[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: cost estimates



Hi Elke:

I will try to address the questions individually, below:


Elke-Caroline Aschenauer wrote:

>Dear George and Zisis,
>
>by trying to come up with the final numbers for the budget, we over run by
>2M$ in procurement and an other 1.5M$ worse of manpower some questions
>came up.
>
>And please don't count any of my questions as an attack against SiPMTs or
>the bcal or ....
>I only try to get numbers straight and answers to unpleasant questions
>which will come.
>  
>
Come now Elke, it's only money and it's other people's money :-) ........

>1. do we really need CFDs for the outer barrel, what is the advantage for
>   the TOF
>  
>

Not at all.  In fact, in my logic that I had expressed in a couple of 
reports earlier, was that CFD's only make sense for the first few inner 
layers where charged particle and critical photon reconstruction 
information is derived from.   It makes no sense, whatsoever, to 
instrument all channels with CFD's.  If we go with flash ADC's, all the 
other channels past the first 5 layers don't even need TDC's.  



>2. currently we instrument the bcal with 4224 channels SiPMTs 1920 in the
>   inner and 2304 in the outer,
>   we say we combine 2 channels in the other for the electronics so we
>   have
>   1920 + 1152 channels of CFD (100$/ch), F1TDC(100$/ch) and FADC (250$/ch)
>   of course combining things even more in the outer BCAL, will safe
>   money, but of course and I know we don't have a good answer yet what is
>   possible.
>  
>

The number of 1920 corresponds to five inner layers.  I feel 
uncomfortable if we base all our photon trajectory reconstruction on 
five layers only.  Maybe I am overly conservative but perhaps adding a 
sixth layer to bring us to a 12 cm individually instrumented depth, is 
something we may wish to consider.  This is a minor point and we will 
need serious MC simulations to see what we lose or gain, if any.

As one can see from the energy deposition profiles we have simulated for 
the BCAL, even for a 1 GeV photon entering perpendicular to the BCAL (a 
situation as limiting as it gets) the number of photons collected past 
15 cm depth is much smaller than what has been collected between 4-12 
cm.   If we combine every two read-out cells in the outer layers we will 
certainly not lose trajectory information because most photons of some 
energy to penetrate that deeply will also come at forward angles, 
predominantly.   In fact, past 16 cm depth or so, in other words past 
the eighth cell from the inner face, even four cells combined will most 
likely not degrade us at all.  When we meet for the SiPM workshop, I 
will show all these profiles and I will also send them to you by the end 
of the week, I hope, so we can discuss this with more info in front of 
all of us. 

The bottom line is that combinations of two and/or four can be done and 
should be pursued not only for cost savings but for simplicity of data 
handling as well.  For the layers past 15 cm, there will be so few 
photons per cell most of the time, we are wise to combine outputs of 
four cells.


>3. now the LV for the SiPMTs, we need bias voltage LV<50V and 10-20 nuA/ch
>   the questions to answer are do we need in individual bias for each
>   SiPMT array?
>   if no how many can we combine, are there measurements which show gain
>   for several individual SiPMTs?
>   If we could combine LV-channels what is the granularity? do we need
>   a resistor change to correct for different bias voltages needed, do the
>   gains change with time, so we have to adjust the adjustments.
>
>---> I know that are a lot of questions and I appreciate answers are very
>     difficult to give but the LV is worse 4224*150=634k$
>
I had answered this question already to Fernardo, I think.  The answer 
is we don't need to control the voltage to each array.  The process 
developed by SensL is such that the variations in breakdown voltage 
between different batches of Si wafers is less than 1 V and the real 
range of variation is closer to 0.1- 0.2 V.  So, the idea is that we 
select arrays with breakdown voltage within that small tolerance level 
and we control them with a common supply voltage.   I suspect that we 
will have a large number of arrays, each requiring the same voltage, but 
we clearly have to limit the number per P/S for other reasons.  Each P/S 
will be identical to all the rest with only a very small adjustment 
range of say 2V.  How may we will need?  I think this is a question of 
the electronics guys to tell us based on current and stability 
requirements.  The ILC group at DESY has developed such systems for the 
thousands of 1 mm2 SiPM's they use and with much greater voltage 
adjustment requirements, can't we get some info from them on their design?

If we can control 16 arrays per P/S, then the number drops down to 264 
and I believe we should eb able to drive 16 arrays for each, right?  By 
the way, the SensL arrays are more in the 30 V range not as high as 50 V.

>
>4. there is also the question where to but the cfds and where to do the
>   summing. Cables are also an issue.
>   If we would but the CFDs next to the SiPMTs we need double the amount
>   of cables 8000 instead of 4000, I think that is not an option, summing
>   the outer BCAL channels by 2 directly next to the SiPMTs will only safe
>   1000 cables.
>   ----> so I think we have to put the CFDs after the cables
>   ----> comments
>  
>
With the outer layer SiPM arrays combined in two's and/or four's, we 
don't even want to use CFD's but we certainly want to discriminate them 
since most of the cells will add zero signal but will contribute DR 
P.E.'s.  This discrimination can happen after their combination at some 
level, say 10 P.E.'s (about 260 keV in energy deposition).   For the 
inner layers is there such a problem with cables?  If yes, then we can 
indeed discriminate them when the analog signal cable reaches the 
electronics racks.

I hope I have answered most of them in an informative way,

Cheers,

George