Making RF Power and Frequency Measurements with Spectrum and Signal Analyzers

This application note explains how fundamental RF power and frequency measurements are made with both traditional spectrum analyzers and with more modern FFT-based analyzers. The document also covers instrument controls and how they affect your measurement results.

Introduction

Frequency-domain measurement of RF power is one of the most basic measurements performed by traditional spectrum analyzers and modern vector signal analyzers. For example, interoperability of communication systems requires these systems to comply with standards-based limits on power transmitted in the assigned bands as well as undesired or spurious emissions in neighboring bands -- one person's music is somebody else’s noise. Unless appropriate measurement techniques are applied, results can incur significant errors. The goal of this application note is to cover some instrument and measurement basics to avoid such errors.

If you are designing or producing an RF communication device, whether wireless or wireline, you will likely encounter standards-dictated spectral mask test requirements. A spectral mask test requirement for the DOCSIS 2.0 cable modem standard is shown in Figure 1.

The example in Figure 1 will be referred to throughout this document. As the figure indicates, the goal is to ensure that the spurious tones and noise generated by the cable modem are below the limit set by the standard. To configure the spectrum analyzer to make the measurement, you must set a number of frequency and amplitude controls (shown in Figure 2).
The picture shows a simplified block diagram of the input stage of an RF analyzer. This paper will refer to this diagram throughout the rest of this paper.




The following paragraphs describe the key amplitude controls and how they can affect your measurement results.

Reference Level (Ref Level): Reference Level sets the maximum input range of the spectrum analyzer. Reference Level controls the y-axis behavior of the instrument, similar to Volts/Div on an oscilloscope. Reference Level should be set to just above the maximum power level expected in the measurement. If the reference level is set too low, the input signal saturates the instrument and creates distortion products in the instrument signal chain, which results in inaccurate measurements. If the reference level is set unnecessarily high, the instrument sensitivity and dynamic range are reduced. For example, raising the reference level on most spectrum analyzers adds more attenuation in the instrument signal chain. There is a direct relationship between this attenuation and the instrument noise floor; the more attenuation is added, the higher the noise floor becomes. The optimum reference level is a balance between minimizing the instrument distortion and the instrument noise floor. In some cases, it may also be advantageous to intentionally set a low reference level (creating instrument distortion) for broadband noise measurements. Doing so improves measurement sensitivity as long as you are cognizant of the distortion products and ensure that they are not included in your measurements. You must also make sure that the input signal does not exceed the maximum input level of the instrument, as this risks permanent damage to the instrument. Most RF analyzers can safely withstand a maximum input signal of +20 dBm without attenuation.

The input range of the instrument also can be set by the Attenuator Settings control. This control refers to the mechanism by which the input range is adjusted inside the instrument. High-power signals are measured by setting the attenuator settings to higher levels. Normally this control is set to AUTO so that the software adjusts the attenuation level based on the Ref Level setting; however, you can change the Attenuation settings instead of Ref Level. Traditional instruments like spectrum analyzers tie the display's y-axis to the Reference Level or Attenuator settings in firmware, but virtual instruments are not constrained -- the display's y-axis can be decoupled from these controls if desired. This feature is useful for visually zooming in on the interesting portion of the spectrum without affecting the amplitude level setting of the instrument. Notice that the Ref Level and Attenuator settings both affect the programmable attenuator shown in Figure 2, so only one of these controls should be set.

The following list describes warning signs of an incorrectly set amplitude control.

• Too high a reference level reduces the instrument sensitivity and may mask out a low-level signal or report an erroneous higher power-in-band measurement by integrating the noise floor of the instrument as opposed to integrating the noise floor of the device under test (DUT).
• Too low a reference level saturates the instrument signal chain with high-power signals at the input, thus creating distortion products in the instrument signal chain and inaccurate measurements.
• For signals where little is known a-priori about their characteristics, you can make multiple measurements (at different reference levels) to determine if the reference level setting is too high or too low. For example, Adjacent Channel Power Ratio (ACPR) measurement measures the power interference generated in the neighboring channels by the transmitter. If the adjacent channel integrated power measurement value drops if you increase the Ref Level, the higher reading was probably corrupted by distortion generated inside the instrument generated by too low a reference level setting. On the other hand, if the adjacent channel integrated power measurement value increases as you increase the reference level, you are likely integrating the noise floor of the instrument instead of the DUT.

Detection Mode is another amplitude control that is only applicable to traditional swept-tuned spectrum analyzers and not FFT-based analyzers. Before we discuss Detection modes, it is important to understand the frequency controls.

Frequency Controls

The following paragraphs describe the key frequency controls and how they can affect your measurement results.

Center Frequency: Center frequency, as the name implies, controls the center frequency of the measurement. Along with frequency span, center frequency defines the frequency range that you see on the front panel of the instrument.

Frequency Span: Frequency span defines the total amount of frequency spectrum captured by the instrument. This span is centered around the center frequency, so the frequency spectrum that is acquired is from Frequency A to Frequency B, where Frequency A equals the Center Frequency minus one-half the Frequency Span, and Frequency B equals the Center Frequency plus one-half the Frequency Span.

In FFT-based signal analyzers, there are two distinct signal acquisition modes that are set by the size of the frequency span. Frequency spans up to the Realtime Bandwidth specification of the instrument are acquired by downconverting a single frequency block and digitizing the downconverted signal. For frequency spans that are larger than the Realtime Bandwidth specification, spectrum blocks are sequentially downconverted and digitized. For example, the Realtime Bandwidth specification of the NI PXI-5660 RF Signal Analyzer is 20 MHz. For frequency spans greater than 20 MHz, the PXI-5660 RF Signal Analyzer performs a sweep of the spectrum by re-tuning its block downconverter in 20 MHz steps and digitizing each 20 MHz block. For frequency spans less than 20 MHz, the block downconverter does not re-tune and performs fixed frequency dowconversion.

Resolution Bandwidth (RBW): Resolution Bandwidth controls the frequency resolution of the frequency axis. The name for the control comes from the way traditional spectrum analyzers operate. In traditional analyzers, a narrow band filter is effectively swept across a frequency span to create the spectrum display. The filter bandwidth determines the frequency resolution across the frequency axis, hence the label Resolution Bandwidth. In FFT-based analyzers, however, there is no analog filter. Instead, the FFT and its associated windowing parameters determine the frequency resolution or resolution bandwidth. In these instruments, a more appropriate label might be Frequency Resolution. There are two commonly used methods for defining the RBW: the -3dB definition and -6dB definition. These definitions refer to whether -3dB or -6 dB points are used are used to define the bandwidth.

Those familiar with FFT analyzers and FFTs may question how the RBW frequency resolution parameter relates to the bin size of the FFT . Table 1 below shows the relation between the bin size (ÄF = Sample Rate/ Number of samples) and the -3dB and -6dB. For example a ÄF = 1 kHz is equivalent to RBW(-3 dB) of 0.88 kHz and RBW(-6 dB) of 1.2 kHz if no windowing is used.


Notice that FFT-based analyzers such as PXI-5660 RF Signal Analyzer offer a choice of windows to limit spectral leakage and improve the resolution of closely spaced tones in frequency domain. A traditional spectrum analyzer does not offer this capability.

The measurement time (sweep time) of a traditional swept analyzer is inversely proportional to the square of the RBW because of the analog filter settling time. As you lower the RBW to improve frequency resolution, the sweep time increases exponentially. By contrast, an FFT signal analyzer is performing a longer acquisition and larger FFT calculation as you reduce the RBW. With the current technology in faster digital signal processing (DSP) devices, this DSP approach has significant measurement speed advantages for higher frequency resolution (narrow RBW) measurements.

Detection Mode and Effect on Power Measurements

Detection Mode is arguably the least understood control on a spectrum analyzer, yet it can significantly alter any power measurement. An integrated power measurement can be off by more than 10 dB if the appropriate detection mode is not used. The detection mode (normal, peak, sample , or negative peak) determines how the spectrum analyzer handles reducing or compressing spectral information. When the spectrum data points are more than the spectrum analyzer display can handle (display is usually limited to 1,000 points or less), the spectrum analyzer must choose a data reduction strategy.

Remember that the number of spectrum data points (count on the x -axis or frequency axis) is calculated according to the following formula:For example if the Frequency Span = 100 MHz, display points = 1000, each display points represents 100 kHz of frequency interval. Thus for RBW = 10 kHz, the display limitations forces discarding 9 out of 10 data points. The Detection Mode selects which sample from each frequency interval is selected. For the purposes of this discussion, the proportionality constant k has been dropped from the analysis for simplicity.

Figures 6 and 7 below show the effect of Detection mode on the displayed spectrum.



The various detection modes and their impact on integrated power measurement is summarized in the table below.






To illustrate the measurement result variation, Figure 8 shows the variation in results for a single-tone nominal 0 dBm signal measured with a traditional swept-tuned spectrum analyzer and a FFT-based analyzer. Notice that the readings are closer to the actual power for small Frequency Span/ RBW Ratio (i.e when the instrument does not have to do a lot of data reduction). The variance is quite large for larger Frequency Span/RBW Ratio, as shown graphically in Figure 8.


A traditional spectrum analyzer uses a frequency sweep between the start and stop frequencies. This frequency sweep is generated by an analog ramp, and the start frequency of the span is directly synthesized from a high-accuracy timebase reference. The accuracy of frequency measurements at most points in the display is limited by the accuracy of the analog ramp signal. The frequency accuracy is also limited by IF filter center frequency accuracy. FFT-based analyzers are not constrained by this limitation, since there is no analog ramp sweeping a filter. The accuracy of frequency measurement is uniform across a given frequency span, therefore no particular optimization such as placing the signal of interest in the middle of the span to improve accuracy is required. Accuracy across the frequency span is solely dependant on the timebase accuracy and the measurement algorithm, thus better frequency accuracy and repeatability are much easier to obtain.

In a traditional swept analyzer, the main contributors to frequency errors are:

• Reference Frequency Error
• Frequency Span Accuracy (typically 0.5% of the span)
• Resolution Bandwidth (typically 15% of the resolution bandwidth)

In contrast, the main contributors to frequency error in an FFT-based analyzer are:

• Reference Frequency Error
• Resolution Bandwidth (Depends on the measurement algorithm. Can vary from > 50% of RBW to < 10% of RBW)

To compare these errors, it is appropriate to ignore the reference frequency error, as this can typically be compensated for by using a highly precise frequency reference such as a rubidium source. For example, to measure a 100 MHz oscillator to 10 ppm accuracy requires frequency accuracy of better than 0.25 kHz (using a test accuracy ratio of 4). Frequency Span above 50 kHz and RBW settings greater than 1 kHz compromise the measurement on a traditional swept-tuned spectrum analyzer unless optimized techniques, such as placing the 100 MHz tone in the center of the span, are taken. The constraint on using smaller RBW translates to longer measurements because of the sweep time of the filter. Remember that the sweep time grows exponentially as you reduce RBW in a traditional swept-tuned spectrum analyzer. A typical spectrum analyzer has a sweep time of 150 ms to 200 ms for the given example.

An FFT based analyzer’s accuracy is going to be limited primarily by the measurement algorithm, which in turn can be more or less dependant on resolution bandwidth. The algorithm used in the Spectral Measurements Toolkit , the measurement software accompanying the NI PXI-5660 RF Signal Analyzer, applies a patent pending 3 point interpolation technique to measure the frequency of a tone at a much higher resolution than that indicated by the Resolution Bandwidth. Power measurements in 3 adjacent FFT bins are weighted to determine the tone peak frequency measurement. In the above example, setting the RBW as large as 2 kHz does not have a problem achieving accuracy better than that required.

The fact that FFT- based analyzers can use comparatively larger RBW settings to measure frequencies accurately without using accuracy-optimized measurement techniques translates to faster measurements or more accurate measurements in the same test time. The NI PXI-5660 RF Signal Analyzer can perform the above measurement example in less than 20 ms-- more than 6X improvement over traditional spectrum analyzers.

Conclusion

We have discussed the basics of power and frequency measurement techniques with both traditional spectrum analyzers and FFT-based signal analyzers. It has also been shown that unless appropriate instrument settings are applied, a wide variance in measurement results can be expected even from the same instrument, which can complicate measurement comparisons between two instruments, such as an instrument used for bench characterization versus one used for production test. Therefore an understanding of the instrument operation is essential for properly setting up the instrument for a given measurement.

Reference
Patent Application 20030040876 Estimating a plurality of tones in an input signal 20030040876
Related Links:
NI PXI-5660 RF Signal Analyzer Product Page

Reader Comments | Submit a comment »