Feature Story

SDR MEASUREMENTS POSE CHALLENGES

Software Defined Radio (SDR) is a very powerful concept. Its technologies provide software control of a variety of modulation, interference management and capacity enhancement techniques over a broad frequency spectrum (wide and narrow band), while ensuring secure communications management. There remains a key issue which has yet to be resolved — how to verify the performance of an SDR, especially with waveforms that don't exist yet. An important part of the need for this verification is that radios have many limitations that are independent of the software running on them. For example, the software may be able to tell the radio to tune to a particular frequency and set the output power level, but if the radio's amplifiers, matching networks, and antennas don't have those capabilities, the radio isn't going to work. This dichotomy creates a critical need for a way to verify the underlying analog capabilities of an SDR.

Presently, there is no single-instrument "SDR transmitter tester" or "radio conformance tester." Testing an SDR platform requires both a test signal source and a wideband signal analyzer, so engineers are creating their own solutions that combine a signal analyzer, arb-based vector signal generator and analysis software. Both instruments, when used in SDR testing, must provide enough frequency range (0.5 MHz to 3 GHz) to cover the radio's range, while also having enough instantaneous bandwidth to handle different SDR modulation formats. In addition, the signal analyzer may need to measure harmonics and spurious signals to even higher frequencies.

Engineers also need to establish SDR verification philosophies based on certain criteria that include both hardware and software tests. When establishing the verification philosophies, engineers need to consider some important RF measurements, including power, frequency response, phase noise, and spurious signals. In addition, modulation quality measurements, such as EVM, can be done on a variety of signals to verify the baseband and IF signal processing capabilities of the radio.

Power

A traditional measurement that carries over to SDR applications is average power. However, there are new considerations that make power measurements more difficult. Digital radios can have very high peaks. For example, multi-carrier WCDMA or OFDM signals can have peak power of >10 dB. Utilizing "peak reduction" technology is helpful but it may adversely affect the error rate. Other considerations when making power measurements are the advanced amplifier technologies used in SDR development and that some communications standards define test signals with known peaks, such as WCDMA Test Model 1, while other communications standards have not developed such test signals.

Even measurements as simple as signal power require thought. One key factor that must be considered is if the signal is bursted - that is, does it turn on and off? If so, what can be used for a trigger signal? Other factors include the signal's bandwidth and if external components are between the test equipment and Device Under Test (DUT). Many modern signal analyzers and power meters can make these measurements, but engineers must be aware that measuring a particular kind of signal may be beyond a specific instrument.

For example, measuring bursted signals with an average power meter is problematic because it doesn't allow the power level to be known while the signal is just on. It also doesn't allow for a particular part of a signal to be measured—such as a preamble or other "training sequence" that is sometimes used as a power reference. A peak power meter or signal analyzer is a better choice for measuring the power on bursted signals. Some signals even require synchronization to a particular part of the demodulated signal, which makes a signal analyzer the instrument of choice.

An example of a simple power measurement using a signal analyzer is shown in Figure 1. In this case, the signal is not bursted so a simple swept spectrum of the signal with averaging can be viewed, and then the power can be determined by integrating across the frequency band of interest. This is the Channel Power measurement shown below the spectrum.

Figure 1: Measurement of an SDR's power using a signal analyzer

Frequency Response

In SDR applications, receiver equalization compensates for frequency response. However, frequency response variations can have an impact on the effective coverage area of a network, because these variations reduce the margin at the receiver. For example, a 1 dB frequency response variation might cause as much as a 1 dB, or about 20%, change in coverage area. This is a very important factor, as coverage area is critical for commercial networks, as it affects the number of expensive base stations that need to be deployed.

Another factor engineers must consider is that SDRs may utilize spread spectrum operation. The result is greater importance in the RF frequency response requirement due to the wider range of RF frequencies that may be used.

This frequency response can be measured in a variety of ways. If you are measuring the components used in the radio, a Vector Network Analyzer (VNA) is often the instrument of choice, as the VNA provides extremely accurate measurements. If the entire radio needs to be measured, however, a signal analyzer or power meter is the correct instruments to use. In the case where you are measuring components or subsystems and you need additional measurements such as modulation quality on the same DUT, a signal analyzer and vector signal generator can be a good choice for the frequency response measurements.

Depending on the particular conditions, you might measure the level of CW or modulated signals at various frequencies, measure a modulated signal and use a measurement equalizer, or use specialized test signals such as chirps, noise, or noise-like signals. Figure 2 shows an example of measuring the CW signal from a signal generator through a fairly flat amplifier over the frequency range of 100 MHz to 1 GHz. In this case, the signal generator was stepped in 1 MHz increments, and the signal analyzer was set to the Max Hold trace function to capture the entire frequency range. There is approximately 1 dB variation peak-to-peak across the frequency range.

Figure 2: A signal generator's CW signal level measured through an amplifier over a 100 MHz to 1 GHz range.

Phase Noise

In order to determine phase noise requirements in an SDR, engineers must consider both the transmitter and receiver local oscillator requirements, along with the receiver's timing recovery. This is difficult, considering the timing recovery is usually unspecified.

Phase noise is a common measurement for oscillators, and modern signal analyzers often have optional capability for making these measurements, and showing the results in graphical form. The offset range, sweep type (FFT or swept), and amount of averaging are usually adjustable, depending on the testing parameters. By overlaying multiple measurements, comparisons can be made of device performance under various conditions, such as carrier frequency, bias voltage, or temperature. This creates a much more accurate determination of an SDR hardware performance.

An example phase noise measurement of a 2 GHz synthesized oscillator, with a jitter measurement of 102 femtoseconds between 100 Hz and 100 kHz offset can be seen in Figure 3.

Figure 3 - Phase Noise and jitter measurements of a 2 GHz synthesized oscillator.

Spurious Signals—Harmonics, Intermodulation and More

Measuring the harmonics, intermodulation and other spurious signals of an SDR is extremely complex. Typically, third or fifth order harmonics must be measured but in some designs even higher order harmonics need to be measured. Harmonics and intermodulation are usually dominated by the power amplifier, and vary depending on the power level and amplitude statistics. Spurious signals can come from a variety of places, including oscillator leakage and intermodulation products. Intermodulation can cause many unwanted responses, including adjacent and alternate channel interference. Multiple transmitters sited near each other can also intermodulate, causing interference signals to appear at a wide variety of possible frequencies.

Measuring harmonics and intermodulation is best done using a signal analyzer or spectrum analyzer so that the level of the intermodulation signal in the presence of the transmitted signals can be seen. The dynamic range of the signal analyzer is critical to these measurements. For some signal types, however, the measurement requirements exceed the capabilities of even the best signal analyzers. In these cases, special measurement techniques are required, such as measuring the signal before a transmitter's output filter, or using filters to eliminate the large transmitted signal from the input of the signal analyzer.

Figure 4 is an example sweep showing harmonics and spurious signals on the same 2 GHz synthesized oscillator. The relative and absolute level of the harmonics and major spurious signals are clearly seen. For example, the 2nd harmonic is at -52 dBm, or about 49 dB below the carrier.

Figure 4: Harmonics and spurious signals on a 2 GHz synthesized oscillator

Linearity

As mentioned in the spurious signals section, the performance of RF components varies with power level, as well as amplitude statistics. Traditional FM waveforms, for example, are very easy for amplifiers, as the amplitude is constant. The constant amplitude allows the use of class-C amplifiers, which can be extremely efficient. However, many types of digital modulation can have a very large variation in amplitude. These variations place large demands on the analog components, especially the transmitter's power amplifier. Two key measurements that can help are linearity (gain versus level, also known as power sweep, or gain compression), which is usually measured with a VNA, and the Complementary Cumulative Distribution Function (CCDF) measured with a signal analyzer or peak power meter.

The CCDF measurement shows the probability of a peak occurring in a signal, and is useful for two reasons. First, the size of a signal's peaks can be seen, allowing for a comparison of the peak size to the linearity measurements made on a VNA. This will indicate if the signal will be compressed and if so, how often. Secondly, a comparison of the CCDF can be done, to determine if any compression is happening. If there is compression, it will also indicate how much and how often.

Modulation Quality—Error Vector Magnitude (EVM)

The best measure of a radio's ability to send high-quality signals is to directly measure the modulation quality. There are a variety of measurements that all indicate the relative level of the signal to everything else—noise, phase noise, jitter, distortion, and in some cases frequency response errors. Unfortunately, EVM is somewhat specific to the particular modulation parameters being used at the moment. Just because EVM is low for one symbol rate, for example, it is not necessarily also low for another symbol rate. Thus, EVM is an excellent check of the radio's analog performance for a particular set of conditions, but it is not necessarily indicative of another set of conditions.

In addition to EVM, and the related overall metrics of Rho, RCE (Relative Constellation Error), and MER (Modulation Error Ratio), there are a number of other measurements that can be used for troubleshooting tools. For example, measurements of IQ imbalance, carrier leakage, and quadrature error indicate the gain match, offset, and phase match of analog modulators. EVM versus time can show amplitude droop or frequency drift. EVM vs. frequency can help find IF frequency response errors and interfering spurious signals.

When measuring modulation quality of a complete transmitter the signal analyzer is the correct tool to use. The signal analyzer can emulate the receiver in an SDR. A set of measurement functions that verify the performance of an SDR's transmitter, such as channel power, adjacent-channel power ratio (ACPR), occupied bandwidth (OBW), and modulation quality measurements for the QAM and PSK signals often found in SDRs, can be conducted with this type of instrument.

If you have components or subsystems to measure, the combination of signal analyzer and vector signal generator are ideal for modulation quality measurements. In this case, the combination signal analyzer/vector signal generator allows real-world testing scenarios to be created. The vector signal generator can act like the software-programmable transmitter portion of an SDR, allowing the generator to be programmed for existing and emerging waveforms. If the generator has a dual memory design, you can also use it to emulate a pair of waveforms, allowing you to add controlled noise or interference to the test signal, allowing measurements under real-world conditions.

Conclusion

Test tools for SDR RF measurements should effectively emulate the behavior of the radio's receiver and transmitter sections with adequate frequency range, wide modulation and demodulation bandwidths, wide dynamic range, and excellent level accuracy to handle existing and emerging SDR designs. These tools must include flexible test instruments that allow for standard modulation and custom modulation analysis to be performed.

Engineers also need to develop verification philosophies that incorporate both software and hardware issues in order to verify the overall performance of transmitters and receivers.

Return to: 2007 Feature Stories