Calibrating Wattmeters at RFC

Gordon Hardman

Accurately measuring RF power is a tricky business, to put it mildly. Doing so accurately in a meter, such as the AP4510, that can be sold at a reasonable price and used in a typical radio shack for years is even trickier. One of the keys to the dependability of the AP4510 and others in the family is the careful calibration that each meter undergoes during production. As we were breaking down and moving the production setup for the RF Concepts wattmeters to our new location last week, I was reminded of this and thought it would be worth sharing a little of it with you.

Power, defined as the rate of doing work, is not a fundamental quantity, such as energy, that can be referred to a single fundamental standard. The best that can usually be done is to use some proxy, such as temperature, to compare one power reading against another. In many practical meters for use in radio power measurements, it is really voltage that is being measured. Since we know the impedance of the system in which the voltage measurement is being made (often 50 ohms), the power can be inferred from the voltage measurement.

At the heart of the 4510 series of meters is a rugged directional coupler constructed out of strip line in a four layer printed circuit board. This coupler is designed to maintain its high directionality (ability to discriminate between forward and reflected power) over time. The coupled arms are terminated in high stability resistors, and the voltage across these resistors is what is measured by the instrument. The RF voltage is turned into DC using wide dynamic range logarithmic detectors. In addition, a sample of the RF voltage on the "through" arm of the coupler is used to estimate the frequency of the power being applied to it. These are all digitized by the RF microprocessor (there is a second "display" micro as well). One final parameter that is sensed by the micro is the temperature of the strip line PC board in the vicinity of the coupler.

In a more simple wattmeter, there would be generally only one adjustment to calibrate the power sensor at a particular power and frequency. This would be the case, for instance, in the venerable Bird 43 wattmeter. If you want to calibrate, say, the "2-30 MHz 2.5kW" sensor, you would have to pick the power and frequency at which you want to do this. If done with care, the power at this calibration point could be "dead on". At other powers and frequencies, you would have to rely on the basic design of the instrument to yield a reading that is as accurate as the designer could make it. There is no ability to compensate for errors due to non-linearities in the detector, errors due to different coupling factors at different frequencies, and finally, no compensation for errors due to temperature changes. The 4510 compensates for all of these, not just by hardware design, but by applying mathematical adjustments to the measurements to yield readings of improved accuracy.

The 4510 series is specified over a wide frequency range (1.8-30 MHz), a large range of power (300mW to 3,000W) and over the -25 to +65C temperature range. Each meter is calibrated to determine a set of coefficients which form the basis of the mathematical corrections that are applied to the readings taken by the micro. This could be a formidable set of data if it was in the form of a table at every frequency, power, and temperature combination. Fortunately, this is not necessary. It was determined that a set of so-called fourth-order-polynomials is sufficient to turn a set of raw readings of voltage, temperature and frequency into a highly accurate measurement of forward power, reflected power and SWR. About twenty coefficients are sufficient to describe these polynomials. The coefficients are determined in two stages during manufacturing.

The first step is frequency calibration. After being assembled and undergoing a basic functional test, each RF board is connected to a computer automated test (CAT) setup. A computer controlled signal generator applies power to one port of the board, and an accurate power meter with a computer interface is connected to the other. The computer communicates with the micro on the RF board over its serial interface. The CAT computer steps through from 1800 kHz to 29700 kHz in 100 kHz steps, and records the forward and reflected power readings. These data are stored in a file.

The second step is the temperature compensation. In this test, a group of six boards is placed in a temperature chamber, and again the CAT computer controls everything. The temperature of the boards is first taken down to -25C. Once their temperature (as reported via their own individual on-board sensors) has stabilized, then the readings from all the boards are recorded in files. Two different power levels are recorded here, since one of the major effects we are measuring is the drift of the log detectors intercept point with temperature, and a minimum of two measurements is needed to determine this. The temperature is then increased in steps up to a high of +65C, and measurements are recorded along the way.

The third step is to "crunch" all this data and distill it down to the twenty coefficients needed by each board when it is part of a meter. This calibration data is stored in yet another file on one of the CAT computers. In order to keep all these files from getting mixed up, we use a unique electronic serial number (ESN) which we get by using a specific temperature sensor which has a unique code embedded in it. All files for a particular board have this ESN as part of their filename.

The fourth and final step is to upload the coefficients into the board. Again, the CAT computer makes sure that the ESN of the target board matches the filename of the coefficient file, and passes the data over the serial port to the RF board. There it is stored in non-volatile EEPROM, and the calibration is complete.

We have been using this process for many years, and found it to be quite reliable. The equipment is being cleaned, re-calibrated and installed at our new location, and so wattmeter production will continue there to the same high standards as at the old location.

Gordon, WØRUN