CERN Pt. 2: "Just What do You Think You are Doing, Dave?"

After visiting the four major LHC experiments at CERN (and a fun Saturday teaching) all that was left was to head to the Computer Center to make some photos.

You think you have archiving issues? You have no idea. More after the jump.






46,000,000,000,000,000 Bytes Can't Be Wrong

Forty six petabytes.

That's the current combined HD and tape storage attached to the large roomful of computers at the CERN Computer Center. Suffice to say, shooting everything in 16-bit RAW would not be a problem for them.

That storage (12PB of spinning hard disc space and another 34PB of robotically accessible tape storage) is what is needed to store and process the insane amounts of data that the LHC will throw off when it is up and running.

Basically, they first have to look at everything to see what is statistically likely to be able to be thrown away. At each of several computer screening processes down the line, the data is further culled in sort of a needle-in-a-haystack-to-the-Nth-degree process.

If they do it right, when they are done they will be left with strong evidence of the Higgs boson -- the Holy Grail of particle physics.

My hosts, Peter and Andras, said that they could use some new photos of the Computer Center. And since I happen to know a liiiittle bit about tape-drive mass storage back from my days as a proud owner of a Commodore 1530 Datasette tape drive, you gotta think I'm pretty well-suited for the job.

The robotic IBM tape drive seen at top was completely encased in a cabinet. There was a window at each end, and some small windows up top. There was some cool ambient (literally -- it was blue) floating around in there so I decided to go with it and gel the strobe with a CTB.

The CTB (tungsten-to-daylight conversion) gel is my favorite of the blues. It is natural looking, unlike some of the freaky blues that populate the sample books. It is also pretty efficient, not sucking up too much of the flash's light output. And last, it is easy to contrast with the corresponding CTO gels I keep with me all of the time.

We were not able to get inside the machines, which were running all of the time. That robotic head swoops around constantly at very high speed -- and high torque. As in deadly high torque if it catches you inside when it wants to do some tape hunting.

So, we could shoot through one of the end windows. But that would leave us with few options for lighting. We stuck an SB-800 with a CTB gel at the other end and originally aimed it away, at a nearby wall for a nice, clean light source. But Aaron suggested we try turning it around and firing it right into the innards of the machine. We tried it, and it looked better.



So, how do you expose for the inside of the machine? What power flash setting do you use?

It always comes down to solving your most important problem first. And my most important problem was to get some depth of field going, so more of those tapes would be in focus. The backlight would throw cool speculars off of them, so you want as many of them to read as possible. More power equals more depth of field.

Thus, first step is get some decent power into the flash. I set the SB-800 on 1/2 power, manual pop, and chimped the various apertures until it looked best. This is not rocket science here, right?

Next step: Balance the ambient. I have two sources to consider -- the inside lighting of the machine and the LED status light of the robotic head.

My flash power is set. My aperture is set. All that is left is to walk the shutter speed down -- 1/30, 1/15, 1/8, 1/4 -- until the LED and blue ambient start to look good.

A flash meter would be of little help to me in this shot. As soon as I get my arm in there with a Minolta Meter IVF, that thing would come to life, swoop in and lop it off, prolly.

But that's okay -- I have not used a flashmeter in years and still have two arms to prove it. You don't need a flashmeter. As long as you identify your primary problem to solve, that solution leads you to all of your other settings, one at a time, as described above.

Frankly, our biggest problem turned out to be the unpredictability of the positioning of the robotic head. It was about as uncooperative as your average three-year-old hyped up on a bowl of Frosted Flakes.

I wanted to light the head from the top with an amber-gelled second flash coming in from a top window. But the thing just would not cooperate, and we had to get upstairs to do the hard drives pretty soon. So we nixed the top light and went with a mix of backlit strobe and ambient interior lights.


Another Floor, Another Tunnel

Finished with the tape jukebox, we went upstairs to shoot the main computer room. The CERN guys are pretty good at assembling some serious processing power straight off of the rack, and this room is basically a ton of Linux boxes and hard drives strung together for massive processing and storage capabilities.

It was pretty awesome to spend time there -- rows and rows of racks and racks of computers and hard drives. With almost no place to hide a light.



Okay, so take a look at the top view of the shooting alley, which is essentially two enclosed rooms. There are glass doors at each end (which we opened) and a glass roof. And we do not have enough light to nuke the whole room, so we will have to use the ambient.

So, first problem is to tame the ambient. It is a (thankfully) consistent fluorescent environment. For our purposes, this ambient light will come in from the top.

Set the camera to FL balance. Move the WB adjustments around until it looks best. Then underexpose the room by about two stops. This is because we will use the ambient as fill lighting to carry the whole room.

Since we are shifting the white balance, we will have to gel the strobes to compensate. No problem there -- just green 'em up.


So, where to put the strobes?

The subject would be Andras, who actually wrote the drive burn-in program that was causing all of the blue lights to blink on and off. It was like being on the bridge of the Enterprise. Or maybe in 2001: A Space Odyssey. I love technology, but I could see myself at night being a tad creeped out by all of these computers, HDs and tape drives going nonstop all around me.

(Not that it bothered Peter, our co-host and CERN computer whiz, who had promptly fallen asleep on one of the tables when we were shooting the tape drive monster.)

So the juncture between the rooms was a logical place to hide the key light, which was greened and fired through an umbrella. That would do two things: Light Andras from camera back right, and paint a nice specular on the left rack as it reflected back to me.

I put it on 1/8 power and adjusted my aperture until it looked best. Each time I moved my aperture, I would also have to compensate with my shutter speed to keep that minus two stop ambient exposure dialed in.

So that was easy enough. Now wold come the hard part -- how to fill Andras and pick up some of the spiffy hardware with light at the same time?

Position-wise, we had few choices. The most obvious being right next to the shooting position.

I stuck a second, greened SB-800 right at my left side. Just a foot or so away. This would fill Andras and throw some speculars off of the cables and Linux boxes. But it would also nuke everything in the foreground.



That's easy enough to fix with a grid spot on the flash. Everything along the edges that is close is also at the wide part of the flash beam. So knocking that out with a grid allows us to push the light down the tunnel and keep it from nuking the near sides at the same time.

How did we choose our power setting? Easy -- we already had the aperture chosen by the key light. The shutter was chosen to set the ambient exposure at a couple of stops under. Since those do not change, we simply adjust the power of the nearby "fill" strobe until we get the look we want.

No rocket science, no flash meter. Just good 'ol chimping.

Now all that was left was to make sure to catch some of the blinking hard drive lights. No big secret there, either. Shoot lots of frames and go for the law of averages. Worked like a charm.