Why Thermal Management is a Hot Topic in 5G Design

October 21, 2020 by Dr. Steve Arar

As 5G barrels forward, engineers are increasingly talking about one of the major design challenges of this technology: thermal management.

5G technology is expected to provide wireless communications with less than 1 ms latency, 100 times network energy efficiency improvement, and data rates up to 20 gigabits per second (Gbit/s).

Recently, IDTechEx published a report about innovation and growth opportunities of 5G technology from a thermal management point of view. According to this research, GaN-based power amplifiers (PAs), die-attach solutions like pressure-less silver sintering, and thermal interface materials can play a significant role in addressing thermal management issues of the 5G technology.

In this article, we’ll discuss why thermal management is important in 5G and some ways to resolve this issue. 


Why Does 5G Require More Efficient Thermal Management?

A key technology to make 5G a reality is massive MIMO with full-dimensional adaptive beamforming. MIMO systems employ an array of antennas to reduce inter-user interference, increase network capacity, and achieve beamforming. The figure below shows a system with a 4×4 array of antennas.  


Depiction of how a highly-integrated package can create design concerns

Depiction of how a highly-integrated package can create design concerns. Image courtesy of Rick Sturdivant


With digital beamforming, each of these antennas should have its own RF transceiver. A typical RF unit consists of several different blocks such as an LNA, a PA, two ADCs and DACs, and some filters and mixers.

To avoid signal integrity issues at the 5G frequency range, it's important to integrate different circuit elements of an antenna into a single chip and place this transceiver chip close to the antenna. Hence, with a 4×4 array of antennas, there are 16 transceiver chips on a single board.

This level of complexity leads to a power-hungry system where thermal management is of paramount importance.

For example, such a system designed to operate at 30 GHz can have a heat density of about 1 W/cm2 (4 W of heat being generated by a 4-cm2 board). This might even be considered a comparatively low-power application.

Future 5G networks are expected to employ massive MIMO with hundreds of antenna elements to compensate for large propagation loss and achieve efficient frequency usage. The thermal management of these networks will pose serious challenges.  


GaN: A Fundamentally Better Fit for 5G

Power amplifiers are the most power-hungry building blocks of an RF transceiver and can account for as much as 75% of total dissipation when transmitting. The local heat flux of mm-wave PAs can be as large as thousands of watts per square centimeter.

PA device technology along with innovative circuit structures is necessary to make 5G a reality. From a device selection point of view, GaN-based solutions may be the best options. These devices have superior characteristics such as low output capacitance, high output impedance, high power density, and high breakdown voltage.

These features allow us to have high-power PAs with improved efficiency. The following figure compares output power and efficiency of published PAs. 


Depiction of how a highly-integrated package can create design concerns

Image courtesy of Rui Ma


As you can see, GaN PAs can deliver a higher level of output power at very high frequencies. Besides, GaN technology allows us to have a higher efficiency over a wide frequency range.


Choosing a Thermal Structure

Although GaN-based PAs have the potential to offer higher efficiency and output power, thermal management is still challenging even with these high-performance devices. In fact, without an efficient thermal structure, the generated heat can put stress on a GaN device and limit its RF performance. For example, a thermally-limited GaN device can have reduced gain, output power, and efficiency. Further thermal stress can ultimately lead to reliability issues. 

Depending on the heat density of an application, one can choose an appropriate thermal structure. For example, with a heat density of about 1 W/cm2, cooling configurations based on natural convection phenomenon might be applicable. At higher heat densities, forced-air cooling or liquid-cooling configurations might be required. 


Research on Embedded Cooling

In addition to these conventional methods, there are advanced techniques that attempt to reduce the thermal impedance between high-power chips and the coolant to achieve a more efficient thermal management solution.

In fact, researchers are developing chips with embedded cooling where thermal management is achieved by pumping a heat-extracting dielectric fluid into the chip through microscopic gaps as wide as a single strand of hair (~100 μm). Inside the chip, the liquid coolant extracts the heat and turns into vapor-phase. The vapor is then transferred to the outside of the chip where it recondenses and dumps the heat to the ambient environment.


Diagram of a pumped two-phase cooling loop

Diagram of a pumped two-phase cooling loop. Image courtesy of Pritish R. Parida


Interestingly, the employed dielectric fluid can even come into contact with the chip's electrical connections. As a result, this technology can be used to cool 3D chip stacks where a heat sink or cold plate might not be an effective solution.