How to prevent damage of frequency divider IC?

I am trying to upgrade my 2GHz frequency counter with /8 prescaler IC probe to measure up to 10GHz. Probe is similar to manufacturers evaluation PCB, with two SMA connectors and power input. I used laptop USB 5V to power the IC. My first attempt ended with damaged IC, it becomed very hot, and power consumption was above 200mA (70mA in datasheet). But i am unsure why it burned out, because many attempts to measure some counting was done before found out it is not working at all. As i had only one working IC left i started to check things very carefully. First of all i made power connection through ampermeter to ensure that consumption was below 60mA.

Laptop and frequency counter connected to 220V AC. I found that frequency counter BNC connector ground and laptops USB ground have 5V AC voltage between them. Regardless of laptop connection to 220V. So i did not connect them. Then i tried 220v to 12v DC switching power adapter, connected to LM2596 DC-DC. AC voltage between grounds was around 38V. Then i tried lithium battery pack with 12v output. There was no AC voltage, but there was audible noise from prescaler IC, and power consumption was strange (40mA or higher than 60mA). Then i tried old 220v to 12v power supply with transformator - which must be totally isolated. All was good until i made longer 2meters cable to connect 12v power supply output with LM2596 DC-DC input. There was again audible noise from prescaler, pretty loud, and power consumption was increased to 100mA. Happily it is still works. Now most safe way seems to use 9v battery with 5v linear regulator. Another way i see is to open frequency counter, find it's power supply and make 5v output, and maybe put prescaler inside too.

Prescaler maximum allowed limits are (10dBm for <7GHz, 0dBm above 7GHz). Is there any possibility that connecting two grounds with AC difference gives surge above prescaler limits?

Quite possible the voltage is high enough to cause overload but surely you are AC coupling the signal to the counter input and at frequencies >2GHZ you are using relatively small capacitors for the job. Their reactance at SMPS frequencies should be very high so coupling would be highly attenuated.

I'm thinking 10pF coupling capacitor @ 2GHz = 7.9 Ohms
same capacitor at say 50KHz = 318K Ohms.

Brian.

I am afraid that you have a severe PCB layout or/and decoupling problem. Prescalers, as many other circuits, can get in self-oscillation:

http://www.cel.com/pdf/appnotes/an1014.pdf

I never got audible noise from a prescaler, but some DC-DC supplies might enter in a state that generate audio noise.

Finally i measured DC-DC with oscilloscope. Without load there was some noise around 5V with frequency peak a little less than 60kHz and amplitude around 100mV if i remembered it right. I connected old opAmp pcb and noticed very high noise, i think it was same thing around 60kHz. After a while i realized that input capacitor before LM2596 is absent! I forgot to solder it. After placement noise still exists, but does not increase with loading. Next time i will pay more attention to power supply, with oscilloscope. Thank for you replies. I have idea to replace capacitors to 10pf on output and around 2 pf on input, although evaluation PCB states that all caps are 100pf.