<html>
<head>
</head>
<body bgcolor="#FFFFFF" text="#000000">
Kei,<br>
<br>
In the abstract it states:<br>
<blockquote>Due to the addition of light guides, improvements are
seen when these results<br>
are compared to extrapolations of the results of a previous
experiment that used the same lead glass blocks.<br>
</blockquote>
<br>
How do we justify this statement? Is it simply by comparing the
test beam measurements with the final resolution from Radphi? I
would counter that the light guides are not the only difference
there. Back when we were setting up Radphi, Scott Teige reported on
a test that he did in the context of the old E852 calorimeter where
he directly compared the resolution of blocks with and without the
"cookie" and (according to him) the difference was negligible. I
never saw data demonstrating this point. You might ask Scott, since
he is still right there in Bloomington.<br>
<br>
Apart from that claim, I am skeptical that the addition of the
cookie explains the big improvement shown in Fig. 8 over what we
measured in Radphi. I suspect that radiation damage in the blocks
in Radphi is more important than the small improvement (maybe 25%
more photoelectrons?) in photostatistics in determining the
resolution at low energy.<br>
<br>
The improvement is measured, so that is fine. Attributing it all to
the cookie needs to be proved, I think. Here are some other effects
that probably contribute at a similar level.<br>
<ol>
<li>poor online gain balance in Radphi -- done using the flasher
and plexi sheet, which was itself rad damaged and very
non-uniform in illumination. You can correct this to some
extent offline, but not completely.</li>
<li>weaknesses in the gain calibration -- we had no clusters of
"known" energy to use for this calibration, so we had to
bootstrap the calibration off the pi0 mass peak. This technique
works, but it produces biases because of edges and dead blocks,
whose effects propagate out into the intervening volume as you
iterate the calibration procedure.</li>
<li>radiation damage -- a lot of this was already there before we
started the main physics run. We had a few beam tuning
accidents during commissioning of the experiment. Hence the
lead wall, which was added before we started the main physics
run in 2000.</li>
</ol>
<p>I don't mean to say that the data we took with Radphi was junk.
We achieved 7.2%/sqrt(E) in spite of all of these effects. But we
should not neglect them all and say that the final resolution
would have been what you saw in test beam, had we not left out the
cookie. That seems to be what is implied in the abstract.<br>
</p>
<p>-Richard Jones <br>
</p>
</body>
</html>