I will like to follow up on the comments written on this. As James pointed
out we do in fact have quite a bit of knowledge
already; This is in part documented in the CDR as well as in several
articles on the prototypes RICH.
I have two comments on the ADC response.
i) As Borge points out the adc values in a gicen pixel in most cases only
reflect the spread of one p.e. (you can estimate the chance of having
2 pe.e in a given pixel from one ring - it is small), and therefore fits
should only go by made by hits.
ii) There is some cross talk among neigbouring pixels(in the same 2*2
matrix) , possible 10-15% or so.
One should estimate if this is in fact the origin of a 'small' signal and
decide whether to weight it or
not I.e. we might reject pixels in the fit if they are deemed to be from
crosstalk..
On fitting; In the cdr analysis essentially we filled histogram vs R and
found the maximum yield vs radius and determined the radius
With single tracks it worked exceedingly well- but for possible multiple
hits and a less well determined radius it should be reconsidered
also thinking about 'background' hits that contaminate the spectrum
It is crucial to be prepared to deal with multiple rings , and it may or may
not be possible to have centers known
For the good tracks the centers are approximately known, but background
tracks (e-conversions ) have no corresponding center (in theta, phi)
known, and may intersect the 'real' ring.
The cromatic aberations should also be considered, when the fits area made
(in terms of weight).
regards
Flemming
------------------------------------------------------
Flemming Videbaek
Physics Department
Brookhaven National Laboratory
> Hi all,
>
> Concerning ring finding algoritms, I think one should be careful assuming
> fixed radius in our case. We have pions, kaons and protons, many of them
> close to the Cherenkov threshold, i.e. not saturated rings.
>
> To use Julia's method would mean first to determine the momentum and then
> assume in sequence that the particle is a pion/kaon/proton, do the ring
> finding and fitting to a circle and then at the end produce probabilities
> for each of the hypotheses. This is indeed a viable method, but did does
> introduce systematic effects. Adding knowledge about the centre of the
> circle is just one more parameter in the game. This kind of method is
> reported as one option in the paper I quoted.
>
> The other type of method relies a little more on knowing the centre, but
> is probably less biased when it comes to pre-knowledge about the
> radius. It is basically Claus' concept: start with a centre, histogram
> radii from the centre and determine a best radius - possibly followed by a
> refitting of the hits in the found ring to a (slightly) shifted centre and
> radius.
>
> In either case, the worst problem is to suppress background hits not
> belonging to the ring. It pays to have a good understanding of the
> background sources: white noise, other reconstructable tracks, not
> reconstructable tracks, etc. The background is typically not flat, because
> the geometry in the RICH is limited, i.e. you are always close to the edge
> of the detector.
>
> Another thing relating to Claus' mail. Gain matching is of limited
> value. All you really need to know is if the signal is above threshold to
> be counted as a photon. Trying to count photons in a given channel from
> the measured pulse height is difficult to impossible due to the signal
> spread of a single photon. I am not sure, Claus, if your dR signals were
> weighted with ADC count. I think they shouldn't be.
>
> Cheers,
> Borge
>
> +------------------------------------------------------------------------+
> | Børge Svane Nielsen E-mail: borge@nbi.dk |
> | Technical Coordinator http://www.nbi.dk/TECHI |
> | Niels Bohr Institute Phone: (+45) 3532 5433 |
> | University of Copenhagen Fax: (+45) 3532 5465 |
> | Blegdamsvej 17, DK-2100 Copenhagen O, Denmark |
> +------------------------------------------------------------------------+
>
>
This archive was generated by hypermail 2b30 : Mon Sep 10 2001 - 22:23:28 EDT