This is part one of the interview. You can read part two here.
Tell us something about yourself.
I have a background as an aircraft mechanic and commercial pilot.
Why machine vision?
To be honest, we didn’t start out in machine vision. We were building OEM patch panels and active interface devices for Cisco and HP. We were good at it, but not the best in the world and we decided that we needed to be great at just one thing. I thought that patch panel and interfaces should be that one thing the company focused on. It was a colleague that made the case for machine vision.
How did you start to learn about machine vision?
At the beginning, we didn’t even know what Camera Link was. It started when we got a request to build some of these cables for a customer, but after we made the cables, he passed away. So we looked at the assemblies we had built for that customer and wondered “what’s machine vision?” So we decided to learn.
It was around 2005, and I remember going to trade shows, and we’d always see cables falling just apart. They were not molded and looked as if they were built in someone’s garage. I thought “We could really make an impact in this industry.” Most companies didn’t seem really serious about the cable assemblies at that time. We wanted to be the best. We had to walk around machine vision shows with a suitcase full of cables and one of our colleagues would always have a right cable. It was like magic.
What were the industry challenges at that point?
When we started in this business, in 2005-2006, it was chaos. Every interface/connector was different, and it was very labor intensive to come up with each solution. We used LVDS interfaces, which would start at 300 MHz, then go to 600 MHz to go faster.
Around that time Camera Link came out. The industry had done a really good job of consolidating around a single standard—it really helped the industry grow. Everyone adopted rapidly, which was good for a while.
The happiness didn’t last?
The industry wanted to go faster, which made it more complicated. At the start, we could build 10 meter Camera Link cables at a given bandwidth and speed of about 40 MHz. But people almost immediately wanted to push Camera Link to 66 and finally 85 MHz. They wanted cables that were faster, longer, more reliable and cheaper. It’s good to have a standard interface, but due to the industry demands, we had to make dramatic improvements to give them t what they want- longer distance, faster speeds, zero errors. This keeps the industry moving forward.
Or people start to create alternative standards…
Yes. And that’s why, really, there’s a reason for every standard. Each is a solution to a certain type of problem. Today the rest of the system lags behind the camera. The cameras are so fast, with so many pixels, that the data rate is just awesome. We can go faster and bigger, but the camera can do so much more. I think we’ll see 12.5GB/s soon and in a few years we’ll have 25GB. But if you want that, you’re still in the frame grabber business, and needing special cables.
So how do you choose?
You have to make the best of the situation. High data rates generally mean you have to use shorter cables. Still, CXP and GigE go pretty far. So does CLHS. We’ve been able to get USB3 Vision copper cables to go 25meters. With fiber optics, we can transmit up to 100meters. GigE / Ethernet is economical, and very common. It’s everywhere. That’s good, but it’s a problem in industrial environments. The RJ45 used on gigabit Ethernet cables wasn’t’ designed for them. It’s too fragile. It goes back to how it was developed from the very start. The standards body only involved the cable guys at the end. That was an oversight.
USB3 is high speed and is a common interface on most computers. It makes sense to adopt this standard, but there was a length limitation in the beginning which the standards committee has now addressed. Firewire was a good high speed attempt, but had length limitations as well. CoaXPress and CLHS are great standards for the pinnacle of the market but I wish there was only one standard for that market segment. Still, it’s not like the bad old days where everything was only custom-made. People have a lot of choices today for industrial machine vision cables.
But all this choice comes with cost and complexity.
Today there are so many varieties and options; we carry 150,000-200,000 different configurations. That drives the price up. I’d like to find a way to reduce that. But it’s only really a problem if you don’t think about cables up front. People often think of cables last. Cables are less than 5% of the total cost of a vision system. It can be hard to explain why they should pay twice the price for an industrial cable instead of an off the shelf consumer type cable. We realized that we had to go out and train our distribution channel to talk about why you choose a fifty dollar industrial cable instead of a 5 dollar consumer cable.
It comes down to what you are doing. Everyone has a special situation. Cables are so important because they are often the most vulnerable parts of systems that operate in specialized industrial or medical environments. These are brutal environments (dust, noise, heat, vibration, motion) and with high certification standards, especially inthe medical and military markets.
For cables, we have to think about how they will be used, and assume that they are mission-critical. Whether it will be used in stereotactic brain surgery or sitting connected to a computer on a desk? We have to spec and test it accordingly. The cost comes down to risk. For almost any system, it’s not worth it to try to save a few dollars. If it’s the wrong cable, if there are any problems in terms of quality or compatibility with the system, it could bring the system to its knees.
[To be continued…]