Billion Dollar Baby (The CMS Silicon Tracker)

By John Conway | July 26, 2007 1:09 pm

The CMS Silicon Tracker

I got here to CERN a few days ago, and things are quieter than recently. It’s the start of vacation season and so the cafeteria is not so full and parking is plentiful. Lots of summer students are hanging around in the evening, and some of the meetings are sparsely attended.

In the CMS experiment, the folks who have labored long and hard to build the world’s biggest and most complex silicon-strip detector have earned a bit of vacation. As of about August 1, this billion-dollar baby will begin buttoning up for its journey underground into the heart of the experiment.

So I found myself alone in the room with this thing, in the Tracker Integration Facility, and it is hard not to hold it in awe. “It” comprises thousands of flat detectors, each a thin rectangle of silicon crystal with microscopic aluminum strips embedded in it. These strips sense the passage of charged particles like pions, muons, electrons, and so forth, passing the tiny charge they collect to custom-designed readout chips which send the data in digital form to an army of processors which gather all the information from a single 25 nanosecond “beam bunch crossing” into a nice tight wad for later processing.

If you unrolled and laid out all the silicon detectors in the CMS Tracker you could tile a tennis court. It’s mind bogglingly complex, the product of hundreds of people workng for a dozen years. But there it is, and you know what? It works. Its been put through its paces on the surface and it’s just about show time downstairs at LHC Point 5.

I call it a billion dollar baby but it’s hard to say what it really cost. That number takes into account at least some of the labor by engineers, technicians, physicists, and students all over the world, but probably not all of it.

With the testing of the tracker at an end and the transport not yet begun, I am here with my CMS pixel colleagues to take advantage of our last (and first, actually) chance at seeing if our detector will fit neatly inside the tracker.

We are working on the innermost detector in the CMS experiment, the pixels. Once you get so close to the collision that charged particles are millimeters apart, you can no longer use long strips to detect them but have to go to arrays of pixel sensors. Our detector may be a lot smaller than the tracker, but it has even more individual readout channels, around 45 million in total. Each pixel is 0.1 by 0.15 mm (100 by 150 microns) and is read out by a custom chip developed at PSI in Zurich, and bump bonded to the silicon sensors. (Enough jargon yet?) The data from pixels with charge above a preset threshold are sent out on a serial line, converted to an optical signal and digitized upon receipt in high speed electronics in an adjacent cavern underground, then sent up the data acquisition stream.

Anyway, when all this mumbo-jumbo is done we have a big set of three-dimensional space points along the trajectory of the same particles that have also passed through the tracker. But the pixel points allow us to see what happened very close to the primary proton-proton collision vertex. Combining the pixel hits along a track we can project back to the vertex with a resolution of about 10 microns. Given that the pixels are much larger than that this is quite a trick: we rely on the fact that particles split their signal among adjacent pixels, and we can use a sort of averaging trick to get much more precise than a single pixel width.

My task here, though is much more practical. We want to make sure that the pixel detectors we are building will fit inside the tracker, the two halves of the detector meshing neatly at the end of their grooves inside the tracker. Up until this point it’s been an engineering project, but now we have real hardware and we need to be sure that no parts will interfere mechanically with each other.

Stay tuned, and in a few days I will post some photos of that and tell you how it worked. Gulp!

  • Quasar9

    Cool, tiling the bathroom with a mosaic should be a doddle after that.
    Happy bump’ bonding the pixels and silicon sensors
    A sort of averaging trick to get much more precise than a pixel width, must give a pretty accurate average for a billion bucks, eh?

  • Myhatma Gander

    WHAT! They let you go into that room alone?! What if the Higgs had taken over your brain, and caused you to cut some wires, see ? Of course you will deny it. What lax security…or have *they* already taken over the CERN security people?!

  • Lab Lemming


    How do the pixels convert the passage of a single charged particle into a signal that your electronics can detect?

  • Yvette

    Ok, I have a silly question but I’ve been wondering for a little while: how do you plug this stuff in? I’m guessing a bit more than a three-prong chord is required. :) Thanks!

  • Ellipsis

    Yvette — see, e.g., for a variety of nice and expensive choices for how to plug an expensive detector in.

  • aquariid

    Once built my own cloud chamber, and it worked. I only needed to buy a chunk of dry ice which cost a couple of bucks. A billion dollars- sheesh!
    But seriously, what are these pixel sensors? Are they light detectors like a ccd?

  • Coin

    So does this mean once the thing’s running it will be generating neat bubble-chamber looking images, only in 3D?

  • astromcnaught

    Nice post.
    Given 14 TeV collisions, won’t all the materials surrounding that disintegrate within seconds?

  • perturbed

    Wow. And we thought *we* had a cabling problem…
    – Your friendly ATLAS neighbors

  • a cornellian

    I assume they are using depleted Si diodes. By reverse biasing the diodes you can fully deplete the bulk. When something, like a high energy particles (or x-rays or photons or electrons), fly through it interacts with the valence band electrons and promotes some to the conduction band. Due to the electric field in the depleted region the electron-hole pairs are swept apart. One or the other (I would guess the holes) is the charge collected by the pixel front end.

    Is it really fair to say that it is digitized in an adjacent cavern? From what I could find poking around the nets the PSI chip (PSI46) puts out the addresses of pixels that have been triggered by particle paths. Any analog information is squashed by the comparator in the pixel front end.

    Is there a connection between this detector group, the MediPix detector group and the Pilatus group?

  • JoAnne

    What if it doesn’t fit inside the tracker? It’s not like you can just shave off a few pixels…

  • eugene

    great cable management :)

  • Galileo

    What if it doesn’t fit inside the tracker? It’s not like you can just shave off a few pixels…

    Well, when I was in high school robotics, all we needed to meet the size requirement was a file and some hammers… 😉

  • Aaron F.

    Hey, wait! Galileo couldn’t have done robotics in high school! That comment must be from an imposter!!! I’m sure the last comment this blog recieved from Galileo was real, though. Because… yeah. *Shifty eyes.*

  • Coin

    Aaron, when he says “robotics”, he means “bronze sphere rolling down an inclined plane”.

    Very simple robotics, basically.

  • Ellipsis

    Joanne (since John apparently is too busy to answer his questions…):
    If there are problems like that, they tend to be (at least one hopes) structural rather than active elements (the head of a bolt, say, or a cable connector) that could be refitted without disturbing the silicon itself. It would have to be a pretty big error if the pixels themselves actually overlapped with the strips (if the latter were the case, I think the Japanese refer to what would result as “seppuku”). :/

    Most of the time these sorts of things do fit without any problems, though.

    But note that just a couple years ago, the beryllium inner cylinder of the Babar tracker was scraped by an errant bolt head on the insertion of the silicon detector (very luckily the damage was patchable)!

  • Alexey Petrov

    Is there a reason for some wires being green and some – orange? Or someone is just a fan of the University of Miami? :-)

  • Ellipsis

    Alexey — green and orange are the most widely-available colors of the shielding of commercial fiber optic cable. They are chosen to be very easily visible, because fiber optic is less happy about being stepped on, for example, than regular copper cable.

    I’m not sure which lines the CMS tracker has chosen to be green and which to be orange. (I’m actually on Atlas — of course that doesn’t make me any less anonymous than the name “Ellipsis” — same is of course true for CMS.) Possibly they are using orange for trigger info and green for the event info that has passed the level-2 trigger. Or it could be green for TIB and orange for TOB? A CMS tracker person would have to answer.

    Anyway, it is often interesting to note (you know this, but the average member of the public doesn’t) that in most modern tracking detectors, the _vast_ majority of the data coming out of the detector is actually for the trigger decisions and is subsequently thrown away, as compared with the much smaller amount of data that actually is saved (for the events that pass the trigger). (About 10 Terabytes / s, vs. 1 GB / s for both Atlas and CMS, at the nominal beam luminosity.)

  • Mike Saelim

    Hey! I’m in the University of Michigan REU at CERN this summer, and one of the people in our program is working at the TIF. Look around for a girl named Teesa from MIT!

  • five 0h

    please corrrect me if im wrong but 1 tev is about 1.6 x 10 ^ -7 joules, making 7 of them (LHC design beam energy) 1.12 x 10 ^ -12 joules. now assuming that a mouse weighs 200 grams and can sprint at a rate of 1 m a second (running from kitty) its kinetic energy would equal ~ 10 ^ -1 joules.

    what we have here is kinetic energy of the proton beam being ~11 orders of magnitude less than a sprinting mouse.

    my point : the tracker will be fine.

    the silicon however, that will need to be replaced eventually since the radiation will damage the lattice over time.

    if the beam were to spill into the silicon, this process would be shortened dramatically 😛

  • Rando M

    Not quite. That’s the energy per particle, and the beam has several of these particles. 600 megajoules worth.


Discover's Newsletter

Sign up to get the latest science news delivered weekly right to your inbox!

Cosmic Variance

Random samplings from a universe of ideas.

See More

Collapse bottom bar