Diode Ring Modulator PCB Layout

[kml183]

Hello!

I wanted to build a simple PCB for a diode ring modulator to run some experiments on. The local oscillator input will be at max 455 Khz, while the RF input will be between 7.5 to 30 Mhz. I am matching the traces for a 50 Ohm impedance. I will be using JLCPCBs JLC7628 stackup. The signals for my layers are:

Top: RF
First: GND
Second: RF
Bottom: GND

I was hoping for some feedback for best RF practices on my layout. Questions I had:

1. How necessary is an RF Fence? I see this a lot in in other RF devices and they seem to be used as a Faraday cage of sorts.
2. Should I have my traces that are terminated by an SMA Connector exposed?
3. Is there a rule of thumb for spacing between the RF section say a via array that stitches the ground planes together?

I am new to RF PCB Layouts so any advice would be appreciated

[profdc9]

You can see how I did it in my project

http://www.github.com/profdc9/RFBitBanger

I placed all four diodes in a ring and the transformers nearby.   The transformers are hand wound, and the entire top of the board is a ground plane, and the bottom are the traces.

[iMo]

At those shortwave frequencies the 4 layers board and 50ohm traces impedance matching are not really necessary (your max freq 30MHz wavelength is around 10m). Almost any layout will work fine - like mentioned above - one side of the pcb the ground plane and the traces on the other side.
Your pcb design above is rather a "GHz design".. 

[Odysseus]

Your layout does a good job minimizing direct coupling between ports. However, the matching of your individual diodes will dominate isolation. I suggest binning them for a matched set, or using a ring diode quad like the BAT15-099R.

Also, I think you could add some vias near or inside the SMA ground pads. Currently, the return current path under your top layer traces has to go all the way out to your via stitching before it can reach the SMA connector.

[kml183]

Your layout does a good job minimizing direct coupling between ports. However, the matching of your individual diodes will dominate isolation. I suggest binning them for a matched set, or using a ring diode quad like the BAT15-099R.

Ya that is a good idea! Thanks!

Also, I think you could add some vias near or inside the SMA ground pads. Currently, the return current path under your top layer traces has to go all the way out to your via stitching before it can reach the SMA connector.

You are totally right I will update for that!

At those shortwave frequencies the 4 layers board and 50ohm traces impedance matching are not really necessary (your max freq 30MHz wavelength is around 10m). Almost any layout will work fine - like mentioned above - one side of the pcb the ground plane and the traces on the other side.
Your pcb design above is rather a "GHz design".. 

Ya ha this is super overkill, Ya I think I am going to to update to a two layer board. Save some money!

You can see how I did it in my project

http://www.github.com/profdc9/RFBitBanger

I placed all four diodes in a ring and the transformers nearby.   The transformers are hand wound, and the entire top of the board is a ground plane, and the bottom are the traces.

Very Cool! Thanks!

[Warpspeed]

Before you get too excited about laying out a four layered board, I suggest you get a quotation from the JCL website.

I did a 100 x 100 four layer board for a very high speed video graphics generator. 
Five of those boards delivered to Australia cost $85.00
That is slightly higher than the $2.00 cost of just five plain ordinary two layer boards of the same size.

A better bet might be a commercial diode ring modulator in a metal can.
Certainly cheaper, and with guaranteed performance.