RF Connector Ground Isolation

I went and made a PCB for a UWB pulse generator I posted about before (https://www.edaboard.com/threads/206991/), and it appears that I should have thought more carefully about ground isolation in my design.

I have a SMA connector that receives input of a 3.3V pulse (100 MHz) into a comparator (P pin is directly connected and N pin goes through a delay line first to generate a short pulse output). The comparator has 50 ohm internal terminations to ground for the P and N pins. My problem is that the 3.3V pulse goes through the terminations and to ground, and then it gets into the output of the circuit that feeds into my pulse detector connected via a shielded line.

So, I'm about to start on a revision of the PCB, and I want to have separate ground planes for the input and the output of my pulse generator. How should those planes be connected together to prevent the input pulse from interfering with the output? I've seen designs that use inductors while others suggest using a narrow trace to connect the planes. Is there a good source of information on what to do use in my situation?

Also, why do I see in some designs RF connectors having a large block of copper on the top layer of a PCB that connects to the ground of the connector and then that section of copper has a few vias to an internal ground plane? Is this the same principle of helping with ground isolation?

From my experience, a nice common mode choke can work very well to isolate the common mode current and prevent it from propagating back down the shielding of a transmission line. These chokes can vary by necessary frequency content to be filtered, but nonetheless should be a good starting point for you!

Have Fun!

The term "ground isolation" sounds odd in RF design, particularly when talking about GHz frequencies. Closing the shield seems rather the way to go. Of course, if separate subcircuits or e.g. distant antennas are connected by cables, unwanted common mode waves may be an issue. In this regard, common mode chokes can be helpful.

FvM said:

Closing the shield seems rather the way to go.

Click to expand...

What do you mean by closing the shield?

What do you mean by closing the shield?

Click to expand...

Just take the words strcitly. High frequency interferences can't be blocked by ground isolation, they are radiated and crossing a gap. They are better blocked by a surrounding shield.

FvM said:

They are better blocked by a surrounding shield.

Click to expand...

Sure, but that's not the issue I'm having. My problem was more to do with the ground of my RF connector connecting to the ground plane of my circuit and noise from that messing with the components in my circuits. If I had a shield, would I not want to have some sort of isolation between the shield and my circuit's ground plane, ie separating the analog ground and the digital ground?

I expect, that "ground isolation" will cause more interferences, not less.

I tend to agree with FvM.

Are you sure that the issue is related to ground plane? how did you test it? Did you respect the typ rules for RF circuits (proper grounding, decoupling and so on)?
A picture or a drawing of the circuit will help to understand better.

Mazz

I must admit, that I wanted to express my general doubt about ground isolation and separated grounds in RF and high frequency circuits in the first place. Nevertheless, there may be cases where isolation of circuit parts is reasonable. But I didn't hear an indication, that it would be with your design.

The original question is about isolation of a RF connector's ground. Why? The coaxial cable shouldn't be expected to carry common mode currents. If it does though, common mode chokes would be the appropriate means. The more interesting point is about the signal current sourced from the connector, it's forward and respective return path in the circuit. If it's not designed properly, interfering currents may be injected into the common ground plane and possibly emerge at other places. At least, it can be a source of radiated interferences. In case of logic signals, differential signalling is the ultimative means to get rid of this kind of problems.

FvM said:

The more interesting point is about the signal current sourced from the connector, it's forward and respective return path in the circuit. If it's not designed properly, interfering currents may be injected into the common ground plane and possibly emerge at other places.

Click to expand...

That is my experience, too.
Some designers focus on the signal path but forget about the return path. Both is equally important.

It is a really common problem nowadays. The RF chips available are pretty robust, so digital engineers think they can lay out their wireless boards without the help of an RF engineer. In doing so they put cuts and traces in the ground planes in bizzare ways that make an RF guy only roll his eyes in disbelief. You usually get called in when "we can not meet the FCC emissions testing, but we put in the filters like the application note said to". You find their entire PCB board is now an unintentional antenna.

I went ahead and made a simplified schematic to help illustrate what I'm doing. **broken link removed**. On top is the pulse generator, and on bottom is the pulse detector. They are implemented on separate boards with an internal ground plane and power plane. The input and outputs are SMA. I currently have the output of the pulse generator hooked to the input of the pulse detector via a shielded cable. If I only hook up my external pulse generator (using my oscilloscope's logic output) to the signal pin of the SMA connector on the top board, then things seem to work ok (pulse is reported when I send a pulse into the generator board, but I don't have the equipment to tell me anything about the pulse being detected). If I also hook the external pulse generator's ground to the ground of the SMA connector, then the peak detector constantly signals a pulse detected even with no pulse being fed into the board. I don't know why. This is where my original question came from in that should I have separated the grounds on the pulse generator board for the input SMA and output SMA and put a choke between them?

Another issue I'm having is that if I hook the output of the pulse generator to my oscilloscope (200 MHz) and don't power the pulse generator board at all while sending pulses into its input, the oscilloscope shows a small 100mV pulse corresponding to the rise of the pulse fed into the board and a 100mV negative pulse when the pulse fed into the board ends. I don't really know if I can trust this though.