RF power measurements with sensors multipath

Some considerations on how to improve the accuracy of power measurements on RF signals using modern sensors multipath (multi-path) connected to a PC or measuring instruments via the USB interface.

From a classical approach to the power sensor integrated

Traditionally, the power meters of radiofrequency signals were realized using a combination of a 'base unit and a sensor external power connected by a cable.
In this type of power meters, the RF signal is converted to a voltage signal from a power sensor, amplified and then converted into a digital signal and displayed on the base unit.
The connection between base unit and sensor uses a technique of purely analogue transmission between the power sensor and the base unit. This approach has the advantage that you can choose the power sensor more suitable for your application without having to worry too much about the characteristics of the base, which remains the same. However, the disadvantage of this configuration is natural that a power sensor can not be used alone, ie without a base unit.
More recently, the technological development has led to the increasing miniaturization of components, as well as increased performance, so much so that today there are cheap and efficient small number of processors.
Therefore, today a power meter can be easily implemented as a small integrated unit connected directly to a PC or a base unit through a common standard USB port.
In this case, the base unit will not perform any processing of the analog signal. His job is essentially to feed the power meter and display the measured values.
It is a solution that offers an obvious advantage: the integrated power meter is no longer made ​​up of many components and can be completely characterized during the manufacturing process. This eliminates the need to calibrate the measuring system, consisting of the combination of the sensor and the base unit, with a reference signal prior to its use.
A further advantage offered by a power sensor USB is that the signal processing is less sensitive to unwanted noise, as is performed within an integrated device and the reset becomes necessary only for the detection of amplitude signals very small.

Sensor Technology

The power meters are available based on various technologies can cover a frequency range that extends beyond 100 GHz and a power range from 100 pW measurable to several tens of watts.
The technologies currently most frequently used to realize the power sensors are:

• Thermal detectors
• Multipath detector diode (multi-path)
• Sensors broadband or peak diode detector
• Sensors CW (continuous wave) with diode detector and logarithmic detectors integrated

The thermal sensors convert the RF power applied in heat using a resistor. The RF power can then be calculated based on the difference in temperature between this resistor and its immediate vicinity. The main disadvantage of this type of thermal sensors is their low speed of measurement and, therefore, the consequent inability to view the envelope of power.
Furthermore, the thermal sensor may be used to measure powers starting from 300 nW and its dynamic range is therefore limited.
To overcome the limitations of thermoelectric sensors is used to power sensors based on diodes, which can ensure a dynamic range up to 90 dB. Depending on their embodiment, the diode sensors can also measure the envelope of power up to a bandwidth of several tens of MHz
A power sensor based on diodes converts an RF signal into a voltage signal using an RMS detector (rms).
For power levels less than -20 dBm, the detector maintains a linear relationship between the RF signal and the output voltage. This area is known as square-law region.
In the region to the quadratic response, the diode detector behaves more or less as a heat detector and is largely immune to harmonic and the amplitude modulation.
For signals of higher level, no longer exists a linear relationship between the RF signal and the output voltage of the detector, therefore, precise measurements of power are possible in this region only if the bandwidth of the signal is less than the bandwidth of the detector. In addition, each measured value must be linearized numerically before it can be used for further calculations.

Power sensors with diodes multipath

E 'need to use multiple techniques to fully exploit the advantages of a diode detector in the construction of a power meter that can offer a universal' wide dynamic range of measurement.
Firstly, more diodes are connected in series to form the so-called "stack", or stack. This configuration results in an improvement of the dynamic range of 10dB • log (N), where N is equal to the number of diodes. Furthermore, in the detector are integrated two or three measuring paths (path) independent with different levels of attenuation. Depending on the power level of the RF signal input, you choose the path that offers the best performance.
The sudden switching between different paths inside the detector is a possibility, but it is associated with the phenomenon of 'hysteresis. But you can take more gradual transitions, as happens within the sensor NRP-Z of Rohde & Schwarz.
This latter approach offers many advantages, including the prevention of steps of signal, better reproducibility thanks to the elimination of hysteresis and the ability to measure the envelope of power without interruption. There is also an improvement of the ratio S / N of up to 6 dB in the transition region.

Figure 1 - Improving accuracy in the transition region due to the weighting of the internal signal paths

Figure 1 shows the measurement uncertainty linked to 'uncertainty in the region of transition between two paths for the different cases of sudden switching and gradual transition.
The blue curve represents the path of more sensitive measure that works at the transition point at its upper limit of measurement. In this working point, the uncertainty of measurement increases rapidly due to the harmonics or modulation.
The red curve represents the behavior of the path less sensitive. This path is functioning at the lower limit of the switching point. The uncertainty of the measurement increases due to the noise and zero drift when the signal level decreases.
Thanks to the gradual transition between the two paths in the region of signal switching, you get better performance and increased measurement speed in the transition region.
The considerable efforts made in the development of power sensors with diodes multipath has reached a certain degree of success. These sensors have now reached the accuracy of the thermal sensors by providing, at the same time, a better dynamic range and a higher measurement speed.
The integration of multiple paths of measurement that simultaneously function has been made ​​possible by the realization of the power meters integrated.

Measurement Accuracy

The quality of a sensor is reflected in the accuracy of the measurement. For power sensors, the measurement accuracy can be obtained is usually specified in the reference conditions.
Consequently, it is very important to be familiar with the manufacturer's specifications, in order to determine what additional sources of errors exist and how they can affect a particular type of measurement.
The user should also pay attention to the following:

• Connectors securely fastened
• Resetting generally must be performed with the RF signal off
• Proper impedance matching to the device under test (DUT)
• Correct setting of the RF frequency

If it is the user who is responsible for the configuration of the measurement system, you must ensure that all parameters are set correctly. One of the most important parameters is the length of the moving average filter (averaging).
Increasing the length of the filter reduces the noise level, but it increases the measurement time. The manufacturer's specifications in the Data Sheet to help determine the optimum value.
Based on the example of NRP-Z21 sensor of Rohde & Schwarz, this relationship can be illustrated in the following table. In this case, a CW signal must be measured at 5 GHz and-40dBm (100nW).
Here, the sensor uses the path detector more sensitive. The Data Sheet of the producer may be consulted for the signal given by reading the 'absolute uncertainty. This value includes the inaccuracy of calibration, the non-linearity and the influences of temperature.

Absolute uncertainty			2.2%
Derived from zero (zero offset)	100 pW (typical, without zeroing)		0.1%
For noise	40 pW for measuring time of 10.24 s
Noise measurement times for certain	1 s	128 pW	0, 13%
	0.5 s	181 pW	0.18%
	0.1 s	405 pW	0.41%
	0.01 s	1.28 nW	1.28%

The zero drift path in the more sensitive is specified with a value of 100 pW and should be set in relation to the signal level. This error factor can be ignored in this example. Consequently, no manual reset is required.
The noise level is specified in the Data Sheet for an integration time defined and must be converted as required by the user. In this example, the multiplier is sqrt (10.24 / T _meas).
In the case of an amplitude modulated signal, the integration time must be an integer multiple of the period of the signal itself. If the period is unknown or variable, a considerable improvement in accuracy can be obtained by multiplying the integration window with a bell curve. This technique is known as "smoothing" (leveling) with Power Sensors R & S NRP-type Z.

Fig 2: Measurement of a GSM signal with a time interval active (timeslots), 0 dBm power of the burst, with an integration time of 10 ms or of an exact period.

Figure 2 shows an example of the effects of different settings on the accuracy of the measurement as a function of the measurement time. In the case of a repetitive signal, it is always necessary to measure out of at least two integration windows.
This allows the sensor to switch the polarity of the analog signals between two adjacent measurements.
This technique is known as "chopping". It effectively eliminates the offset voltages during the processing of the analog signal together with the influence of the 1 / f noise.

Impedance mismatch

Finally, we must deal with an issue that too often is ignored in daily practice: the impedance mismatch.
The mismatch between the power sensor and the device under test is usually the factor that most influences the accuracy attainable.
A power sensor is factory calibrated to display always the 'amplitude of the incident power. The calibration takes into account the internal losses as well as the reflected power.
If the connected source was ideal, the reflected power from the sensor power would be absorbed completely. In this case, the result would be displayed correctly.

Figure 3: The power meter shows the power of the incident wave (Pi).

However, the sources of the real signal reflect a part of incident power to the power sensor. This component is superimposed on the power emitted by the source and may cause a higher or lower extent depending on its phase angle with respect to the main signal. The measurement error due to disadattamente can be roughly determined by the following formula:
± 200% • | ΓG | • | ΓL | o ± 8.7 dB • | ΓG | • | ΓL |
The modulus of the complex reflection coefficient of the source (ΓG) or the load (ΓL) can be calculated from their ROS (VSWR):

If a power sensor which has a ROS (VSWR) equal to 1.15 is used with a device under test that has a VSWR of 1.6, then you will get a measure affected by an error of ± 0.14 dB or ± only 3.1% for the effects of mismatch.
It is a mistake that alone is already higher than the absolute uncertainty specified in the technical data of the sensor in the previous
You can follow different approaches to try to avoid this type of error:

• Use a power sensor as much as possible adapted
• Adjust the source optimally and if necessary insert a small attenuator
• Correct the results of measurement by the gamma correction

In the simplest case, the adaptation of the device under test can be improved by inserting an attenuator with an attenuation factor of between 3 dB and 10 dB. This expedient alone reduces the error due to mismatch of a factor of 2 to 10.
If the complex reflection coefficient of the device under test is known, then it is also possible to correct the numerical value measured.
The power sensors R & S NRP-Z perform this conversion automatically because the reflection coefficient of the sensor is re-washed directly in the factory during the characterization process and then is immediately available. What remains to be done is to determine the user the reflection coefficient of the device under test and communicate this to the power sensor.

In summary

The measurement precision RF power is primarily determined by a proper choice of the measurement. This is particularly true when it is necessary to perform fast measurements over that precise, as is usual in production environments.
Technical progress in recent years has made ​​available a wide range of power sensors integrated small, robust and accurate.
The power sensors diode multipath are currently widely used in many application areas. Offer an accuracy comparable to that of thermal sensors and are practically independent of the type of modulation applied to the signal. In addition, they also offer the widest dynamic range of any commercially available power meters.