signals in the GSM cellular band that are higher than -70 dBm. Moreover, in an even more extreme scenario, GPS signals in the 1.57 GHz band might operate at power levels of -157 dBm and below.

When choosing an analyzer, be sure to check the range of attenuation that it offers. The combination of maximum attenuation and dynamic range determines the minimum signal level that can be viewed. In some cases, RF instruments have optional preamplification. A preamp can provide for analysis of extremely low-level signals by amplifying them with a low-noise amplifier.

Dynamic Range
The dynamic range is a specification that describes the maximum and minimum signal amplitudes that you can measure simultaneously. The maximum signal level is determined by the attenuation applied to the signal, and the minimum signal level is determined by a number of different factors. These factors include noise introduced by the amplifier, spurs and harmonics, or even carrier (LO) leakage. More specifically, defining dynamic range is the ratio of the largest signal that can be measured relative to the power of the greatest distortion, noise, or spur. This parameter is specified in dB, with a larger range being more desirable.

Spurs and noise can be introduced almost anywhere in the RF signal chain. The nonlinear characteristics of components such as mixers and amplifiers often result in distortion products, each of which can produce spurs in the frequency domain. Moreover, the bit resolution of the ADC also can affect dynamic range. As a general rule, the greater the bit resolution of the ADC, the better its dynamic range is.

Dynamic range is an important specification for low-amplitude measurements. It is even more important when measuring a low-power signal level next to a high-power signal. Because the reference level of the instrument cannot be set below the maximum power of the high-power signal, the dynamic range of the instrument determines the minimum signal that it can view next to a high-power signal. This is illustrated in Figure 5, which shows a low-power signal adjacent to a high-power GSM signal. As the figure demonstrates, a signal analyzer must have a dynamic range of at least 60 dB to measure the smaller signal.

Averaging Methods
With averaging methods, you can improve the accuracy of measuring low-level spurs by reducing the affect of noise on a signal. By averaging over several periods of a signal, you can eliminate random or white noise from the signal and converge to its real value. In this section, we describe root mean square (RMS) and peak-hold averaging.

There are several different methods to perform complex averaging. You can determine the RMS average to average the power or energy of the signal by calculating the weighted mean of the sum of squared values. With RMS averaging, your instrument can detect low-level signals. While noise might typically obscure these signals, averaging enables the periodic noise components of the signal to average out, leaving only the desired signal. This is illustrated in Figures 5 and 6, which show the FM band with and without RMS average and demonstrates more accurate detection of low-level peaks.

Peak-hold averaging is another averaging method you can use to combine multiple spectrum acquisitions into a single spectrum measurement. Peak-hold averaging keeps the peak of each bin through multiple FFT calculations. As a result, peak-hold averaging raises the noise floor because it takes the highest amplitude of all signals measured for many averages. However, while peak-hold averaging raises the noise floor of the measurement, it also enables identification of transient signals by showing peaks of subsequent spectrum measurements on the same graph. This is illustrated in Figures 7 and 8, which show the 885 MHz GSM cellular band with and without peak-hold averaging enabled.

It is important to note that averaging should be used with caution because it can affect the accuracy of carrier-to-noise measurements.

Displayed Average Noise Level
As described in the last section, the apparent noise floor of an RF analyzer depends on much more than the noise introduced by the RF system. The averaging mode that you use can significantly affect the average noise floor. This section describes how the resolution bandwidth (RBW) of the signal can affect the average displayed noise floor of the instrument as well. To illustrate this concept, we have measured a 20 MHz bandwidth with a single peak. As Figures 9 and 10 show, reducing the resolution bandwidth actually lowers the displayed noise floor of the instrument.

You can see that the displayed average noise level (DANL) of the instrument is highly dependent on the resolution bandwidth being used. This is an important specification because it provides an indication of the smallest detectable signal that the instrument can display. Because the DANL is dependent on various settings of the instrument, it is typically specified along with the conditions at which it was measured. For example, a typical DANL specification might read something like: -115 dBm between 1 GHz and 2.7 GHz with resolution bandwidth (RBW) set to 1 kHz with 0 dB input attenuation at 25deg C. Because the apparent noise floor of the instrument increases with a wider resolution bandwidth, the noise floor is often normalized to a common RBW, which is often 1 Hz.

When comparing the DANL between two manufacturers, it is important to ensure that both measurements are normalized to the same bandwidth. To make a fair comparison, the easiest technique is to normalize both instruments to 1 Hz RBW. Simply subtract 10 log (RBW) from the given noise floor measurement. For example, an instrument that shows a noise floor of -115 dBm at a 1 kHz RBW will show a noise floor of -145 dBm at a 1 Hz RBW. By normalizing both instruments to the same bandwidth, you can compute a fair comparison of instrument performance.

Note that many traditional spectrum analyzers normalize measurements to a 6-MHz video bandwidth. With some simple math, you also can take any measurement normalized to 1 Hz and normalize it to 6 MHz as well. Simply add 10 log (6 MHz), which is 67.8, to the measurement that has been normalized to 1 Hz. In this scenario, a measurement of -145 dBm normalized to 1 Hz will be -145 dBm + 68 dBm = -77 dBm.

The important take-away is that the displayed noise floor of an instrument depends on the bandwidth being used. Thus, when comparing multiple instruments or making noise measurements of the device under test, be sure to normalize the signal level to the appropriate bandwidth.

We hope you have enjoyed this three-part series on RF instrument specifications. For more tips and techniques on making RF measurements, visit the RF Developer's Network, a semimonthly multimedia publication designed to teach basic RF measurement fundamentals.

Related Articles

• Use spectrum/signal analysis to limit RF power and spurious noise emissions
• Using Real Time Spectrum Analyzers to test passive RFID tags
• Testing security and association for Certified Wireless USB