The Tech Blog

Antennas have been the driving force of the modern wireless industry for a long time, especially the antennas on PCBs. Microstrip patches, slot antennas, and F-antennas printed on substrates like FR4 give us reliable performance below 30 GHz. They let us tune frequency by adjusting patch dimensions and feed geometry. When we push into bands above 30 GHz, losses from copper roughness, substrate modes, and dielectric loss make PCB antennas inefficient and bulky.

To overcome the limits of PCB antennas, we are moving into two new approaches: Antenna-in-Package (AiP) and Antenna-in-Chip (AiC). In AiP, think of the antenna as part of the chip’s casing. It’s a miniature multi-layer printed-circuit board that just happens to have a bare RF chip solder-bumped into its centre and then gets shipped as a single surface-mount part. By printing or mounting the radiator onto the package substrate using low-loss organic laminates or ceramics, we ensure the materials do not absorb the signal. This lets us build small beam forming arrays and pack amplifiers, filters, and phase shifters right into the module.

AiC takes the idea all the way onto the silicon die. Here, we etch antenna shapes into the top metal layer of a CMOS chip. I’ve seen designers thicken that metal, add patterned ground shields beneath it, or insert artificial magnetic conductors to bounce more energy out of the lossy silicon. Some even place a thin cover (a superstrate) above the chip to help steer or focus the beam. Since silicon has a high permittivity and loses energy, it can’t get more efficient than a certain point.

AiC is an ultimate miniaturisation play, and AiP is the performance play. If you’re designing a smartphone or radar sensor today, AiP is the practical path, it gives you gain, beam-forming head-room and a cleaner PCB layout. You only pick AiC when absolute footprint has to be the size of the silicon itself, and you are pushing beyond 100GHz where antenna is only a few hundred microns long and PCB interconnects become impractical. To date, only a handful of functional antenna-on-chip prototypes have been demonstrated.

I was planning to do some home networking and happened to run across SFPs. Let’s see those in detail today. SFP stands for Small Form-factor Pluggable, and it’s essentially a hot-swappable transceiver module that plugs into a switch or router port to adapt to different media types.

All of would know the classic RJ45 connectors, that can be used for connecting Copper cables(Cat5e) to network interfaces. As networks grew beyond 100m and speeds pushed past 1 Gbps, a better interface became necessary. Around the early 2000s, major vendors agreed on the SFP, so different manufacturers could supply compatible modules. In SFPs, you get a slot where you can choose a copper module for short runs or an optical fiber module for longer distances, even mixing multimode and single-mode fibers. That was a game-changer for cost and flexibility. SFP is a port that eases the life a bit for the designer, as they really don’t need to take care (to an extent) if it’s Copper or Optical cable on the end. No board respin needed.

Physically, the SFP “cage” connector is larger than RJ45. The SFP module fits inside the connector and has the interfacing circuitry required, freeing the host board from that footprint. It’s hot-swappable, meaning if a copper port fails, you replace just the module without powering down the device.

In practice, you will use for a copper SFP module when I need to extend a cable-run just beyond the 100 meter RJ-45 limit. For anything over a few hundred meters, or if you need immunity from electrical interference, fiber SFP (or SFP+) is the answer.

SFPs come in different types considering the speed class they cater to you. Look up on wiki for the full list. There are variations like QSFP (Quad SFP) and QSFP28, catering to 100 Gigabits of transfer speed. These are the things that move data around in data centres. In essence, SFP allows for a host of things as a pluggable interface.