strange behavior on 10 GHz gain path

Thread Starter


Joined Jan 12, 2013
hey all,

I was just wondering if anyone could provide some input on this 10 GHz design I'm working on. I essentially have a 10 GHz CW source which is pushed through two different 10-16 dB amplifiers before going through a bandpass filter to the output. The full setup is like this

DDS -> 12 dB gain -> 3dB pad -> 10 dB gain (high P1dB) -> high pass -> lowpass

The behavior I'm seeing is that my output power actually goes up by 1-2 dB when I increase the 3dB pad to a 6 dB pad! Does anyone have any idea what could be causing this?

Does anyone have experience in designing PCBs at these frequencies? Is it standard to always put some attenuation between gain stages? Should the directivity of the amps be sufficient to prevent standing waves if they are unconditionally stable?



Joined Dec 13, 2015
Hello Dom,

It sounds like it could either be that the matching between stages is changing or one (or both) of the amplifiers are behaving non-linearly. It is hard to say without more information.

Are you working in IC development like with a foundry? If so, what technology are you using?

Can you provide a schematic and part numbers for your design? Are these measurements based on simulations or have you made a full PCB already?

I have experience designing PCBs in 50 MHz to 6 GHz and RFICs in 20 - 30 GHz ranges.

Attenuators are good for impedance matching by reducing the amount of reflected power from the next stage, and improving amplifier linearity by ensuring an acceptable backoff. I don't know if it is common but it can be helpful.

Designing in 10 GHz is not an easy task. I would recommend using 0201 or smaller as the parasitics of the 0402 package starts to dominate after about 6 GHz. You must make pad sizes as small as possible if they are larger than the trace width. You should consider all traces as transmission lines because the wavelength in a rogers material is approximately 15.6 mm (615 mil). Also, consider using a high performance substrate. At 10 GHz the substrate losses are pretty substantial.