Designing a 5.8G AGC using HMC346 (negative voltage control attenuator)

I'm designing a 5.8G AGC system. use HMC346 as voltage control attenuator. When the input power change from -10 to 0dBm, I let HMC346 change from -14dB to -4dB loss. Now the output power can be controled in 1dB range.
But the problem is that HMC346 is a negative voltage control attenuator. When loss change from -14 to -4dB,the negative voltage change from -2V to -2.5V. So I use a voltage inverter AIC1660 to create negative volgate -5V and then use OP to do RF detect output voltage adjust and positive to negative.
Because voltage inverter AIC1660 has a fixed osc frequency 10kHz. It will be carried on the negative voltage and then input in the HMC346. Finally the RF output signal will carry this 10kHz modulation.
So any one can help me to solute this problem or is there any other solution to avoid this problem? (5.8G VCA I can only find HMC346)

For 10dB dynamic range that you are looking for this Skyworks VVA using positive control voltage, it will work at 5.6GHz:
http://www.skyworksinc.com/products_...2.asp?pid=8791

but I need it work at 5.8GHz, Its operating range is 2.7G to 4.0G. How about its function at 5.8GHz?

On the AIC1660 output you have 10uF capacitor as filter, check ones again this point. Try to put in parallel and several smaller capacitors with 10uF for example one 1uF and one 100nF.
Post here PCB design, maybe some wires are too long.

As I said, for 10dB dynamic range SKY12146 it will work at 5.8GHz. Try it.

thank you for your help，I will order sample to test it。
And any other help also is welcomed!

Added after 1 minutes:

I have try this way, but the modulation is too high beside the main carrier (about -20dBc) to filter.

Yes, a positive controlled vva will eliminate a voltage inverter.

However, there are literally hundreds of millions of microwave communications systems that have switching supplies in them. They all work fine. first of all, the op amp you use has power supply ripple rejection. Find an op amp whose ripple rejection extends to 10 KHz--you should be able to effectively reduce ripple by 60 dB in the op amp alone. Then, as suggested above, do some series R shunt C filtering of the 10 KHz ripple. You should be well withing any emissions mask requirements at this point.

Also, make sure the circuit board layout around the voltage inverter is proper--no large ground loops, adequate power supply bypassing.

First thanks for your advice. But my experience lack on the "OP select". What's mean ripple rejection? Can you give me some advice about which OP to select ? I use AD8092.

On the other hand , the AGC system need to be controled in 1us. So any R or C filtering will delay the switching time. I do not want to poor it.

Look at this op37 data sheet, on page 3:
http://www.analog.com/en/prod/0,,759_786_OP37,00.html

Power Supply Rejection Ratio PSSR: 1 uV / V

1 volt of ripple on the power supply rail will cause 1 uV of ripple on the op amp output, or 120 dB of rejection. The only trouble, especially with lower speed op amps, is that the rejection rapidly rolls off with frequency, so you need to add tantalum and ceramic coupling capacitors at the power supply leads. In fact, for this op amp, on page 10, you can see that this excellent rejection rolls off to only 50 dB at 10 KHz! Maybe you can run your voltage inverter a little lower in frequency?