Calibration Report W7ZOI Power Meter built by W2GNN. Clifton Laboratories Jack Smith K8ZOA 12 March 2010

Transcription

1 Calibration Report W7ZOI Power Meter built by W2GNN Clifton Laboratories Jack Smith K8ZOA 12 March 2010

2 Overview of W7ZOI Power Meter Designed by W7ZOI and published in June 2001 QST. Uses an AD 8307 log amplifier to provide an output voltage proportional to the logarithm of input signal. Hence the output is proportional to signal input in dbm. Output drives analog meter and auxiliary voltage output, approximately 60 mv/dbm slope. Wideband, from 1 MHz to above 500 MHz. Excellent dynamic range, from - 70 dbm to +13 dbm. This unit is built by Gary, W2GNN. 2

3 What is to be calibrated? Establish input return loss over usable frequency range. Return loss is an important driver of power measurement accuracy. If the power meter does not represent a good match to the 50 ohm driving source, Agilent s multipart Application Note 1449 (parts 1 through 4) is an excellent starting point in understanding power measurement. Part 3 discusses the effect of sensor return loss on power measurement. Power that is reflected from the power meter due to imperfect termination is not measured. And, the power delivered to the power meter is a function of the meter s reflection coefficient. No adjustments are made in the calibration for the W7ZOI s return loss. Up through 150 MHz, the return loss is excellent and in view of other much larger uncertainties in the calibration, no adjustment is appropriate. 3

4 What is to be Calibrated (cont d) The analog voltage output is linearly proportional to input signal in dbm over a wide part of the instrument s range. The desired calibration provides the slope and intercept values relating output voltage to input power in dbm. The equation is thus: Input(dBm) = a + bx where: a is the intercept, i.e., the input power corresponding to 0 output volts. b is the slope, i.e., the dbm per volt relationship of the output voltage. The linear relationship between output voltage and input power in dbm breaks down at low input levels due to noise and at high input levels as the AD8307 saturates. In general, the coefficients a and b will vary with frequency. A more sophisticated analysis would include frequency in the equation. Fortunately over the range 1 MHz 148 MHz, the frequency variation is relatively minor and a and b coefficients determined at some convenient frequency, such as 10 MHz, can be used with reasonable error. 4

5 Typical Calibration Curve Extended Range Note the voltage output is quite linear over most of the range. This justifies a linear regression relationship between output voltage and input power in dbm. However, linearity breaks down above +13 dbm and below about -70 dbm. The noise floor appears to be around -80 dbm. To obtain the most accurate regression fitting, only power levels between -60 and +10 dbm are used in the analysis in this report. 5

6 Return Loss Measured over the range 300 KHz 500 MHz with an HP8752B vector network analyzer. Over the range 300 KHz 144 MHz, the return loss is 38 db or better, excellent performance. The return loss degrades significantly by 450 MHz to 9.7 db. 6

7 Input Impedance Smith Chart A Smith chart plot of the W7ZOI power meter shows all data points below 144 MHz clustered tightly around the 50 ohm center point. Data taken with HP8752B vector network analyzer. 7

8 Calibration Setup Calibration setup is as shown. The 437B/8482A power meter is used to verify the HP8657A signal generator output. The HP power meter is based upon heating effects and has a limited range, with the best accuracy being found between -25 dbm and +20 dbm, far less than the W7ZOI design. Hence, as a check at lower signal levels, the 8657A output level was compared with the level measured with an HP8568B spectrum analyzer. The HP power meter calibration is checked against three standard power sources, an HP435B power meter (50 MHz, 0 dbm), Advantest R3463 spectrum analyzer (30 MHz, -10 dbm) and HP8568B (20 MHz, -10 dbm.) All agreed within 0.1 dbm after the 8482A is calibrated against the 437B s internal 50 MHz, 0 dbm source. It would be possible to set the 8657A to a certain output power level as determined by the HP power meter and then use a step attenuator to vary the power delivered to the W7ZOI power meter. This is very labor intensive process. Instead, the 8657A s output is connected to the W7ZOI meter and the signal generator s level is varied under program control. Two regression analyses are included in this study. One is based upon 8657A commanded power levels. The second is based upon the 8568B s measurement of the 8657A s output. A separate comparision between the 8657A s output as commanded and as measured by the HP437B/8482A in the range -25 to +15 dbm established that the 8657A s output is generally quite close to the nominal value. 8

9 HP8657A Level Output Compared with HP Power Meter over range -24 to +14 dbm Frequency (MHz) Mean Error (dbm) Standard Deviation (dbm) In general, mean error is small, and standard deviation is consistently in the 0.15 to 0.2 db range. The largest errors are seen at 29 MHz. This gives confidence that the 8657A will provide reasonable power calibration levels for the W7ZOI power meter. This assumes, of course, that the 8657A attenuator does not introduce additional error

10 Comparison of Signal Generator to Spectrum Analyzer Frequency (MHz) Mean Error (dbm) Standard Deviation (dbm) The figures have been corrected for cable loss between the signal generator and the spectrum analyzer. (The HP power meter and W7ZOI power meter are both connected directly to the HP8657A signal generator output.) The standard deviation is not all that much different than seen with the HP power meter. Mean error is greater, but this is acceptable considering the much greater dynamic range of the spectrum analyzer

11 Error in Calibrating Equipment Based upon the foregoing, an error budget for the calibrating equipment is on the order of 0.5 db below 148 MHz, although most measurements will be closer to 0.25 db. 0.5 db corresponds to a power error stated in watts of approximately 12%, and 0.25 db corresponds to about 6% error stated in watts. These errors seem excessive, but in fact they points to the difficulty in making accurate RF power measurements over a wide power and frequency range, even with reasonably sophisticated test equipment. 11

12 -100 dbm to +10 dbm Plot. Y axis is commanded output from HP8657A Signal Generator 12

13 8657 Truth Regression Analysis The first regression analysis assumes the commanded output from the 8657A signal generator is the truth. The commanded values are used as the Y value in data plots and the regression calculations. 13

14 Plot over Range Used for Regression Analysis 14

15 Regression Line Plot 15

16 Residuals 16

17 Regression Parameters 17

18 8568B Truth Regression Analysis The second regression analysis assumes the 8657A signal generator output as measured by the 8568B spectrum analyzer is the truth. The spectrum analyzer measured values for each commanded signal level / frequency from the 8657A signal generator are used as the Y value in data plots and the regression calculations. A spectrum analyzer has greater error than a wattmeter such as the HP437B/8482A but provides a much greater dynamic range. The correlation between commanded signal generator level, HP power meter (within the range -25 dbm to +10 dbm) and the spectrum analyzer suggest the spectrum analyzer data is only marginally worse than the HP437B/8482A power meter measurements. 18

19 Plot over Range Used for Regression Analysis 19

20 Regression Line Plot 20

21 Residuals 21

22 Regression Parameters 22

23 Wide Frequency and Amplitude Range View of Output Voltage 23