Date: Mon, 01 May 2000 12:09:41 -0500 (CDT) From: Marek Zielinski Subject: calorimeter noise expectations for Run 2 To: dean@sbhep.physics.SUNYSB.EDU, Leslie Groer Cc: Marek Zielinski , Pierre Petroff , touze@lal.in2p3.fr, John Krane , Bob Hirosky , Laurent Duflot , Vishnu Zutshi , Serban Protopopescu Dear Dean, Leslie, The work on setting up simulation in D0sim of the calorimeter noise expected in Run 2 has reached the stage where we would greatly appreciate your further input and help in the most specific terms currently possible. The Orsay group has made significant progress on writing the extension to the D0sim pileup package to handle the noise; I enclose below the message from Pierre Petroff on this. Please take a look at his note to make sure that the described prescription correctly represents the relevant aspects of the real detector. We will need to feed in concrete numbers for the electronic and uranium noise components into this simulation. John Krane has just completed his study of the Run 1 noise (assumed to be dominated by the uranium noise, and thus providing a measure of this component for Run 2) along the lines that we discussed at the calorimeter software meeting during the last collaboration meeting. I am looking now at his histograms to make sure that things look reasonable to me. If you want to see his results, the summary plots of the Gaussian fits to calorimeter cell energies (summed over phi) for non-zero-suppressed zero-bias events (with cuts to eliminate remaining real interactions) can be found on d0chb in /home/jkrane/cal_noise/second_try/cal_noise.ps versus ieta for each ilayer. QUESTIONS: - - From previous discussion I assume that these values can be used for an estimate of the sigma for Run 2 uranium noise, after scaling by 1/sqrt(5.5). Is this correct? - - Given the simulation algorithm described by Pierre below, noise is generated for each contributing crossing, including in-time, pileup and -396ns BLS contributions. Shuld one use the full Run 1 sigma from John K or scale it down by another 1/sqrt(2) or so? (I'm wondering if the Run 1 number already has the BLS effect in it that broadens the distribution by combining in quadrature the in-time and subtraction components.) We still need to understand what size of the electronics noise to combine with the uranium part. It would be extremely helpful (especially under current time pressures for simulation deadlines) if you could suggest specific values or parametrizations that could be used for the next round. Needless to say, this is a shot at having a more reasonable simulation than D0 had so far, but the preliminary nature of such numbers is understood and expected. What we will need in the end is the electronics noise value in MeV per cell, similarly to the uranium noise. To avoid any miscommunication, it would be ideal to either get your estimates in this format, or if you tend to work with ADCs, please enclose what ADC to GeV conversions should be used. QUESTIONS: - - Is a simple 1/sin(teta) representation for the eta dependence of the noise that Pierre mentions below sufficient for the present stage? If so, does it apply to CC only, or to EC as well? - - What are the proper levels of electronics noise for each layer? The numbers below for EM1-3 seem quite large to me. - - Again, given the simulation algorithm, please specify if any provided noise values can be generated independently for in-time, pileup, and BLS, without a risk of double counting. We will appreciate your help -- and as I said, the more specific is the more effective! I certainly do not understand the subtleties of the actual setup enough to be confidend that I could correctly translate a general advice into code... If substantial extra time is needed, please let us know what timescales to expect. To set the scale, the production version of D0sim is scheduled for 5/15 release (and we would need some time for testing) -- I hope we can meet this date! Thanks, Marek - ------- Forwarded Message From: Pierre Petroff To: Marek Zielinski Subject: Re: calorimeter noise and the pileup package Hi Marek, I apologize for this delayed answer. The method to add the noise in the calo is the following: 0) we assume that the noise is not varying with phi 1) build up a map eta vs layer with mean value and sigma for electronic noise 2) same for Uranium noise 3) at each crossing choose randomly electronic noise and U noise and build up a caldatchunk "noise" 4) merge this new chunk with minbias+hard chunk The code is beeing written by Francois Touze ..but the main problem is to estimate in a right way both the electronic and U noises: The baseline is substracted by substracting the previous crossing (which includes noise) with our method pileup+noise must peak at zero The main point is the determination of the noise ! Could you discuss with Leslie ? I was a little bit confused concerning how to extract the noise from his (and Dean) measurements !! If I take the Dean's number: EM1,2,3 = 40/50 Mev ? EM4= 10 Mev ? Hadronic=70 Mev ? 1 ADC count=3 to 5 Mev for EM and 10 Mev for Hadronic What is the sigma value ?? is it for eta=0? must we scale in 1/sin(teta) ? We shall start with this values .. Concerning the U noise what is the situation did you get the numbers from John Crane ? It is difficult for us to get the good values beyond the sea ..I will propose you to take this job in charge and we take in charge the writting of the code .. kind regards PP ------- End of Forwarded Message Date: Mon, 01 May 2000 19:33:40 EDT From: "Dean Schamberger, SUNY at Stony Brook" <"cusb::dean"@SBHEP.PHYSICS.SUNYSB.EDU> To: marek@d0mino.fnal.gov Subject: RE: calorimeter noise expectations for Run 2 HI Marek, Here is a "first pass" reply. I do not have any detailed numbers. 1) - - From previous discussion I assume that these values can be used for an estimate of the sigma for Run 2 uranium noise, after scaling by 1/sqrt(5.5). Is this correct? Yes....and no. If you are working in ADC counts (ie charge seen by the preamp, then yes. If you are working in MeV, then it depends how different the sampling weight will be do to changes outside the calorimeter. Since that depends on if/whether you use the preshower "analog" signals in addition to the LAR signals. Since that might change with analysis, the ONLY correct way is to convert to ADC counts. I will again remind you that the "energy" returned in the "standard" run 1 Calorimeter was NOT the signal seen in the argon for that layer, but increased to (on average) correct to enery lost before entering the calorimeter. So to get the "correct" uranium noise to scale from run 1 to run 2, you must first conver the run 1 energy back into ADC counts, divide by sqrt(5.5), and then convert those ADC counts back into energy using run 2 sampling weights......which ONLY can come from the monti-carlo, since we do not have any testbeam data. 2) - - Given the simulation algorithm described by Pierre below, noise is generated for each contributing crossing, including in-time, pileup and -396ns BLS contributions. Shuld one use the full Run 1 sigma from John K or scale it down by another 1/sqrt(2) or so? (I'm wondering if the Run 1 number already has the BLS effect in it that broadens the distribution by combining in quadrature the in-time and subtraction components.) Yes, every number you have ever seen for the D0 calorimeter includes the "BLS effect" in its measurement. So both the electronics noise (as measured for the new electronics by me) and the uranium noise (extracted from the run 1 data and extrapolated to the run two system) is the TOTAL noise, including the effects of BLS. Pile-up on the other hand must be handled in a way similar to what Pierre says. 3) - Is a simple 1/sin(teta) representation for the eta dependence of the noise that Pierre mentions below sufficient for the present stage? If so, does it apply to CC only, or to EC as well? Yes, the uranium noise is proportional to the exposed readout area, so goes like 1/sin(theta) for CC. (since layers are at a fixed perpendicular distance from the beamline) That changes into 1/cos(theta) in the EC, because a given depth is a fixed Z (ie pads are largest farther from the beamline). To be more precise one should really use the appropriate readout pad sizes, but the analytical function is close enough. HOWever, electronics noise scales like the total capacitance on its input. Part of the capacitance is the detector pad size, and that dominates in XXXX (forget after the comma) The cable capacitance is of about equal value, and in the endcap there is significant trace capacitance, which is NOT independant of phi. So attempting to "scale" the electronics noise, and have it independant of phi, is WRONG. 4) - - What are the proper levels of electronics noise for each layer? The numbers below for EM1-3 seem quite large to me. I have no idea where the numbers Pierre listed came from. The Em4, and Had number are not obvoiusly wrong....but seam larger than I expected. The EM1-3 are simply wrong. Again, because there is not a unique conversion of ADC counts to GeV, I can only quote electronic noise in ADC counts. I will attempt to find enough time to solidify the numbers I showed at the colaboration meeting. 5) - - Again, given the simulation algorithm, please specify if any provided noise values can be generated independently for in-time, pileup, and BLS, without a risk of double counting. I am not sure I understand your question. But to repeat, fluctuations do to minbias events of "no interest" must be handled as Pierre outlined. but both electronics and uranium noise fully account for all effects of the electronics baseline subtraction technique, so only a single value is needed (one for uranium, another one for electronics) for each readout of the detector. I hope this helps......more to come. Dean ------- End of Forwarded Message From: "Dean Schamberger, SUNY at Stony Brook" <"cusb::dean"@SBHEP.PHYSICS.SUNYSB.EDU> To: marek@d0mino.fnal.gov Subject: Re: calorimeter noise expectations for Run 2 OK, The two questions you ask in your latest message were: To avoid any chance for confusion, I'd like to ask directly if noise contributions need to be piled up from previous crossings (ie., specifically, generated randomly for all crossings considered, and combined using same fractions based on the transfer function as for the min bias MC events). NO, Both the numbers for uranium noise (after scaling) and electronics noise (when I give you the info) will properly include all "history" effects, so one a single number for each source should be generated. Second question was: Since we are planning to use John Krane's measurements for the uranium noise (after scaling), the geometric effects should be already included. Agree? I agree. It is the electronics noise (where I do not have the entire detector instrumented) that one will need to "extrapolate" from the few channels I have looked at, to the entire detector. Also, my electronics noise MUST be in ADC counts. The run 1 conversion factors were actually different for each of the 50,000 channels, so I could not generate those number without a fully functional run 1 D0 analysis (VMS) system. There were however "typical" values, which were used in the online monitoring Examine program, since extracting the "real" numbers from the offline database was too slow. Unfortunately those number were (as far as I know) only documented in the fortran code used by the online calorimeter examine program.... Since the death of the online VMS cluster, I no longer know how to access the source (I do not even know which module actually had the numbers in it, or which of the many online libraries the module was released to). Dean ------- End of Forwarded Message Date: Wed, 3 May 2000 16:36:51 -0500 From: Bob Hirosky To: Marek Zielinski cc: Pierre Petroff , touze@lal.in2p3.fr, Laurent Duflot , Vishnu Zutshi Subject: Documentation on ADC to GeV conversion in Run1 Marek et al., On d0chb if one looks in: /d0library/calor_off/source/cahits.doc There is a nice summary of how ACD counts were converted into GEV. The conversion contains a number of components from various data bases. Ultimately, however there is a number multiplied by ADC counts for each channel in the Run1 Code. This happens in the routine: CAEPFL /d0library/calor_util/source/caepfl.for The SF and gev to adc info is used to make the CGEV bank. Someone could hack into the routine 'cgevfl.for' (or one of the routines it calls) and print out ADC to GEV conversion factors and sampling weights for each calorimeter cell to a text file. This will give a snap shot of the information we want for a particular set of RECO sampleing weights. That's probably the easiest thing to do. All the names associated w/ this code are now more familair to Lucent, etc than to D0, but it is possible to extract the information after studying the code a bit - although it is a little cryptic at first glance... There may be a slight twist here because there are two gains X1 and X8 to contend with. For the purposes of noise addition, however, I expect the X1 gains are the important ones. What is the exact information we want? ADC to GEV factor per cell? ( ie total conversion factor only? ) Samp Fracts Per layer ? etc? Where will it be applied in the Run2 code? Please distribute this to any other interested parties. cheers, -Bob ................................................. UIC: 312-996-5648, FNAL: 630-840-5404 www-d0.fnal.gov/~hirosky E-VoiceMail : Dial 1-888-Excite2, access box 847-882-5262, or V-I-R-T-U-A-L-B-O-B E-Fax: 603-761-0339 ................................................. Date: Wed, 03 May 2000 18:05:03 -0500 From: John Womersley Subject: Re: calorimeter again... To: Marek Zielinski Cc: Marek Zielinski Marek, the numbers that I encoded in the C++ calorimeter unpacking were (I think!) exactly the same numbers that Dean is referring to, namely the "nominal" ADC <-> GeV conversions that were only ever written down in the fortran code (which is, by the way, archived on unix, so it isn't completely lost). The nominal numbers used a test-beam-based estimate of how many ADC counts you would get per GeV of deposited EM energy in a given cell, based on the sampling fraction in that cell and the nominal preamp gain. They do not include: * gain corrections - you have to multiply by actual gain/nominal gain for each channel, which (as Dean says) have to come out of the database * or layer weights to optimize resolution (as far as I recall, though in typing this I am not 100% sure, since these were also derived from the testbeam). John Marek Zielinski wrote: > > John, > > We are in the process of discussing with Dean issues related to proper > simulation of calorimeter noise for Run 2 in D0sim. > This led to the problem of translating his ADC estimates to D0GSTAR GeV, > and his comments at the bottom of the message may not be the best way to > pursue. Can you point to a different source of these conversion factors? > Please add any warning or comments about possible convention differences > etc to check about (I am -- still! -- confused what any given Run 1 ADC > conversion includes: dE/dx layer weights, optimized weights, versions of > the normalization, online vs offline etc) > > Thanks, > Marek > > ------- Forwarded Message > > Date: Wed, 03 May 2000 13:02:34 EDT > From: "Dean Schamberger, SUNY at Stony Brook" > <"cusb::dean"@SBHEP.PHYSICS.SUNYSB.EDU> > To: marek@d0mino.fnal.gov > Subject: Re: calorimeter noise expectations for Run 2 > > >Since we are planning to use John Krane's measurements for the uranium noise > >(after scaling), the geometric effects should be already included. Agree? > > I agree. It is the electronics noise (where I do not have the entire detector > instrumented) that one will need to "extrapolate" from the few channels I > have looked at, to the entire detector. > > Also, my electronics noise MUST be in ADC counts. The run 1 conversion > factors were actually different for each of the 50,000 channels, so I could > not generate those number without a fully functional run 1 D0 analysis (VMS) > system. There were however "typical" values, which were used in the online > monitoring Examine program, since extracting the "real" numbers from the > offline database was too slow. Unfortunately those number were (as far as > I know) only documented in the fortran code used by the online calorimeter > examine program.... Since the death of the online VMS cluster, I no longer > know how to access the source (I do not even know which module actually had > the numbers in it, or which of the many online libraries the module was > released to). > > Dean > > ------- End of Forwarded Message