CERN 2006: Low Voltage and Alignment

I visited CERN from 28 November to 16 December to gather information about projects that Texas A&M is currently involved in or may soon get involved in. The best way to put the pictures I took into context and convey what I've learned is to write a blog.

Click on any image to enlarge. The ones with thick borders are movies: they give more detailed information.

Monday, 27 Nov, 2006

Just to put it in the right place chronologically, I wrote an internal note about generalizing tau identification on the plane. I decided to add this to this page because at CMS week, I went to the tau identification meeting.

• A Generalization of CDF Tau-Finding for CMS

Wednesday, 29 Nov, 2006

Low Voltage Test Stand at ISR

Where the teststand lives

The teststand is in building 375, the old ISR ring. Go through the first door and take a left.

This building has been converted into a warehouse and testing facility. The following sequence of images is a rotation near our teststand. You can see the teststand on the right of the last image.

Here are three views of the teststand, starting with Sasha zooming away from it. The third should be good for zooming in.

Here we are walking around the teststand to the back.

The CRB board

The teststand consists of two peripheral crates (PCs, I won't use the acrynym), one above the other. The green part across the top is our CRB board, viewed face-on. (The board is oriented vertically: we're looking at its top face.)

Here's what all the parts are (in the images above):

• The two red (orange?) tubes on the far left and right are water tubes for cooling.
• The four thick wires screwed into the left of the CRB are Maraton power (output of Maraton, input into CRB): +6.5V from yellow and yellow/black, +4.5V in from red and red/black. The partly black wires are grounds; they are screwed into the two rightmost mounts.
• The board which is barely visible on the right (under the right water tube) is the PCMB card (with ELMB mounted on it). (This is easier to see in the movie below.)
• The horizontal line across the center are the regulators (output of CRB, input into the cards we're servicing). The black cards below the CRB board are analysis cards for RPC chambers.

Below is a movie where I describe all of this and point at things, then Sasha corrects me.

Here's the other side of a CRB board, the side which is not visible when in the crate. Sorry about the motion blur; I didn't notice it at the time.

Here are some broken CRB boards. The one depicted in the third picture has a broken plastic mount. These mounts are easy to break, so one should be careful when connecting and disconnecting them from the peripheral crate.

The PCMB and ELMB boards

Here are pictures of the PCMB (bottom) and ELMB (top) boards, next to a Swiss Franc (about the size of a U.S. quarter). In the second picture, I flipped over the PCMB board, but forgot to flip the ELMB.

Here are the parts of the PCMB:

• Left edge: 60 pins (three rows of 20) are input into the PCMB (output from CRB).
• If you zoom into the second PCMB picture a lot, you'll notice that 10 of the solders are square or cross-shaped. These are ground pins.
• The PCMB monitors 9 channels (numbered 1 through 9) of 3.3V to TMB, 9 channels of 3.3V to DMB, and 9 channels of 5.0V to TMB and DMB. There are 9*3 = 27 total.
• There are 6 more channels labeled "====" in the PCMB log files: I should ask Sasha again what these are. It's four of one kind of thing plus two of another.
• There's one more somethingorother that is not in the PCMB log file; again, I need to ask Sasha again.
• On the top of the PCMB board (first picture), you'll see 12 solder points running horizontally through the center, labeled 1, 2, 3, 4, 5, 6, 7, 8, 9, 0, CC, 1V5. These are control lines.
• Four more of these are for monitoring alarms: ACfail, SYSfail, etc.
• If you add all of these up, you get 60 pins.
• On the far right (first picture), you see two Ethernet ports. These power the PCMB/ELMB (three wires in the Ethernet cord) and supply the CANbus interface (one wire).

Now here's a movie where we plug the ELMB into the PCMB (which must be done when the Ethernet power is off, apparently). The hand on the left is mine, the hand on the right is Sasha's, saving the board from being plugged in while hot.

Here are two more views of the PCMB and ELMB boards (respectively), with natural lighting instead of a camera flash. The Franc on the left has been digitally inserted, but trust me, it's the right scale.

Optical cable from teststand to computer

In the images below, note the orange cable going from the front of the peripheral crate into the personal computer (PC, I will not use the acronym) on the left. This is an optical link which connects one crate to another in the real system, and one crate to a computer in the test system.

Here I zoomed in so that you can see the connectors. The third image in this sequence shows a whole roll of optical cable, which Sasha says "is probably not ours."

The cooling teststand

Here's another project: an apparatus to test the CRB's ability to cool itself and its load. The first picture shows a commerical water cooler, the second and third show the apparatus, and the fourth shows the apparatus with a PCMB/ELMB card attached.

• The CRB board is now horizontal, with its back face pointing up at the ceiling (see the orange, er, bronze handle?).
• Its two water nozzles are attached to each other through a red tube.
• A load, simulated by an array of big yellow resistors, is cooled with blue tubing. (Is this connected to the water cooler? I can't tell.)
• A PCMB/ELMB board fits into the slot on the right (same as in the crate, but more visible here).
• The load connects to the testing CRB board through a dummy CRB board. The load is soldered directly to the dummy CRB board which passes the signal through the same kind of plastic mounts that are found in the real crate. The test CRB board snaps into this with its own plastic connectors.

The last point is easier to see in the following video:

Other things stored nearby

Just taking pictures in the hallway, we saw a few more items of note. Here are at least four completed(?) crates, each with a CRB board (with self-connected water tubes).

Here are chambers labeled "CMS Barrel Muon Chamber," presumably the drift tubes.

And here, finally, are the Cathode Strip Tubes! (Verified by Sasha.) In boxes, fresh from China. There are a lot of them.

Integration Meeting

Here's what I remember of what was said at today's integration meeting.

• Tomorrow morning, YE3 will be lowered into the pit. Sasha and I will be there, to take pictures and possibly help. The lowering will take 13 hours and will start at 9 AM CERN time.
• We don't have a plan for how things will be manuvered with all three endcap stations downstairs (which will make space tight!). Wolfram will be asked to make a plan.
• A major reason for lowering now is to encourage Systems to install the infrastructure (patch panel, gas) we'll need with the endcaps down there.
• A temporary solution for getting three-component gas mixture downstairs is to pipe it down the shaft. This will be ready in time.
• We're considering connecting the endcaps together as soon as they're downstairs. This put more constraints on manuverability: they may be separated only four meters. We're ready for this: is everybody else? (We don't want to disconnect them.)
• The faces of the endcaps may not be accessible. We discussed ways of making them accessible and whether anyone would need this. Getting another cherry picker is one idea, building scaffolding is another. Scaffolding is thought to be a bad option because it would block a door. (The endcap is blocking the same door, but not so close that a human couldn't squirm through.)

Of the people that I know, Fred, Oleg, Peter, and Dick were there (as well as Sasha and me). I met Armando (sp?) and Yum-Wook (sp?). I also recognized two others from Purdue, but I don't know their names.

Thursday, 30 Nov, 2006

EMU Lowering

We got to Point 5 just after YE+3 disappeared down the hole. I got this picture by holding the camera over my head. The one on the right is from the underground webcam, much later in the evening.

Two movies to try to give a sense of the environs.

Alignment Modules and DCOPS

Alignment Modules

I found some alignment modules on the YE+2 disk (the one next to the one going downstairs).

Here are the lower ones, up close.

DCOPS sensors

I came close to the lowest DCOPS sensors (without touching them!!!).

In context on the YE+2 disk

I tried to get more of an overview of the sensors on the disk. I asked Dick about this later in the evening, and he says that they have the following configuration: 18 CSCs (9 visible due to overlaps) have 6 sensors, equally-spaced and identically oriented on the CSCs. (Identical up to parity.) It's a pretty clever arrangement, considering that the lasers can't all overlap in the center, because this is where the beampipe goes.

This is an STM-eye view, straight up.

Meeting with Frederic Ronga

Frederic e-mailed and we got together to talk about CommonAlignment over coffee. We looked at the changes I had made and pointed out the need for one more on the geometry side.

We talked about ways of associating Trajectories with Tracks. In the temporary producer I wrote to provide Trajectories, I do not re-fit the tracks so there is a one-to-one relationship between Trajectories and Tracks. In the more general case, tracks are re-fit, so there can be fewer Tracks than Trajectories (due to fit failures). In this case, I suggested a method I have been using to match Tracks from different collections: I check to see that the tracks from different collections share at least one hit. This relies on the assumption that tracks within a collection have disjoint sets of hits. He thinks we should ask the tracking folks about that assumption.

Alignment Meeting

After some difficulty finding the room, I ran into Oleg and followed him. (I was in the right building. Sasha wasn't there because he thought the meeting was cancelled (no e-mail).)

The meeting reminded me of the very dire need for information about track-based alignment. The data samples needed for track-based alignment would determine the needed scope for the hardware alignment project, since the hardware alignemnt is sensitive to short-term changes while the track-based alignment will be final for long time periods.

I brought up the point that the two methods have very different systematics, but this was rejected as being very important, since anything that track-based alignment is not sensitive to, track-fitting is not sensitive to.

Another point that wasn't raised is that hardware alignment would be able to prevent misalignments from causing trigger inefficiencies (L1+HLT). (Such misalignments would need to be huge.)

So I really need to get going on the track-based alignment! In particular, I'll be presenting a talk at CMS Week on Tuesday. In addition to software progress, I should do some calculation to find out how many tracks are needed to do a track-based alignment, since important decisions in hardware alignment seem to be waiting for that answer.

After the meeting, I had dinner with Oleg and Dick in the cafeteria.

I get internet access!

Finally! Thank goodness!

Friday, 1 December, 2006

No pictures today!

Determining the number of tracks needed for alignment

The proper way to answer this question is to finish the software development and do a full-blown test, but we should be able to estimate a rough number from geometrical arguments. A number with 20-50% uncertainty is better than no number at all.

In CMS NOTE 2006/016, there's a table of alignment resolution per track for MB1, determined by toy Monte Carlo. For rphi and z translations, their model predicts a contribution of ~8 mm of resolution from each track. For R translations (along the line drawn from the IP to the detector), each track contributes ~80 mm (much less sensitive, as predicted). Angular misalignments were expressed in local coordinates, in which x is the rphi direction, y is parallel with the beamline, and z points out of the detector. Alignment is most sensitive to phiz, because residuals (in x and in y) are proportional to sin phiz. A single track contributes ~10 mrad to the determination of phiz. The other two directions, phix and phiy, can be determined to ~70 and ~50 mrad, respectively, because y and x residuals are proportional to 1 - cos*phix* and 1 - cos*phiy*. (The difference between phix and phiy is due to two factors: MB1 is longer in y than in x, and about two and a half times more sensitive to x than y.)

Here's a summary (I put it back into table format):

Rotations in phix and phiy aren't particularly important because they lead to shifts in the chamber's local z direction, which track-fitting is insensitive to. Misalignments in phiz matter because it shifts x and y measurements at the edges of the chamber by 25 (x) or 20 (y) mm. This 20-25 mm is independent of the length of the chamber, because if you have a chamber which is twice as long, we'll be twice as sensitive to phiz (misalignments of only 5 mrad) but it will matter twice as much (we get 20-25 mm again). So I can extrapolate this result from MB1 to all the other DT chambers. CSC chambers have a higher intrinsic resolution: 3.4 times better according to E. Torassa's Review of the CMS Muon Detector System. So here's a table of the typical misalignments after a hypothetical track-based alignment procedure using one track:

At the alignment meeting, Duncan said, "we need 150-350 microns in barrel, 75-200 microns in endcap." Scaling 20 mm to 300 microns in the barrel would require 4000 tracks per chamber, and 6 mm in the endcap to 150 microns would require 1600 tracks per chamber (1/sqrt(N) scaling).

The point was raised at the alignment meeting that track distributions would be broadened by multiple scattering. Track residual distributions will be broadened not just by intrinsic resolution and misalignments, but also by uncertainty in the track propogation through material and non-uniform magnetic fields. The study I used above applies to MB1, the innermost muon chamber, so I must ask the question, by how much does the track position uncertainty inflate when we propogate to the outermost muon chamber?

I used some of the code I developed for software infrastructure to answer this question. I propagate tracker-fitted tracks into the muon chamber and print out the uncertainty in rphi at the innermost and outermost chamber, for each track. It is important that I don't use tracks fitted with muons, because each hit would shrink the error matrix. (This study is particularly relevant for fitting the muon chambers to the tracker.)

Here are the results, compared with the uncertainty in d₀ for context.

As you can see, the uncertainty is much larger when you propagate through ECAL and HCAL than it is when you only propagate through a beampipe, and the uncertainty is about 30% larger when you go all the way through the muon chamber. It is also good to see that uncertainty is larger at low p_T, as expected from multiple scattering, and the scale is about right (it seems natural to design a chamber with 50-170 micron intrinsic resolution when track propagation has 40-110 micron uncertainties).

So if we take this effect into account and inflate the uncertainty in resolution by 30%, we find that we now need 6800 tracks per DT chamber and 2700 tracks per CSC chamber.

Tuesday, 5 December, 2006

Using the following figure, I miscounted the number of DT and CSC chambers. Alexei caught my mistake and I corrected my talk minutes before I presented it (by looking at real photographs of the real detector: very reliable). It is now correct, though— I presented the right thing.

To continue the calculation: we'll need 60 times 6800 = 410,000 barrel tracks and 80 times 2700 = 220,000 endcap tracks for the desired precision. (There are 60 sets of DT chambers which don't overlap and 80 sets of CSCs which don't overlap.)

Teruki, Alfredo Gurrola, and Chi-Nhan very promptly determined the eta distributions of decaying Ws and Zs at 0.9 TeV and 14 TeV, and helped me determine that the following figure is correct for cross-sections:

(Except that below the break at 4 TeV is p-pbar and above the break is p-p. I used p-pbar numbers for the 0.9 TeV calculation, which is rougher anyway. The W and Z curves look smooth across the break, suggesting that W and Z production rate is the same for p-p and p-pbar. (They are demonstrably the same at 4 TeV.))

Here's a table of the cross-sections (sig) and cross-sections times branching ratio to muons times eta acceptance (sig ep):

As you can see, the eta distribution at 14 TeV is much flatter than at 0.9 TeV. (At 0.9 TeV, eta acceptances are 67% in barrel, 29% in endcap; at 14 TeV, eta acceptances are 30% in barrel, 30% in endcap.) This is in accordance with Teruki's intuition. At 14 TeV, the time required to accumulate muons for the desired precision is limited by the barrel. (That is, the endcap has enough muons for its precision goals while the barrel is still accumulating.)

Here's the talk I presented at the CSC meeting. Here's the punchline (as far as the above calculation is concerned):

• 2 days of 10³⁰/cm²/s at 0.9 TeV:

  4 mm resolution in barrel, 3 mm in endcap

• 10³³/cm²/s at 14 TeV:

  17 hours for 200-300 microns in barrel (410,000 muons) 10 hours for 100-150 microns in endcap (220,000 muons)

After the talk, Andrey Korytov pointed me to this slide which is the source of the "2 days at 0.9 TeV" figure. There will be 28 days of tests, of which 7% will be usable for physics.

Monday, 4 through 8 December, 2006

CMS Week! I attended many talks and took notes on some of them. I don't think it would make sense to transcribe my notes: it would take a long time and the original slides would probably be more meaningful to anyone other than me. I'll keep the notes in a notebook at my desk at A&M.

Here are links to the slides:

• The master CMS Week Agenda
• The CSC/ME11 meetings on Monday and Tuesday
• The Muon Alignment meetings on Tuesday
• I also went to the CMS Startup workshop, which included talks on CSC commissioning and alignment
• I went to Tracker tau tagging on Thursday, which included a talk on likelihood-based tagging (full top-down tau tagging, not the small extension that I wrote about). Alexei joined this meeting by VRVS, but had to leave before the likelihood talk.
• I went to the early part of the CPT Physics meeting on Friday.

Of course, I also went to the Christmas party.

Friday, 15 December, 2006

In the past week, I continued working on the LV monitor program (adding Maratons) and track-based alignment. (I met with Frederic once again, and he showed me work he has been doing to get Trajectories from the Event. I only need to follow his example.)

Today, on Friday, Sasha showed me the whole LV system, from beginning to end. That will be the basis of my final entry in this blog.

This is the famous Green Barrack (batiment 6593). It was a control room for the MTCC tests.

In the room behind the control room are racks of electronics. Near the floor of the first rack is the first step in delivering LV to peripheral crates, PCMBs, and CSCs.

What we're looking at here are rectifiers (deliver power to peripheral crates and CSCs) on the bottom row, and power supplies for PCMBs on the top row. They both serve the same function: to convert 220 V AC into DC current. The PCMB power is on a completely separate line from that for the peripheral crates and CSCs.

This is a rectifier, taken out of the rack. DC comes out the back.

Here they are in situ again; I'm holding the DC lines that come out of the recitifier. These lines are immediately bundled together into a thick, single black cable. Sasha and I have agreed to call this a "380 V" cable because it carries 380 V. This is technically still low-voltage (the cut-off is at 600 V).

The "PCMB cable" (goes between the PCMB power supply and the PCMBs) is also black and thick. It has a different connector, though. The second picture shows Sasha pushing it back into the PCMB power supply. You have to wiggle it.

This is the top of the rack, where there are fuses (left picture). As for the second picture, I don't remember what these connectors are, but they're also at the top of the rack.

Further back in the room are the computers that send CANbus signals. This is a computer with a CANbus card. It can feed into four PCMB cards, but only one is being used at present. At this end, they're ordinary DB25 and DB9 serial connectors.

Now we take a walk from the Green Barrack to the Assembly Hall (SX5). (Put on your helmet and security shoes!) This is at the bottom corner of YE+1 (it hasn't gone downstairs yet, but will soon). We're looking at a conspicuously empty Patch Panel.

Someday, it will be filled with connectors like these two:

The top is a distributor for PCMB power and signals. The thick black cable (PCMB cable) plugs into the connector on the left. It carries both power and CANbus signals. These are distributed to all the PCMBs through a chain of ethernet wires. (Remember how the PCMB boards are powered through the same ethernet cables that deliver CANbus instructions and read-out voltages?) The bank of six RJ-45 ports on the upper right are for endcaps 1, 2, and 3, out and in. Ethernet wires can support signals over distances of a kilometer if the bit rate is sufficiently low (he said 100 bytes/s, but that may be an illustrative number). That's it for the path from the power company to the PCMB boards! Now, on to the peripheral crates and CSCs!

The bottom box in the picture is a convenience connector for 380 V cables. A 380 V cable goes in one connector (bottom right) and out the other, with its load unchanged. Whenever an endcap disk is moved, the cables must be disconnected, so this just makes that process more convenient. (The alternative would be to cut and repair the cable every time!)

The 380 V cables then go into a splitter box, which is unfortunately not available for viewing, and then into a Maraton. The splitter and Maraton will sit right next to each other, in the tall racks mounted on the sides of the endcap disks.

Below is a Maraton (we're in the trailer now). On the upper right of the front view is the place where you would plug in the short cable from the splitter. On the back are the twelve outputs to peripheral crates (the bolts with wires connected to them). The existing wires are a feedback mechanism to keep the voltage constant. When we connect them to the peripheral crates, we will screw another connector to the same bolt. These are now truly low voltage, 4.5 V and 6.5 V if I remember correctly, so the current is huge (Amperes). Those DB9 serial ports visible in the closeup are for chaining Maratons together (to control the Maraton?).

Now we're back on the endcap, on the minus side this time. Here's a peripheral crate in situ.

In these two photos, I'm pointing at the grounding wire (left photo) and the two power inputs to the CRB board (right photo). Everything that I said on 29 Nov regarding the CRB board inputs applies here. The only difference is that this CRB board is installed on the detector, rather than being in the test-stand.

Here are a few pictures from the top of the detector, showing the mini-racks on the top. They seem to be the same kind of rack as in the test-stand: two peripheral crates each.

Here's Sasha with a larger rack, the kind you can find on the side of the endcaps. These look like they could hold three peripheral crates, but they will have two crates plus additional components (Maratons, splitter boxes, or equipment unrelated to our work).

That's it for the path that leads from the power company to the peripheral crates! One last destination: the CSC on-board electronics. These are powered by the Maraton through a junction box, home-made at the University of Wisconsin. These input through the side and split the current into 18 chambers (2 sectors). Each has a manual fuse that must be replaced by hand if blown. Of couse, if the fuse is broken at this point in the chain, we'd want to check the CSC on-board electronics anyway.

Here's a picture in which I happend to catch some of the CSC on-board electronics. The data cables are light blue and low voltage are dark blue. (High voltage are red.) The junction boxes have not been set up yet, so there are a lot of dark blue bundles of cables waiting.

As a final treat, here's a very rudimentary (but colorful!) introduction to the EMU LV system, using pictures of the actual components, rather than rectangles.

• The jpeg images are called map##.jpg in pictures/map_with_pictures/.
• The tex source is pictures/map_with_pictures/presentation.tex.
• The GIMP file that can be used to recreate this in another order is pictures/map_with_pictures/map.xcf.

That's all!