How to Calculate Peak Power Measurement Uncertainties
By Sook Hua Wong, Agilent Technologies Inc.
The primary uses of an RF power meter are to ensure accurate absolute RF power measurements and to add traceability to a test system. Whether it is a test setup in a production line or on a troubleshooting test bench, or radar and mobile communication measurement in the field, power meters remain the reference equipment to ensure traceability to international standards laboratories. This i's to ensure that power measurements can be duplicated at different times and in different locations.
Even though power meters are considered the most accurate equipment for power measurement, measurement uncertainties can still occur when using these instruments. Many power meter manufacturers publish papers, articles or tools to help users analyze measurement uncertainties, however this is often limited to average power measurements. This paper will explain the measurement uncertainties associated with peak power measurements. A brief introduction to peak power meter measurement techniques and average power measurement uncertainties will also be provided as background to enhance the user’s understanding.
Peak Power Meter Acquisition and Measurement Processing
The front end of a power meter consists of a sensing element, which the most commonly used is a diode sensor. The sensor will convert a measured signal from power to volt. At the acquisition subsystem, the analog voltage signal will be filtered, sampled and converted from analog to digital format. The output is a digital representation of the input analog signal.
There are two paths in the acquisition subsystem: one performing CW (and average) only measurements and one for wideband peak measurements.
The wideband peak path consists of high bandwidth precision amplifiers and a high dynamic range digitizer. The signal envelope is tracked by the sensor and is sampled by a continuously clocked ADC at high speed (typically at 80 MHz or more). This produces a real time sampled version of the envelope power of the signal under test. In order to sense high dynamic range power signals with high accuracy, two parallel ADCs with offset are typically used. Pre and post triggered samples are captured and stored. The acquisition is controlled by a trigger engine to determine the timing of capture. Once a capture is completed, all the samples will be transferred from acquisition memory to DSP for correction and measurement processing. The uncorrected, digitized measurements from the acquisition block will be processed by the DSP and firmware in a number of ways: zero/calibration correction, range and bandwidth correction, or measurement (peak, average, CCDF, trace) processing.
For the CW analog path, the meter works in the same manner as the traditional average power meter; a new measurement is generated after an integration period of at least one chop cycle. Measurements are filtered to reduce noise and produce an acceptable measurement update rate of 20/40/400 or more readings per second.
A thermistor is also built into the sensor. It measures the temperature of the detector, and is required in order to perform temperature based linearity corrections.
Average Power Measurement Uncertainties
In power measurements, as in all measurements, there are many sources of uncertainties or errors. In general, there are three major sources of uncertainties, namely sensor and source mismatch errors, sensor errors, and meter errors.
Sensor and source mismatch errors are typically the greatest and they are caused by the addition and subtraction of incident and reflected waves that create a voltage standing wave pattern on the transmission line. This results in a portion of the source power never reaching the sensor, meaning that it cannot be measured.
The second greatest source of error is the uncertainties associated with the power sensor. Not all the power that reaches the sensing element will be measured. Some will dissipate in other parts of the sensor. The sensor only measures the power that the sensing element dissipates. A calibration factor is used to correct for the imperfect efficiency of the sensing element.
The third is errors associated with the power meter’s electronics such as calibrator source uncertainties, amplifier gain uncertainties and circuit nonlinearities. With the introduction of USB power sensors, this source of uncertainty has been eliminated and is considered is part of the sensor’s calibration uncertainty specification that combines linearity, calibration factor uncertainties, temperature specifications and uncertainties associated with internal calibration processes.
At low signal level measurements, more sources of uncertainties such as zero set, zero drift and measurement noise will appear. These errors can be analyzed and combined using the GUM (Guide to the Expression of Uncertainty in Measurement) method to give the overall uncertainties of power measurements (refer to Table 1). The GUM method has been adopted by all major national measurement institutions and standard laboratories.
Peak Power Measurement Uncertainties
Peak power is essentially average power over a short time period. The sample measurement uncertainties calculation in Table 1 is valid for both peak and average power. The main difference is the noise. Because generally only one sample is looked at in peak power measurements, you should use the value of the “noise per sample” specification. If you are measuring average power over a certain time period (time-gated), you may calculate the noise over the time period. The noise will gradually reduce when you measure over a longer period due to an averaging effect.
For a power meter with a sampling interval of 12.5 ns and noise per sample spec of 2.5 µW, the noise over 5 µs period is 125 nW.
For peak power over the same period, you can simply use the noise per sample, which in this case is 2.5 µW.
RF power meters and sensors are important equipment used to add traceability to test systems for making accurate absolute RF power measurements. The methods used to calculate the uncertainties of power meters and sensors are well established and are typically based on the GUM method. This method can be applied for both average and peak power measurements. The key difference is in terms of the type of noise being used. In average measurements, free run measurement noise will be used while in peak or time-gated average measurement mode, the noise per sample specification will be used. Agilent offers a wide variety of peak and average power meters and sensors. Uncertainties calculators are available for all Agilent power meters and sensors, and can be downloaded from www.agilent.com/find/uncertainty_calculators.
Agilent Technologies Inc.
this article to a friend!