An example of this is the mixed signal asic for industrial controllers. These controllers cover many industries and come in many forms. The basic premise, though, is for a system to monitor and react to a defined set of performance parameters. This usually incorporates some sort of feedback loop which the control system will act upon.
It is here, in the measurement and sensing part of the system that the mixed signal asic has become prevalent as tighter constraints and less hysteresis becomes desirable. Input signals can be anything from positional information to temperature, pressure, gas detection etc.
Today, the world is thought of as digital, however in reality the sensors are analogue and there is signal conditioning and conversion to the digital domain to be performed to allow them to interface with the control systems. Due to the nature of where the sensors are placed (often in confined and difficult environments) it becomes an obvious conclusion to design and deploy an asic. These systems are typically in operation for many years and again the additional benefit of obsolescence protection by using an asic becomes an important factor for lifetime maintenance and spares.
Today, it is possible to integrate substantial functionality into an asic, especially at the smaller 22nm and even 14nm CMOS geometries.
However, these geometries are best suited for predominately digital designs because analogue design becomes more challenging as the voltage can be reduced and the area benefits for the analogue part quickly tail-off. So for industrial mixed signal designs the favoured geometries tend to be in the 350nm to 110nm range.
These can still achieve a high level of digital integration, including DSPs and microprocessors as well as offering the flexibility for the analogue part of the design. Power devices and high temperature solutions are also possible, opening up new application areas for integration onto the asic.
Another benefit is that the fabrication costs of asics has fallen year on year making the return on investment (ROI) achievable in a much shorter time frame even with the relatively low annual volumes typical of this type of application.
It is worth noting that even with low annual volumes, due to the extended lifetime of asic developments, the lifetime product volumes are often the business driver. The additional benefits of optimised performance, obsolescence and IP protection make this route very attractive.
The nature of the sensing elements has also changed as material science has bought CMOS sensing elements which lend themselves to integration either on chip or in the same package.
Fig 1 – Platform type ASIC
The difficulty is not in the implementing on the asic but it is in selecting the appropriate interface. As the data rates tend to be low, the most common are the Serial Peripheral interface, (SPI) or the I2C. Using supply based communication is also possible, as minimising pin count is usually desirable. Also if the sensor is to be deployed as a non-rechargeable battery powered device, then the power usage profile has to be carefully considered, and a low power asic with careful design can achieve long battery lifetimes.
In some cases, the solution is for a platform approach, where one design will be used in different scenarios within a product family, and it will support many sensor interfaces and communication ports. It may also process information local to the sensor (i.e. fat client type architecture).
There is always a balance between performance and costs and the key is to have clearly defined goals at the project outset. An asic design is not simply implementing an existing PCB design, there are different techniques used for analogue asic design, and good system knowledge and experienced designers are essential for a successful project.
In many cases, this comes from an amalgamation of the asic design team and the system designers, so it is crucial to build this relationship early in a project. An example of a platform type of architecture is shown in Figure 1.
This type of architecture can perform processing independently from the main controller and report back. Being generic allows it to interface with many different sensor types, and the controller communications can be wired or wireless. It can also be programmed, so has some flexibility. However, as previously discussed, the optimum solution will incorporate many different factors.
Another key element is the yield and manufacture of the device. This is built on established relationships between asic providers and foundries. So a provider should be able to demonstrate not only all the necessary design skills but also the ability to handle delivery and logistics together with processes and procedures for failure analysis, so that if a field failure does occur, it can be quickly resolved and addressed.
In summary, mixed-signal asics are prevalent throughout the industrial controller market even though at first glance this market does not exhibit the features people associate with undertaking an asic design. This is due to the use of targeted processes which offer versatile, cost effective solutions with significant benefits over discrete solutions.
Carl Hudson is business development manager at Swindon Silicon Systems