A mmWave Phased Array Transceiver comparison

The essential technology that enabled the placement of a mmWave radio into a handset is the beamforming phased antenna array and the phased array RF Transceiver that it interfaces with. As such, phased antenna arrays are being deployed in Antenna in Package (AiP) modules that contain the antenna array itself and the phased array RF Transceiver die.

The mmWave radio architecture that has been observed in the smartphones torn down by TechInsights to date is represented in the simplified block diagram of Figure 1. A super heterodyne transceiver is partitioned between two separate components. A phased antenna array and an RF<->IF Transceiver is housed within an AiP. This ensures the high frequency RF lines are routed a short distance from the antenna array to the transceiver. The lower frequency IF lines are then routed to a separate component that houses IF transceiver die that converts to and from IF to BB. The BB is then routed to the modem.

Due to the lack of any discernable die markings, this die will be referred to as the “Samsung die”. This die is functionally the equivalent to the HG11-PG660-200. TechInsights has performed an architectural level of analysis of both phased array transceivers (ARC-2111-801 and ARC-2010-801). The following is a brief comparison of what was discovered.

Figure 3 is a polysilicon layer die photo showing the rough floor plan of the Qualcomm HG11-PG660-200 RF Transceiver. The die can crudely be divided in half with small sections along the top and middle of the die. Most of the die contains two super-heterodyne transceivers (TRX) that are located left and right of center. Each TRX consists of a pair of TRX tile quads and a shared IF front end. The center section contains the phase-locked loop (PLL) that generates the local oscillator (LO) signals for the transceivers. Each TRX tile quad contains four-unit TRX tiles with two unbalanced transmit/receive bidirectional pads. These pads are then coupled to the antenna elements within the module

Figure 4 illustrates the newly discovered Samsung mmWave RF Transceiver. The top and bottom edges of the die contain an interleaved series of four types of RF amplifier sections: RX type A, TX type A, RX type B and TX type B. Each RF bond pad is connected to an RX and TX amplifier block. The center area contains two identical IF RX/TX sections which include the mixer for frequency conversion. Both RX and TX mixers use the same local oscillator (LO) that is generated from the PLL located in the area between the two IF sections. The IF blocks connect to the RX amplifier blocks directly. To the TX amplifiers they connect through some intermediate RF gains stages.

Table 1 shows a brief summary of the two die.

There are multiple phased array RF Transceiver architectures, depending, on where the phase shift occurs in the signal chain. As with any circuit design there are performance trade-offs to any design. Let’s examine the architectures of the two phased array transceivers analyzed by TechInsights to date.

The Qualcomm HG11-PG660-200 architecture is shown in figure 5. It can be seen from this schematic that the HG11-PG660-200 implements the phase shift at the local oscillator (LO) stage. As a result, the phase shifting is not done on the main RF signal path but rather on a constant envelope LO signal thus relaxing the linearity and bandwidth requirements of the phase shifter. The die has a single PLL that generates a clock that is fed to a transceiver quad. This clock is most likely multiplied up and fed to the 4 individual transceivers within the quad where the multiphase LO signals are generated and fed to the mixers within the transceiver blocks. This architecture requires a mixer at each TRX tile. This means more circuitry and likely more power consumption however the mixers themselves can be down sized as they only drive a single PA. There will however be more stringent requirements on the RX mixer linearity as the signal combining is done after the down conversion and as a result the RX mixer will be exposed to large signal interference. [1,4] The LO phase shifting offers high accuracy and phase resolution. [2]

The Samsung mmWave RF Transceiver die architecture is shown in figure 6. The phased array transceiver architecture seen here performs the phase shifting at the RF level. The phase shifters are placed within the TRX RF Amplifiers block. This is verified via the transistor level circuit reverse engineering of the power amplifier chain shown in CAR-2111-802. The RF signals from 4 TRX blocks are then combined, and frequency converted in the IF back-end block. The RF phase shifting architecture requires a smaller amount of circuitry as there is only a single mixer in the IF stage. The single mixer eliminates the need to distribute the LO signals thus noise coupling can be minimized on the LO. The single mixer however drives multiple PAs which requires an additional RF amplifier in the TX path. The linearity of the RX mixer is relaxed as the signals are combined in the RF domain before entering the mixer. High frequency phase shifters are always a challenge to implement. Phase shifters at mmWave frequencies do not have constant loss over different phase settings, thus requiring an amplifier to compensate for the gain variation. This would explain the VGA at the last stage of the RX amplifier path before the phase shifter. [1,3]

Traditionally the RF phase shifting architecture is the most common as it requires less circuit components that translates to potentially lower cost and power. This is at least true in discrete implementations. It remains to be seen which architecture will dominate in the integrated phased array RF transceivers located in mmWave enabled handsets. Only two such chips have been analyzed to date.

The Qualcomm QTM545 AiP has recently been found in the Samsung Galaxy S22 smart phone. This module contains the new Qualcomm HG11-PR590-200 RF Transceiver die. Which architecture will it contain?

References

• UC Berkeley PHD Thesis by Jiashu Chen “Advanced Architectures for efficient mmWave transmitters “ Fall 2013.
• 5G mmWave Radio Designs for Mobile: Kamal Sahota, Vice President Engineering Qualcomm Inc. http://www.5gsummit.org/hawaii/docs/slides/D2_%239_Sahota_5g_060617.pdf
• RF Architecture Analysis of the Samsung RF Front-End Transceiver Die in the Murata SS1707051 mmWave Antenna Module (ARC-2111-802)
• RF Architecture Analysis of the Transceiver Die on the Qualcomm QTM535 mmWave Antenna Module (ARC-2010-801)