Re: GlueX-doc-926

[Elton Smith <elton@jlab.org>]

HI Andrei,

Your procedure is clearly not what is commonly refered to as "pedestal
subtraction". However, it might be a valid indication of what spectrum
would be achieved IF there was an independent event-by-event method of
separating pedestal from signal. An example of this is when the width of
the pedestal is dominated by 60 Hz noise and one can subtract a different
pedestal value depending on the phase to line current. In this case the
phase of the event relative to line current is the additional independent
piece of information needed.

How do you believe that you can remove the pedestals events from your
sample without cutting on the actual data?

Cheers, Elton.




Elton Smith
Jefferson Lab MS 12H5
12000 Jefferson Ave
Suite # 16
Newport News, VA 23606
elton@jlab.org
(757) 269-7625

On Thu, 13 Dec 2007, Andrei Semenov wrote:

>
> Elton:
>
> Actually, the way you should treat pedestals depends very much on the
> goals that you chasing in your analysis. If the counts in your ADC
> spectrum correspond to the energy depositions of the real particles, and
> IF YOUR PEDESTAL IS NARROW ENOUGH, the event-by-event subtraction of the
> mean pedestal value is perfectly correct. If the pedestal distribution is
> wide, the simple event-by-event subtraction of the mean pedestal value
> will lead (most probably) to the measured energy distortion, but (in the
> most cases) you have nothing else to do.
>
> Our goal is quite different: We suspect that our center-panel ADC spectrum
> contains both real-particles amplitudes AND some "pedestal" contamination
> (viz., "empty-SiPM" events) that leaks through not-very-strong analysis
> cuts (our choise in the purity/statistics trade). Our analysis task is to
> answer the question: "What will be the ADC spectrum shape after removal of
> the "pedestal contamination"? To have such an answer, you must subtract
> normalized "pedestal" spectrum from the "real particles+pedestal-mixure"
> spectrum (i.e., you should remove from the spectrum NOT just the constant
> shift but the SHAPE). And this is very important, otherwise you will be
> confused by the remaining half (above the mean pedestal value) of the
> pedestal contamination.
>
> Surely, in your result of "shape subtraction", the "zero" amplitude will
> be not located in "zero" histogram channel, so you might want to make this
> additional shift on the mean pedestal value. Or you may not do this final
> shift, and just keep in the mind that "zero" amplitude sits (lets say) in
> the channel 90, and make your conclusions accordingly (viz., you should
> study that is the separation of your spectrum from the channel 90). I do
> not think that this last shift is a "matter of of life and death" IN SUCH
> AN ANALYSIS.
>
> Thank you,
> Andrei
>
>
>
> On Thu, 13 Dec 2007, Elton Smith wrote:
>
> >
> > Hi George,
> >
> > I have a question about your report on Si PMs regarding Fig 11. It seems
> > that you have taken the middle spectrum and subtracted the fitted shape of
> > the pedestal from the top spectrum to produce the bottom spectrum. The
> > resulting spectrum shows a displacement from the origin. But this is not
> > the way one should subtract pedestals. These need to be subtracted event
> > by event (not in a distribution) and the result will be a spectrum
> > identical to the middle one but with the peak (currently approximately in
> > channel 90) centered on zero. This distribution does not show a clear
> > separation of signal from pedestal.
> >
> > Am I missing something?
> >
> > Thanks, Elton.
> >
> >
> >
> >
> > Elton Smith
> > Jefferson Lab MS 12H5
> > 12000 Jefferson Ave
> > Suite # 16
> > Newport News, VA 23606
> > elton@jlab.org
> > (757) 269-7625
> >
> >
>
>