26-06-2020 | | By Robin Mitchell
This week Nokia has announced its new collaboration with Broadcom to solve their ongoing 5G issues. Why has this move been needed, and what problems do FPGAs provide for designers?
One of the key design challenges for any engineer when designing a product is to know what components to use. Constructing a circuit from individual parts such as logic components, signal processors, and processors, has its advantages over customisability, but can be problematic. For example, applications involving strict timing conditions may not be possible using multiple discrete components on a PCB which could also be subjected to external interference. In these scenarios, single-chip solutions often provide the best result with designs having two options; FPGAs, and ASCIs.
FPGAs, or Field Programmable Gate Arrays, are extremely useful devices for engineers who may want to make updates to the design on the fly without having to either replace the IC or change its fundamental silicon layout. Since FPGAs are programmed in the software realm, they are easy to prototype with, and can provide turn around solutions incredibly fast. ASCIs, on the other hand, provide designers with the ability to create an entirely custom piece of silicon that only contains the circuitry needed for the design. While ASCIs are far more expensive for prototypes, they are significantly cheaper than their FPGA counterparts when used in a mass-produced product. Since ASCIs only contain circuitry relevant to the design, they consume much less power per unit function than their FPGA counterparts, making them far more ideal in low-power applications that need to maximise signal processing. ASICs, however, are incredibly complex to design and often require a designer with VLSI knowledge to design the circuity from the transistor level.
Nokia, a mobile technology, decided to choose FPGAs as their bedrock when designing their range of ReefShark chipsets. Their reasoning for choosing FPGAs came from a mix of uncertainty and time. During the conceptual stage of the ReefShark chipset, Nokia had noticed that 5G standards had not been fully decided with potential volatility in its specification. If Nokia decided to use ASCIs as the underlying technology, it would be unable to make last-minute changes to their hardware if 5G standards were altered. The second reason for not using ASICs is their poor time-to-market; ASICs take months of development just to get into the simulation stage before a single prototype can even be produced. Thus, Nokia chose FPGAs for its ability to provide flexibility while also providing power and speed of production.
However, this week Nokia have announced their intention to phase out the use of FPGAs and turn to an SoC based design instead. The company that Nokia will be collaborating up with, Broadcom, has a long history of producing such hardware with one of their most famous SoCs being the heart of the Raspberry Pi range of computers. The reason for their change of heart comes from several angles. The first, and most important, is that Nokia has reported a fall in profits in the third quarter of 2019, which is a result of the expensive nature of FPGAs. The second is that now that 5G is established, Nokia can now invest in an SoC that, when produced in volume, will provide a significant cost saving to their 5G chipset range. While some of their products are already shipping with custom SoCs to replace FPGAs, Nokia intends to make all 5G related products SoC only by the end of 2021. This is evident in their hiring of 350 employees dedicated to SoC development.
Being an engineer in a world where standards can change quickly is incredibly difficult. If a company does not make decisions early on about what technology to use, then it may find itself behind in the competition. It is easy to say in hindsight that Nokia should have use ASICs, but at the time, 5G standards had not been properly established, and had they changed, all ASCI devices may have been made redundant costing Nokia a small fortune to rectify the problem.