MEMS | ARTICLE

IoT Sensors - why your R&D Partner could also

be your ideal manufacturing partner HOWARD RUPPRECHT, CEO VTT MEMSFAB LTD

10

MEMS | ARTICLE

Microfluidics | Custom Labware | Titre Plates | Micro Molding

Experience. Knowledge. Tenacity. Passion.

‘A trillion sensors, a million applications and a fragmented market' In a hyper-connected IoT world it is not hard to imagine a trillion sensors collecting rich amounts of data that will provide new insights that will shape the way we live in our intelligent homes or commute in our selfdriving cars. There is certainly no shortage of market research predicting healthy growth in the market for such sensors. This has led to much discussion about how to economically manufacture large volumes of sensors. Some people predict the demise of batch-based wafer processing and the potential need for new manufacturing technologies, such as rollto-roll printing, to meet these demands.

Ask us what we’ll bring to your next project.

While there will undoubtedly be a need for high volumes in some applications, a closer look at the market predictions tells a different story – one of deep fragmentation where there is tremendous diversity in both markets and applications within those markets. Take for example gas sensing with optical MEMS spectrometers; the range of gases and concentrations that need to be detected can be very broad meaning that the characteristics of each sensor need to be customised or tuned accordingly – this is compounded by differing application needs where the environmental demands of an automotive application will differ significantly from an indoor sensor application for buildings. Essentially the opportunities for electronic sensors ultimately comprise of thousands of niches where annual volumes can range from a few hundred to tens of thousands of sensors per year – with very few applications with requirements for millions or billions of sensors of the exact same type.

Why does this matter? What this means is that the demand for new MEMS sensors can be mostly satisfied by existing semiconductor processing & packaging techniques and in many application cases, the volume of wafers required will actually be quite low. In MEMS manufacturing, size is important; most sensing elements are quite small so if we were to take a typical 150 mm wafer with a sensing element that is 4 mm2, a single batch of 25 wafers can produce close to 90,000 sensors (assuming an 80% yield). This volume is enough to satisfy the annual needs of many companies. The paradox is that such low wafer volumes are not interesting to most contract manufacturers and the actual production costs will likely be dwarfed by the R&D investment to develop the sensor platform, and to then potentially transfer the technology to a new facility. continued on page 12

commercial micro manufacturing international Vol 10 No. 4

PDC

+1 480.596.9380 | plasticdesigncorporation.com

e: [email protected]

11

MEMS | ARTICLE

Producing a single batch of 25 wafers of a custom design might cost between €50 & €100k to manufacture and could yield upwards of 50,000 sensors which would satisfy a whole year of production for many companies.

12

MEMS | ARTICLE

Sensor development versus manufacturing costs The process of developing new MEMS, micro or nano-electronic sensor elements can be an expensive business. Depending on the complexity of the sensor and the maturity of the technology platform, development costs can run into millions of euros and take upwards of three years. While some companies do all of the work in-house, many companies are cannot justify this vertically integrated approach and are increasingly partnering with R&D organisations to develop new products. This allows them to leverage public funding and utilise expensive national research infrastructure, often using an existing technology platform that has reached a level of maturity (e.g. the research has taken the technology platform to a TRL level of 6 or 7) which makes the realisation of multiple individual products more viable. To use an example from VTT Technical Research Centre of Finland’s own research activities, the company has developed a hyperspectral sensor manufacturing platform that can be used to produce MEMS based spectrometers for a range of sensing and analytical applications. The collective investment has been more than €5M and has taken over five years to realise. The groundwork and development has been funded by many successive projects, both private and publicly funded. The work has been carried out in the Micronova fab in Espoo which contains two large semiconductor cleanrooms and boasts a comprehensive range of analytical and processing equipment worth tens of millions of euros. Most SMEs do not have the luxury of these resources in-house but can easily access the results through contract research to produce bespoke products, based on the technology platform, customised for their specific application needs. With many research organisations it is also possible to license the necessary IPR so companies can offer unique competitive products that are well protected in the market.

Producing a single batch of 25 wafers of a custom design might cost between €50 & €100k to manufacture and could yield upwards of 50,000 sensors which would satisfy a whole year of production for many companies. If these sensors are subsequently integrated into an instrument that sells for €1,000 per unit, then a €50M business can be enabled and sustained from a very low wafer manufacturing volume. In this example, it would be impractical for any company to bear these costs and amortise a €5M, multi-year development cost over a single batch of 25 wafers each year (€66k per wafer over three years!)

Tech transfer or not? Historically, the prevailing logic has been to take the developed process and then transfer to a production facility but does this always make sense? With CMOS components, there are many foundries offering essentially the same process for different designs, this has led to many fabless semiconductor companies and a wide range of silicon foundries that can offer compatible manufacturing services. Conversely, manufacturing processes for MEMS sensors are notoriously specific to the process flows, recipes and equipment on which they were developed and undertaking a technology transfer project to a new facility can be expensive, risky and time consuming. Ultimately it becomes a question of volume versus marginal cost gains - if the subsequent annual volumes are low, it does not make economic sense to redevelop processes for a new facility as these costs may well exceed the ongoing low-volume manufacturing costs.

Many sensors but few wafers Manufacture in an R&D Fab When ongoing production needs are relatively low (