For the PDF version of this article, click here.

A few of the benefits offered by RF detectors in modern wireless communication systems include lower emissions, improved reliability, and longer talk time. In wireless base stations, RF power detectors monitor power amplifiers to ensure that transmitted RF power is not excessive, reducing emissions and protecting the PA from overheating. In wireless handsets, RF detectors enable transmitted power to be minimized. This, combined with their low current consumption helps to maximize battery life.

Several different technologies are available to measure and control the power level of RF signals. The simplest and most primitive is the diode detector. With its simple topology, it offers a dynamic range that may be suitable for spot signal level measurements, but its accuracy over temperature is generally not good. The diode circuitry can be expanded to compensate for temperature, but this only helps over a small portion of the already small dynamic range. In contrast, monolithic logarithmic amplifiers (log amps) can detect RF power over dynamic ranges of 60 dB or more. Technological advances allow log amps to guarantee measurement accuracy to better than ±0.5 dB across the dynamic range and temperature window. TruPWR rms-responding detectors offer a third alternative for RF power measurement. Advances in packaging and process technology have tightened the temperature stability of rms detectors, and their dynamic ranges exceed 30 dB. In addition, rms detectors are insensitive to changes in the peak-to-average ratios, whereas diodes and log amps are waveform dependent.

Applications requiring high power transmission are dramatically affected by even by the smallest margin of RF power detection error. For example, a 50 W cellular transmitter would expend 25% more power given a +1 dB uncertainty in RF power measurement. Improvements in detector technology offer many solutions for better accuracy, but new applications continue to emerge and push the performance envelope. Calibration and temperature-compensation techniques implemented in RF power management systems can reduce the measurement error to levels approaching 0 dB.

Three barriers to achieving accurate power measurements are variations in manufacturing processes, temperature and waveforms. An RF power management system will attempt to mitigate or eliminate the adverse effects of these variations.

Figure 1 illustrates a typical RF power management system using an rms detector. A small portion of the forward RF power from the PA output is sampled through the directional coupler. The attenuated RF signal is introduced to the rms detector where it is converted to a dc voltage through a linear-in-volts response. The dc output is fed to a digital signal processor (DSP) via an analog-to-digital converter. If the digitized signal level of the PA is unacceptable, the DSP will adjust the RF power level by correcting the gain in the transmit signal path. The system will reach steady state once the measured power level matches the desired signal level. This RF power management configuration is not limited to a particular application. Both base stations and handsets may incorporate variations of this same RF power management system.

As the peak-to-average value of the RF signal varies, the output response of a log amp will vary. This introduces an uncertainty that must be compensated by the DSP. Diode detectors also respond to peak power, not rms power, so they, too, are affected by peak-to-average ratio changes.

The RF detector used in the power management system in Figure 1 is a mean power detector (rms detector) whose response, unlike that of diodes and log amps, is independent of waveform. It achieves independence from peak-to-average ratios by computing the square, mean and root functions of an rms calculation. The RF input is fed to one of two identical squaring cells. The squared signal is then averaged through a low-pass filter network. The signal is fed to a high-gain error amplifier that has the second squaring cell in its feedback path. This feedback loop performs the square root function, thus completing the rms calculation. The output is a linear-responding dc voltage whose conversion gain has units of Vdc/Vrms.

The rms-responding detector allows the RF power management system to monitor and dynamically adjust the transmitter's output power as the peak-to-average ratio of the transmitted signal changes. Figure 2 illustrates the accuracy in measuring various waveforms (each with a different crest factor). The method used to calculate the error is discussed later.

The waveform independence is particularly useful in W-CDMA and WiMAX communications systems whose waveforms vary over time. W-CDMA waveforms will become increasingly complex and have higher peak-to-average ratios as a result of increased call loading. WiMAX uses various combinations of OFDM signal modulations to balance the inversely related data rate and reception reliability. To optimize the quality of the link, the signal modulation is dynamically adjusted as users move toward or away from the base station. The composite signal envelopes of the data bursts may have significant peaks that can drastically change over time.

The accuracy of the RF power management system is dependent on the calibration and compensation techniques adopted. The DSP actively adjusts the transmitted power by adjusting the gain of the transmit signal path. It can react to environmental changes if it knows how they will affect the RF detector measurement. Compensation routines integrated in the decision-making software can help to significantly reduce or eliminate two of the error variants — temperature and waveform. For example, if a power detector consistently drifts in one direction with temperature, a compensation algorithm can be implemented to remove the expected error.

Calibration of the RF power management system is performed once in production. At this time, a lookup table that covers all combinations of variants is stored in non-volatile memory. The system can then accurately determine the transmitted signal level by adjusting it if necessary. In general, the one-time calibration is only done at a single temperature (usually ambient) due to test time and cost constraints. The RF power management system is, therefore, dependent on the repeatability of the RF detector's temperature characteristics. If the detector's response is not repeatable from device to device, then the effects of temperature changes cannot be predicted and temperature compensation is not viable.

The simplest method of calibration uses two points, as shown in Figure 3. During calibration, two RF signal levels are driven through the transmit signal path and the corresponding RF detector outputs are recorded. In the case of Figure 3, the two points are highlighted at -12 dBm and +7 dBm. The slope and intercept are calculated using these two points and are used to characterize the ideal response line, shown in green. The response at +25 °C is not perfectly linear and varies slightly from the ideal response line. The error is graphed on the primary y-axis and represents the variation from the ideal response line, scaled in dB. Because production calibration is done at one temperature, the responses of the detector at -40 °C and +85 °C are compared to the same ideal response line. The accuracy of the RF power management system is determined by the proximity of the error curves to the 0 dB error axis.

The simple two-point calibration technique employs no temperature compensation. This method depends solely on the stability and repeatability of the temperature performance from device to device. Figure 4 shows the performance of 55 devices drawn from multiple manufacturing lots over temperature. The error distribution plot helps illustrate the expected accuracy of the rms detector using only a two-point calibration technique. Over temperature and process, the distribution of devices stays well within ±0.35 dB across the majority of the dynamic range. The stability of the detector's manufacturing process dictates the width of the distribution bands at each of the temperatures. In the case of this rms device, the bands have a spread of about 0.15 dB.

The performance of the RF power management system can be improved by recording more points at the time of calibration. In doing so, the minor nonlinearities are eliminated, producing a near perfect measurement at the calibration temperature of +25 °C. Figure 5 shows the error distribution of 55 devices using multipoint calibration. In this case, the error at the various temperatures is calculated with respect to the calibration model, i.e., the +25 °C ambient response. This error calculation method differs from those used in the previous case where an ideal linear model at ambient temperature was used. Because the imperfections at +25 °C are calibrated out, the error at ambient temperature becomes equal to 0 dB.

Over temperature and process, the distribution of device error for detected RF power stays within ±0.25 dB across the dynamic range. In the higher power range, the rms detector has an equally tight error distribution at the convergence or cross-over point, as shown in Figure 5. The high accuracy range centered on +2 dBm offers a 7 dB window of ±0.1 dB detection error over temperature. This level of accuracy becomes useful in high-power amplifier (HPA) applications where accuracy is required over a reduced input range. In contrast to the previous example of the 50 W transmitter expending 25% more power, a 0.1 dB measurement uncertainty would result in expending 2% more power.

The trade off to achieving this level of accuracy in an end application requires calibration at multiple points in the device's operating range and more test time. The simplicity of this method depends on the independence of the convergence point over process variations and temperature changes. Consequently, this level of performance can be achieved without a temperature sensor.

Adding a temperature sensor to the RF power management system and encoding the expected temperature drift in a lookup table empowers the DSP to compensate for temperature variations and to improve the overall system accuracy. As in the case of multipoint calibration, the compensation scheme must remove the ambient nonlinearities. It must also remove the temperature drift associated with the distribution.

The temperature drift of the detector can be characterized by examining a distribution of devices. The average error introduced by temperature variations is described as temperature coefficient with the units of V/°C. As temperature moves away from 25 °C, more drift takes place and more compensation is required. The level of temperature drift is not constant across the entire dynamic range. This makes a single temperature coefficient ineffective. Consequently, the dynamic range must be split into small subsections, each with its own temperature coefficient. The complexity of the algorithm can vary depending on the accuracy desired. Figure 6 shows the performance of 55 devices over temperature using a compensation scheme with a resolution of 15 subdivisions.

Over temperature and process, the temperature-compensated distribution of devices stays within ±0.1 dB across the dynamic range. In this case, the RF power management system is attempting to eliminate the adverse effects of temperature variants. If the modulation of the measured signal is known, then a similar concept can be adopted when variations from waveform need to be removed. However, if the response of the detector is not repeatable over process variations, then there is no method to predict the effects of the variants and compensation becomes ineffective. Compensation schemes are dependant on repeatable device performance.

Depending on the complexity of the calibration and compensation methods used, the accuracy of RF power management systems can approach 0 dB detection error. Even though waveform dependence and temperature drift can be compensated for, there is little that can be done for manufacturing process variations. Repeatable temperature or waveform behavior over process is required for any compensation technique.

More advanced compensation algorithms can be introduced with increasing complexity. As more variants are introduced, the required lookup table grows exponentially. New standards like HSDPA and HSUPA integrate a large combination of signal waveforms. The permutations of the temperatures and the various waveforms demand for prohibitively large lookup tables. TruPWR rms detectors offer a route to simplifying the complexity by reducing and in some cases eliminating at least one dimension of the lookup table.

Carlos Calvo is an applications engineer in the Advanced Linear Products Division at Analog Devices. He received his Bachelor of Science and Master of Science in Electrical Engineering from Worcester Polytechnic Institute.

1 2 3 4 Next

RSS    Save to Del.icio.us  Digg This