Crosstalk in power meters - any suggestions?

I'm building a power meter using the AD8307 log amplifier. The circuit itself is almost as simple as the one they show in the datasheet (figure 32 page 17, http://www.analog.com/static/importe...ets/AD8307.pdf, the only difference is that I added 100K resistors to ground near C1&C2 (between the caps and the IC) to eliminate charging of the caps from internal bias currents. I also added a 0.1uF cap from pin4 to ground. The layout is kept pretty minimal with the input being a small loop as possible after an isolation transformer (if needed I will post screen captures of both the schematic and layout)

What I am trying to measure is a multi-channel RF amplifier which outputs signal at 480KHz with AM. The amplifier is connected to the power grid through an isolation transformer that prevents leakages to the ground at the output freq. The outputs (2 in this case) are hooked up to a 30dBm attenuator which goes to an Agilent power meter (for verification = gold standard) and runs in series to my power meters. My AD8307 meters are housed in septate aluminium boxes (Hammond 1590B style) and are fed from isolated power supplies. Internally I use a DC-DC converter which offers another 1500V isolation.

The amplifiers are outputting about 20W max and there is only a few % crosstalk seen on the Agilent power meter. However on my power meters I can see as much as 100% crosstalk.

Everything is hooked up using BNC cables so shielding is pretty good. Does anyone have any suggestions about where to look for crosstalk source? I checked the AD8307 meters and they work very well on their own, however not when two are hooked up.

Thanks!

You have to give us the full schematic for any meaningful comment.

If you are not driving the AD chip differentially, then just short one of the caps to ground, as in fig 17. The .1 uf at the output of the log amp might make it unstable! That is a very high speed op amp at the output. I would buffer the .1 uF with a series resistor, at least 100 ohms.

I have used their dual log amps in designs, and they really work well. Absolutely no crosstalk from 4 to 16 Ghz. I suspect it is the rest of your system, like those "isolation transformers".

If I were using this log amp, and hooking it to an AC mains, i would be worried about all sorts of stray RF pickup. I would put a 480 KHz bandpass filter at the front end.

I did not see any charge build up on the caps that would require a resistor to ground. I can see using one on the other side of the series caps (th non chip side) so as to not screw up the internal biasing.

Thanks for the quick comment. I will try to upload the schematic later today (don't have it here). I admit not having experience with RF circuits, which is making this project a bit hard. I thought the small caps might help in giving short impedance paths to HF signals on the output line (which is DC). I also put an R-C filter near the output BNC that helped with the noise (tuned to 100Hz).

I am running the circuit after an isolation transformer (Coilcraft WB1010 1:1 http://www.coilcraft.com/pdfs/wbt.pdf) which sits inside the box after a insulated BNC jack. This is needed because I want to hook the output of the log amp into a PC for sampling and that is hooked up to ground so I have to isolate the RF amplifier signal.

Regarding the 100K resistor at the input, you say that there is no need for them between the 0.1uF caps and the IC innput but maybe between the 50 ohm resistor and the cap? Why?

Here is the schematic of the main circuit:

* The resistors after the 0.1uF caps at the input are 100K and not 52.3 ohm as shown.

What in not shown is the power supply which is a DC-DC converter 5V in, 12V out + 5V linear regulator (I know this sounds a little strange but there are some design limitations in the entire system). I have 10uF cap before and after the regulator. The DC-DC converter is 1500V isolated. The power input to the circuit is from an isolated power supply.

The entire circuit is housed in a Hammond 1590B aluminium (cast type) box. The input BNC has a plastic jack so it won"t short with the output BNC jack which is standard metal type and shorts to the box. This BNC is hooked up to an NIDAQ card which can read the input in both single ended or differential mode.

corrected schematic:

Maybe your DC-DC causes some spurs, use a very good power supply to replace it, and use RC pi filter for power.

10uF is too small, try to use 1000uF.

I will check the circuits tomorrow with a battery. If they run OK on battery power that means that the RF is getting there by conduction and not radiation or induction... why use a huge cap?

Sorry, the 1000uF for power supply EMC filter. Not for RF.