[b1_ana] Fwd: Independent Technical Advisory Report

[O. A. Rondon]

For fixed total time on the floor, we gain sqrt(n) by having shorter,
more frequent cycles, at the cost of increased polarization build
up/removal overhead that takes away from data statistics.

If there is no anneal between, say pairs of 5+5 hour cycles, the price
is that the second pair will have worse Pzz. If there is an anneal after
each pair of cycles for each target, the price is the anneal's overhead.

I suggest we give Dustin's question a quantitative answer, by
calculating the total error, with the drift error decreasing as sqrt(n)
cycles but the statistical error increasing as t_beam = t_total -
t_overhead, with t_overhead = the sum on table 4 of the proposal,
suitably adjusted by n.

And I also suggest reducing P_zz for a second cycle with the same target
without an intervening anneal by, say 5% absolute(?).

Someone (Narbe, Ellie, .. ?) could make this optimization in a
spreadsheet, where one can change the number of cycles and the anneal
frequency.

Cheers,

Oscar

Dustin Keller wrote:
> Yes I would like to reach a consensus on the best that we can do
> via ~sqrt(N).  ei. what is the final word on number of cycles
> per point.  Anneal are likely not necessary at the 6 hour point
> but we do loose time by having to repolarize twice as much.
> 
> dustin
> 
> On Thu, 30 May 2013, O. A. Rondon wrote:
> 
>> Hi Karl,
>>
>> I posted a draft of my comments about the pf and target length. All are
>> welcome to read/comment, etc.
>>
>> The LaTex and pdf files are here (name tac-itac*)
>> http://twist.phys.virginia.edu/~or/b1/?C=M;O=D
>>
>> It bears repeating that the uncertainties in Azz due to time dependent
>> systematics are per cycle period. They will be further reduced in the
>> error of the mean Azz for each x point  by the square root of the ~ 10
>> cycles per point. Otherwise, nothing would be gained by repeated
>> measurements. At least that is what I understand from Bevington's text
>> on errors.
>>
>> So the error bands based on dA_zz(xi) = 3.7E-3 are actually too
>> pessimistic. A realistic error would be 1/2 to 1/3 of that for the
>> combined data per x point, considering that the 3.7 factor might be a
>> bit bigger.
>>
>> Cheers,
>>
>> Oscar
>>
>> Karl Slifer wrote:
>>> A more pessimistic read by the iTAC.
>>>
>>> -Karl
>>>
>>>
>>>
>>>
>>> ---------- Forwarded message ----------
>>> From: Susan Brown <sbrown at jlab.org>
>>> Date: Thu, May 30, 2013 at 12:23 PM
>>> Subject: Independent Technical Advisory Report
>>> To: Karl Slifer <karl.slifer at unh.edu>
>>>
>>>
>>> Dear Dr. Slifer,
>>>
>>> Attached please find a copy of the Independent Technical Advisory report
>>> for your PAC submission.
>>>
>>> Cheers,
>>>
>>> Susan
>>>
>>>
>>>
>>> ------------------------------------------------------------------------
>>>
>>> _______________________________________________
>>> b1_ana mailing list
>>> b1_ana at jlab.org
>>> https://mailman.jlab.org/mailman/listinfo/b1_ana
>>
>>
>>
>>
>>
>> _______________________________________________
>> b1_ana mailing list
>> b1_ana at jlab.org
>> https://mailman.jlab.org/mailman/listinfo/b1_ana
>>
> _______________________________________________
> b1_ana mailing list
> b1_ana at jlab.org
> https://mailman.jlab.org/mailman/listinfo/b1_ana
>