Click here to go to our main MMIC design page
Click here to go to our main page on MMIC switch design with some examples
Click here to go to MMIC Switch Design Example 2
New for May 2019! Continuing our series on MMIC switch design, we have posted a video of Liam Devlin from Plextek RFI explaining first-hand how to design, layout, and perform EM simulation of a DC to 20GHz SPDT (single pole double throw) switch MMIC. If you go back to "Example 2" you will see that the designs are very similar, but in Example 3 you will hear first-hand how the design was done. The design is a series/shunt/shunt/shunt topology (six FETs total, two control voltages).
DC-20 GHz SPDT switch MMIC design by Liam Devlin
Some notes from the Unknown Editor...pretending he was invited to Liam's final design review. At a design review, always be respectful and don't act like you know everything. Bring up your concerns, and let the designer talk to them. You will soon find out there are often hidden reasons for the way the design turned out. Learning is a two-way street! And drop in some kudos for features that you really like. Don't talk over anyone else. To summarize, don't be a jerk.
(Reaches for second donut...) Liam did a nice job of cramming many hours of work into a short video! Here are a few key things I want to point out, but none of them will be the source of future tragedies.
For first-pass success to be assured at the very high end of the band, the FETs themselves will have to be EM simulated. As a general rule, a PDK model of a switch FET can't be trusted, especially if you change how RF enters and exits it in the layout. This gets a lot worse when you try to design for millimeterwave frequencies.
I really wanted to reach into the screen and give that series line between the first and second shunt FETs a loop, to decrease the circuit area.
Isolation could be increased with a via wall between the two switch arms. There seems like plenty of room.
The topic of FET fingers parallel or orthogonal to transmission lines could make for an interesting M101 page someday. Generally, this falls under the category of: "if I tell you, I'll have to kill you"...
The second and third shunt FETs are close together, you could consider making one larger FET in a series/shunt/shunt topology. You could flip one FET and they would be on opposite sides of the transmission lines.
I usually like to stop the mask from going out the door by demanding that port numbers be added to the RF and DC pads. Liam beat me to that, bravo!
An alternate set of control line pads could be added on the north and south side of the chip. This could prove useful in laying out future modules.
Liam used some small (500 ohm?) resistors on either side of the DC control lines crossing over RF lines. This is always a good idea to prevent resonances without over-analyzing the problem.
Power handling was suggested at 0.5 watts. The culprit is the series FET, it will start to saturate a high RF current. Learn about that here.
The switching speed of the device can be estimated by the RC time constant of the DC control lines and FET off capacitances.
Want more examples like this? Try:
Example on shunt-shunt and series shunt-shunt X-band switches
Example on series shunt-shunt-shunt 18 GHz switch
Or check out our MMIC Design page.