Extended Analyzer Capabilities Blur Distinctions

Signal analyzer is a generic name for many types of test instruments. Among them are a spectrum analyzer, a vector signal analyzer (VSA), and a real-time spectrum analyzer (RTSA). There are differences among the capabilities that each of these instruments can provide, but many of the techniques they use today are similar. Figure 1 summarizes operation of these three types of instruments.

Figure 1. Block Diagram-Level Comparison of Signal Analyzer Architectures Courtesy of Tektronix

To view a larger image click here

RF and microwave signal analyzers may use two or three stages of down-conversion and filtering to ensure low spurious response and wide bandwidth for the resulting IF signal. Analog down-conversion is based on the nonlinear mixing of the input signal with a local oscillator (LO). Mixing produces both the sum and difference of the input and LO frequencies.

LO and IF planning can be very complicated: The second- and third-order products resulting from the input mixing with LO harmonics must be considered as well as the desired first-order sum/difference frequencies. An IF filter selects the frequency band to be processed.

Although most stand-alone instruments include the LOs, modular instruments may be partitioned so that you have more control of the type of LO. For example, you could elect to use a synthesized signal generator with millihertz frequency resolution and fast switching.

A basic spectrum analyzer measures signal power vs. frequency. You can think of a fixed-width filter being swept through a selected band of the instrument’s frequency range. The filter width is called the resolution bandwidth (RBW) and the selected range the span.

In operation, the LO is set to the appropriate frequency and linearly swept through a range of frequencies corresponding to the length of the span. The resulting IF signal is applied to the RBW filter. In a traditional spectrum analyzer, the LO sweep is driven by an analog sawtooth waveform, but regardless of the circuit techniques used, many modern spectrum analyzers retain the swept designation and display results in a familiar way.

The instrument does not provide phase information although many spectrum analyzers implement the more narrow RBW filters via FFT techniques. This is faster than an actual analog filter, and sharper filter shapes can be achieved. The widest RBW usually is only a few megahertz. If a large signal is close to the frequency of a much smaller one, it’s the job of the RBW to block the large one while allowing the smaller one to be measured. Depending on an instrument’s detailed architecture, phase information may be available and can be accessed through digital demodulation options.

A VSA down-converts a band of frequencies and digitizes the band for subsequent processing. Although some spectrum analyzers also do this, a VSA separates the in-phase I and quadrature Q signal components during analysis of the sampled IF data. This information is used to perform modulation analysis such as error vector magnitude (EVM) calculation and develop a constellation diagram. Because a band of frequencies is simultaneously down-converted, the LO need only be set to multiples of this bandwidth to cover the entire input range.

An RTSA uses VSA techniques but implements the necessary signal processing very quickly via DSP algorithms. Tektronix is closely associated with the term RTSA, and the company’s newest instruments actually have two distinct signal processing paths.

A separate real-time engine processes I and Q digital data to build a composite spectrum display representing about 30 ms worth of time samples. These frames of data are displayed one after another in real time. Much as the digital phosphor technology used in the company’s TDS oscilloscopes presents a real-time overview of signal activity in the time domain, the real-time engine in an RTSA generates a similar display in the frequency domain.

Having discovered unexpected signal behavior in either type of instrument, you then can set up sophisticated triggering states to catch the suspect condition. RTSAs are ideal for exposing time-related problems, especially involving complex modulation or signal interference. Their special capabilities are not particularly relevant to component characterization where you might instead want the lowest possible noise floor and most linear down-conversion performance.

Under the Hood

Signal information is determined by the modulation applied to the carrier. To measure the modulation without distortion requires an instrument with two specifications. The RF or microwave frequency range relates to the highest carrier frequency for which the instrument is suited and is a number of gigahertz generally less than or equal to 26.5.

Typically, manufacturers create families of similar models with different RF frequency limits such as 3 GHz, 6 GHz, and 12 GHz. The modulation or instantaneous bandwidth is the other specification and ranges from several to a few hundred megahertz.

The signal analyzer architecture reflects the need to satisfy both specifications. An RF or microwave input signal first encounters signal conditioning that may include attenuation, amplification, and filtering. However, the most significant operation performed on the analog signal is down-conversion to a lower IF frequency.

The bandwidth, linearity, and noise of the down-conversion process are critical to the analyzer’s overall performance. Ideally, the bandwidth must be larger than the modulation bandwidths of whatever communications standards you may be working with.

Phase Matrix
Figure 2 shows block diagrams of the five PXI modules Phase Matrix developed in support of synthetic instrument (SI) down-converter applications. The input signal is attenuated and switched to one of two paths depending on its frequency. The low-band path is further switched to either a 250-MHz bandwidth bypass or a 40-MHz bandwidth down-conversion channel. The relevant LO frequency is provided by the external LO module or a separate low-band LO driven by the 100-MHz reference oscillator in the external LO module.

Figure 2. Block Diagram of Modular 26.5-GHz PXI Down-Converter SystemCourtesy of Phase Matrix

To view a larger image click here

The high-frequency path within the RF input-conditioning module includes a high-performance YIG filter preselector that defines the band of frequencies applied to the microwave band module. Here, the signal is down-converted to an IF center frequency of 250 MHz with a ±175-MHz bandwidth. The IF output module switches the 250-MHz IF straight through or provides further narrowband down-conversion. Detector stages are included in this as well as the microwave and low-band modules.

Obviously, you may not require the level of customization supported by this family of modules. For example, it addresses both wideband 8-b low-resolution and narrowband 16-b high-resolution applications. This kind of flexibility is key to synthetic instrumentation.

The company envisions this suite of instruments working with successive generations of higher performance digitizers, for example. It’s important to note that within the simplified Figure 1 block diagrams many of the detailed operations shown in Figure 2 still must be accomplished. They’re just not shown.

National Instruments
Following down-conversion, the IF signal is digitized. Recent improvements in commercially available high-resolution ADCs have had a noticeable effect on signal analyzers, according to David Hall, National Instruments (NI) RF and communications product manager. He explained that without a sufficiently wide-bandwidth ADC, an analyzer can only handle part of the signal at a time. Some products create a wider spectrum by stitching together the analysis results from separate acquisitions. This approach combines data that occurred at different times and is slower than acquiring the entire band in one acquisition.

In contrast, a wide-bandwidth ADC that simultaneously captures the entire signal of interest ensures the fastest operation and reduces errors. NI’s latest VSA, the Model PXIe-5663, has a 10-MHz to 6.6-GHz frequency range and a 50-MHz, 3-dB instantaneous bandwidth. High-resolution ADCs such as the 16-b device used in this instrument are available with sampling rates greater than 100 MS/s. For example, Linear Technology’s LTC2209 samples as fast as 160 MS/s and Analog Devices’ AD9461 at 130 MS/s.

DSP following digitization also has progressed. NI has combined PXI Express (PXIe) serial data transport technology with multicore host parallel programming via LabVIEW. The result is high performance in a compact instrument. It’s also interesting to note the significant reduction in both phase noise and noise floor in the PXIe-5663 compared to previous 2.7-GHz NI instruments.

At this year’s NIWeek, BAE Systems’ Technical Fellow Robert Lowdermilk described some of the signal processing work his company is basing on NI’s technology that facilitates FPGA programming via LabVIEW. Once the IF signal from the Phase Matrix PXI down-converter modules has been digitized, virtually any type of analysis can be accomplished via DSP techniques. BAE Systems is augmenting host-based parallel processing with FPGA-based hardware algorithms to achieve the necessary high speed in a programmable SI approach to RF signal analysis.

Keithley Instruments
The Model 2820 with an RF frequency range from 400 MHz to either 4 GHz or 6 GHz and a 40-MHz modulation bandwidth is a VSA that also performs as a spectrum analyzer. A design emphasis similar to that of a software-defined radio is responsible for the instrument’s flexibility. That is, the hardware has been kept as generic as possible with the specific functionality determined by software. In this case, an FPGA interfaces data from the 14-b, 100-MS/s ADC to the waveform memory, digital up/down-converter, and 500-MHz DSP as shown in Figure 3.

Figure 3. Block Diagram of Model 2820 VSACourtesy of Keithley Instruments

To view a larger image click here

Running the ADC at a constant 100 MS/s means that aliases always fall in the same band, and a precision fixed filter can be implemented. However, that rate is inappropriate for many applications, so digital up- or down-conversion decompresses or compresses waveforms, establishes a range of signal sampling rates, conserves memory, and speeds signal processing. The DSP runs algorithms that analyze and demodulate the waveform.

One key benefit that results from this type of architecture is fast measurement speed. According to Ron Rausch, senior marketing manager at the company, “A complete set of WCDMA measurements can be made in 60 ms, and it takes about 38 ms for 802.11b WLAN. Although only 10 measurement results are reported, hundreds of analysis measurements are needed to produce them. For example, measuring the rms and peak EVM requires measuring the EVM of 52 subcarriers, calculating the rms value, and then displaying it along with the peak value. For WiMAX, similar measurements are made on up to 2,048 subcarriers.”

Rohde & Schwarz
The recently introduced Rohde & Schwarz Model FSV with 3.6- or 7-GHz bandwidth is simply called a signal analyzer. The 28-MHz bandwidth covers most of today’s wireless standards including LTE’s 20-MHz bandwidth. A 40-MHz option deals with future WLAN 802.11n wideband technology that allows two 20-MHz bands to be bonded together.

Three stages of mixing and filtering are used to develop IF signals centered at 8,410 MHz, 730 MHz, and 90 MHz. The final IF output is digitized to 16-b resolution with a 128-MHz sampling rate. A 200-MS I/Q memory depth is provided to ensure a long recording time even when testing high-bandwidth/high sample rate systems. The FSV will perform wideband modulation measurements during development of chipsets and mobile stations as well as in the development, maintenance, and installation of infrastructure.

Anritsu
Anritsu’s Model MS2690A Signal Analyzer can include a vector signal generator (VSG) option. Keithley Instruments has taken a similar approach. Both companies are going beyond offering only a VSA or a VSG to provide a complete test setup with which you can generate a complex digitally modulated signal as well as analyze how your target DUT responds to it.

Sean Grisier, field applications engineering manager at Anritsu, said, “Amplifier testing is a primary application addressed by the Series MS2690A Analyzer. Often, customers use VNAs for this, but the MS2690A has comprehensive demodulation functions for the latest wireless standards. In particular,” he continued, “LTE uplink and downlink demodulation options support amplifier testing for the base transceiver station (BTS) and user equipment (UE) markets. Both LTE analysis and VSG capability have proven to be critical factors in this application.”

A 31.5-MHz IF bandwidth is standard with a 125-MHz option. It’s becoming common for users to be given access to raw I/Q data following digitization, and Mr. Grisier highlighted the usefulness of this facility. “Especially if the signals do not conform to commercial wireless standards, the user can elect to have the data post-processed off-line. It can be imported to an off-line tool, processed on the MS2690A, moved to a USB memory device, or transmitted to an external PC via the high-speed eSATA interface.” The IF output signal is centered on 874.988 MHz for spectrum analyzer operation and at either 875 or 900 MHz for VSA operation depending on bandwidth.

Three models are available to address frequencies up to 6, 13.5, or 26.5 GHz. Built-in calibration oscillators ensure total level accuracy within ±0.5 dB up to 6 GHz.

Avoiding the need for a pre-selector below 6 GHz is one of the factors the company cites in achieving the low measurement uncertainty. In addition, an integral phase calibration oscillator compensates IF filter performance and helps provide accurate modulation measurements.

Agilent Technologies
The Model MSX N9020A Signal Analyzer features a large number of distinct measurement applications for specific wireless standards as well as the capability to run the company’s more general-purpose 89601 VSA software. Four models have frequency ranges of 20 Hz to 3.6 GHz, 8.4 GHz, 13.6 GHz, and 26.5 GHz.

Five separate bands are listed for the MSX Series as well as for the Anritsu MS2692A to cover the full 26.5-GHz range. Anritsu has a wide first band up to 6 GHz, which avoids switching between bands when working with both 2.5-GHz and 5-GHz carriers such as used by WiFi 802.11a/b/g. In contrast, the first band of the Agilent MSX Series extends to only 3.6 GHz.

Standard IF bandwidth is 10 MHz with a 25-MHz option. A 14-b ADC running at 90 MS/s digitizes the down-converted signal and feeds a completely digital IF section. Depending on the selected frequency span and RBW, a swept digital RBW filter may be a better choice than FFT filter processing, and the instrument can be configured to automatically choose the better approach for the fastest sweep time.

A separate analog baseband capability also is available with dual-channel I/Q 16-b 100-MS/s digitizing. Converted data has I/Q corrections applied to improve accuracy and is digitally up- or down-converted as required and finally stored in a 500-MS memory. By running 89601 VSA software on your MSX platform, you have access to more than 50 demodulation formats with which to verify the baseband signal. And, all MSX instruments can be upgraded to include the baseband option.

Tektronix
Because of the emphasis Tektronix has placed on speed in the RTSA instruments, unfamiliar capabilities have been provided in addition to those usually found on spectrum analyzers and VSAs. One in particular, overlapping FFTs, is an operating mode that provides greater insight into the behavior of very short, fast events. The company has backed up the innovative nature of the RTSA products with in-depth technical information to help users understand these types of spectrum displays.

The RTSA processes digitized data in 1,024-sample frames, producing a 512-point FFT from each. In the spectrogram display, time progresses up the screen, the frequency span is linearly displayed from left to right, and color or intensity indicates signal amplitude. If one spectrogram line corresponds exactly to one frame, then at the maximum 50-MS/s I/Q data rate, each line represents about 20.5 µs.

Very short events do not contribute significantly to the spectrum developed for a single frame. For those events near the ends of a frame, windowing further reduces their effect. Because coherent sampling cannot be guaranteed, windowing must be applied to the FFT frame data, and this intentionally attenuates samples near either end of a frame. Overlapping the FFTs solves these problems although it introduces additional new ones in the process.

Instead of starting a new spectrogram line—that is, performing a new FFT—only after the 1,024-sample buffer is full, the FFT is computed sooner using partly old data and partly new data. If the overlap is large so that 90% of the data is old and only 10% is new, then the spectrogram line rate will be 10 times faster: 10 lines will be used to display the same spectral information that previously was shown on one. Naturally, this stretches the spectrum display vertically in time.

However, beyond the disadvantage of the unusual appearance this causes, the actual FFT operation now can correctly deal with infrequent, very short events. If an event occurred near the end of the first complete frame, with 90% overlap it now will show up 10%, 20%, and so on into successive frames. This means that it will not be excluded by windowing nor will it be missed because its contribution is so small.

Summary

As with many things, and certainly test instruments, the devil is in the detail. It should be obvious that few of the instruments being sold today by major test and measurement vendors really are purely spectrum analyzers or VSAs. They are based on conventional RF/microwave front-end technology, but after the IF signal has been digitized, the instrument features are limited only by the designer’s imagination. Microprocessors, FPGAs, and DSPs are all used to perform digital algorithms that mimic traditional spectrum analyzer and VSA performance as well as offering new capabilities.

Deciding which instrument best meets your needs can be approached by considering either the input RF/microwave specifications or the backend digital performance. After all, the frequencies seen by the front end of a signal analyzer are much too high for direct digitization, so the specifications remain similar to those of traditional spectrum analyzers and VSAs. Unless your application requires very fast measurement speeds, front-end performance could be the deciding factor. Many of the instruments allow the user to access the I/Q data directly, and several companies provide off-line analysis software.

If speed is important, then consider the demodulation options available for most instruments, which support a wide variety of modulation standards. And, don’t underestimate the usefulness of a built-in VSG if you must develop the stimulus as well as measure your DUT response. Finally, make sure you understand how the manufacturer thinks the instrument will be used. Chances are you will be happier with a signal analyzer designed to address applications like yours than with one that wasn’t.

November 2008