Skip to main content

Let the software do the work

If the size of a market were in direct proportion to the number of suppliers, then everyone would be using smart cameras and the traditional frame grabber manufacturers would be in trouble. But, of course, the number of suppliers is more indicative of the growth rate of a market rather than its actual size. Frame grabber makers are under some pressure though; it just means they are having to run a little faster to keep ahead these days.

The deal everyone in the vision industry was waiting for was a major camera maker buying a major frame grabber company, and that happened in 2005 when Dalsa bought Coreco for a total consideration of $72 million. Industry sources suggest that the logic was that these two companies would combine their expertise and Coreco’s massive software library to create the killer smart camera.

Two years on that has not really happened; Dalsa and Coreco are doing well but its competitor in the Canadian province of Quebec, Matrox, has simply moved the goalposts. It believes that success in the imaging market is about the software rather than the hardware. Its major move even before the Dalsa/Coreco deal has been to move from highly-customised hardware to standard hardware with extremely powerful software libraries and tools. And if you want it in a smart camera, well it can do that too. The company will still be in the hardware business, but it expects to move to something more of a hardware platform that OEMs can adapt to their vision application. At the same time its OEM customers are asking it to use its expertise in integration to build a bigger part of the solution. The frame grabber as we know it is growing up – not going away.

François Bertrand, VP of sales and marketing, says the element that had surprised people at Matrox most about the Coreco deal was the price as a multiple of its earnings. But he feels that Dalsa, as a publicly-quoted company, had different pressures from a privately-held company like Matrox, so its route to growth was going to be different.

He says: ‘I think we are seeing an industry that has to consolidate – and one of the larger changes is the acquisition of Coreco by Dalsa. Dalsa is a public company and so it has to answer to shareholders. It has a historical growth rate of 10 to 15 per cent in revenue and operating profits. It is also using public money, so it has a bit more flexibility than a lot of the other players out there. It was not a surprise as it makes sense for Dalsa to provide more of the elements of the pie to customers. We are pursuing this as well but we are taking a systems approach.’

Bertrand says that all the evidence available shows that the vast majority of cameras in use in the market are analogue, but that this is starting to change. The priority for Matrox is to make sure that it is in a good position to continue delivering steady growth in three or four years’ time, when the analogue/digital mix will change.

He says: ‘We have never sold as many frame grabbers as we have this year. There is more of a skew toward Camera Link, but if those numbers are going to change, then it has not happened yet. We expect it to come, though. Our revenue per unit is going down, but it’s part of being in the high tech business that you are constantly expected to reduce the cost of your product. But the number of applications we are operating in does compensate for that. As the price goes down, there are more applications. Some companies don’t like that but, if you take a forward-looking view, then it is better to reinvent yourself than to let a competitor do it first. We are not afraid of participating in a market where prices are lower.’

François Bertrand, VP of sales and marketing at Matrox

One of its biggest moves has been into the Gigabit Ethernet market. With standard PCs now being shipped with GigE interfaces built in, Bertrand believes that this interface is becoming more important in areas such as metrology. He says: ‘Ultimately you cannot sell someone something that is more powerful than the customer really needs. In some applications GigE or even USB 2.0 is adequate. Unit prices are coming down, but that is not such a problem if you are actively looking for new markets for your products; and this is what we are doing.’

Traditionally Matrox’s business is selling hardware to OEM manufacturers. It invested large amounts of money in custom integrated circuits in order to deliver the maximum amount of functionality on the frame grabber board. Over time, two things happened. Firstly, as prices drop, you have to sell more units to recover the cost of developing the application specific integrated circuits (ASICs) and, secondly, standard chipsets have become more powerful. Matrox was faced with a watershed as to whether it was still worth the investment in custom chips and decided that the future was more likely to be in field programmable gate array (FPGA) technology.

Bertrand says: ‘We have taken our highly skilled people, who did the ASIC design, and they are now working very hard on our FPGA products. If you look at the latest Intel Core 2 Quad, it is a very powerful processor and the FPGA pricing model has come down significantly. It’s the same kind of HDL (Hardware Description Language) knowledge base. The last ASIC we did was north of $1m for our supplier to even pick up the phone.

‘If you look at some of our recent ASIC products, it would be impossible to put all that onto the present generation of FPGAs, so there is still a sizeable amount of engineering and knowledge that we can only offer the market through an ASIC product. But when it comes to developing new products to be in production in 2010, then it is not impossible to think that we can put our most complex ASIC into an FPGA. This has been a very significant change for us, over the past few years. The good news is that we are well equipped to produce these products.’

This development has more to it than cost reduction. FPGA technology means that the hardware is effectively defined by the software (firmware) loaded onto it, and the connection between cost and volume is broken. It suddenly becomes worth making small volumes of highly specialised devices at much more realistic prices.

Bertrand believes this approach will appeal to some customers, but not others. They could ask Matrox to take its vision processor and only implement a subset of its functions onto an FPGA on the board.

He says: ‘Some customers will be willing to share their intellectual property with us, but a lot of customers will not. The real way to leverage this technology on a low-cost and customisable platform is to provide the customer with tools it can use to make its own FPGA. That is the strategy we have adopted. Our flagship product is our Matrox Imaging Library (MIL), a toolkit of some 300 image processing and analysis functions. Out of this we have created an FPGA development kit. This takes the most suitable MIL commands and packages them into an FPGA bit stream. This allows the customer to pick and choose what they want. For example, you can have a Bayer filter that is optimised for speed or optimised for space, and the customer will choose what is assembled into the FPGA.

‘If they have specific intellectual property that they do not want to share with us, they can use the Altera Quartus 2 tools to add their own IP and connect it to the Matrox building blocks by themselves. We think this is a tremendous opportunity for our customers. They can configure the FPGA to do whatever their application requires.

Bertrand believes that the FPGA toolkit is the most important product for Matrox this year. It is investing in training its customers to use the development kit so they can create their own specialised hardware. He says the development tools market is not yet mature and there are many companies pursuing the dream of being able to programme something in C and being able to download it to an FPGA (known as ‘C to gate’).

He says: ‘This is new technology and at the moment it is expensive. There is a lot of investment at the beginning, but looking three or four years down the line we have a roadmap for this development and we think we have a very compelling story.

‘Every product we have launched in the last year has an optional FPGA and this will continue to be a feature of all our new products. We believe we are doing a little missionary work here, as this is not a common offering. It means we are having to do a little bit of education in the market. But the customers with whom we are working regularly show a lot of interest in this.’

Bertrand believes that the core value in Matrox is its knowledge and intellectual property built up in MIL, and that is where the future of the company lies, rather than the more conventional business of a frame grabber manufacturer, which is shipping boards. But he thinks that it is possible that in the future the hardware platform will become a standard volume product, pushed out like PC motherboards to be loaded with MIL routines.

‘This is further out in the future. We have opened up our software to be used with IEEE 1394 and GigE Vision interfaces. But there is a lot more to it. What I am seeing in the market is that people are concentrating on what they do best and are outsourcing more work to their suppliers. I think the need for vision suppliers to provide system-level products is more of a short-term trend than the commoditisation of hardware. Customers are asking us to do camera products, embedded systems and so on because they want us to produce more of the solution.’

He says that another of Matrox’s strengths is that it also works on the display end of the process with its graphics products division. He says customers often have issues when it comes to integrating their vision systems with display systems and Matrox has often provided consultancy services. He thinks that in the future this will be more important to customers who want to concentrate on their expertise in the application field, and leave it to their suppliers to tackle the interfacing issues by providing a more complete unit.

He says: ‘I find that so long as you price reasonably and are not trying to use your position as the supplier of a more critical component, business comes your way from existing customers.’

He says that the margins on each unit will not be as fat as they were in the glory days, when the frame grabber had several ASICs and plugged into the back of a PC/AT. The future for Matrox is providing a value proposition in a different way.

Customers are also in a position of facing greater pressures in delivering applications to their customers and supporting them. Matrox does not intend to match its customers’ knowledge of the applications or provide support to end users across the globe, but it can provide a larger part of the vision system that is the crucial component at the heart of the application.

Matrox has been investing in its own capacity to deliver larger pieces of the final solution. It has recently built a clean room at its headquarters, because a customer wanted Matrox to integrate the imager into the product. This customer wanted a particularly small camera for a high volume application, but it was interesting enough for the company to extend both its facilities and expertise.

Matrox has already dipped its toes in the smart camera market. Bertrand agrees that this is becoming a very crowded market and there will have to be some consolidation. What will make or break the smart cameras is their software capabilities rather than there hardware.

He says: ‘You can have the smartest camera in the world, but if there is no software to run on it, it’s not going to solve any of your OEM’s problems. If you look at other areas of the software market, it is very hard for a latecomer to enter the marketplace. That has been a reality check for many of these guys. That is one of the big problems that is holding back that market.

‘We have taken a different approach. We were one of the latecomers in launching a product, but we wanted to ensure it was fully x86 and Windows CE compatible, and fully supported in Microsoft Visual Studio. We are leveraging the Intel low power processors. What our customers are looking for is the SSE processing capability. We decided to wait until we could use our MIL library across platforms so the customers can choose what is the most suitable platform for their applications, without having to make an investment in new software.’

Bertrand says the market for the smart camera is applications like factory automation. This market is struggling at the moment because of the meltdown in the US automotive industry, which means there is very little new investment. But the low power requirement, combined with the extensive software library, is opening new opportunities particular in portable devices.

He says: ‘Between the laptop and the PDA, you are going to start seeing some interesting technology coming out that is very low power and runs on batteries. We are keenly watching this, because we think that an evolution of this is going to be useful in machine vision. We want to use technology within the Intel family, where we are using industry standard components and software. This may only be a $5m market, but it is likely to grow and will be a very interesting market for Matrox.’


Read more about:


Editor's picks

Media Partners