Wednesday, March 28, 2007

Multitasking or tunnel vision?

Judging from your feedback, my posts on concurrent programming, “Are hyperthreads good for you?” and “How many cores are good for you?” have been of interest to you, so here is a post on human multitasking.

We have all learned to appreciate multitasking, especially through threads or lightweight processes, because when a process idles waiting for the user to type the next key, RAM serving a fetch due to a cache miss, a program segment to be paged in, etc. the CPU can be used to process other tasks.

Even unrelated tasks benefit from multitasking. For example, instead of using an I/O driver to read and write from disk, a program can use the operating system’s virtual memory manager, which can collect a number of I/O requests from all threads and optimize them before doing a series of absolute disk accesses.

As you are a citizen of the blogosphere, I assume you are also doing quite a bit of multitasking, like maybe writing code while listening to a podcast, reading email, chatting with a colleague, and keeping an eye on a stock ticker in the corner. Are you getting the same benefits from multitasking as your computer does?

Not so, according to the article “Slow Down, Brave Multitasker, and Don’t Read This in Traffic” (subscription required) Steve Lohr wrote on 25 March 2007 in the Business section of the New York Times. Reviewing recent research at the universities of Michigan, Vanderbilt, Oxford (sponsored by Hewlett-Packard), Illinois, New York, and Boston, as well as research at Microsoft and MIT, Mr. Lohr concludes that in us humans context switching is so slow, that multitasking is not advised, except in rare situations. This is so also for people whom in the post on “Research in transition” I characterized as having a field dependent cognitive style. Such people are very good at parallel thinking, and masters for example in deconstruction (yes, Jacques Derrida is the philosopher of the blogosphere) or in esoteric Buddhism (yes, Kukai got there 1,200 years before Derrida). However, their parallel thought streams are all related or dependent.

When I consider the number of emails I receive each day, with the 15 minute context switching overhead reported by Mr. Lohr from Microsoft’s research, I would not get anything done at all. I am on email since 1980, and ever since retrieve my messages manually from the POP server only three times a day, when I get in at work, after lunch (OK, I multitask email and digestion, but do not use my brain for the latter), and before returning home.

Therefore, you cannot get instantaneous answers from me and your blog comments may take a while to show up online, but that is the why email is so much more convenient than the telephone—it is asynchronous. I do not use chatting, because in my view it is like walkie-talkies, not an efficient collaboration tool.

An interesting result quoted in the New York Times article is that younger people are not better at multitasking than older people. However, this is not due to the context switching overhead time, but to older people having a better tunnel vision, or capacity to focus on the task at hand: “The older people think more slowly, but they have a faster fluid intelligence, so they are better able to block out interruptions and choose what to focus on.”

To conclude, let me get back quickly to the comments to this blog. If you read it this month, you may have noticed a disparate variety of topics. The reason was for me to sound out were your interests lie. The only form of feedback I get are your comments, and this is the only information on which I can decide what to write on. Feel free to comment openly; if you use a screen name, I have no way to uncover who you are, nor can Corporate Communications, so I cannot even bribe somebody. The humans in the loop approve all comments not violating HP’s Standards of Business Conduct.

PS: as usual, since our software does not support links in comments, I am adding the links here

Wednesday, March 21, 2007

Virtualization and performance

Color science has always required considerable computing power to calculate complex appearance models for every pixel in an image. For a while, like everybody else, we were riding along with Moore’s law giving us faster and faster machines. However, the intersection of a number of paradigm shifts is making our life particularly hard at the moment.

In this post I will list a few of these disruptive changes. The first is that in our nomadic society everything has become distributed and for example CIELAB is no longer an adequate model for color proofing. Today a print job may have to be soft-proofed on a display remotely on a different continent. This entails the use of more sophisticated appearance models like CIECAM02.

Moreover, today it is no longer sufficient to consider aperture color, we must account for complex color or natural scenes. This makes the models even more computing intensive, as you can appreciate in Wu and Wardman’s paper with the CIECAM02-m2 proposal in the latest issue of Color Science and Application, Vol. 32, N. 2, pp.12- 29.

Concomitantly, in the same issue Nayatani and Sakai propose a new concept of color-appearance modeling they call Integrated CAM. This new model can predict tone and nuance in NCS, which all of you who painted your living room in yellow know, is important to extend colorimetry to color design.

Moreover, when calculating inks, we no longer can just compute the black component. Today we must separate into 12 or more inks. And to take into account the different direction of paper grain in Europe and the U.S., and the brighteners in papers and inks, the calculations have to be spectral.

Printers are not the only devices requiring complex calculations. Today we know that video-conferencing only works well when we have a very high quality video and audio channel, and when we reduce cost by processing multiple video channels through a single PC, this requires writing very tight code and tuning the performance to the last cycle.

Another disruptive change is that in the enterprise, computing is moving from the desktop to the server, and with the modern nomads wanting to compute on the go without schlepping a suitcase with the battery, this trend is happening also for the consumer. And now the system is multi-modal, it no longer processes just spreadsheets but also shows picture, movies, and audio-video conferences to the antipodes.

And in the new mega-sized utility data centers the trend is towards virtualization, to keep the energy requirements in check while disentangling functions so that a failure does not become a catastrophe. Therefore, we have to understand virtualization. And because CPUs now are multi-core and hyperthreaded, we have to brush up our concurrent programming skills and learn performance well enough not to be surprised by those 25% lost MIPS common to hyper-threading.

In our case this is even more complicated by the ulterior paradigm shift of opening up the GPUs to developers and moving imaging operations to the graphics card. Using a GPU effectively requires concurrent programming skills beyond those required for multi-threading on multi-cores.

The latest issue of the Computer Measurement Group’s MeasureIT online magazine has a good introductory article on these issues, which can get you started with pointers to the latest results appearing in the literature.

PS: as usual, since our software does not support links in comments, I am adding the links here

Monday, March 19, 2007

AIC Judd Award 2007 to Alan Robertson

The equivalent of the Nobel Prize in color science is the AIC Judd Award. The AIC (Association Internationale de la Couleur), has just informed its member societies that the majority of the Judd Award Committee voted for Dr. Alan R. Robertson as the recipient of the AIC Judd Award 2007.

Since 1973, when Betty Judd proposed to establish an AIC award in memory of her husband, Deane B. Judd, to recognize outstanding work in the field of color science and technology, the International Color Association has been carrying out the process of selection of the recipients for this award every two years. The selection is an arduous procedure that includes nominations presented by AIC members, and analysis of antecedents of the nominees by a Committee mostly composed by previous recipients of the award.

This year there were six nominations, thus it was a difficult “competition.” Seven persons served in the Judd Committee, which was coordinated very successfully by the AIC Vice President, Berit Bergström. The award will be presented to Robertson at the AIC Midterm Meeting in Hangzhou, China, next July 12-14, where he will deliver the special Judd lecture.

Dr. Alan R. Robertson obtained his B. S. and Ph. D. degrees in physics from the University of London, England. He worked at Canada’s National Research Council from 1965 until 2000, first as a research scientist in the Photometry and Radiometry Group specializing in color science and measurement, and later as Director of the Chemical and Mechanical Standards Section. He is now partly retired but continues to work on various color science projects as well as acting as an advisor to the Director General of the NRC’s Institute for National Measurement Standards. Last year he received the ISCC Godlove Award. He is an Associate Editor of Color Research & Application.

Tuesday, March 13, 2007

How many cores are good for you?

In my recent post on hyperthreads I wrote they are not really that useful because the art of concurrent programming has been mostly lost and therefore they do not give you added performance. Most computers today are available with multicore processors. Do you need them, and if so, how many cores do you need?

I use my computer mostly for programming, so in principle one core should be sufficient. However, every couple of days I need to reboot my computer because it cannot keep up with my typing speed. A quick look at the Task Manager reveals a process coectl32 that hogs the CPU. It does not let its priority to be turned down and I suspect it has to do with the corporate Common Operating Environment that keeps my computer patched, so I do not dare to kill it.

A second core would be nice, because it can take care of this CPU hog. Another resource intensive program over which I have no control is the antivirus program, which fights with the patch program and can render my machine unusable when it does its weekly disk scan (the scan can take a day or more when it fights for cycles with the patch program).

Three cores would be the sweet spot for the way I use my computer as a programmer. If you create multimedia or frequently burn CDs, you may want a fourth core for that task.

This is not necessarily bad news, because you still have bottlenecks on the bus, the memory, and the virtual memory. Since the programs no longer have to fight over processor cycles, you could get away with 100 MHz or slower cores. However, in the corporate environment we often have to use video for executive communications or for video conferencing, and sometimes the link has to be encrypted. Therefore, 1.4 GHz may be a more realistic requirement.

This is good news, because you can run four 1.4 GHz cores at 50 W total, much less as a conventional 3+ GHz processor. The PC will be quieter and require less air conditioning for cooling.

PS: as usual, since our software does not support links in comments, I am adding the links here

Friday, March 9, 2007

Research in transition take 3: funding

The first post on this topic was my view, the second post was that of company executives, and this third post is that of grantors. We will see how their role has become more difficult and when they should blow the whistle.

We have learned that the field dependent cognitive style has become an impediment, while quick decision making and short term thinking have become the key desirable cognitive style. We have also seen how today's R&D executives view their technical workers as shock-troops of innovation and intellectual mercenaries with tattoos on their tonsils. So, let's look at the real troops.

The first world war was marked by the trench war along the Rhine, with soldier-to-soldier close-body combat. In the second world war, we had flying fortresses, missiles, and a large research investment in the Manhattan project. The war in Viet-Nam saw a big research involvement of think tanks like SRI and the national laboratories. The first Iraq war was fought from afar with cruise missiles guided by sophisticated imaging algorithms that could steer them into bomb shelters with the precision and skill of a pocket thief.

The second Iraq war is very different. It is a door-to-door urban war with soldiers using the latest personal communication gadgets. DARPA now finances short term research that has immediate application in the battlefield, a far cry from J.C.R. Licklider's days of doling out large sums of money to support large research efforts.

While there is still a fair amount of funding available for research in medicine and homeland security, that for research in science is much tighter.

Yet, there is still a large amount of people pursuing a Ph.D. in science and dreaming of a career in research or academia. Maybe they do not mind tattoos, but they have not realized that if they want to make money (today's metric for success), they have to stay far behind the front line and pursue an MBA so they can go in sales and then climb up the corporate ladder. Or at least study law, which is still very well paid although not as generously as an executive.

Because of this, there is a large group of researchers vying for dwindling funding while seeing their MBA and law colleagues bringing home fat pay checks. This makes life hard for those giving out and supervising grants.

Gone are the days when Lick had a vision, identified researchers able to realize that vision, and could throw a big chunk of funding at them, confident they will carry through. Today this kind of research is done by a committee defining a framework, then researchers get together in a consortium and apply for a grant.

From my experience in different countries, I have learned how the harder it is to get funding, the more researchers with less scruples are tempted to play political cards or personal contacts to get funding. This makes it hard to review grant requests, because traditionally reviewers like to give advice to the requestors and point out weak spots. However, any speckle on a review can get the request denied.

So we reviewers had to adapt and give raving reviews to the requests. However, the moment of truth is when the project comes up for review. The natural tendency is to be very severe and to judge the results very narrowly, i.e., making sure that the project has followed precisely the course described in the request. After all, we reviewers think by granting the funds we have taken a personal risk and we are liable, and we think of the people who were denied grants.

As University of Cambridge's William Byrne pointed out in a recent discussion, this way we also suffocate research, because by nature it is unpredictable and if we constrain it too tightly it can never bloom. We need to leave breathing room and let the recipients sufficient wiggle room to try out radical new things. Reviewers need to take risks so the projects they fund can take risks.

This now brings a new dilemma. In every profession there are the good and the bad. What should we do when less scrupulous people become good at the game and then pervert the whole system?

In Switzerland, every child learns how to deal with such a situation in high school, but many then forget. In fact, every student has to read, study, enact, criticize, etc. Max Frisch's play Biedermann und die Brandstifter. They learn that family and friends are more important than conformism and social appearances. They learn, that when a wrestler starts carrying around gasoline barrels, you do not dither and blow the whistle right away. They learn that the Far East belief of doing good so good things will happen to you does not hold in the West were its is an eye for an eye and a tooth for a tooth

Here is the end of scene 3 with the morale:

Chor
Strahlt die Sonne,
Wimper, o göttlichen Auges,
Aufleuchtet noch einmal
Tag
Über den traulichen Dächern der Stadt.
Chorführer
Heute wie gestern.
Chor
Heil uns!
Chorführer
Nichts ist geschehen in der schlafenden Stadt.
Chor
Heil uns!
Chorführer
Immer noch nichts…
Chor
Heil uns!
Man hört Verkehrslärm, Hupen, Straßenbahn
Chorführer
Klug ist und Herr über manche Gefahr,
Wenn er bedenkt, was er sieht,
Der Mensch.
Aufmerkenden Geistes vernimmt er
Zeichen des Unheils
Zeitig genug wenn er will.
Chor
Was aber wenn er nicht will?
Chorführer
Der, um zu wissen was droht,
Zeitungen liest
Täglich zum Frühstück entrüstet
Über ein fernes Ereignis,
Täglich beliefert mit Deutung,
Die ihm das eigene Sinnen erspart,
Täglich erfahrend, was gestern geschah,
Schwerlich durchschaut er, was eben geschieht
Unter dem eigenen Dach: —
Chor
Unveröffentlichtes!
Chorführer
Offenkundiges.
Chor
Hanebüchenes!
Chorführer
Tatsächliches.
Chor
Ungern durchschaut er's, denn sonst —
Der Chorführer unterbricht mit einem Zeichen der Hand.
Chorführer
Hier kommt er.
Der Chor schwenkt die Front.
Chor
Nichts ist geschehen der schlafenden Stadt,
Heute wie gestern,
Um zu vergessen, was droht,
Stürzt sich der Bürger
Sauber rasiert
In sein Geschäft…
Auftritt Biedermann in Mantel und Hut, Mappe im Arm.
Biedermann
Taxi! … Taxi? … Taxi!
Der Chor steht ihm im Weg.
Was ist los?
Chor
Wehe!
Biedermann
Sie wünschen?
Chor
Wehe!
Biedermann
Taxi! … Taxi? … Taxi!
Was ist los?
Chor
Wehe!
Biedermann
Das sagten Sie schon.
Chor
Dreimal Wehe!
Biedermann
Wieso?
Chorführer
Allzuverdächtiges, scheint uns,
Feuergefährliches hat sich enthüllt
Unseren Blicken wie deinen.
Wie soll ich's deuten?
Fässer voll Brennstoff im Dach —
Biedermann schreit.
Biedermann
Was geht das Sie an!
Schweigen
Lassen Sie mich durch. — Ich muss zu meinem Rechtsanwalt. —  Was will man von mir? — Ich bin unschuldig…
Biedermann erscheint verängstigt.
Soll da ein Verhör sein?
Biedermann zeigt herrenschaftw Sicherheit.
Lassen Sie mich durch, ja.
Der Chor steht reglos.
Chor
Nimmer geziemt es dem Chor,
Richter zu sein über Bürger, die handeln.
Chorführer
Der nämlich zusieht von außen, der Chor,
Leichter begreift er, was droht.
Chor
Fragend nur, höflich
Noch in Gefahr, die uns schreckt,
Warnend nur, ach kalten Schweißes gefaßt
Naht sich bekanntlich der Chor,
Ohnmächtig-wachsam, mitbürgerlich,
Bis es zum Löschen zu spät ist,
Feuerwehrgleich.
Biedemann blickt auf seine Armbanduhr.
Biedermann
Ich bin eilig.
Chor
Wehe!
Biedermann
Ich weiß wirklich nicht, was Sie wünschen.
Chorführer
Daß du sie duldest, die Fässer voll Brennstoff,
Biedermann Gottlieb, was hast du gedacht?
Biedermann
Gedacht?
Er mustert den Chor.
Meine Herren, ich bin ein freier Bürger. Ich kann denken, was ich will. Was sollen diese Fragen? Ich habe das Recht, meine Herren, überhaupt nichts zu denken — ganz abgesehen davon meine Herren: Was unter meinem Dach geschieht — das muß ich schon sagen, schließlich und endlich bin ich der Hauseigentümer!…
Chor
Heilig sei Heiliges uns,
Eigentum,
Was auch entstehe daraus,
Nimmerzulöschendes einst,
Das uns dann alle versengt und verkohlt:
Heilig sei Heiliges uns!
Biedermann
Also. —
Schweigen
Warum lassen Sie mich nicht durch?
Schweigen
Man soll nicht immer das Schlimmste denken. Wo führt das hin! Ich will meine Ruhe und meinen Frieden haben, nichts weiter, und was die beiden Herren betrifft — ganz abgesehen davon, daß ich zurzeit andere Sorgen habe…
Auftritt Babette in Mantel und Hut.
Was willst Du den hier?
Babette
Stör ich?
Biedermann
Ich habe eine Besprechung mit dem Chor.
Babette nickt zum Chor, dann flüstert sie Biedermann ins Ohr.
Biedermann
Natürlich mit Schleife! Das spielt doch keine Rolle, was es kostet, Hauptsache, daß es ein Kranz ist.
Babette
Sie verzeihen, meine Herren.
Sie entfent sich.
Biedermann
… kurz und gut, meine Herren, ich habe es satt. Ihr mit euren Brandstiftern! Ich geh an keinem Stammtisch mehr, so satt hab ich's. Kann man eigentlich nichts anderes mehr reden heutzutag? Schließlich lebe ich nur einmal. Wenn wir jeden Menschen, ausgenommen uns selbst, für einen Brandstifter halten, wie soll es jemals besser werden? Ein bißchen Vertrauen, Herrgottnochmal, muß man schon haben, ein bißchen guten Willen. Finde ich. Nicht immer nur das Böse sehen. Herrgottnochmal! Nicht jeder Mensch ist ein Brandstifter. Finde ich! Ein bißchen Vertrauen, ein bißchen…
Pause
Ich kann nicht Angst haben die ganze Zeit!
Pause
Heute nacht, meinen Sie denn, ich habe ein einziges Auge geschlossen? Ich bin ja nicht blöd. Benzin ist Benzin! Ich habe mir die allerschwersten Gedanken gemacht — auf dem Tisch bin ich gestiegen, um zu horchen, und später sogar auf dem Schrank, um mein Ohr an die Zimmerdecke zu legen. Jawohl! Geschnarcht haben sie. Geschnarcht! Mindestens vier Mal bin ich auf den Schrank gestiegen. Ganz friedlich geschnarcht! … Und trotzdem: — Einmal stand ich schon draußen im Treppenhaus, ob Sie's glauben oder nicht, im Pyjama — vor Wut. Ich war drauf und dran, die beiden Halunken zu wecken und auf die Straße zu werfen — mitsamt ihren Fässern! — eigenhändig, rücksichtslos, mitten in der Nacht!
Chor
Eigenhändig?
Biedermann
Ja.
Chor
Rücksichtslos?
Biedermann
Ja.
Chor
Mitten in der Nacht?
Biedermann
Ich was drauf und dran, ja — wäre meine Frau nicht gekommen, die fürchtete, daß ich mich erkälte —  drauf und dran!
Er nimmt sich eine Zigarre aus Verlegenheit.
Chorführer
Wie soll ich's abermals deuten?
Schlaflos verging ihm die Nacht.
Daß sie die Güte des Bürgers mißbrauchen,
Schien es ihm denkbar?
Argwohn befiehl ihn. Wieso?
Biedermann zündet seine Zigarre an.
Chor
Schwer hat es, wahrlich der Bürger!
Der nämlich, hart im Geschäft,
Sonst aber Seele von Mensch,
Gerne bereit ist,
Gutes zu tun.
Chorführer
Wo es ihm paßt.
Chor
Hoffend, es komme das Gute
Aus Gutmütigkeiten,
Der nämlich irrt sich gefährlich.
Biedermann
Was wollen Sie damit sagen?
Chor
Uns nämlich dünkte, es stinkt nach Benzin.
Biedermann schnuppert.
Biedermann
Also, meine Herren, ich rieche nichts…
Chor
Weh uns!
Biedermann
Rein gar nichts.
Chor
Weh uns!
Chorführer
So schon gewohnt ist er bösen Geruch.
Chor
Weh uns!
Biedermann
Und kommen Sie nicht immer mit diesem Defaitismus, meine Herren, sagen Sie nicht immer: Weh uns!
Man hört ein Auto hupen.
Taxi! — Taxi!
Man hört, wie ein Auto stoppt.
Sie entschuldigen.
Biedermann geht in eile weg.
Chor
Bürger — wohin!?
Man hört, wie ein Auto losfärt.
Chorführer
Was hat er vor, der Unselige jetzt?
Ängstlich-verwegen, so scheint mir, und bleich
Lief er davon,
Ängstlich-entschlossen: wozu?
Man hört, wie ein Auto hupt.
Chor
So schon gewohnt ist er bösen Geruch!
Man hört das Hupen in die ferne.
Weh uns!
Chorführer
Weh euch.
Der Chor tritt zurück, ausgenommen der Chorführer, der seine Pfeife nimmt.
Chorführer
Der die Verwandlungen scheut
Mehr als das Unheil,
Was kann er tun
Wider das Unheil?
Er folgt dem Chor.

Thursday, March 8, 2007

The end of JPEG

This is what today's Microsoft announcement at PMA suggests. Textually, they stated "HD Photo offers compression with up to twice the efficiency of JPEG, with fewer damaging artifacts, resulting in higher-quality images that are onehalf the file size. In addition, HD Photo offers increased image fidelity, preserving the entire original image content and enabling higher-quality exposure and color adjustments in the image. This new format offers the ability to decode only the information needed for any resolution or region, or the option to manipulate the image as compressed data." Is this the end of JPEG?

If you missed the announcement at PMA, you can read the press release. Microsoft claims its new format allows lossless or high-quality lossy compression and in addition also floating-point high-dynamic-range image encoding. Furthermore, it allows the efficient decoding for multiple resolutions and subregions. If you combine this information with their claim that it delivers compression quality comparable with JPEG-2000, it is possible that HD Photo is similar to JPEG-2000, and based on wavelet encoding. At this point I do not know and will have to wait for more information from Microsoft.

Will HD Photo replace JPEG? An indication may be that JPEG-2000, which has a similar feature set, did not replace JPEG. This is because JPEG is easy to implement and extremely efficient, especially if the hardware has circuits for the butterfly operation. Camera manufacturers need all the cycles they have for image processing, and they may not be willing to give those cycles up for HD Photo. Introducing a new format late in the game is difficult as we learned with FlashPix and JPEG-2000. But at this point we should be positive and give it a spin with the downloadable Photoshop plug-in.

Tuesday, March 6, 2007

Leonard Euler's 300th birthday

…is coming up Sunday 15 April 2007. In the San Francisco Bay Area the birthday will be celebrated at the Exploratorium. See http://www.swissnex.org/our-actions/events/euler/view for details.

Research in transition take 2: intellectual mercenaries with tattoos on their tonsils

Kemal sent a pointer to a special report in The Economist's latest issue on the rise and fall of corporate R&D with a different reason than the end of cold war. It also has interviews with our supreme leaders. You can read the full article at http://www.economist.com/science/displaystory.cfm?story_id=8769863.

The authors report that now the big corporate laboratories are either gone or a shadow of what they were, stating that the approach to R&D is changing because long-term research was a luxury only a monopoly could afford. They quote Eric Schmidt, a Bell Labs and PARC researcher now at the helm of Google, saying the “smart people on the hill” method no longer works. Instead, researchers have become intellectual mercenaries for product teams: they are there to solve immediate needs. And Intel's Sean Maloney with research is better “the closer the development is to the brutal market reality. Our people have that tattooed on their tonsils.”

About HP Labs they write:

The distinction between development and research is intentionally blurred at HP, epitome of the researchdriven organisation. At its base in Palo Alto, the offices of its founders, “Bill” and “Dave”, are preserved in the laboratories where they pioneered products and principles like open-plan cubicles for employees and “management by walking around”. Among the company's senior scientists is Bernardo Huberman, a physicist and former PARC researcher, whose work bridges computer science, economics and sociology

Instead of looking at fundamental questions about the universe, Mr. Huberman's prolific papers identify how “prediction markets” can be used to identify successful projects inside companies; how to price a unit of gridcomputing (broadly the harnessing of the collective processing power of many computers) and how to viebest for internet users' attention. The old model of research, of “putting people in a bubble”, is over, he says. The most interesting research is now done “where technology touches people”

Having researchers work more closely with customers pays off in other ways. For example, HP's work for DreamWorks Animation SKG, a film studio, required a highly sophisticated video-conferencing system so executives could regularly talk face-to-face without having to leap on an aeroplane. What the HP laboratories came up with was so successful that the company commercialised the system as a product, called Halo. This is now used by other companies, including PepsiCo and AMD, a chipmaker. Such schemes represent the primacy of “D” over “R”. “Is that a bad thing?” retorts Shane Robison, who oversees HP's technology strategy, with every pore visible on his three-foot-long face over the Halo teleconferencing system. Decisions about investment research are made by betting where the industry is heading, he says. HP still does some basic research, adds Dick Lampman, who heads HP Laboratories, but when people “celebrate the other model, they lose sight of what it takes to take a good idea and make it into an exciting product.”

Continuing with their image of researchers as tattooed mercenaries, they conclude "the new model of R&D turns researchers into the shocktroops of innovation." Well colleagues, are you ready to rattle your sabers?

Monday, March 5, 2007

HPL-97-164. Professional Portrait Studio for Amateur Digital Photography

HPL-97-164. Professional Portrait Studio for Amateur Digital Photography

The information in technical papers is very different from information in product brochures, magazines, or how-to Web sites. When you search information on a product, you can count on the vendor giving you current and relevant information. When a product is updated, the information about it is also updated. When a product is obsoleted, its information is made a little harder to find, and when the product is no longer supported, most information about it disappears.

This is not so for technical reports and papers. Research is built on previous research and that previous research must remain available forever so people can always go back to previous work and reproduce results. When you do a search, your engine will give you all links it finds, but does not evaluate them for you. If you are not a researcher be aware on how researchers use links. They typically use a search engine like Medline, Web of Science or Google Scholar to find all papers citing that first paper, and then consider only the last result, not that in the first paper they find.

Now to Professional Portrait Studio for Amateur Digital Photography. In 1996 our project did the technical demos at the HP Analyst event in Montreal. Among others there was portrait studio for executives manned by a famous local photographer, using HP's new first generation digital camera with 6 second shutter lag. People with big egos are very hard to portrait, and there is no tolerance for failure. The the portrait has to be flawless.

Compared with the cameras you buy these days, the first generation cameras were really bad. The 6 second shutter lag meant that the photographer had to think 6 seconds ahead, and then everything had to be perfect.

At that time, cameras really had only about 6 bits per pixel, the rest was just sensor noise. This means that the illumination was critical. There had to be plenty of light and the difference between highlight and dark shadow had to fit in the 6 bit dynamic range.

The only light technology that would allow a consumer camera of 1996 to get professional quality was to use a high frequency D65 light source as it is used in TV studios for example for the news. With today's cameras you do not do it that way. You no longer need a high frequency light source, you no longer need D65, and you have 12 or more bits per pixel to play with.

So please, do not use the instructions in this technical report to build a portrait studio! There used to be about 150 downloads per month, but all the sudden in February it jumped up to about 500. That is scary!

When a year ago I had to shoot some important portraits in difficult light situation for my neighbor, I went to the nearest camera store to find current literature. I bought a DVD made by world-renowned photographer Joe McNally. I am sure you can find it in a photography store near you. The title is The Speed of Light and I highly recommend it. If you came here through a link, please use the comment field below to let me know about it, maybe I can put the genie back in the boc or at least muzzle it.

Mini review. Camera Lenses: From Box Camera to Digital

When buying a digital camera or camera phone, most people consider the price per pixel, i.e., the price and the number of pixels, to be the main quality factor. If you are in the market for a new camera or phone, this book can help you find the best camera for your application and your budget.

Wait a minute, why am I telling you to read a lens book when what you want to buy is a camera? Well, the lens is the device that creates the image. The sensor just scans it and the image processing chip corrects the lens and sensor's shortcomings. There are three elements that determine the image quality, and in order of impact they are the lens, the image processing software, and the sensor.

Gregory Hallock Smith is an optical engineer and lens designer. He designed all of the camera lenses for JPL's Mars Exploration Rovers that have been taking pictures on Mars since January 2004. He knows best how to design an imaging system given a set of constraints, and this is why I recommend this book for the informed buyer of a digital camera.

When I was a young lad, I used to shoot black and white and make my own enlargements. In my first phase I was shooting mostly with a 24 mm lens on HP- 4 film I used to push it to the extreme by heavily underexposing it and then putting it through a soft physical development in Emofin. This was giving me a nice big grain while I still had a full tonal range. Then a friend showed me what he can get out of my enlarger from his Pan-F negatives. I was amazed and switched to Technical Pan, which has 320 line-pairs per mm, and was stupefied about the quality of my enlarger and my lab skills. However, in this second phase I had to completely change the way I see, because my 24 mm lens had become a clunker. I was now limited to photographing with a 55 mm micro lens and a 500 mm catadioptric lens, which are diffraction-limited.

Because of the blur disk from the point spread function (PSF) has a fixed size for a given lens and aperture, when the sensor resolution is increased, the pixels become smaller, but you just get more pixels inside the same disk, so they are not capturing more image information.

In section 16.4 Smith gives you a numerical example. His 4 megapixel pointand- shoot camera has a 1/1.8" sensor, hence a pixel pitch of 3.16 μm. The lens' PSF and its ƒ/8 Airy disk (the diffraction limit) are 10.7 μm in size, i.e., the blur disk is already 3.4 pixels across, i.e., he has already more pixels than he can use.

You really want less pixels, because then for the fixed sensor size you get larger pixels, which are more sensitive and give you more bits per pixel. If your sensor can for example give you 12 bits per pixel, then if the camera has good imaging software it will do some retinex magic to map those 12 bits into the 8 bits of a Jpeg file and give you an image with much better tone reproduction.

This book will help you understand the optical limits of a lens and let you find out the values typical of your lenses from a compilation of the most important lens designs. With this data you can then find the pixel size required to match your lenses and for a given sensor size, the number of pixels required.

Gregory Hallock Smith, Camera Lenses: From Box Camera to Digital, SPIE Press, 2006 http://bookstore.spie.org/index.cfm?fuseaction=DetailVolume&productid=660181

Looking in from outside

Usually I write about what I see when I look out of my window. Today I am stepping out on the terrace and looking in.

I assume quite a few of you readers will be at PMA in Las Vegas March 8–11, 2007. When you will exchange business cards with my colleagues, you will notice a new project name. Yes, we have been reorganized to better address our top leader's vision of being aligned with the divisions. Our new project name is Computational Color Research and there are more of us. In the summer we are also getting spacious brand new labs on the floor downstairs. In summary: these are good times for us.

Of course I cannot tell you what we are up to, as this is confidential. But at PMA our top leadership's vision for color will be announced and you will read about it all over the press. If you are not able to attend PMA, you can follow it on their blog. You will not see anyting on this blog, because all these new happenings will be covered in detail on our Newsroom Web site.

Saturday, March 3, 2007

Research in transition

When I look out of my window at HP Labs in the Stanford Industrial Park, I can see other research labs, like Google Labs, IBM Almaden, Intel Research, Microsoft Research Silicon Valley, SRI, Sun Labs, Yahoo! Research, and others. When I walk around the terrace, I see more, like FXPAL, PARC, Ricoh Innovations, and so on…

There are also the ghosts, like Canon Information Systems, DEC SRC, DEC WRL, the IBM Scientific Center just across the street, Olivetti, Philips Multimedia, Schlumberger Research, and many more. Not to mention that the existing labs are smaller than they used to be.

In the streets of Palo Alto I easily bump into old ex-researchers. They make do with all kinds of odd jobs, like marketing, scrambling software for missiles, due diligence for venture capitalists, draft patents, install home theatres, herd sheep, teach at local colleges, tinker in their garage on inventions that hopefully attract venture capital,…

Yes, wasted brain power. Why? Business people say the times have changed, that it is cheaper to do mergers and acquisitions, outsource research in emerging countries, or contract it out. Some say today anybody in the company can invent, or they use contests and blogs to get new ideas from customers.

I do not believe this is true. I think the problem is structural, and this explains why the researchers out of a job are all old, the same people that used to be revered with monikers like greybeards and greybacks. Let me first get out of the way the confusion between research and innovation. Here are the definitions from the Oxford Dictionary:

research
the systematic investigation into and study of materials, sources, etc., in order to establish facts and reach new conclusions
innovation
bring in new methods, ideas, etc.

Research is about creating new knowledge, doing new science. Innovation is about harvesting new research results elsewhere and adopting them. From a strategic point of view, research is hard and time consuming, therefore it generates high profits (it takes the competition a while to catch up) and differentiation. Innovation instead requires an open mind and a good nose, but the innovation has to be cheap because the competition can also get it.

Now that we know the difference between research and innovation, we can also appreciate that research is much more tricky, expensive, and risky. So why did companies set up labs in the first place? We can go back to history and read the section on our inception on page 134 of Mr. Kirby's interview with Mr. William Hewlett:

Kirby
In '66 HP established what it called the Hewlett-Packard Laboratories, a central research organization. Can you tell us what led to the formation of HP Labs and how it was organized and staffed?
Hewlett
Well, I can tell you what led to its organization. Dave and I were concerned that our engineering staff was too concerned with the day-to-day products and they were not looking ahead far enough. We actually told them we'd like them to take one day off a week and try to be creative for the future — or maybe it was one day every two weeks. It didn't work. The guys had projects they had to get out on time schedules. They were important. They were tomorrow's bread and butter and they couldn't afford to take time to do other research. So we decided we really needed a forward-looking group and we set up HP Labs. We had a certain budget — and Barney Oliver was the first head. I don't remember how it was organized beyond that.
Kirby
So they were doing research that was much farther out than the divisions were doing?
Hewlett
They did three types of operations. (1) Where a division did not have the technical skills to pursue its field. A good example was F&M. F&M had no electronic capability, so the Labs could look at chemical analytical products from a long-term standpoint and I think the mass spectrometer is a good example of that. It was a Lab development.
(2) The second area was to put us in new fields. The best example was the desktop calculator. That was developed in the Laboratory, as I remember, and we realized that we had trouble injecting things into a division because they already had their own priorities, so we decided where it was going to go — which was Loveland — and we got Loveland to come out here and work with the Lab people so they felt it was their project. Then it was a smooth transition. Another one that was not so successful was the surveying equipment.
(3) The third thing was just doing advanced research, anticipating what was going to be needed. So there were three categories: starting new businesses, undergirding divisions that were weak technically in electronics, and doing basic, forward-looking work.

This argumentation should still hold today, so why then is research disappearing and why are experienced people pushed out of the door first? As mentioned earlier, the problem is structural. The change — or as it is sometimes called corporate R&D transition — is not happening now but happened in 1988. The trigger was the fall of the Berlin Wall.

During the cold war, we researchers were bred to outbrain the Russians. As long as this was the mission, people like J.C.R. Licklider could dole out large sums of money to support large research efforts. Many projects in the research labs were funded because the Feds, as we called them affectionately, had put out a request for proposals and our mother companies wanted to bid. For example, when I implemented the first commercial color management system, my the first taker was the State Department and my first customer was Mr. Carlucci.

The critical thing to understand here is that selling to the Feds is very different than selling in the commercial market. The Feds first decide exactly what they want, then they take the best bid and pay cost plus, i.e., whatever the thing costs plus a profit margin. In the commercial market, instead, there is a price that the market can bear. The vendor has to assess this price and then allocate R&D funds that can meet this price.

The larger a company is, the larger the market for a new product has to be to make it worthwhile for a company to enter that market. Usually, a larger market means that the technology is already established and there are other players. Therefore the price is already tight and there is just margin left for innovation, not for R&D.

We now understand why the research labs shrank and why the process started in 1988. But why is it affecting mostly old researchers? The reason is that our brains get wired in part genetically and in part when we get educated, but they cannot get rewired. To be successful in the cold was era, a researcher's brain had to be wired differently than in today's open market economy. Today's researchers think differently. How so?

creative professions

In this illustration of creative professions, in the cold war era a successful researcher, i.e., one who could aspire to become a greybeard and later a greyback, was a speculative designer, because he had to be a master both in creating new science by going back to first principles, and be able to implement a new idea. Of course nobody is perfect, and one aspect of managing researchers is to pair up a speculative designer with a scientific penchant to a designer with an engineering penchant.

In the current era of free markets, instead, the successful researcher is the one with an entrepreneurial mind, who combines a different cognitive style, and this is why I stated that today a successful researcher has a differently wired brain. Since brains cannot be rewired, the problem is structural, i.e., old researchers cannot be recycled. I should add that in the figure, greed is intended in the positive sense of being moved by a very strong desire.

I am indebted to Kemal A. Delic for sparking this discussion.