More 5G-capable devices will begin to appear in a variety of form factors, beginning at the base station level and filtering down to the edge. At the handset level, extended battery life is always in demand, which requires running devices at lower power than pre-5G devices. Despite the goals of running at lower power, thermal management will remain a top priority in new designs, especially in mmWave handsets.
At the base station and data center level, off-loaded data processing tasks will consume progressively more power as more data is collected and processed in the cloud. At both levels, successive 5G rollouts will continue to move beyond 6 GHz, requiring greater use of GaN devices to power mmWave 5G devices and networks.
Managing heat and temperature in these systems is critical to extending useful lifetime, particularly for components. At high temperatures, electromigration also occurs at a faster rate, noise is more intense, embedded antennas radiate less strongly at the desired frequencies, and a host of other problems arise in 5G systems. Solutions to these challenges must be addressed at the board level, circuit design/operation level, and through the use of active thermal management solutions. Any solution will need to work within device form factor requirements without interfering with core functionality.
Anyone familiar with computer systems generally knows that traditional thermal management solutions can be bulky, and both electrically and audibly noisy. Form factors in the variety of devices that make up 5G infrastructure and devices will dictate which thermal management strategies are suitable, as well as their effectiveness.
The solutions implemented at the base station will not match those implemented in a handset. Obviously, given the size of current and future handsets, we can’t place powerful cooling fans in consumer cell phones. Additionally, we have the issue of unwanted EMI from thermal management solutions (fans and heat sinks), which can interfere with low level signals received at a base station.
At the data center level, the thermal workload placed on rack-mount servers can’t be overlooked. However, as these units are constantly monitored and are not subject to the same constraints as base stations, more creative active thermal management solutions are feasible in this environment. Fig. 1 shows a summary of the potential thermal management strategies that can be implemented at the data center, base station, and handset/IoT levels.
Figure 1: Potential thermal management strategies that can be implemented at various levels in 5G products.
Handsets carry such small form factors that active management strategies (cooling fans or liquid cooling) are simply not possible. The same applies to the majority of IoT devices (wearables, smart home systems, etc.), although this category is so broad that it encompasses a huge range of products. In these products, implementing passive thermal management techniques at the board level alongside circuit optimization is key to preventing overheating of critical components. Just this past summer, the first wave of 5G handsets was found to fail in summer weather, where the Qualcomm X50-based 5G modem switched off due to overheating, forcing handsets to run at 4G . In the near-term, one can expect other systems and networks relying on 5G modems (V2X, edge devices, etc.) to suffer similar problems.
In IoT devices that are large enough to accommodate bulky heat sinks or large cooling fans, combined passive and active cooling measures can bring temperatures low enough to allow persistent 5G operation. New thermal interface materials (TIMs) with very high thermal conductivity can help remove heat from critical components and dissipate heat to the device enclosure. In addition, new high frequency laminates and unique substrate materials may offer promise in equalizing temperatures throughout rigid boards, although this solution is infeasible in boards for handsets as they move to an all-flex architecture. In mmWave 5G handsets, the solution involves continuous optimization of RF chipsets and power electronics, as well as the use of newer TIMs for heat dissipation. Bulky, flat copper heat sinks are already present on the enclosure of some smartphones, but more needs to be done to prevent the aforementioned failures in summer weather.
Figure 2: Advanced CVD diamond TIM from ElementSix .
One place to start is in optimizing power regulators for various functional blocks in a handset. In switching regulator circuitry, using smaller discrete components is desirable as this increases the rolloff frequency at the output. By running power regulators at a higher PWM switching frequency with faster edge rates, thermal losses can be reduced in the regulator, which reduces heat generated near critical components. The edge rate of a PWM signal is an important determinant of the heating losses in power regulators. Using a faster edge rate forces MOSFETs used in these regulators to fully switch off, which reduces conduction and power losses, thus allowing smaller components (both physically and in terms of component values) to be used in regulator circuitry.
In terms of EMI, using a higher frequency in a switching regulator creates a new danger. The layout for these circuits needs to be optimized alongside the overall topology and switching signal in order to minimize switching noise in downstream circuitry. The aforementioned use of smaller components simultaneously aids layout minimization, efficiency, and thermal management.
The power regulator optimization steps for handsets can immediately be mirrored in base station equipment. Other solutions implemented at the data center level can be mirrored at the base station level. Base stations contain racks of equipment that are typically cooled with air conditioning, internal fans, and heatsinks on critical components. Major telecom carriers and base station OEMs are now looking to liquid cooling to remove waste heat, effectively copying solutions being brought into data centers. This solution allows equipment to be made smaller and lighter by removing bulky (and electrically noisy) cooling fans. Advanced thermoelectric assemblies are also available for outdoor deployment, including at base stations. These assemblies are an ideal replacement for compressor-based air conditioning systems as they can force cool air directly into rack-mounted base station equipment.
Figure 3: 5G Base stations are taking a cue from gamers and data centers by adding liquid cooling for thermal management.
Within base station equipment, thermal demands at the board level motivate the use of different substrate materials in conjunction with high frequency laminates. The existing crop of high frequency laminates from companies like Rogers and Isola are useful for providing a low loss, mmWave-compatible substrate, but the thermal conductivity of these substrates can’t compare with composite, ceramic, or metal-core PCB substrates. These alternative options, particularly ceramic PCB substrates, provide thermal conductivities reaching up to a factor two-hundred times larger than that of standard FR4 cores and laminates. Ceramic substrates also do not suffer from fiber weave effects, which create some interesting signal integrity and EMI problems at mmWave frequencies.
There is still plenty of innovation to come in 5G, which will bring about new solutions and challenges for thermal management. Innovations in beamforming, hybrid digital/analog base station design, the possibility of radio-over-fiber, and more advanced materials for components are just a few areas that require rethinking thermal management strategies.
Si power amplifiers used for RF transmission and signal processing at mmWave frequencies are notoriously inefficient, reaching only up to 30%. These components may have been appropriate for earlier incarnations of smartphones, but newer devices in handsets and base stations require a more power efficient solution. Major semiconductor companies are now building out their manufacturing capacity for III-V semiconductors in preparation to satisfy greater demand for sub-6 GHz and mmWave 5G components.
Power amplifiers and other RF components built from GaAs, GaN on Si, and GaN on SiC can see efficiencies well above 60%. Although these materials are in production and can be purchased at volume, the costs involved are still higher than the cost of equivalent all-Si components. In the mmWave regime, GaN on SiC is the best choice for mmWave components, including power amplifiers, ADCs , and analog front-end arrays , thanks to its broader bandwidth. GaN is already ideal for passively cooled base station electronics and other applications, and this material will inevitably find its way into mobile handsets. Another candidate III-V material that is receiving attention is InGaAs .
A new paradigm for delivering broadband wireless access services at microwave/mmWave frequencies is ROF. In ROF, a radio wave is upconverted from a baseband signal and used to modulate an infrared light source (laser diode). This modulated signal is then transmitted over a fiber link. This optical communication method enables multi-gigabit transmission to base stations over long distances with SMF or MMF optical fibers with DWDM. This methodology provides higher bandwidth and can be used to replace multiple lossy coax cables. An alternative method, which is still a subject of research, involves generating a mmWave carrier frequency by beating two modes from a mode-locked laser .
Figure 4: Standard ROF link architecture.
Despite the promise of this method, conversion between digital and RF analog pulses and back can create a new source of heat that needs to be removed from sensitive optical components. The methods for base station cooling mentioned above are applicable to radio-over-fiber, although the appropriate choice of FET laser diode drivers and other optical components is also critical. Laser diode driver circuitry, PIN detectors, the lasers themselves, and other optical components are very sensitive to changes in temperature. Power output and sensitivity generally drop as temperature increases. FET drivers for long range links should, at minimum, include a grounded heat sink that is attached with a highly thermally conductive TIM. The thermal demands only increase when larger mode-locked lasers are used for pulse generation, and it remains to be seen whether these systems will integrate with the existing 5G architecture.
New design software solutions are already allowing innovators to experiment with novel 5G-capable systems and deploy them quickly by taking a fully modular approach. Design tools, like ARM’s DesignStart and Geppetto from Gumstix, allow innovators to develop new systems-on-chip (SoC) and single-board computers for IoT products, respectively. A modular approach is a great design strategy for electronics engineers who may not have design expertise in a particular area, such as single-board computer design or SoC design.
These tools will allow designers to move their processing workloads away from data centers and perform them at the edge, which relieves the corresponding processing workload at base stations. These tools also enable new embedded products for applications in machine learning/AI, robotics, smart infrastructure, medical, automotive, and much more—all of which can be made 5G-capable as needed. The thermal management strategies outlined above can already be implemented in products created with these design tools, allowing new designers to quickly begin innovating novel 5G-capable products and solutions.
At NWES, we've partnered with industry leaders in 5G, and we have the experience you need to create your next 5G-compatible product. We're here to help electronics companies design modern PCBs and create cutting-edge technology. We've also partnered directly with EDA companies and advanced PCB manufacturers, and we'll make sure your next layout is fully manufacturable at scale. Contact NWES for a consultation.