Date: Thu, 19 Sep 2002 10:50:29 +0200 (CEST) From: Laurent Duflot Subject: Re: [CTF] Data files Hi, here an updated version of the plots from a study by Tuan Vu Anh and myself (sent to the JetID conveners, we should have sent it to caltf as well). We studie the jets in different periods of data taking and with different versions of reco to look at the effet of NL and zero suppression. The preliminary conclusions were that 1.5 sigma zero suppression gives far too many jets (bad and "good"), that n90 indicates that these are made from noise, with CHF indicating that the Coarse Hadronic layers contrinute a lot to the noise. If nothing is done, doing physics with 1.5 sigma ZS seems extremely challenging. Your observations, based on using the same release with different ZS levels, go in the same direction. I also noticed online that the ET distribution in layer 15 for zero bias event is very different from that of other layer, so the noise might not even be gaussian. With 1.5 sigma ZS, one would have to fine tune the jet reconstruction. For exemple, one could take into account the energy in CH only if there is significant energy in the preceding layers. In any case, having jets that are too sensitive to noise will make the data analysis very difficult. This is in particular a challenge for the JES. Event using p11.11, so 2.5 sigma ZS, there are significant differences with the pre-1.5 sigma online period (p10.15, that has no NL corrections): the CH fractions is still large for good jets and the jets experience significant split/merge which indicates that even good jets swallow jets made of noise. The offline 2.5 sigma ZS is supposed to do the same as the online ZS, except that : - the sigmas used offline may not correspond to the current running conditions, hence the noise may not be efficiently suppressed, in particular in CH - the NL corrections may enhance significantly the noise it might be interesting to have a reconstruction of this run without NL corrections and 2.5 sigma ZS. cheers, Laurent