Luminosity V_1.0

Procedure:

From Aug 8th, 2001 to Mar 7th, 2002: Based on the Online measurements using the "hit counting" method
From Mar 8th, 2002 to October, 2002... : Based on the Online measurements using the"empty crossing" method

The Online measurements are done in the standalone CLC 6U crate which are read out into the CLLD bank.

Lummon looks at the CLLD bank for events within 1 run section and takes the difference in live integrated luminosity between the 1st and last event to calculate the total integrated luminosity per run section. This value is then entered into the Online Luminosity column in the CDF II Datafile Catalogue.

We apply corrections to the Datafile Catalogue Online values, as described below, and fill the Datafile Catalogue Offline values.

These Offline values are what users should access to calculate their dataset luminosity

We estimate the uncertainty in the total integrated luminosity for this version to be +/- 10% dominated by the absolute normalization. This will be reduced in future versions with better MC simulation and further data studies.

Details on the Online "hit counting method" :

Used the two CLC outer layers
Used Xilinx code with a fixed # of ADC counts as the hit counting threshold
Required at least one hit per CLC module (East/West) in coincidence before counting
Used the following calculation: Lum = (f / Sigma_clc) * (Nhit / <Nhit>_1)
f = tevatron's frequency = 36/(159*132 ns) = 1.715 MHz
Sigma_clc = effective CLC cross section for an East-West hit coincidence (for hits above the ADC threshold)
Nhit = Avg. number of hits per bunch crossing, measured in 2,000 Tevatron cycles, sampled over about 7 seconds
<Nhit>_1 = Avg. number of hits (East+West) in a single p=pbar interaction
The table below shows the values we used for Sigma_clc and <Nhit>_1. The values changed "dynamically" to keep a consistent lum value when PMT gains changed.

Details on the Online "empty crossings method" :

Used the two CLC outer layers
Used Xilinx code with a fixed # of ADC counts as the hit counting threshold
Required at least one hit per CLC module (East/West) in coincidence before counting
Used the following calculation: Lum = (f / Sigma_clc) * mu
f = tevatron's frequency = 36/(159*132 ns) = 1.715 MHz
Sigma_clc = effective CLC cross section for an East-West hit coincidence (for hits above the ADC threshold)
mu = Avg. number of p-pbar interactions per beam crossing
mu is estimated from the formula: P(0) = exp(-epsilon*mu)
P(0) is the. probability of finding (experimentaly) and empty crossing
epsilon = Sigma_clc/Sigma_tot
This formula works well for luminosities .le. few 10**31. For higher luminosities a correction is needed.
The table below shows the values we used for Sigma_clc The values changed "dynamically" to keep a consistent lum value when PMT gains changed.

Period covered	Beg. store / end store	E/W Hit counting threshold (ADC)	Discrim. threshold (mV)	Sigma_CLC_2lyr (mb)	<Nhit>_1

XXX xxx th, 2001 -> Jul 26th, 2001	??? / ???	500 / 500	10 / 10	???	???

Jul 27th, 2001 -> Jul 30th, 2001	585 / 593	500 / 500	10 / 10	38.90	17.60

Jul 31st, 2001 -> Sep 11th, 2001	617/691 / ???	500 / 500	10 / 10	36.50	13.80

Sep 12th, 2001 -> Nov 27th, 2001	700 / 750	500 / 500	10 / 10	37.78	16.00

Nov 28th, 2001 -> Jan 8th, 2002	784 / 889	500 / 250	10 / 8.3	37.78	16.00

Jan 9th, 2002 -> Jan 21st, 2002	900 / 917	500 / 250	10 / 8.3	37.78	15.20

Jan 22nd, 2002 -> Feb 5th, 2002	919 / 955	250 / 250	8.3 / 8.3	36.37	14.51

Feb 6th, 2002 -> Mar 7th, 2002	965 / 1048	250 / 250	8.3 / 8.3	36.61	13.94

Mar 8th, 2002 -> Apr 4th, 2002	1063 / 1154	250 / 250	8.3 / 8.3	37.50	14.50

Apr 5th, 2002 -> May 16th, 2002	1169 / 1329	250 / 250	8.3 / 8.3	36.60	14.10

May 17th, 2002 -> Aug 8th, 2002	1332 / 1634	250 / 250	8.3 / 8.3	35.80	13.60

Aug 9th, 2002 -> Aug 26th, 2002	1654 /1691	250 / 250	8.3 / 8.3	36.20	14.30

Aug 27th, 2002 -> Sept 24th, 2002	1692 /1787	250 / 250	8.3 / 8.3	34.98	12.87

Sept 25th, 2002 -> Oct 31st, 2002	1795 /1929	250 / 250	8.3 / 8.3	34.32	12.48

Corrections to the Online measurement:

Filled gaps for run sections with zero Online luminosity due to Lummon failure

If for a given run more than 10% of the runsections in the Online Datafile Catalogue were zero then the Offline value of these runsections is set such that the total run luminosity is in agreement with the ACNET value (taken from the Run Configuration database). In the Aug-Oct '01 period this happened for ~25% of the runs with a luminosity greater than 1 nb in ACNET. For the later periods LumMon was more stable and it happened only in about 5% of the cases
If less than 10% of the runsections are zero we used a linear interpolation from adjacent non-zero runsections to compute the Offline luminosity for these runsections. If the run starts(ends) with runsections having zero luminosity then these are set to the value of the first(last) non-zero runsection.

For runs between 122255 and 126686 (between July 31st, 2001 and Sept 12th, 2001) we had an effective cross section such that the Online luminosity was measured 20% higher than it should have been.

We multiplied the Online Data catalogue values by 0.83 for these runs

Correction due to CSL not sending all events to Lummon

Lummon needs the first and last event of a runsection to calculate the full integrated luminosity. The CSL in many cases does not send these events and on average Lummon underestimates the luminosity by 1-2%.

Correction due to PMT gain drift between Sept 14, 2001, and Oct 6, 2001

The effective CLC cross-section and the number of hits per single interaction change if the PMT gain changes (due to the fixed Xilinx thresholds in the Online measurement). Offline corrected for this by looking at the single particle peak and applying thresholds relative to these peaks. We find a correction in the form of: L_offline = 1./(1.035 - 0.0062 * t [days]) * L_online

Correction due to PMT gain drift between Oct 6th, 2001, and Oct 31th, 2002

Oct 6th, 2001 - Jan 21st 2002: The effective CLC cross-section was changed dynamically in the online code such that no correction was needed offline.

Jan 22nd, 2002 - Feb 5th, 2002: Multiplied Lum_online by f_corr=1.01 to get Lum_offline

Feb 6th, 2002 - Mar 7th, 2002: Multiplied Lum_online by f_corr=0.98 to get Lum_offline

Mar 7th, 2002 - Apr 4th, 2002: Multiplied Lum_online by f_corr=1.02 to get Lum_offline

Apr 5th, 2002 - May 15th, 2002: Multiplied Lum_online by f_corr=1.01 to get Lum_offline

May 16th, 2002 - Aug 9th, 2002: Multiplied Lum_online by f_corr=0.99 to get Lum_offline

Aug 10th, 2002 - Aug 26th, 2002: Multiplied Lum_online by f_corr=1.02 to get Lum_offline

Aug 30th, 2002 - Sept 15th, 2002: Multiplied Lum_online by f_corr=1.00 to get Lum_offline

Sept 16th, 2002 - Oct 31st, 2002: Multiplied Lum_online by f_corr=1.01 to get Lum_offline