Because I sometime do the same thing multiple times, I'm often pretty careless with filenames, particularly for plots. So, I've decided to organize all of my plots by date; I'm usually not stupid enough to make the same filename muliple times on the same day.

So, I create a directory in my blog:
$(HOME)/www/blog/upload/2006/07/24/
That is, my main blog directory, an upload path, and the current date (24 July 2006) in a well-formated fashion.

For one-off uploads, I use the built-in MoveableType file uploader (which ignores the "day" part of this path and just uses month). But I make a lot of plots; for that, I like to use a ROOT macro, which I explain below...

Rint.Logon:     /Users/tagg/.rootrcdir/rootlogon.C

(Note that the path above is to my machintosh home directory, which is where I do all my work.) Then I make this hidden directory ".rootrcdir" in my home directory, and put in a file called "rootlogon.C". This file does all sorts of things, like set up default pallette and colors and so on. Here is a link to my current version. Note that I include another file in the same directory, called publish.C

Now the "publish" command is available from any ROOT or Loon command line.

Using it is very simple. When I have a plot in my main Canvas that I want to show to the world, I simply give it a name and call the script:

root [5] publish("20cm5f2cm") 

URL is: 

<a href="http://minos.phy.tufts.edu/tagg/blog/upload/2006/07/24/20cm5f2cm.pdf">
<img alt="20cm5f2cm" src="http://minos.phy.tufts.edu/tagg/blog/upload/2006/07/24/20cm5f2cm.png" width= "505" height= "321"/></a> 

Info in <TCanvas::Print>: file 20cm5f2cm.png has been created 

Info in <TCanvas::Print>: pdf file 20cm5f2cm.pdf has been created 


C++ Macro file: 20cm5f2cm.C has been generated 

Warning: No xauth data; using fake authentication data for X11 forwarding. 

20cm5f2cm.png                                                                           100% 7905     7.7KB/s   00:00    

20cm5f2cm.pdf                                                                           100%   13KB  13.3KB/s   00:00    

20cm5f2cm.C                                                                              100% 4405     4.3KB/s   00:00    

Then I simply cut-and-paste the link it prints out into my blog entry. The script automatically sets up the new directory-of-the-day if it doesn't yet exist. It creates a .png, .pdf, and .C macro version of the plot. Then it uploads all three versions. The image gets pasted into the blog, and clicking on the image will give a high-quality PDF version that can be used in LaTeX or otherwise. This example gives:

(20cm5f2cm) 5 fibres spread across a 20 cm wide strip, as before, but make the strip 2cm thick (instead of 1 cm).
(20cm6f) 6 Fibres spread across a 20 cm wide stirp.

(20cm5f2cm )14.7% (compare to 20.5% for a 1cm-thick).
(20cm6f) 23.5%

As usual.. increasing the number of fibres helps (as always) and having a thicker counter helps. But you're better off having two 1cm counters than a single 2cm counter.

And, of course, these multifibre strips are probably a pain in the ass to manufacture.. but I'm no expert on that.

---Nathaniel

The cases here are
(8cm1f) An 8cm-wide strip with a single central fibre
(8cm2f) An 8cm-wide strip with two fibres, like two MINOS strips joined together

(20cm4f) A 20cm-wide strip with 4 fibres at +/-2 and +/- 6cm.
(20cm5f) A 20cm-wide strip with 5 fibres, like 5 MINOS strips stuck together.

The efficencies are:
(8cm1f) 11.2%
(8cm2f) 19.8% very similar to the standard MINOS strips
(20cm4f) 17.0%
(20cm5f) 20.5% again, very similar to the standard MINOS strips

Again, statistical uncertainty is roughly +/- 0.2%.

New: Two more configurations. Like (e), but:
(f) three fibres at +/-1.2cm and 0cm
(g) three fibres at +/- 1.0cm and 0cm.

The efficencies are:
(a) 19.1%
(b) 28.2%
(c) 31.9%
(d) 29.8%
(e) 40.3%
(f) 41.9% new
(g) 41.6% new

The statistical uncertainty is about 0.2% in each case.

Below the fold is an illustration of how far the photons have to go before they hit the fibre, and the pathlength vs transverse position.

The next two plots show how far photons have to go before they hit the fibre, as a function of where they started. The first shows the 2-d plot (with a logarithmic Z scale). Note the bins along the bottom, which dominate the picture. The second is a profile plot. (Errors indicate error on the mean, not the spread) which shows how this smallest bin pulls the distribution. These are for the standard MINOS strips (case a).

These can be compared the number of bounces before the photon hits the fibre, shown in the follwing plot. Clearly, in my optical model, it's the number of bounces, not the attenuation length, that dominates the game:

The results:
+/- 3.8 cm: 8.9% each fibre
+/- 3.0 cm: 9.7%
+/- 2.0 cm: 10.0%
+/- 3.0 cm: 9.8%

Thus, it's pretty much insenstive to fibre position, although my prediction is that you do slightly better if you put your fibres about 1-2 cm from the edge of the strip.

Fibres at +/- 3cm.

Fibres at +/- 2cm.

Fibres at +/- 1cm.

How well do these two cases compare?

Now, here's my double-wide strip. I've put the two fibres in the of the strip, at +/- 3.8 cm from the middle.

Performance:
The absolute performance of the single strip is 18.9%.
The absolute performance of the doublewide strip is 17.8% (2 x 8.9%)

Discussion:
This indicates that this design actually loses very little light. In fact, you no longer have the few-mm gap between the two strips, which may actually increase your absolute efficiency.

However, this is an attendant loss of tracking resolution. This could be mitigated by doing a light-sharing analysis of the two fibres: i.e.

y ~ (fibre1 - fibre2)/(fibre1+fibre2)

How well would this work? Depends on poisson statistics. In a MINOS strip, you get about 3.5 p.e. of light at 4m per strip end (for a normal-going MIP). That means that in this case, you would get only about 1.7 pe per fibre in the double-wide case. Not really enough to provide a good tracking resolution; you're better off with the single strips.

The total efficiency isn't very good.. only about 10%.

Interestingly, the tails of the distributions are well modelled by a Breit-Wigner form. At the very edge of the strip it doesn't work (since the sides of the strip reflect some back, raising the outer part of the tail). The central peak isn't well modelled: this is due primarily to photons that undergo only 0-1 bounces before hitting the fibre.

Note that this particular simulation doesn't include the fibre acceptance; this is just the odds you'll hit the fibre. The disribution looks very similar for the more rigerous case; statisitics are just a lot worse.

The bottom line is:
Photons tend to travel further in MINOS strips, which isn't too surprising.
Photons tend to bounce more often in the MINOS strips, which again isn't too surprising.

Update: Rebuilt to get pathlength right.

OPERA:

MINOS:

Bounces:

OPERA:

MINOS:

Below, I show what happens to a MINOS strip if you move the fibre to 1.5cm off the axis. Interestingly, the total response of the strip drops only a small amount, though obviously the position response changes substantially.

Blue is obviously the strip where I've moved the fibre to the side

Here's the 2D picture, which looks as you would expect:

This time I've got the full fibre model included, as well as the scintillator optics.

For both cases, I took the cladding radii to be 88% of the outer radius for the core, and 96% of the outer radius for the inner cladding boundary.

For the OPERA style scintillator. Total efficiency is 2.00%

Double_t zh = 0.53;
Double_t yh = 1.315;
Double_t xh = 400.0;
Double_t fibrewidth = 1.0*Munits::mm;
Double_t fibredepth = 2.0*Munits::mm;

For the MINOS style scintillator. Total efficiency is 1.9%. The blue band around the outside is a binning artifact.

Double_t zh = 0.5;
Double_t yh = 2.0;
Double_t xh = 400.0;
Double_t fibrewidth = 1.2*Munits::mm;
Double_t fibredepth = 1.3*Munits::mm;

To compare the two, I've plotted the survival probability of the photons as a function of the starting Y position (i.e. across the strip). Red is the narrower OPERA strip; black is the wide MINOS strip.

I think the spike in the MINOS curve is high because the strip is narrower, so you get more light reflected off the nearby surface. The OPERA wings are higher because light doesn't wander into the corners to get trapped.