HMC424ALP3E Attenuation Response

I am testing the attenuation response of the HMC424ALP3E and am getting unexpected results past 16dB attenuation. My input signal is at 6.5GHz and the below graph details the response I get as the attenuation is changed from 0dB to 31.5dB. It seems that after 16dB the attenuator does not respond correctly. I am measuring the output power on a spectrum analyzer so I know the power being measured is only the power of the 6.5GHz tone and not any other broad band spurs or harmonics.

I have also verified the control voltages of V1 through V6 at each attenuation setting and the control voltages are correct in regard to the HMC424ALP3E truth table. The measured control voltage levels at the attenuator control pins are -1.9V for a logic low and -4.5V for a logic high. 

The bottom figure is a schematic of our circuit. The UP_V1 through UP_V6 inputs are connected directly to 3.3V digital control lines and the zener diodes have zener voltages of 5.6V (PN: BZX84C5V6TS-7-F).     

  • After spending the day troubleshooting I believe the root cause has been identified.

    Since we do not have an x-ray I started with suggestion two by connecting the control pins directly to the 0V and -5V rails. This did not show any change. Then I tried different input and output coupling capacitors (10pF and 47pF) and again did not see an improvement. Finally I decided to pull the chip to inspect the ground pad. Lo and behold the exposed ground pad was not connected! The amazing thing is the chip exposed ground pad had solder on it but the PCB pad was completely unconnected! I have never seen this before and I suspect the assembly shop (who I will not mention) did not get the board hot enough during the solder re-flow process. The really disturbing thought is how many other chips on this board have exposed ground pads that are not connected..... what a nightmare.

    Anyway after re-flowing the chip back on I now get the expected attenuation response. Thanks for your help!