<html>
<head>
<meta content="text/html; charset=ISO-8859-1"
http-equiv="Content-Type">
</head>
<body bgcolor="#FFFFFF" text="#000000">
Hi Andrei,<br>
<br>
In order to keep of record of these discussions as well as make sure
no one is excluded from the conversation, I am copying this message
to the halld-cal listserver.<br>
<br>
I believe Zisis has answered the issue of bias distribution. It
seems you are proposing to do what we intended all along for the
production.<br>
<br>
As far as TDCs, there is no problem in using saturated pulses into
the discriminator. (Of course the pulse to the fADC should not
saturate). These pulses effectively give you smaller rise times. The
timing resolution from leading edge discriminators is proportional
to the (pulse height/threshold), so the higher gain allows setting
the threshold to a higher value for the same time resolution. That
gives more flexibility.<br>
<br>
Thanks, Elton.<br>
<pre class="moz-signature" cols="72">Elton Smith
Jefferson Lab MS 12H5
12000 Jefferson Ave
Suite #16
Newport News, VA 23606
(757) 269-7625
(757) 269-6331 fax </pre>
<br>
On 5/22/12 5:15 PM, Zisis Papandreou wrote:
<blockquote cite="mid:4FBC01E0.5090102@uregina.ca" type="cite">
Hi Andrei:<br>
<br>
Thanks for the discussion. Let me make some comments on the bias
distribution.<br>
<br>
The plot that you attached indeed shows how the distribution was
set up for the beam test. This is not how we will have it for the
production run. (Elton and I discussed it at length this
afternoon).<br>
<br>
1. The bias distribution was designed with a several things in
mind. Load on the line was a key one, ability to power SiPMs in
sums individually, space restrictions in getting cables into the
SiPM enclosure, and of course cost. With 4 biases, each line
powers 10 units, and these were distributed as in the picture
Fernando drew (attached to your slide 9 and also attached below).
Bias 1 load is driven mostly by cells 1 and 2, with the remaining
two rows adding a smaller amount. Bias 2 is symmetric in load to
Bias 1. Bias 3 and 4 loads should be less. One could argue that
in the outer rows (see picture) arrangement should be 1234 (from
top to bottom) instead of 4321 as is now. But probably this makes
little difference. Fernando, please correct me if I am wrong.<br>
<br>
2. I am not sure I understand the muon/cosmics requirement for
powering in layers. If we want the whole inner layer on, we power
both lines 1 and 2. We can then leave 3 and 4 off. Bias 1 powers
cells 1 and 2, and the second row in the double sum (9-12). Bias
2 powers cells 3 and 4, and the first row in the triple sum
(13-16). So there is no conflict in operating individual rows in
sums. For muons I would expect nominally all units to be on, with
the exception of runs were we want to sample the individual ones.
Layer biasing looks nicer but I don't see what we lose as it is.
Obviously, the two arrangements lead to different patterns (e.g
the current system cannot power the inner layer and the 1st row in
the double sum at the same time), but the current one respects
load on line Fernando can probably provide numbers on load. Am I
missing something that cannot be done?<br>
<br>
3. I agree with you that what we want is to have both SiPMs on any
given cell be on at the same time. The bias distribution has the
flexibility to do this as it is. What we want to do effectively
is a "label matching", so that U1-D1 power the opposite ends of
cells (instead of U1-D2 as it was for the beam test). This would
be less confusing to remember than the beam test system. For the
beam test, not all combinations of biases on both sides were taken
(ie there probably was no U1-D2 combination to illuminate both
ends of the same cells). The ones taken were a minimal set (if I
can call it so) to show that each SiPM individually in a sum was
operating ok. The production assignment can be accomplished
without redesigning the board. It is a cabling and channel
assignment issue with the distribution system. <br>
<br>
Cheers, Zisis...<br>
<br>
<blockquote
cite="mid:46607.142.3.164.95.1337713934.squirrel@webmail.jlab.org"
type="cite">
<div class="__pbConvHr">
<div>
<div><img src="cid:part1.05040701.00090600@jlab.org"
height="25px" width="25px"></div>
<div> <a moz-do-not-send="true"
href="mailto:semenov@jlab.org">semenov@jlab.org</a></div>
<div> <font color="#9FA2A5">22 May, 2012 1:12 PM</font></div>
</div>
</div>
<div __pbrmquotes="true" class="__pbConvBody">
<div>Elton and Fernando:<br>
<br>
I do not agree with Elton's plan, and I did try to explain
my ideas in the<br>
last slide of my yesterday's talk (but it looks like nobody
was willing to<br>
spend 5 more minutes to discuss very serious issues :)<br>
<br>
OK, one more time.<br>
<br>
The last paragraph in my talk was: "Gains in the timing
channels should be<br>
big enough to provide enough efficiency on the low edge of
the dynamic<br>
range with minimal stable thresholds of discriminators, and
should be<br>
small enough to not saturate the signal shape and not to
burn<br>
discriminators on the high edge of the dynamic range. The
test of the<br>
chosen final timing settings on cosmics is required."<br>
<br>
To be specific, I propose to set the threshold of 50 mV as a
conservative<br>
estimation of "minimal stable thresholds of discriminators";
with this<br>
threshold, the gain in timing channel should be the same as
in FADC<br>
channel to cover the low edge of our dynamic range. Because
we want not<br>
just triggering of discriminator (that is OK if the
threshold is just<br>
below of the peak amplitude) but good timing (that requires
the threshold<br>
to be at least in the middle of the pulse front), I
recommend the gain in<br>
the timing channel 2 times bigger than the one in the FADC
channel. Taking<br>
into account 2-Volt dynamic range of FADC, we will have 2x2V
= 4V pulses<br>
for the events in the high edge of the dynamic range, that
is not very<br>
small but (probably) manageable amplitude for the
discriminator.<br>
<br>
If we go with Elton's plan, for the events in the high end
of dynamic<br>
range, we will have pulses of 5x2V = 10V amplitude in the
input of<br>
discriminator (sometimes even more), and I don't think this
is a good idea<br>
unless we intend to burn our discriminators.<br>
<br>
I do understand the origin of the desire to put bigger
signals to the<br>
timing channels: indeed for PMT operation, the bigger
portion of charge<br>
from PMT anoge to the timing channel means more stable pulse
front (just<br>
because of statistics) => better timing. Unfortunately,
these<br>
considerations don't work for our case: we discussing the
amplification of<br>
already existing pulse, and bigger gain just means bigger
amplified<br>
unstable pulse (viz., the same shape, just bigger
amplitude). No win here,<br>
just disadvantages (if we remember about pulse distortion
because of<br>
saturation effects).<br>
<br>
Now about the bias distribution: Fernando, I'm terribly
sorry but your<br>
statement on the meeting about the bias distribution was
incorrect. If you<br>
have a look on pages 1 and 2 of your document "BCAL Readout
Tests -<br>
Milestone" (GlueX-doc-1951), you will see that the bias
distribution for<br>
upstream and downstream boards in Hall B test were exactly
the same =><br>
with U1,D1 bias (for example), the most inner biased cells
look on<br>
different cells of BCAL (because of the mirror flip).
Moreover (just to be<br>
sure that this is not some error in the paper but the real
thing we had in<br>
the tests), see the event display (in the attachment) for
the run 587<br>
(U1,D1). In the first (inner) layer, the biased upstream and
downstream<br>
readout cells look on different areas of BCAL. If we will
use U1,D2 biases<br>
(that we did not in the test), the inner layers will be OK,
but the rest<br>
of the module will not be seen from both sides again.<br>
<br>
Again: I do propose change the biasing distribution to the
layer-wise.<br>
If we will make 2 types of boards (viz., upstream boards
with a mirror<br>
bias distribution in respect to the downstream boards), it
will be better<br>
than our present situation, but there are 2 disadvantages
here:<br>
1. We will need to develop 2 types of boards;<br>
2. In the cosmics tests, it's very useful to have whole
layer powered (to<br>
select muon passage) but not just a half of the layer.<br>
<br>
Hope, you will agree with my arguments.<br>
<br>
Andrei<br>
<br>
<br>
<br>
</div>
</div>
<div class="__pbConvHr">
<div>
<div><img src="cid:part1.05040701.00090600@jlab.org"
height="25px" width="25px"></div>
<div> <a moz-do-not-send="true"
href="mailto:barbosa@jlab.org">Fernando J Barbosa</a></div>
<div> <font color="#9FA2A5">22 May, 2012 6:17 AM</font></div>
</div>
</div>
<div __pbrmquotes="true" class="__pbConvBody">Hi Elton,
<br>
<br>
Some updates/comments:
<br>
<br>
3. The fADC250s have signal amplitude dynamic ranges of 0.5V,
1V and 2V. We will have to set the gain of the readout to fit
within one of these ranges (currently 2V) while allowing for
Vover operation. I will need to know the new gain setting as a
percentage of the present configuration.
<br>
<br>
4. The changes have already been implemented (for the final
implementation) so that the new heat spreader (we already have
these on hand) will cover all the electronics; heat from the
heat spreader to the outside cover (ambient) has been
implemented via standoffs and modeled by Jim to provide
efficient heat removal; the cooling plate/system has enough
capacity to handle any residual heat. The new transition board
is practically complete. All these changes will make assembly
a snap, literally.
<br>
<br>
Best regards,
<br>
Fernando
<br>
<br>
<br>
<br>
</div>
<div class="__pbConvHr">
<div>
<div><img src="cid:part1.05040701.00090600@jlab.org"
height="25px" width="25px"></div>
<div> <a moz-do-not-send="true"
href="mailto:elton@jlab.org">Elton Smith</a></div>
<div> <font color="#9FA2A5">21 May, 2012 7:06 PM</font></div>
</div>
</div>
<div __pbrmquotes="true" class="__pbConvBody">Hi all,
<br>
<br>
We need to provide Fernando with input to the next (final)
iteration on the electronic boards, so these can be prototyped
and tested in preparation for production. The following
changes are planned for the next set of boards:
<br>
1. TDC gain (relative to fADC). Presently the boards have been
adjusted with a x5 gain (down from x10 for the original
boards)
<br>
2. Operating point for the SiPMs (i.e. nominal overvoltage).
<br>
3. Electronic amplification of the fADC (depends on the
operating voltage).
<br>
4. Mechanical changes to the assembly to efficiently extract
heat from the electronics, including changes to the transition
board.
<br>
<br>
I think we need to provide this information soon, otherwise
the next iteration cycle will start impacting the overall
schedule. Here is a straw proposal for each of the above,
which I offer for discussion and comments:
<br>
1. I suggest we stay with the x5 amplification. We can take a
look at the runs with different thresholds, but we still have
quite some flexibility in what thresholds to use, as we are
presently using a nominal threshold of 100 mV.
<br>
2. I suggest we use Vover=1.2 V. We want to make sure that we
can operate the SiPM over a range of bias voltages and this
would allow us to run between Vover=0,9 to 1.4 V spanning a
fairly broad range. It also gives us higher PDE with a nominal
increase in dark rate. If we run at 0.9 V, we loose PDE and I
do not think we will want to run much lower than this, so we
would be limited at the low end of adjustability. Yi and
Serguei can comment on the dark rate/PDE trade-off.
<br>
3. Pick the electronic amplification for the Vover=1.2 V
setting that reasonably matches the dynamic range at the 5 deg
setting.
<br>
4. These changes will simplify assembly and likely make heat
removal more efficient. They improve the overall design.
<br>
<br>
Comments and feedback are welcome.
<br>
<br>
Thanks, Elton.
<br>
<br>
</div>
</blockquote>
</blockquote>
</body>
</html>