Power Meter: AD8307

2013/7/27 (finished @ 2013/9/9)

A power meter with wide frequency range and dynamic range and relatively accuracy is a key factor to make home brew activity feel professional and get a beginning  RF measurement lab.one of reason is most of  RLB( return loss bridge) type measurement need detect a low power signal.(might - 40dBm ~ 10dBm).  a high dynamic range power meter to help all those thing out.

The most popular design of Power Meter is the AD8307 scheme by By Wes Hayward, W7ZOI, and Bob Larkin, W7PUA on QST June 2001.

And Bob Kopski, K3NHI  published another very interesting project base on this meter: <An Advanced VHF Watt meter > on QEX  Many/June  2002。 which provide versatile signal processes.

1) RF section is same as W7ZOI's meter
2) D1 in Ic2A make a perfect diode and avoid op-amp's low voltage output. [input 0 output 0.6 at lest, but after that diode, still 0 volt ]
3) the good method to get a negative voltage.

How to use:
A) to scope: 10mV/dB, offset use to shift the output, i.e. adjust the scope Y position
B) to DVM: 1mV/dB with sign
C) offset use to select the 10dB range for analog reading.

Note by jane@qrp.pops.net:
AD8307 specifications allow a typical +/- 0.3 dBm "ripple" in the input to output transfer characteristic. Bob, K3NHI verified this ripple in his lab. Therefore; depending on the position of a signal in the transfer curve, a move between 2 different power levels may yield as much as 0.6 db peak difference error. Figure 8 of the Rev D datasheet shows this. Figure 24 plus associated text tells why this occurs: the transfer characteristic is a chunk or segmented realization of a log transfer characteristic. It's really good — just not perfect.

The AD8307 input resistance = 1100 Ω , so we must place a resistor in parallel with the input to establish 50 Ω. Many builders just shunt a short-leaded 51 Ω resistor from input to ground like Wes did, however, you might also see builders place a 1% tolerance 53.2 Ω R in that slot to derive an input Z of 49.9 Ω. (  if use 50R shunt to ground, parallel with 470+1.1k ,  error will less than 0.1dB if use  W7ZOI's schematic , this not a problem)

My Meter , RF section

en, it seems there is error  for a precision 50R, cause the 8307 had 1.1k input impedance.  in HF, you can not ignore the 15pF's impedance.  I going to use 51R carbon deposit resistor, real value might be 51R to 52R. this should be fine for me.

The die-cast enclosure
a quality enclosure with SMA connector make it useful to integrated by whole device or use it alone. and provide good isolation.

Initial testing 

before I begin the design process, had simple testing, here is result :
 signal dBm  Attenuation  dBm to 8307  8307 output A 
 mv/dBm  8307 output  B  mV/dBm
 x         x    0.247    0.247   Noise floor
 -10  0dB  -10dBm  1.8891    1.891  
 -10  10dB  -20dBm  1.6296  25.9mV/dBm  1.630  26mV/dBm
 -10  20dB  -30dBm  13741  25.5mV/dBm  1.378  25.2mV/dBm
 -10  30dB  -40dBm      1.126  
     -50dBm      0.787  Big error( 2mVpp)

Average : 25.5mV/dBm         (max error 0.4mV*90 (dynamic range) ~= 1dB max error, but this is not a lab level testing )

(The Big error meliorate when i finished all circuit and put it into box, but still have 2dB error when attenuation 50dB)

I decide make it base on Bob Kopski, K3NHI's design. prototype board for many resistor is boring, so i like to make a real PCB design[But i does not]
1) a analog meter range is 100uA, not 1mA
2) a digital output given right sign,not integrated DVM, 10mV/dB
3) a sweep output given 10mV/dB
4) have offset 

offset Error analysis

My on hand meter had 2k resistance, i suspect is there a big error while add a offset to it?  lets figure out.

solve this equation, we get 

For my meter, if i use 10uA per 10dB, and scale the signal to 35mV/dBm then we had use a R+2k= 35mV/1uA = 35k. 
and we use 5uA offset , then the I3=fn(Vx)  relation ship we get   this: 

we can find the error is very little compare to 3.5k, it's just 70R,  2% non linear. on analog meter we might not notice this. the fixed offset change a little bit to 4.7uA. 

Meter Scale consideration 

here is my 100 uA meter prepare to use as analog dBm output.   we had total 100 uA to use, so where the -10 dBm(or 0 dBm) located? 

we had noise floor 0.24V, this mean we will use up fist 10dB on meter (not fully used, just close to 10dB). 
 -10 dBm we get 1.891V, and  25.5mV/dBm,  we know from noise floor to -10dBm we get (1.891-0.24)/25.5mV ~= 65dB changes.  so on the zero  of meter should be  -10dBm - 65dB -10dB ~= -85dBm. 

for convenient, we will add 5 dB offset, let's it start from -90dBm.  so  -10dBm (to  -90dBm)   use 8 divs of the meter.  this mean 80uA. if use R=33k, and 35mV/dBm, we know -10dBm will generate current:  1.891V*(35/25.5)/35k ~= 75uA,  this result given the offset do the work, shift it to 80uA. 

What happen if no offset? 
can we set  -10dBm to 75uA?  we can, i think,  but our estimation not accuracy enough to make it perfectly happen, what if when get 35mV/dBm, but meter pin not perfectly point to 75uA?  we stuck. 

Scale design Important note
cause first 10dB is noise floor and to -10dBm, we had 65dB change, so point to -10dBm, we will use at least 7.5 divs of the meter.  so don't put the -10dBm to less than 75uA for my meter. 

scale design : 85c1 

Design analog Amplifier to adopter 85C1

The left is the 100uA Meter driver,  right one is the offset generator with 14X gain.

The simple negative  power  supply quick prototype
[Warning: i build 2 of this module, one of it make OP amp Crazy! (capacitor position different)]

The Signal Process board, dead bug prototype design.

Final testing and alignment before put it into box

When I  ailment with the power board before box it,   everything works fine.  then, i made another power board, but this time ,OPs became crazy.  the 'noise' output is 0.5V while nothing input), seems oscillating, but oscilloscope show nothing.  Do the test before box it....

40 dB Tap

Click FOR high resolution image.

40db tap

Test  UHF

verify 0:  DC attenuation
input: 1.4850V, output: 15.2mV   ====>  -39.8dB
input:1.9372  ,  output : 19.90mV  ====> -39.8dB

Test 1: calibrate the 40dB tap
10Mhz -10dBm signal source   === 40dB tap === AD8307 Power Meter
bypass 40dB tap:  -9.96dBm
insert 40dB tap:  -50.6dBm
so @10Mhz, it's   40.64dB attenuation.

Test 2 : UHF test:
handy transceiver ICOM-91A  430Mhz(fully charged battery)    =====   40dB tap  === AD8307 Power Meter
reading in meter: -3.6dBm, ==> 37dBm output from ICOM91A === 5W 
this seems very fine for all of system.

but  when i use 470Mhz,  I get this:-7dBm,  i very double this result, what happend?   this won't stop me, thanks god, i draw a chart:

Now things became interesting, that's output filter of PA!!  here is the sch from service manual,  en, most of moden handy transceiver don't have a tuned output filter.

New update @2013/9/12
Above VHF power character is wired, i re-test it, with shortest cable, it's better now:

So, a bad coax cable really affect the result, i think it's because of strong power leak signal to detector, affect it's accuracy.