SoC Technologies in Lighting - Today’s Landscape and Tomorrow's Practices

Since the early 2000s, system-on-chip solutions have had a tremendous impact on embedded and portable technologies. The concept of integrating multiple specialized chips and functions into a single piece of silicon set new standards in computer miniaturization, driving the smartphone revolution and paving the way for new applications. With the advent of the LED era, SoCs entered the lighting component market, promising to revolutionize the analog lighting control industry. This transformation is underway, accompanied by chaos and uncertainty, but there is no doubt that system-on-chip technologies will remain part of the lighting landscape. However, there are still many challenges to overcome. The adoption rate of connected lighting is not particularly impressive and things seem to be stuck at the early adopter phase. We’ve been hearing about smart lighting benefits for several years, but technology fragmentation hasn’t reduced at all, wireless systems can’t match wired solutions in terms of reliability and scalability, and there are still more questions than answers with respect to future business models and industry standards. Are SoCs part of the problem? Let’s take a closer look at system-on-chip technologies to find out whether future developments in SoC design can contribute to widespread adoption of connected lighting in commercial spaces.

SoCs Basic Architecture and Structures

A system-on-chip contains a number of hardware components. At the heart of every SoC is a CPU, which is where information processing takes place. Depending on the type of processor used, different types of SoCs can handle information at different rates. The efficiency of this process is extremely important from the perspective of commercial lighting systems. A mesh topology, which is considered essential for connected lighting applications, imposes high requirements on wireless networking infrastructure. Such networks keep buzzing with activity as hundreds of data packets constantly travel in all directions, relayed between individual nodes and repeated a number of times to prevent packet loss. They are never silent, even when lighting conditions remain unchanged in a given space. This is particularly true for adaptive lighting systems that use sensors to drive efficiencies, while sending various types of data to the cloud. Connected lighting systems simply require SoCs to process significantly more information than typical IoT applications, which is why computing resources are so important.

Information processing capabilities are also strongly correlated with the ease of development. When working with less efficient CPUs, developers need to spend significant amounts of time on code optimization. This results in higher costs and longer time-to-market as programmers struggle to ensure stability of their software stacks instead of developing robust software features. The same applies to another hardware component found in every system-on-chip, the memory block. SoCs with 8kB RAM and 8-bit processors based on architectures designed in the 80s can still be found in chipmakers’ portfolios. They do offer stable programming tools but require enormous optimization effort due to limited processing and memory capabilities. More recent system-on-chip solutions, e.g. the ones with at least 64kB RAM and 32-bit processors, provide much more efficiency and design flexibility, while enabling the usage of advanced programming techniques, such as thread isolation. From a developer’s perspective, it is incomparably faster, easier and more convenient to build software stacks for such silicons.

In addition to RAM memory, SoCs also include flash memory. This is another critical component for connected lighting applications. Flash memory stores information that is shared by all of the nodes of a wireless mesh network. This includes sensitive data, such as security keys. It should be an absolute requirement for SoCs used in professional lighting to have built-in flash modules since they are resistant to the so-called "trash can attack”, a serious security threat that is very often overlooked. If a discarded device has an SoC with external flash memory, the memory unit can be unsoldered and security keys can be retrieved from it, allowing unauthorized persons to eventually access the network. This cannot be done with SoCs using built-in flash.

Another hardware component found in SoCs is the radio module. In hopes of reaching a broader set of market applications, silicon vendors, more and more often, provide support for more than one radio protocols in their chips. This seems like a reasonable response to technological fragmentation in the wireless communication landscape, although it would be better for everyone if such desperate moves were not necessary. Multiple protocol stacks require more resources and in the world of the IoT - and commercial smart lighting in particular - there is never enough processing power or memory.

SoCs also include a variety of integrated peripherals, sensors and external interfaces - such as UART or PWM. Both are very relevant when it comes to development of products for connected lighting applications, but are commonly found in today’s SoC solutions.

Figure 1: A traditional computer architecture can be seen on the left. All major components are separate chips interconnected via data magistrale. A typical SoC architecture is presented on the right. All SoC components are integrated into single chip and connected internally

Evolution of SoCs

If we look at technological advancements in SoC design over the last couple of years, it is difficult to point out specific groundbreaking developments that could be called real enablers of higher integration between systems-on-chip and lighting products. A number of factors were important from the point of view of connected lighting solutions, but it’s more about gradual improvement and optimization than revolutionary changes. In which areas this improvement was most significant? Data processing, memory resources and energy efficiency, just to name a few. As was already emphasized, more resources translate to higher stability and robustness, while at the same time allowing for quicker development and shorter time-to-market. But as we delve into more specific features of software stacks used in connected lighting products, it turns out that benefits are much bigger.

Security is a perfect example. The whole problem with security in the IoT is that relevant security mechanism needs to be run on small devices with very limited resources. At the same time, attempts to break these mechanisms can be carried out on powerful computing farms. And smart devices have to withstand such attempts. If we expect to see connected lights inside hospitals, airports and schools one day, then security simply must be flawless. In the era of traditional wired lighting control systems, this wasn't even an issue. Security just wasn't needed. But in the world of connected lighting, it’s clearly one of the biggest challenges. To meet these stringent requirements, SoCs need to handle advanced encryption and authentication procedures. Strong processors found in the latest generations of chips are capable of handling modern and effective encryption algorithms without a significant trade off in overall SoC performance. To support encryption mechanisms even more efficiently, systems-on-chip often include dedicated hardware security modules.

Chips with more processing power and more memory resources can also handle software stacks with more functionalities implemented in them. This directly translates into more features that smart lighting products can accommodate. Such features might include, e.g. a time scheduler automatically controlling LED fixtures, a system for real time synchronization, support for battery operation, etc. On the other hand, flash memory resources are very important for the over-the-air firmware update capability. SoCs with sufficient flash can carry out this process without interrupting device operation because they can store two copies of firmware. A device can keep using the original copy while the update is being downloaded and implemented, and then switch to a new version once everything is ready. SoCs with smaller flash resources need to overwrite the original firmware during the over-the-air update process, which makes it impossible to perform such updates over a mesh network. And it needs to be remembered that software stacks for connected lighting are particularly heavy. This again shows that SoCs with more resources are simply more functional and can enable more mature and reliable products.

We mentioned energy efficiency as one of the fields where the biggest progress was made over recent years. For lighting and sensor-based applications, these advancements are priceless. Nodes of a connected lighting network need to be able to process information on a constant basis even though some of them have no fixed power supply, which is why support for efficient battery operation is essential. In addition, even when LED fixtures are turned off, their radios are constantly turned on - so optimization in power consumption helps prevent excessive energy drainage. Over the years, silicon vendors have developed advanced power and resource management mechanisms to maximize energy efficiency and battery life. In the latest chips, peripherals have independent and automated clock and power management so that they can be powered down when not required for task operation. This allows for keeping power consumption to a minimum.

Figure 2: In a traditional computer architecture, encryption keys are copied between storage, CPU and memory (left). In an SoC without a cryptographic module, encryption keys are copied between storage, CPU and memory while not leaving the chip (center). In SoCs containing a cryptographic module, encryption keys don't leave a cryptographic module of the chip (right)

SoC’s in Smart Lighting Applications

Looking ahead, can we expect any spectacular developments in SoC design that could open new opportunities for developers of smart lighting products? A trend we can expect to start witnessing soon is the introduction of cryptographic modules that can radically improve security in connected devices. In addition to accelerating encryption operations, they will act as highly secure storages for network keys. Such cryptocells can carry out all necessary encryption procedures, and keys created as part of these procedures will never leak outside - they won’t be circulating inside the SoC itself even though the chip will be able to use them to perform encryption operations. On the processor side, dual-core CPUs are used more and more often these days in modern SoC designs. The trend is likely to continue, increasing processing capabilities of silicon chips.

But apart from that, we should not expect any hardware revolution in SoC technologies over the next couple of years. What we can expect are relatively small, incremental improvements. CPUs will be getting slightly faster, memory resources will be growing and power consumption will continue to decrease, enabling even longer battery-life in sensor-based applications. All these improvements are certainly welcome, but this is not what will drive widespread adoption of connected lighting in commercial spaces. There are SoCs on the market that can handle even the most demanding software stacks. Silicon technologies are already mature and advanced, and today they give us all the tools we need to build smart lighting networks with wire-like reliability and multiple powerful features. What we still lack is a fully scalable and globally interoperable low-power wireless communication technology. This is where the real revolution is yet to happen, and this is the area that should be watched very carefully right now.

The Future of SoCs in Lighting

This year looks particularly interesting as far as connectivity technologies are concerned. The Thread Group recently announced availability of the first line of certified software stacks and finally launched its product certification program. We should soon be able to verify how well this new protocol performs in connected lighting applications. Chances also are that by the time this article is published, the Bluetooth SIG will have released the long-awaited Bluetooth Mesh specification. Contrary to the application layer agnostic Thread technology, Bluetooth covers all of the layers of the OSI communication model and thus seems well positioned to address the problem of interoperability, arguably the biggest roadblock to widespread adoption of smart lighting. With the arrival of new technologies designed to meet the most recent market trends and customer expectations, we are also likely to see legacy protocols fade away. The upcoming months certainly look very interesting as we might finally see some sort of consolidation among wireless connectivity solutions used in connected lighting.

Whether or not 2017 will be a breakthrough year for low-power wireless communication, smart lighting technologies will be getting more mature and more reliable over time. Eventually, the business will grow to a point where silicon vendors will have reasons to start competing more fiercely for customers in the lighting segment. This is when we might see the first line of SoCs dedicated to connected lighting applications. Again, we should not expect this to spark any revolution, as SoC-for-lighting designs will most likely be focused on further optimization. Crucial components (such as flash memory resources) might be enhanced, while redundant ones (such as interfaces for multimedia or display) might be removed. Developing will get even easier and costs will drop further, although core technologies will be pretty much the same as in today’s high-end SoC solutions.

Conclusions

On a final note, when speaking of hardware for connected lighting, we should not forget about the crucial role of software. As much as we need SoCs to handle complex software stacks required to build professional smart lighting systems, it is the software that defines what these systems will do and how we’ll be using them. Systems-on-chip become another specialized component that is needed to introduce new features and functionalities, but they are just the means we utilize to accomplish these goals. At companies, individuals responsible for key decisions regarding SoC design implementation usually have much more to do with software than hardware. And as time goes by, it is the silicon that adjusts to software needs, not the other way around.

Figure 3: Tiny Silvair Bluetooth Mesh evaluation nodes using advanced SoC technologies

(c) Luger Research e.U. - 2017