The idea and circuit for this power meter go back to a publication of Wes Hayward. Having the surface mount version of Analog Devices AD8307 I changed some parts (capacitors) to SMD which better fit to the SO-8 case. Apart from that the circuit used is very similar to the one published. The gain of the noninverting amplifier is set to G = 2 to get an output signal of about 50 mV/dB which may be displayed by an external DVM or scope. Using an internal 50 uA instrument I used a 100 kOhm multiplier resistor to translate a range from -80dBm to +20 dBm into a displayed value of 0 to 50 uA. The AD8307 will fade out at low signals of about -75 dBm and its maximum input is +16 dBm. A test showed that this instrument is usable with very good results in the range -70 dBm (0.1 nW) to +16 dBm (40 mW).
A friend allowed me to test the power meter using his hp 8640B generator. The first table shows the measured output voltage (in mV) for the power level given in dBm at the input of the power meter; frequency has been 10 MHz during this test. The third column shows value read from the internal 50 uA-instrument. The last two columns are related to a linear regression - a "best-fit" line which has been calculated using measured points from -70dBm to 13 dBm. The last column clearly shows that this line (according to equation (1) below the graphic) is a good approximation which may be used in some PM8307---A/D-converter---Processor setup. V_OUT from the second column of the table is used in equation (1) to calculate the values in column four.
|Level P_IN/dBm||Output V_OUT/mV||Current Display/uA||Calculated Level/dBm according Equ. (1)||Difference "Calculated - Actual Level"/dBm|
The graphic shows the data of the first table. A straight line was established using linear regression for the measured values from -70 dBm to 13 dBm (...well my Casio pocket calculator fx-5500LA did it for me). The correlation coefficient r, which describes the exactness of a linear dependance, was found to be r = 0.99998. r ranges from -1 to +1; the value +1 represents an exact linear dependance. Yes -- both the measured curve (from - 90 dBm to 16 dBm) and the calculated best-fit line are drawn even so you can't see them separately within the linear region because the good fit of the linear regression. The graphic is made available for download as a high resolution PNG Bitmap-file DF7TV-pm8307-test-high-res.png. Equations which describe this line are given below the graphic and may become useful in later applications. You may observe the fading out of the AD8307 at low power levels as the curve shows a S-shape. To not destroy the AD8307 during the measurements the input power has been limited to a maximum of 16 dBm.
Wow! the calculated slope is very, very close to the 50 mV/dB I had assumed during my measurement of a filter curve before I could do this test. For input power levels in between -70 dBm and 16 dBm the measured curve is very close to this straight line. When time allows a second test will be made with a close look to the range below -70 dBm. Nonlinearity in this range could be compensated using an A/D converter and processor so that useful measurements may be possible in the range below -70 dBm.
The second table shows a wideband test. Power level used here was -20 dBm for all frequencies ranging from 625 kHz to 320 MHz and again the output voltage is given in mV.
So regarding the 0.625 to 40 MHz range we got a deviation of less than +/-10 mV (or abt. +/-0.2 dBm) if the output voltage at 10 MHz is taken as a reference.