Amplitude accuracy is a key barometer of spectrum-analyzer performance. Simply put, amplitude accuracy instills confidence in a measurement. In an extreme case, poor amplitude accuracy can lead to shipping a component that did not meet a customer's requirements, and failing a production-line component that performed properly. Understanding how spectrum-analyzer amplitude accuracy is determined and the factors that affect it can help guide the selection process when it is time to choose an RF/microwave spectrum analyzer.
A swept-tuned superheterodyne spectrum analyzer (Fig. 1) mixes input signals with a local oscillator (LO) with signal amplification, filtering, and detection performed at intermediate frequencies (IFs). The preselector filter (sometimes a lowpass filter) prevents high-frequency signals from reaching the mixer and mixing with the LO. The reference level shown on the spectrum analyzer's display is adjusted by the level of gain in the IF amplifier. This amplifier adjusts the vertical position of signals on the screen without affecting the signal level at the input attenuator. The horizontal scale is in frequency; vertical scale is calibrated amplitude, either linear (in volts) or logarithmic (in dB).
Spectrum-analyzer amplitude accuracy is specified in terms of absolute and relative accuracy. Absolute amplitude is the actual power level of a signal, in units of dBm. Relative amplitude, in dB, is the difference between two signal levels, using one as a reference for the other. An example of a relative measurement is a check of harmonic signal level, where the amplitude of the harmonic is measured relative to the amplitude of the fundamental frequency signal. Both absolute and relative amplitude measurement accuracy is improved in a spectrum analyzer by measuring a calibration source with precisely known amplitude and frequency.
In a spectrum analyzer, the front-end signal-processing components are sources of amplitude errors, including the amplifiers, filters, mixers, and LO. In some designs, improved components can reduce the measurement uncertainties associated with some of these components. The PSA Series of high-performance spectrum analyzers
(Fig. 2) from Agilent Technologies, for example, incorporates a full set of digital IF filters to minimize the amplitude variations inherent to analog IF filters. But simply improving some of the components in the signal-processing chain does not eliminate all of the error sources; better understanding how the various components in the spectrum-analyzer block diagram interact makes it possible to minimize amplitude errors and optimize amplitude measurement accuracy.
Why is amplitude measurement accuracy important? For an absolute measurement, for example, some communications standards require the use of modulated carrier signals not to exceed established power levels. For relative measurements, excessive harmonic and spurious signal levels from one communication system can cause interference with other systems. Amplifiers designed for these systems must be tested to determine that they meet linearity requirements and do not contribute to higher levels of harmonics and spurious signals. Filters for these systems must likewise be tested for passband and rejection performance.
The way that a spectrum analyzer's components work together contributes to different sources of error. Although not a complete list, Table 1 summarizes many major sources of amplitude measurement accuracy in a spectrum analyzer. Most manufacturers publish specifications for both absolute and relative measurement uncertainties. Since relative uncertainties affect the accuracy of both relative and absolute measurements, they will be the main focus of this White Paper.
One of the major factors of amplitude measurement uncertainty is the analyzer's frequency response or flatness. This is the relative amplitude uncertainty as a function of frequency over a specified frequency range. It is a function of input attenuator flatness, mixer conversion loss, LO amplitude variations, and input-signal filtering. Frequency-response uncertainty is usually specified for both absolute and relative measurements. The relative uncertainty describes the largest possible amplitude uncertainty over a frequency range relative to the midpoint of that band. It tends to be lower than the absolute specification for the same band. But to obtain the frequency-response uncertainty for relative amplitude measurements within a band, the relative frequency response specification must be doubled to reflect the peak-to-peak frequency response, which is often greater than the absolute frequency-response specification.
The preselector filter, usually a YIG-tuned filter, also contributes to the analyzer's frequency response. The filter must be precisely tuned and aligned in frequency to avoid additional frequency-response variations, and limited to the sweep rate of the LO plus whatever tuning delays and tuning compensation are needed for the YIG filter to keep it aligned in frequency relative to the LO. The front end of a spectrum analyzer usually also employs a lowpass filter to eliminate higher-frequency content from signals falling below the lower limit of the YIG preselector filter (typically about 2 GHz). Although this filter also contributes to the overall frequency response of the analyzer,it tends to suffer less amplitude error than the YIG filter.
Because a spectrum analyzer relies on mixing input signals with harmonics from the LO, the instrument operates in a variety of different frequency bands. Each of these bands has a specified frequency response, and amplitude measurement uncertainties result when switching between bands. A 26.5-GHz (E4440A) PSA spectrum analyzer, for example, operates in five internal mixing bands: 3 Hz to 3 GHz, 2.85 to 6.6 GHz, 6.2 to 13.2 GHz, 12.8 to 19.2 GHz, and 18.7 to 26.5 GHz. Whenever a measured frequency span crosses two or more of these internal mixing bands, and band switching occurs, some amplitude measurement uncertainty will result. When comparing signals in different bands, the frequency response is the sum of the responses of the two bands, along with any band-switching uncertainty. If band-switching uncertainty is not specified, the absolute frequency response relative to the calibration source (instead of the relative frequency-response uncertainty) can be used for each band (Table 1).
Scale fidelity is another source of amplitude measurement uncertainty in a spectrum analyzer. It applies when a signal at one vertical position in measured with respect to a second signal at a different vertical position. Scale fidelity depends upon the detector's linearity, the linearity of the analyzer's analog-to-digital-converter (ADC) circuitry, and the capabilities of the logarithmic/linear and vertical amplifiers to transform different signal voltages into their appropriate relative power (log) or voltage (linear) levels on the display. For most logarithmic amplifiers, log linearity degrades at decreasing signal levels.
For logarithmic (in dB) spectrum-analyzer measurements, scale fidelity is better for small amplitude differences between signals. It can be a few tenths of a dB for signals that are close in amplitude to 2 dB for signals with large differences in amplitude. A typical scale fidelity specification is 0.4 dB/4 dB to a maximum of 1.0 dB. The 0.4 dB/4 dB specification is applied when the two signals are close in amplitude; the cumulative specification applies to signals having larger differences in amplitude.
The adjustable reference level in a spectrum analyzer provides a great deal of flexibility when measuring signals over a wide range of levels, but the adjustment range also contributes to amplitude measurement uncertainty. The reference level, which is a function of the input attenuation and the IF gain, can generally be adjusted from a minimum at the displayed average noise level (DANL) of the analyzer to a maximum at the maximum input level (with attenuation). Setting the reference level essentially adjusts the gain of the IF amplifier, which itself (like any amplifier) has variations in amplitude as a function of frequency. Thus, any changes in the reference level control during a measurement will introduce uncertainty.
The reference level of a spectrum analyzer is usually calibrated with an internal source having known characteristics (although the calibration can also be performed with an external source). In the PSA spectrum analyzers, for example, the internal calibration source is comparable to that used for a power meter, with a known absolute power level of -25 dBm at a fixed frequency of 50 MHz and absolute amplitude accuracy of ±0.24 dB. (Compare this to the absolute amplitude of accuracy of ±0.34 dB for the general-purpose ESA-E Series spectrum analyzers, which also employ a 50-MHz internal calibration source.) When they can be used, these settings—reference level of -25 dBm and 10-dB input attenuation—will provide the highest level of accuracy in the spectrum analyzer since they have been calibrated directly to the absolute amplitude reference.
An example of a reference level uncertainty specification is ±0.3 dB at -20 dBm, with some incremental uncertainty the further the reference level is set from the -20-dBm level. Those comparing different instruments for this specification should be aware of some variations in terminology. For the 8560 Series portable spectrum analyzers from Agilent Technologies, reference level uncertainty is specified as "IF gain uncertainty." For the company's ESA and PSA Series spectrum analyzers, it is called "reference-level accuracy."
Since RF/microwave attenuators also exhibit slight variations in nominal attenuation as a function of frequency (and sometimes time and temperature), the step attenuation accuracy is a function of frequency. Switching the input attenuation setting between a reference level calibration and a measurement contributes uncertainty to the reference level accuracy. The accuracy of most attenuators used in spectrum-analyzer front ends can be expected to degrade with frequency: better accuracy at lower frequencies, decreasing accuracy at increasing frequencies. Typical input attenuator switching uncertainty is ±1 dB.
Because the frequency response of an analog filter is far from ideal, amplitude characteristics can vary from filter to filter. Switching resolution-bandwidth filters in the middle of a measurement can contribute to amplitude measurement uncertainty due to differences in the filter characteristics, especially for analog filters. Digital filters are capable of much tighter filter specifications than their analog counterparts, but are more expensive to implement. In Agilent's ESA Series mid-level spectrum analyzers, for example, the resolution-bandwidth filters are implemented digitally to 300 Hz, with wider-bandwidth filters realized with analog components.
The higher-performance PSA Series spectrum analyzers feature an all-digital IF section with Fast Fourier Transform (FFT) analysis and the digital implementation of a swept IF receiver. (Fig. 3) The approach not only improves amplitude measurement accuracy, but also boosts sweep speed compared to analyzers using analog resolution-bandwidth filters.
Changing the display scale on a spectrum analyzer can also contribute to amplitude measurement uncertainties. Changing the per division scaling factor, for example from 10 dB/div to 1 dB/div or to linear, introduces uncertainty associated with the relative calibration characteristics of the spectrum analyzer's log/linear amplifiers. The uncertainty can be avoided by simply not changing the scale during a measurement. Typical linear-to-log switching uncertainty is ±0.25 dB at the reference level, although this uncertainty may not be a factor for spectrum analyzers that store trace data into memory and display the traces from memory.
The total relative amplitude measurement uncertainty involves totaling the different error sources. Some of these errors result from a change in a control or setting. If any controls can be left unchanged, such as the RF attenuator setting, resolution bandwidths, or reference level, all uncertainties associated with changing these controls can be eliminated, and the total measurement uncertainty will be minimized. In a PSA Series spectrum analyzer, which employs a full range of digital resolution-bandwidth filters, there is no added error when changing a resolution-bandwidth filter during a measurement, as there would be for an analyzer based on analog resolution-bandwidth filters.
Improving amplitude measurement accuracy can be as simple as leaving a few settings and controls unchanged during a measurement. For better amplitude accuracy:
- Don't change the resolution bandwidth setting during a measurement. This doesn't apply, however, if the spectrum analyzer uses all-digital resolution-bandwidth filters, such as in the PSA Series spectrum analyzers.
- Don't change the attenuator setting between a reference-level calibration and a measurement.
- Don't change the per division scaling factor during a measurement.
The signal delivery network used with a DUT may degrade or alter the signal of interest even after a spectrum analyzer has been calibrated for optimum accuracy. The electrical contributions of the signal delivery network must thus be cancelled out. One way to do this is with a spectrum analyzer's built-in amplitude correction function used with a test signal source and a power meter. A typical signal-delivery network suffers losses at varying levels across a frequency range (Fig. 4, left). To cancel out these unwanted effects, the analyzer is used to measure the attenuation or gain of the signal-delivery network, especially at those frequencies where the losses or amplitude variations are the greatest. By building a list of associated amplitude and frequency pairs, and linearly connecting these points to create an error-correction waveform, the input signal can be offset to these corrections. (Fig. 4, right). The same technique can also be applied to antennas, cables, and other equipment used with a spectrum analyzer during a measurement. Corrections can be stored within the memory of the analyzer so that calibration is not necessary every time a test setup is changed.
As an example of calculating amplitude uncertainties for a typical measurement, consider the measurement of a 1-GHz RF signal with amplitude of -20 dBm. To show how different spectrum analyzers will yield different levels of accuracy, the same test was performed with a model E4440A PSA Series analyzer and a model E4402A ESA-E Series analyzer. The settings were the same: 10-dB attenuation, 20-kHz measurement span, reference level at -10 dBm, log scale, coupled sweep time, resolution bandwidth of 10 kHz, and video bandwidth of 1 kHz. Since the ambient temperature was at room temperature (+20 to +30°C), the published specifications for the E4440A PSA analyzer (digital IF) show an absolute amplitude uncertainty of ±0.24 dB. For the ESA analyzer (analog IF), the number is ±0.54 dB. Add to both of these figures the value for each analyzer's absolute frequency response and the sum is a value for the total worst-case amplitude uncertainty for that measurement (Table 2). At higher frequencies, particularly involving harmonics where band switching is necessary, the uncertainties are larger.
The use of digital filters has dramatically increased the accuracy performance of some RF/microwave analyzers, such as those in Agilent's PSA Series. Judicious use of the analyzer's operating controls during a measurement can also ensure that the instrument delivers the full level of that specified accuracy.
FOR FURTHER READING
For more information about PSA accuracy, contact Agilent Technologies at (800) 829-4444 or visit www.agilent.com/find/PSA_Accuracy to download the following application notes:
- "8 Hints for Better Spectrum Analysis," Application Note 1286-1.
- "Agilent PSA Performance Spectrum Analyzer Series: Amplitude Accuracy," Product Note 5980-3080EN.
- "Agilent Spectrum Analysis Basics," Application Note 150.