normalized to their average value. Note that the simulations were performed at 60 GHz, which is slightly different from the IEEE channel 2 (60.48 GHz). Although the measurements also include the composite response of the IC and the flip-chip transition, the measured results are in excellent agreement with simulations—validating both the complex electromagnetic (EM) model-based simulation methodology and the 60-GHz measurement setup.
III. MLO PACKAGES
Although excellent 60-GHz antenna performance was achieved in the previous section, LTCC is relatively expensive
compared to organic packaging, and for many consumer electronic applications, the higher cost is prohibitive. Therefore, in this section, we present a lower-cost AiP solution that uses standard PCB processes. A cross-sectional view of the multi-layer organic (MLO) package is shown in Fig. 6. Unlike LTCC where stacked vias are inherently available, the usage of blind or buried vias would increase PCB manufacturing cost significantly. An open cavity at the bottom accommodates the flip-chip attached IC, avoiding the 60-GHz via structures. The
package has a liquid-crystal polymer (LCP) core surrounded by woven-glass-reinforced laminates (RO4000 series). LCP has emerged as a promising low-cost alternative for mmWave packaging material due to its unique electrical and mechanical characteristics [16]. LCP was selectively used for antenna feed
lines to have minimal insertion loss while RO4000 laminates with Jade adhesives were used elsewhere. Since electrical properties of these materials at 60 GHz were unknown even to their manufacturers, they had to be measured using a cavity resonator and then fed back into EM simulation. While the prototype reported in [17] contains antenna components only,
the final package also contains power distribution networks (PDNs) on metal layers M1 and M3, 60-GHz interconnects and
other high-speed signals on M2, and low-speed signals on M4.
Since there were only two metal layers (M1 and M2) above the chip cavity and M2 was heavily crowded with signal traces, several metal strips were carefully designed on M1 to connect
each power pin to a corresponding power plane on M3. Since
M1 was functioning as a reference plane for high-speed signals
on M2, the strips had to be narrow enough not to cause return path discontinuity, and at the same time wide enough to avoid excessive inductance and resistance in PDNs. Every critical
power domain was modeled using the transmission line method
[18], which allowed various scenarios to be simulated in a short
period of time. For the high-speed signals that needed to be connected to a board, through-hole vias and BGA solder balls
were optimized for discontinuity cancellation [19]. The simulated antenna gain of the MLO packages is similar to that addition, the thickness of the solder mask should be controlled of the LTCC packages. accurately to prevent joint open failure. Fig. 7 shows the first-of-a-kind MLO package fully integrated with the 60-GHz The superior electrical performance of LCP comes with a
phased-array IC. The MLO package has the same footprint
cost. Since it has only recently been used in low-cost PCB
(288 pin, 28 mm x 28 mm) as the LTCC package.
processes, it was time-consuming to find the right recipe. Considerable effort was also made to come up with a reliable Fig. 8 shows a comparison of the simulated gain and flip-chip assembly process for mounting the chip in the cavity. measured power of the 16 Tx antennas over the four IEEE First, the chip cavity dimension needs to be optimized. If it is channels. Again, note that the simulations were performed at too large, a large portion of 60-GHz traces will be exposed to 58, 60, 62, and 64 GHz, which are slightly different from the underfill material whose loss tangent is typically an order of four IEEE channels (58.32, 60.48, 62.64, and 64.8 GHz) at magnitude greater than that of LCP. On the other hand, if it is


which the antennas were measured. Given the complexity of
热门文档