Despite consumer media highlighting some of the more ludicrous “innovations” of the Internet of Things (IoT), including wirelessly connected toothbrushes and cocktail stirrers, the industrial market is forecast to be the largest adopter of the IoT “connected” phenomenon. Germany’s Industry 4.0 initiative is an influential driver in the manufacturing sector, but Industrial IoT (IIoT) applications also embrace HVAC, transportation, smart metering, in fact everything that was once described as machine-to-machine (M2M) communications, and more. The promise of the IIoT is greater efficiency and flexibility in manufacturing, in operations, in asset management and in the service and proactive maintenance of industrial systems.
The building blocks of the IIoT are sensors that detect what’s going on in the world, data converters (analog to digital and vice versa), local processors, memory for data storage, and wired or wireless connectivity to actuators or to remote data collection and analysis centers. Remote processing may utilize powerful cloud computers.
Traditionally, most sensor-related electronic systems are built onto printed circuit boards using discrete components. The on-board microcontrollers that form the processing engines for many such systems now include peripheral functions like data converters, memories, and processors onto the chips. They also come with a variety of connectivity options such as serial interfaces and controller area networking, and some now include wireless connectivity too. However, most systems need many external components in addition to microcontrollers to achieve the desired functionality.
As the degree of automation grows, so do the number of inputs and outputs in industrial electronic systems (see figure 1). The inputs come from switches, keyboards, touch pads, encoders, scanners, and sensors. The sensors may measure pressure, temperature, humidity, air quality, acceleration, or a host of other environmental conditions. Outputs can include drivers for displays, heaters, motors, and actuators or switches. There could also be connectivity – wired or wireless. In between the input and output, there are often a variety of signal conditioning circuits to simplify compatibility between external devices. These may be as simple as voltage level shifters or more sophisticated signal conditioning functions.
Even a modest control board might contain hundreds of components, and these must be specified, sourced, purchased, and held in inventory before they’re assembled onto the board. You often need a lot of components because there is no off-the-shelf integrated circuit that will do the job. Standard components from the major semiconductor suppliers are designed to fit the widest possible range of applications, which means that they usually represent a compromise for every one of them. No two applications are the same.
Where there’s an expectation that products will be manufactured for many years, there’s another risk to using standard semiconductors. They can become obsolete quickly, as some will have been developed primarily for use in consumer electronic products, where life cycles may be little more than a couple of years.
These are just some of the reasons why the idea of creating custom integrated circuits for industrial applications is so appealing. The chips do exactly what’s demanded of them – no more, no less. And they do so pretty much on their own, with minimal external components.
But there are several other benefits to designing a custom chip. Systems are made smaller, lighter, and simpler, consume less power, and are easier to test. They are more secure because it’s much tougher to reverse-engineer a chip than a board, so design IP is better protected. And there’s an opportunity to add product differentiation that may be hard for competitors to replicate. Remember, the custom IC enables you to create exactly the functionality you need or want.
When analyzing potential savings, it’s not just the bill of materials you need to consider. There’s also the cost of sourcing components, the cost of holding them as inventory, board assembly and test costs. In some instances, the ongoing energy savings of creating a custom chip may also be significant. Depending on the individual application, cost savings of up to 80% over the lifetime of a project have been demonstrated, and payback periods can be as little as a year in some instances, primarily depending on system complexity and the quantities to be manufactured.
So, if it were not for the perceived high cost and complexity of developing custom integrated circuits, particularly where product volumes are modest – say, a few tens of thousands per annum – many more industrial companies would explore the option. Let’s examine this cost issue in a little more detail.
Electronics designers know that transistors are getting smaller, the number per chip is going up (in line with Moore’s Law), and the cost of advanced semiconductor manufacturing is rising as processes become more precise and more technically demanding. The cost of mask sets used in chip making can run into many millions of dollars. But this can obscure the fact that most mixed-signal chips – those that combine analog and digital functionality – are not made using the latest sub-micron technology. They don’t need to be in order to achieve the desired performance, and in some instances, it would be impractical because of the relatively high voltages that could be present in various parts of an industrial control circuit. Mixed-signal chips are made using relatively mature semiconductor manufacturing processes, often on equipment that is already fully depreciated, so the costs are a fraction of those experienced when using leading-edge processes.
When doing any cost analysis, it’s important to consider total costs over the projected lifetime of a product. These are the most relevant factors when estimating the cost of a developing a custom mixed-signal chip:
- The number of systems that will be manufactured each year.
- The expected lifetime of the product – for how many years you expect to be making it, or variants based on the same chip.
- The number of discrete data converters on the board, in addition to those integrated into microcontrollers.
- The number of other analog components on the board – transistors, diodes, op-amps, regulators, references
- Whether there’s a processor on the board.
- Processor performance. For example, on a scale of 1–4, a basic 8-bit microcontroller might be a “1” and an ARM 64-bit multicore device would be a “4.”
- How much memory – RAM and Flash – is on the board, both on-chip in microcontrollers and as discrete components.
- Whether the board has RF connectivity. For a first estimate, whether this is Bluetooth, Wi-Fi, or some other standard doesn’t make much difference to the calculation.
From these few facts, you can come to an initial estimate of costs and see how long it would take to break even on a custom chip investment. S3 Semiconductors has developed a free online calculator (see figure 2) to facilitate the calculation. You can access it at: https://tinyurl.com/s3sbomcalculator. It takes less than a minute to enter the data and the results show the total product volume needed to break even, the break-even timeline, and the total cost savings in USD over the lifetime of the product.
About the author
Darren Hobbs is the Director of Marketing and Strategy at S3 Semiconductors. He is a graduate of University College Cork and the Henley Management College. He has over 20 years’ experience operating in the semiconductor industry in roles such as CMO, Product Line Management, Product Marketing and Project Management in a mix of SME and large multinationals.