The feed network carries RF energy between the radar transceiver and the radiating patch, and at mmWave frequencies it is often as important to antenna performance as the patch itself. At mmWave frequencies, line losses, parasitic radiation, impedance discontinuities, and fabrication tolerances become significant fractions of a wavelength, so the feed must be designed for low loss, controlled impedance, and minimal unwanted coupling. On mmWave radar boards, the two dominant transmission-line structures are the microstrip line and the grounded coplanar waveguide (GCPW), often used together.

Microstrip Line Feed

A microstrip line, similar to the microstrip patch antenna introduced earlier, consists of a conductor strip on the top layer of a dielectric substrate with a continuous ground plane underneath as shown in Figure 1-9. It is the simplest and most widely used feed for patch antennas because it is coplanar with the patch and can be etched in the same step, with no extra layers or vias. The characteristic impedance (typically 50 Ω) is set by the strip width, substrate thickness, and dielectric constant. At mmWave frequencies, thin, low-loss substrates are commonly used, giving 50 Ω line widths of a few hundred micrometers.

Microstrip feeds have limitations at mmWave frequencies. The open structure radiates from bends, steps, and discontinuities, which raises sidelobe and cross-polarization levels. The line also couples to nearby traces and supports surface waves in the substrate. Conductor losses increase with frequency and are aggravated by copper surface roughness, since the skin depth at 77GHz is only about 0.24 µm. Because the feed line sits on the same layer as the patch, its radiation also adds directly to the antenna pattern.

Grounded Coplanar Waveguide (GCPW) Feed

GCPW, also called conductor-backed coplanar waveguide (CBCPW), places the signal conductor between two coplanar ground strips on the top layer, with a continuous ground plane on the bottom layer. The top grounds are stitched to the bottom ground through rows of vias, often called a via fence. The impedance is set mainly by the signal width and the gap to the coplanar grounds, so it depends less on substrate thickness than microstrip does. This gives several advantages for mmWave radar boards:

• Field confinement and isolation: The electric field is concentrated in the gaps between the signal and the coplanar grounds. With the via fence, this suppresses radiation and coupling to neighboring lines, which is important where multiple TX and RX channels are routed closely.
• Surface-wave and parallel-plate mode suppression: The via fence, with pitch kept well below λ_g/4 (often below λ_g/8), prevents surface waves and unwanted parallel-plate modes between the top and bottom grounds. λ_g is the guided wavelength along the transmission line.
• Chip and package interface: mmWave radar transceivers in BGA packages present ground-signal-ground (GSG) pad arrangements. GCPW matches this geometry naturally, giving a clean, low-inductance transition from the IC ball to the board as shown in Figure 1-10.
• Lower dispersion and tolerant geometry: GCPW maintains a more stable effective permittivity over frequency, and its impedance can be adjusted by changing the gap without significantly widening the line.

In practice, a typical radar board routes signals from the IC using GCPW and then transitions to microstrip near the antenna, since the patch itself radiates as a microstrip-type structure. This GCPW-to-microstrip transition can be made by tapering the coplanar grounds away from the signal line while keeping the impedance continuous, and placing the final ground vias carefully. A poorly designed transition causes reflections and spurious radiation, so it is usually optimized with full-wave electromagnetic simulation.

The maximum power transfer is of primary concern when connecting the source, feeding element and the radiating patch antenna structure. Thus, the impedance of the source, feed line and the radiating element is matched with each other. At the source side, the package balls sit on the PCB pads from where the feedline starts. This section is impedance matched to reduce the reflections back to the source. At the radiating side, since the input impedance at the radiating edge of a patch is high (often 150–300 Ω), a direct edge feed needs impedance matching. Common approaches include quarter wave transformer, inset feed and stub.

The trade-offs with GCPW are tighter fabrication tolerances (gap widths and via placement must be controlled precisely), higher via count and cost, and somewhat higher conductor loss than an equivalent microstrip line because of current crowding at the conductor edges.

Alternative Antenna Feeding Techniques

• Aperture-coupled feed: The patch and the feed line are on separate substrates, divided by a ground plane containing a slot. Energy couples electromagnetically from the feed line through the slot to the patch. Because the ground plane shields the feed from the radiating layer, spurious feed radiation is very low and polarization purity is good. Each substrate can be chosen independently: a thick, low-permittivity substrate for the patch (wider bandwidth, better efficiency) and a thin, higher-permittivity one for the feed. The cost is a multilayer stack-up with precise layer alignment, and some back radiation through the slot.
• Proximity-coupled (electromagnetically coupled) feed: The feed line runs on an intermediate layer beneath the patch, with no direct contact and no slotted ground plane. Coupling is capacitive. This approach offers among the widest bandwidths of the common patch feeds and keeps the feed line away from the radiating surface. It also requires a multilayer structure and accurate alignment.