Steve Wasserman: While leafing through the wonderful Life of Jos Claerbout website a few months ago, I came across a picture that had very special meaning to me. In the picture, Jos is prevailing on his Popster to explain some of the intricacies of NTSC color television signals so that he might be able to create a better color picker. (The color picker is a tool that Jos wrote to help website designers choose colors that look good on televisions.)
Jos wanted to understand everything about television. This was logical because he was writing a set of articles and tools explaining what a web designer should do in order to make things look good on a TV. When the basic TV signal standard was designed in the '40s and '50s, displaying web content was certainly not envisioned. Because of the compromises in this standard, all sorts of artifacts haunt web pages that look perfectly fine on the higher-quality displays that computers use. Jos wanted to understand each and every one of these artifacts and how to minimize its impact by using good design techniques.
Color television is not a light subject. A complete understanding of the topic draws on many fields of electrical engineering, including signal processing, circuit design, CRT physics, and other esoteric things.
On his quest to perfect a set of instructions for creating WebTV pages, Jos sought me out. Now, I don't want to brag here, but I have a degree in electrical engineering from MIT and many years of experience with television electronics. At our first meeting, things quickly went off the deep end. I introduced Jos to color spaces, dot crawl, crosstalk, high voltage regulation, and all sorts of things. We moved to the lab in order to illustrate some of these topics.
Jos was not content with merely knowing the names that engineers had attached to the various types of artifacts; he was hell-bent on understanding the very core concepts of color television. I did my best to give explanations that Jos could follow, but we ran into many roadblocks. Not knowing Jos very well (or his educational background) at the time, I asked, "Do you know anything about frequency responses?"
I will never forget the answer: "Not really ... (short uncomfortable pause uncharacteristic of a Jos conversation) ... But I know some people who can help me out. How about we meet again in a week or two?"
I'm sure I rolled my eyes. Here was this poet telling me he was going to learn signal processing and Fourier transforms in a week or two. We parted ways and reconvened a week or two later.
At our second meeting, it was immediately obvious how much effort Jos had put in. In that short time, Jos was able to understand enough of the basic concepts that we were able to have a much deeper discussion, which we did for a couple of hours. Jos left the meeting feeling rather proud, and I left feeling enormously impressed. No person has ever impressed me as much as Jos did that day.
Jos and I talked often about TV and other things over the next few months. He quickly got into extremely complex issues for which I had no answer, and he did some very good original thinking on the subject. A lot of this work made it into the color picker.
A few days before he died, we had a meeting scheduled to discuss his latest version of the color picker. He sent me an email to cancel because he had recently introduced some bugs into the code, and he wanted it to be working properly before he showed it to me. We rescheduled for later, but it was not to be.
And then I saw on the website! A photograph of Jos devouring knowledge of television color spaces with his beloved Popster (in full Toessel, of course) so I could finish my explanation!
I cried and cried when I saw it. The image reminded me how much fun it would have been to continue our professional relationship and our budding friendship. It was impossible to have a bad time with Jos. I dearly miss having him brighten my day.
-- Steve Wasserman
I didn't really work with Jos on the strict mathematical implementation of the Fourier transform, but rather on an intuitive sense of it. For example, I could draw two signals on the board and Jos could tell me which had more high-frequency content. (This was based on a neat trick that one of my college professors showed me. You can estimate how fast the spectrum of a signal falls of by counting the number of derivatives that must be taken in order to make the signal into impulses. A step, for example, turns into an impulse after one derivative and falls off as 1/f. A ramp takes two and falls of as 1/(f^2), and so on.) Jos understood that things with sharp edges (like the edge of a colored letter on the screen) would have large spectra that could leak over into other parts of the TV signal. Okay, maybe technically he wasn't doing Fourier transforms, but he certainly understood how they were relevant to the problem he was trying to solve. (I did lots and lots of Fourier transforms in college, and I'm not sure that Jos' approach to signal processing was any less helpful in this case.)
I remember the problem of adjusting the colors very well. This came from an idea that Jos and I worked out to help minimize the dot crawl that would inevitably occur at the edge between two colors, as in colored letters on a colored background. When two colors are separated by many degrees in YUV space, an abrupt transition between them will cause dot crawl. You can see this on a colorbar signal by noting how big the "zippers" are that run up the screen between the color bands. The one in the middle (I think it's the green to magenta transition) has the biggest zippers because the phase of the color carrier changes by 180 degrees there. As I remember the idea, when Jos' code sensed that there was a big phase transition between the two chosen colors, it would suggest new colors by reducing the saturation of one of the colors in order to lessen dot crawl.
There were lots of other problems we were working on.