Idle Diffractions

Are We Learning From Our Mistakes?

As we approach 20 years of the Hubble Space Telescope, the subject of why it broke and how it was fixed continues as a favorite topic on the lecture circuit.  Recently, I was invited to a panel discussion with the broad topic of quality in engineering – in this case software, at DVCon in San Jose.  The audience of a few hundred was there to discuss methods for identifying and maintaining software quality.  When I was invited, I pointed out I know none of the industry acronyms, but, the organizers were undeterred.  Joining me on the panel were a series of what I would call working Vice President’s, meaning they rose up through the ranks as an engineer, as I have.  This made for an excellent mix and perspective relative to the interests of the audience, which from my vantage point appeared to be those with 5-10 years in the “working” world. 

The event was entirely unscripted, other than the moderator having a backup set of prompting questions initially to get the audience engaged and later to step up the pace should audience interest start to lag.  Now, I am not historically a fan of panels and rarely attend a panel discussion format event.  To mention an exception, the yearly Emerging Technology panel at MIT has consistently brought together the best of the best in a forum where they speak candidly on topics that are relevant.  I highly recommend this event.

Now, it seems that between the five of us on the panel, plus the moderator, we established a synergy that emulates the standards I experienced with the MIT panel and a rapport with the audience that encouraged and achieved a true group event.  The audience engaged quickly and asked challenging questions, to which the moderator orchestrated our responses to keep them terse, relevant, and engaging.  The questions represented their real challenges in understanding their real environment.  What was rewarding was, it did seem that we at the front really had “been there and done that” and could offer the questionee a perspective that did cause them, as well as many in the audience, to pause and move ahead in their understanding of the environment that surrounds them on a daily, weekly, and monthly basis.

I was encouraged to use the Hubble 1^st Servicing Mission story (never to be called a “repair” mission) as a context for any comments I might be inclined to make.  As a result, I established early on the evolving theme, which I initiated at the University of Rochester two years ago and at SPIE Annual last summer,  that the days of engineering of any flavor in the ideal world needs to be put away.  Working engineers, especially the evolving generation, need to think and act defensively.  Specifically, while your contribution will of course be perfect, the environment which your contribution will be placed in will be flawed in a many, many ways.  The Hubble primary mirror being an excellent case history, but it is one of so many, that I have to conclude this is a pervasive flaw in our model of the world of engineering as a whole. 

Of course, hindsight, as “they” are known to say, “is 20-20”.  (A digression – if you are an American born and raised engineer, working in America, it is worth realizing that many, if not most of these colloquial  expressions are truly meaningless and quite disorienting to the majority of the audience, who are not in the category described above).  So, rather than dwelling on the mistakes of the past, what can be applied to the future.  In my case, I have directly managed, (usually 1:1 with another PhD in my field and on my staff) over 2,000 projects for over 500 companies.  In addition, I was on the Hubble repair team (but by luck, not the Hubble “breaking” team) and I have in my background other experiences that are even more expensive than the Hubble repair.  From this I of taken the following insights into being a substantial contributor in complex technical environments. 

First, of course, you do all of the basics, obsessively. 

A)     You communicate the parameters that govern the problem you are working on (Table 1 in optical design/engineering speak) in writing, with every external communication.

B)    You determine your assumptions, ESPECIALLY in today’s science, your most basic assumptions.  Does 2+2=4 (not necessarily).

C)    You find two independent ways to check a calculation, preferably with completely different tools

But, and here is what is new, you do not have the luxury to do everything twice (and we are assuming the business model is not the one where it was recognized that repairing a failure, even one you caused, can be a valid source of substantial revenue). 

Where do you focus your attention?

You focus on what is the project doing for the first time.  A consistent failure path is where the new aspect of the project changes an assumption that has held for all time, until the day the project starts, and, you did not “think” about.  I have example after example after example after example … (you probably get the point). 

I give a one hour lecture with some of these examples, and, perhaps I’ll bring some to this to blog space, in the future.

In the meantime, I hear they have posted the entire panel discussion on the Internet, not that I have checked it out.

http://dvcon.org/2011_Panel_Session

I am more than happy to propagate this conversation, should anyone have insights or great stories to add.  Also, if you have a local society or other forum, and this sounds like an interesting talk, I enjoy speaking on this topic, and I seem to be somewhat global these days.

Double Jeopardy? 50 Years After “Lens Design with Large Computers”

In 1966, a forum was held at the University of Rochester titled “Lens Design with Large Computers”, following on the heels of a less formal, and less documented, event in a similar vein in 1964.  Ten years later, I found myself at my first summer job in the optical design department of Perkin-Elmer (where I helped with the tolerance sensitivity analysis for the Space Telescope, Fine Guidance Sensor and test setups for AXAF (what becomes Chandra, in 25 years), little knowing it would not be relevant for a full 20 years), working with authors of some of those 1966 papers.

This is brought to mind by the IBM event of the decade, the Jeopardy Show.  When Big Blue took the chess title in the first challenge, it was immediately obvious that yes, this would affect me directly.  Why?  There are many parallels between chess and optical design.  Sure enough, within 5 years, the global optimizer in CODE V, which was merely a concept for debate at the 1994 IODC conference, was challenging the best designers.  In fact, 2003 was the last time anyone on my staff claimed they were better. 

So, what does the Jeopardy Show mean for lens design?  At dinner with a chaired professor in the interaction of humans with machines recently, he threw out the following disturbing observation.  What is going to happen when the computer becomes the expert, who will teach the computer new things?  Under the assumption that those currently espousing that the field of lens design is no longer necessary, with computational focal planes will only commandeer 60-80% of the application space, who will be teaching who in 30 years. 

Looking at the history, as was written earlier, optics is often found at the leading edge of technology.  James Baker represents the first, and some might say one of the best examples of a lens designer developing a computer-aided optical design environment.  He, along with Robert Hopkins, also a ardent designer, independently led the emergence of computer-aided optical design from 1945 – 1955.  By 1955 there is a mix of mathematicians, like Donald Feder at Kodak and optical designers, such as the founder of ORA, Tom Harris, working to move the software on many fronts, but primarily optimization running up to the 1968 conference.  If we look at the 1968 authors, listed in Table 1, we see the lens design community is the dominant source of papers.

Speaking from the American history alone, from 1970 to about 1990 there were a few smaller companies, including Scientific Calculations (ACCOSV), ORA (CODE V), David Gray (Cool/GENIE?), Doug Sinclair (OSLO), and a few other independents developing optical design codes to support their optical design consulting.  More prominent in this period were the codes under development at major aerospace companies led prominently by what was then Hughes/El Segundo and LMSC Palo Alto Research and the specialty reconnaissance companies like ITEK and Perkin-Elmer.  Beginning in 1980, I was at Perkin-Elmer, where I spent my evenings, for two years, learning the 200-300,000 lines of code and beginning to expand its capabilities to include, for example, circular gratings in support of AXAF.  In this period, pretty much across the board, lens designers were driving the developments.  However, by the late 70’s, things started to evolve (or devolve depending on your perspective) as the accountants started to realize what it was costing to maintain the code, typically $2-$3M per year and companies like ORA were offering a comparable, or often better product for a mere $1000/month, (per designer).  By the mid-80s, I had moved to ORA, seeing the end of software support at Perkin-Elmer, but, notably where I was presented with in retrospect, a historically significant choice, join the software “department”, OR join the optical design group.  Having no formal training in software, I went with the design group. 

So how does this all relate to Jeopardy?  The thought question is, as computer become more and more expert at, for example lens design, the number of experts in optical design will, by necessity (read economic pressures) decline.  Could this number approach zero in 20-30 years?  And if so, where are we?  Will computers intelligence rise to not only make the execution of code something only “they” can do, but, also find paths to new ways to solve problems or adapt to the evolving technology?  And, if they do not, who will be left to teach them?

Recall the events of WWI, when the British and the Americans both woke up one morning to realize the had fewer than 5 optical designers in the entire country and worse yet, even if they could execute a binocular design they had forgotten how to make glass.  It does seem déjà vu may be on our horizon.

Click on images below to enlarge.

 

                           

                   

    

Why Space Is Not Easy

I recently attended a small technical conference on stray light in Toulouse, France.  It was sponsored by CNES (the French space agency) and by several European aerospace companies. It was basically an optical engineering conference on a very specialized area of optics called “stray light.” As you can imagine, this means light that you don’t want showing up on the detectors of your optical system. A familiar example is a “ghost image” that you might see on a photo that you shoot towards the sun. You might see rings or hexagonal shapes or patches in your picture, obscuring part of the actual picture. These are internal reflections from the lenses or apertures that normally would be too faint to see unless the source is very bright (like when the sun is partly or fully in the frame).

In spacecraft optical systems (telescopes or imaging systems for Earth or other planets), this is a big problem. The sensors for these systems are very sensitive, and there are many ways for photons you don’t want to bounce around and find their way to your sensors, even if you are careful and don’t point directly at the sun, moon, or Earth (the brightest sources for most spacecraft). Preventing or reducing stray light is an important part of design and fabrication of such optical systems and can even determine the success or failure of a mission.

A case in point from a paper presented by Mr. Thierry Viard of Thales Alenia Space in sunny Cannes, France. The COROT spacecraft was launched in December 2006 to search for exoplanets with short orbital periods, especially large terrestrial (Earth-like) planets. COROT is a relatively small and low-cost spacecraft, so launching it to a dark, distant Lagrange point (e.g., L2) was not possible. An 800 km polar orbit would have to do, which meant that the very bright Earth would always be nearby (in addition to the sun and moon, though the sun would always be kept “behind” the spacecraft by pointing to a different area of the sky depending on the season).

The “transit” method of detecting exoplanets depends on recording very slight changes in the brightness of rather dim stars as a small planet passes in front of the star as viewed from our direction. The telescope had to be designed with a form that would minimize the chances of light from “off-axis” objects (those outside of the desired pointing area) getting to the detectors. Several optical criteria used in the chosen design form helped with this, in addition to extensive “baffles” (black metal rings that only allow light from certain directions to get in).

But this was not enough for COROT to succeed. The engineers also determined that cleanliness would make or break this mission. The slightest dust in the spacecraft could also scatter light to the detector, and such scattering had to kept to about 10 photons per pixel per second from the nearby Earth (which is putting out about 10^20 photons per second per pixel, a huge, huge number). This required keeping particle contamination (dust) to something under 200 parts/million, which is very clean even for a clean room used to building delicate spacecraft.   And there was no way to even test this without risking even more contamination, so it had to be based entirely on simulations. Pretty scary!

But guess what? The COROT optics team succeeded. When they first imaged the starry sky, it was pitch black except for the bright, distinct stars. Just as designed. They detected their first exoplanet  in May 2007, and the first Earth-like exoplanet , COROT-7b (1.7 times Earth’s radius) in 2009.

Of course there is much more to this or to any spacecraft than the optical systems. This is just an example of the clever engineering, hard work, extreme attention to detail, and ultra-high quality needed for any successful space mission. Space is not easy.

Photos courtesy of CNES.

Some Really Old Photons...

... were recently collected by the newly refurbished Hubble Space Telescope.  John Wood, one of my favorite scientists at NASA, from my days on the Hubble Repair team, sometimes send over some new results that he finds significant.  Since I wandered into optics to develop instruments for astronomers, after getting my initial degrees in Physics and accidently in Astrophysics, I’ve always tried to keep up, a little.  As you may recall, after a lot of politics, the Hubble was granted a new mission and some new instruments were installed.  They immediately went forward to repeat taking the now famous Hubble Deep Field shots, those with the longest possible exposure and in doing so, with the new equipment, resulted in collecting AND DETECTING some of the oldest photons ever, some 13 billion years old, shown below.

This comes to us from

http://spaceflightnow.com/news/n1001/11hubble/

which, of course, you should click on, if you have just 1 more minute. 

This got me trying to figure out how to relate to a 13 billion year old photon.  I didn’t get too far, but, try this.  I have had the misfortune to fly, so far, some 1.5 million miles on United, starting in 1992.  Since it is easy to relate to a mile, now I can somewhat relate to 1.5 million of them.  In a typical year, I fly some 100,000 miles.  Now, in round numbers, it is 25,000 miles around the earth, so, I fly about 4 times around it in equivalent miles in a year, at an average of 500 miles per hour, (not including airport time), thereby spending about 200 hours a year in an airplane.  Now light, on the other hand, inserted into a fiber, goes around the world 7 times, IN ONE SECOND, nice trick (where I’ve conveniently ignored the refractive index of the fiber).  So, light can go in somewhat less than a second, what it takes me 200 hours to accomplish. So, now I can almost relate to what light can do in one second.  A few other round numbers to play with, there are very nearly, pi*10^7 seconds in a year.  (pi=3.14).  So, light can go around the earth about 7pi*10^7  (~210,000,000) times in a year.  Now we’re getting close to one billion.  So, in 5 years, light can go around the earth 1 billion times.  So, here is an irrelevant statistic.  If light would travel around the world (it doesn’t since the it prefers a straight line, but, these days we could put it into a fiber), then, a pulse launched when I was born, would have now circled the earth over 10 billion times.  But, now 13 billion light years is another matter.  Since the 10 billion circlings of the earth only represent 1/7th of a second per circle, versus the distance travel by light in a year.  How to relate to that?  Well, the distance to the sun is slightly less than 100,000,000 miles (93 million actually).  Light that arrives here at earth is about, 10 minutes old.  (568 seconds).  In a day then, light can go to the sun and back, about 125 times.  So, finally, in a year, light can go to the sun and back about 125*365 times, or about 45,000 times.  At this point, I don’t think that helps me relate, but, almost. 

As an aside, Doug Goodman, being Doug, has a bag of poppy seeds, carefully measured out, to contain 1,000,000 seeds – should you every try to visualize 1,000,000 of something.  Then, you only need to visualize 13,000 of those (it is about the size of a sandwich baggie) to get to 13 billion.  Good luck.

When Will Virtual be Reality?

A futuristic-looking Head Mounted Projection Display (yes, it really does work).
Image courtesy of ODALab, Prof. Jannick Rolland.

This week I attended the International Symposium on Mixed and Augmented Reality (ISMAR) conference in Orlando.  This is an annual conference that circles the globe every three years, from the U.S. to Asia to Europe.  Next year it will be held in Seoul.  It started life as an IEEE workshop in 1998, maturing to a full conference in 2002.  In a year when any other conference I’ve been to was down 30% in attendance, ISMAR’s attendance doubled to 400, about the size of the upcoming International Optics Design Conference (IODC 2010) in Jackson Hole, WY in early June (couldn’t pass up the opportunity to plug IODC).

This year for the first time ISMAR attempted, with some success, to mix science and art.  Nearly one-third of the conference agenda was led by educators, performers, and authors.  While in general each group stayed on their own side of the hall, every now and then someone would find themselves in the “wrong” session.  In my case, this happened yesterday.  The speaker, a poet, was running down some recent works developed for augmented reality. Who knew? I found it to be a fascinating topic, well outside of my box.  In most cases, these are video format pieces that are meant to be experienced with a head-worn display (HWD), the latest terminology for this technology.  Which of course does not happen to exist, and therefore this art form is currently experienced though the screen of an iPod.  This is almost embarrassing to those of us who have been working diligently for over a decade to bring a personal HWD to a store near you.  In the piece that is the easiest to describe, the “artist” introduces you to “video archeology.”  You are given a map, a video, and an iPod customized to know where it is via GPS.  As you pass over key points on the map, the iPod displays video content created earlier, elsewhere, of events that took place at a location going back in time.  I found this to be a very exciting concept.  There were a number of other pieces that were more abstract, but nonetheless begin to show how transformational augmented and virtual reality is likely to be.

It was somewhat by coincidence that the talk I gave at the OSA Annual Meeting the week before, in San Jose, was on the technology of head-worn displays.  Coauthored with my wife, Prof. Jannick Rolland of the Institute of Optics at the University of Rochester (and formerly of CREOL), this paper reviewed some current technologies that are being pursued to result in a viable personal display.  The presentation used at both OSA and at a workshop at ISMAR can be downloaded here.*  In addition, there is an excellent overview article on personal displays in particular, authored by Dr. Ozan Cakmakci, the most recent addition to ORA’s Engineering Services Group, and Prof. Rolland, his advisor, in the April 2009 issue of Optics and Photonics News.

* I wasn’t on the ISMAR agenda, but since I had the material I sat in on a panel. It probably didn’t hurt that Prof. Rolland was the conference co-chair.

Ozan Cakmakci, who recently joined ORA’s Optical Engineering Group, sporting the latest in augmented reality eyewear. Image courtesy of ODALab, Prof. Jannick Rolland.

The look of the current generation of prototype see-through displays,
clockwise from the upper left: Lumus, Sony, ODALab, ORA.

Interesting Links

• International Symposium on Mixed and Augmented Reality 2009: http://www.ismar09.org/
• Optical Diagnostics & Applications Laboratory (ODALab): http://www.odalab-spectrum.org/
• OSA presentation, “The Coming Generation of Head-Worn Displays (HWDs): Will the Future Come to Us through New Eyes?”