Live- and Dead-time Analysis

We have performed investigations of live- and dead-time for the Amanda detector using data from 1997, 1998, 1999, 2000, 2001, 2002, 2003, 2004, 2005 and 2006.

For years 1997-1998 only smaller sub-samples have been investigated but "ALL" 1999, 2000, 2001, 2002 and 2003 data have been processed. Since the 2004 data we are using the nano-DSTs, and is thereby up-to-date already during data-taking. Note, however, that although quite complete there are files missing from the nano-DST directories in Madison! It would be valuable to fix this by getting hold of this files (but how?) in order to get the dead- and live-time as correctly determined as possible!

New 07/2003:	GPS clock investigation.
New 12/2003:	Full 2002 sample analysis preliminary released
New 06/2004:	Bad run deselection introduced for 2002 sample
New 06/2004:	Adding some runs which appeared very late!
New 07/2004:	2004 data ready online from nano-DST!
New 07/2004:	Full 2003 data set processed
New 12/2004:	Full 1999 data set processed
New 01/2005:	All 2004 data processed from nano-DST
New 01/2005:	Started 2005 processing from nano-DST
New 01/2005:	Raw data rate investigation
New 02/2005:	VLF effects and high data rates
New 10/2005:	VLF effects seen during runs 9469-9476 (pdf, gzipped PostScript)
New 10/2005:	Raw rates per minute for 2005 data (pdf, gzipped PostScript)
New 01/2006:	Completing analysis for the 2005 data and starting to treat 2006-data. As for 2005 data we have plots of raw data rates per minute (pdf, gzipped PostScript) for 2006 data. Note e.g. huge effects in rates during IceCube drilling and calibration at the start of the year.
New 02/2006:	Missing sequences for runs 9319 to 9562 for 2005 data patched up. Still some missing files outside this run range.
New 04/2006:	Introducing bad file list for 2004 using information from Markus Ackermann.
New 04/2006:	Missing data from 2005 (and some from beginning of 2006) patched up using data brought from the South Pole by Markus Ackermann.
New 01/2007:	Details on VLF-breaks in 2003 and 2004 data added.
New 03/2007:	Patchup of 2004 data from Zeuthen store (seems to add about 3 days of livetime!).
New 03/2007:	For 2000-2004 we now implicitly mark runs outside the default analysis range as bad runs (those are not in the bad file lists). We hope this may avoid some confusion.
New 09/2007:	Added run 7989 for 2004 which somehow have turned up later than the rest (found by Johan Lundberg).
New 06/2009:	Updated 2006 data with some missing files found in Madison by Alfio Rizzo (see details).

Year	Existing run range	Analysis run range	Bad list
1999	0-693	?	Bad list
2000	193-584	199-584	Bad list
2001	3099-3821	3113-3398	Bad list
2002	4075-6574	5462-5959	Bad list
2003	6576-7728	6906-7379	Bad list
2004	7737-9053	7800-8090	Bad list
2005	9055-9597	?	Bad list
2006	9598-10227	?

Web Tools
1999
2000
2001
2002
2003
2004
2005
2006

A small working report in (gzipped) PostScript or pdf format has been written. We do not repeat arguments here but refer to this source for details on methods used and the results of the analysis. As a complement additional detailed information is given in web-pages referred to below.

NOTE that this report really is/was an early working document and may not be completely updated for the most recent developments.
Tests: Output from early test runs:
2000: Systematic investigation of 2000-data (reprocessed)
- Files to be processed at Zeuthen.
- File log with detailed investigation of GPS problems.
- Log-files per run for 2000 data.
- Gzipped PostScript-plots with dead-time plotted for each run vs file number and, at the end, with a final plot of dead-time vs run number.
- Web-tool for live- and dead-time calculations for 2000 data.
2001: Systematic investigation of 2001-data (reprocessed)
- Files to be processed at Wisconsin (see also special web-page).
- File log with detailed investigation of GPS time problems.
- Log-files per run for 2001 data.
- Gzipped PostScript-plots with dead-time plotted for each run vs file number and, at the end, with a final plot of dead-time vs run number.
- Web-tool for live- and dead-time calculations for 2001 data.
2002: OBSOLETE! See next section! Investigation of selected 2002-data
- Selected minimum bias raw data files from year 2002 taken from Wisconsin as they came along.
  - Be careful in interpreting the summary results for this heterogenous sample.
  - File-wise information ought to be ok.
  - You may inspect histograms for each file, same order as in log file and where histogram id equal to 1000 plus sequential file number.Note that superimposed are PAW fits not necessarily exactly the same as our ML-estimates.
  - Remaining GPS-problem seen between first two events (0 and 1) in each run (a known roll-over bug it seems).
  - As always dt-plots for individual files are close to ideal BUT we find at least two cases in run 5696 (file 299 and 303 in listing and histograms 1299 and 1303 in pages 50 and 51 of the plots) where the distribution deviates from a perfect exponential distribution lacking events in the tail and where our dead-time estimate may be not too good. This behaviour is odd and unexpected and can only (?) be explained by changes occurring within a file leding to a superposition of exponential distributions. The same tendency is seen for a few other files and may explain some of the dead-times seen being slightly higher that for nearby data (even if the normal cause is a rate change).
  - Note run 5818, plots 1352-1354 on page 59, where suddenly the rate is 350 Hz as compared to the normal value around 85 Hz at this time. Using the web-tool below either with argument * (all data), or 4000-5817,5819-9999 to exclude run 5818 the overall dead-time for the selected 2002 data drops by 0.27 %, a non-negligible amount.
  - Comments: Checking delta-t's thoroughly as was done for the full 2000 and 2001 for the selected files from 2020 we find some oddities which are listed in a special list.
- Log-files per run for 2002 data.
- Web-tool for live- and dead-time calculations for 2002 data.
2002: Investigation of the full 2002-data set
- The full data sample from 2002 has been processed using the tape robot at Zeuthen. A few files have been excluded due to numerical problems when the rate is extrodinarily high. We will investigate if we still can include them but very likely they will be regarded as bad files later.
- Web-tool for live- and dead-time calculations for te full 2002 data set.
- In the summer of 2002 the default deselection of bad files according to the list by Thomas Becka was introuced.
- At the same time it was noted that raw data files for 2002 hade been added at Zeuthen in the spring of 2004 which was updated.
- Thomas Beckas list includes bad and good runs in the run number range 5435-6536 but do not include early and late runs. Thus use (at least) this as an argument to the script.
- What is worse is that e.g. the last run in Thomas' range 6536 is not present in the raw data! This is probably outside normal analysis but for completeness we should check missing runs and try to find raw data for those to be added!
2003: OBSOLETE! See next section! Investigation of selected 2003-data during data-taking
- Selected minimum bias raw data files from year 2003 taken from Wisconsin as they came along.
  - Be careful in interpreting the summary results for this heterogenous sample.
  - File-wise information ought to be ok.
  - You may inspect histograms for each file, same order as in log file and where histogram id equal to 1000 plus sequential file number.Note that superimposed are PAW fits not necessarily exactly the same as our ML-estimates.
- Log-files per run for 2003 data.
- Web-tool for live- and dead-time calculations for 2003 data.
- Comments: Checking delta-t's thoroughly (as was done for the full 2000 and 2001 data but not regularly for 2002 or 2003 selected samples) we find some oddities. In a special list we show all potentially odd delta-'s and see
  - Occasional time gaps, often 64 seconds, which we assume are off-time du to the VLF-antenna transmitting.
  - Quite a few occurrences of negative dt's compensated by nearby gaps being more positive. This is perhaps due to gps-quality checks failing and suddently the time is taken from another gps-clock (we once investigated all 3 gps-clocks ans saw that they were shifted).

2003: Investigation of the full 2003-data set

Using files stored in Zeuthen the full 2003 data set is now processed (20 July 2004).

As pecularities we may mention:

Runs 6738-6741 had year set to 2034.
Runs 6743-6746 and 6749-6869 has year set to 2030.
Runs 6742, 6743, 6747-6750, 6911 and 6935 have corrupt timing. For 6747-6750, 6911 and 6935 only the fraction of a second is cyclically changing while day and second is stuck.
Runs 6751 (files 72-114) and 6752 (files 23-34) has rates up to 400 Hz causing some computational problems. They are excluded from the analysis.

Run 6925 event 3279060 has suddenly day 41750 and time 34775.013333 when one would expect 49 and about 78896.0.

   EM  3279059  6925    2003 49 78895.987574
   EM  3279060  6925    2003 41750 34775.013333
   EM  3279061  6925    2003 49 78896.023219

Run 7412, 7414 and 7424 are excluded since they caused computational problems. Those runs have short files with varying rates and are most probably bad runs with non-exponential behaviour.

Day roll-over bugs:

Run	File	Event	Change day
7205	110	5838500	196 to 195
7206	110	5817268	197 to 196
7208	106	5661266	199 to 198
7210	107	5692970	201 to 200
7266	28	1502281	241 to 240
7328	8	443114	285 to 284
7341	29	1526316	293 to 292

The second-roll-over bug is frequent and not corrected for! Probably a minor effect on live-time but could eventually be corrected for. Cf special investigations done for the selected 2003 data above.

VLF breaks of about 64 seconds are common and correceted for in the live-time determination. In total we lost 3.57 days of live time in 2003 (maybe slightly more adding a few uncommon, not 64 s, VLF-breaks, cf below).

See detailed list with well defined VLF-breaks as well as other breaks longer than 5 seconds. Note that the latter includes the known "feature" with an artificial time gap between event 0 and 1 for a run but a few breaks not being 64 s match the 900 s VLF cycle. Gaps between runs are not included in this list.

Web-tool for live- and dead-time calculations for te full 2003 data set.

Log-files per run for 2003 data.

2004: Investigation of 2004 data

Using the nano-DSTs in Madison 2004 data is fully processed.
[OK Mar'07] Note, however, that there are a few files and at least one run (7875) which seem to be missing.
As pecularities we may mention:
- Day roll-over bugs (note that in raw data MJD=day+53004 is used):

Run	File	Event	Change day
7744	135	6771519	10 to 9
7757	124	6386602	16 to 15
7759	125	6384391	18 to 17
7811	17	865976	54 to 53
7816	17	860800	59 to 58
7817	17	855798	60 to 59
7832	125	6195401	75 to 74
7849	122	6094350	92 to 91
7851	121	6035583	94 to 93
7854	123	6074142	97 to 96
7855	121	6051310	98 to 97
7882	135	6802105	124 to 123
7892	136	6669487	134 to 133
7896	123	6271691	136 to 135
7956	9	498257	185 to 184
7999	16	867351	226 to 225
8028	120	5890069	255 to 254
8040	119	5953463	267 to 266
8047	120	5964300	274 to 273
8077	130	6656790	303 to 302
8091	120	5984017	311 to 310
9032	92	4845470	356 to 355
9033	150	6979407	357 to 356
9039	150	6824040	360 to 359

The second roll-over bug causing dt's of -1 s followed by +1 s is frequent (see comment on 2003 data above).
This explanation turned out to be wrong!
Seeing a lot of -1 s and +1 s delta-t's we assumed it was the same bug as seen in 2003. However, a more detailed investigation shows that it is a new bug. Occasionally when seconds are updated it jumps one too much causing a dt=+1 s – many events afterwards, at the next second shift, the seconds are not updated causing a dt=-1 s! I.e. a bunch of events are wrongly shifted +1 s.

VLF breaks of about 64 seconds are common and corrected for in the live-time determination. In total we lost about 21 days of running time in such breaks for the full 2004 data set.

See detailed list with well defined VLF-breaks as well as other breaks longer than 5 seconds. A few breaks not being 64 s match the 900 s VLF cycle. Gaps between runs are not included in this list.

Run 7763 and 7772 had sudden corruption with MJD.
For 7763 at event 17283 MJD from 53027 to 62524 and then back to 53027 at event 22688.
Run 7772 had more similar problems. Reset to 53028 which was true for the full run (corresponding to day 24 of 2004).
For run 7782 files 26-37 are missing.
[OK Mar'07] For run 7809 files 92-95 are missing.
[OK Mar'07] For run 7959 file 49 is missing.

Zeuthen store of 2004 nano DST's found and a small update adding some missing files/runs made:
a) run 7809 adding files 92-95,
b) run 7875 adding files 000 to 094,
c) run 7959 adding file 049,
d) run 8075 adding files 012-130
e) run 8076 new with one file 000,
f) run 8077 adding files 000-037,
g) runs 8121 and 8269 short with one file each,
h) runs 8270, 8291, 8328, 8396 (had to skip last 4 files 045-048 due to rates up to above 600 Hz, files 041-044 fishy too!), 8459 and 8532 new.

Johan Lundberg found that run 7989 was missing in the database. It was found in Zeuthen and processed. This was quite short run with 37 files and added only 0.261 days of livetime.

Web-tool for live- and dead-time calculations for the full 2004 data set.

Log-files per run for 2004 data.

2005: The full set of 2005 data has been processed using nano-DSTs. Note, however, that although "almost" complete there are some files missing. I hope to be able to patch this up! Can anyone help with this ... please!? See below!!! I think most of those missing pieces may have found now!

Using the nano-DSTs in Madison 2005 data is processed as it comes in.
As pecularities we may mention:
- Day roll-over bugs (note that in raw data MJD=day+53370 is used):

Run	File	Event	Change day
9060	155	7163026	6 to 5
9088	073	3371372	14 to 13
9095	112	5147621	19 to 18
9109	249	10152485	25 to 24
9111	651	20552902	27 to 26
9128	000	36088	38 to 37
9152	024	1106014	53 to 52
9155	126	5782102	54 to 53
9176	018	804293	67 to 66
9179	017	817145	68 to 67
9180	017	805404	70 to 69
9188	139	6397650	75 to 74
9190	150	6975313	77 to 76
9201	079	3663822	84 to 83
9212	160	7433867	95 to 94
9219	019	886309	99 to 98
9239	019	899552	115 to 114
9242	082	3837652	118 to 117
9253	076	3550626	122 to 121
9273	019	894886	140 to 139
9299	013	629692	158 to 157
9327	140	6525004	174 to 173
9339	051	2398835	180 to 179
9340	061	2785022	181 to 180
9361	041	1921470	196 to 195
9420	004	209538	236 to 235
9476	153	6980807	277 to 276
9477	142	6399563	278 to 277
9479	106	4805221	279 to 278
9486	134	6037759	285 to 284
9491	159	7319710	287 to 286
9506	142	6547295	294 to 293
9511	144	6664991	299 to 298
9517	130	6066566	303 to 302
9520	020	964009	305 to 304
9523	014	667754	308 to 307
9525	152	7168309	310 to 309
9547	106	5020495	329 to 328
9553	105	4957591	335 to 334
9567	017	828250	345 to 344
9568	015	702671	346 to 345
9590	158	7374912	359 to 358

[OK Feb & Apr'06] A future problem to get the 2005 data base complete from nano-DSTs may be to fill in some missing sequences:
Run 9109, files 068-357 and the full run 9110 had very high rates (IceCube drilling going on) and had to be skipped,
Run 9111 files 0-805 skipped (806-812 kept),
Run 9112 files 8-251 skipped,.
[OK Apr'06] Gap between Run 9112 file 363 and run 9124 file 36! (some high rate files skipped)
Run 9124 files 50-52, 75, 79-80, and 154-165 skipped (0-35 missing!),
Run 9125 files 52-55, 107-109, 114-115 and 118-120 skipped,
Run 9134 suddenly got rates up to 330 Hz and files 75-78, 103, and 106-113 were skipped.
Run 9136 files 123-125 up to 260 Hz and skipped.
[OK Apr'06] Run 9198 files 157-164 missing.
[OK Apr'06] Run 9199 files 1-2, 9-40, 42-end (?) missing
[OK Apr'06] Run 9200 only six out of at least 163 files present!
[OK Apr'06] Run 9201 files 1-77 missing (3 there but corrupt).
Run 9220 files 0-3 skipped due to high rate.
[OK Apr'06] Run 9222 files 017, 019-054 empty.
[OK Apr'06] Runs 9223 and 9224 missing (?).
[OK Apr'06] Run 9225 files 0-072 and 142-146 missing.
[OK Feb'06] Run 9343 files 119, 121-139, 149 and 151-156 missing.
[OK Feb'06] Run 9446 file 103 missing as well as some events between files 104 and 105
[OK Feb'06] Run 9459-9461: Run 9459 files 36 and 38-end, all of run 9460 and run 9461 files 0-131 missing "due to Rita"
Run 9469-9476: Huge VLF effects seen with rates up to 500 Hz (see plots in pdf or gzipped PostScript)!
        Should most probably disregard run 9469 files 119-120, run 9471 file 33, run 9472 files 13-18,
        run 9474 files 1-6 and run 9476 files 0-35. Also in later data analysis!
        The total effect is quite small though as seen by comparing the results of a full run to one with selection
        9055-9468,9469(0:118+121:*),9470,9471(0:32+34:*),9472(0:12+19:*),9473,9475,9476(36:*),9477-*.
[OK Feb'06] Run 9510: missing files 63-68, 70-78 and 80-86.
[partly OK Feb'06] Run 9557: missing files 96-end, runs 9558-9563 and run 9564 until file 142.
[OK Apr'06] Still missing files from run 9562 file 135 until run 9564 file 141 (files 0-019 and 042-098 for run 9562 and 042-044 for run 9563 skipped).
[OK Apr'06] Run 9594: missing files 96-106. 108-110 and 112
[OK Apr'06] Run 9595: missing files 85, 93, 95, 97-122, 141, 143-148, 163-167 and 169-170 (akipped files 003-006).
[OK Apr'06] Run 9596: missing files 50-51, 92-96, 98, 100-101 and 133-139
[OK Apr'06] Run 9597: missing files 1-3, 5, 8-10, 12-13, 16-17, 21, 42-44, 46, 48-50, 82-83, 87-90, 93, 117-121, 123 and 125

On 3rd of February 2006 we could patch up missing sequences for runs 9319 to 9562 from a
new directory where South Pole tapes had been staged recently (thanks to John Richards).

In April 2006 we could patch up missing sequences for runs 9112-9124, 9198-9201, 9562-9564 and 9594-9597 with data brought to Zeuthen from the South Pole bu Markus Ackermann.

As for 2004 data quite often when seconds are updated it jumps one too much causing a dt=+1 s – many events afterwards, at the next second shift, the seconds are not updated causing a dt=-1 s! I.e. a bunch of events are wrongly shifted +1 s. This effect seemed to disappeared after run 9069 in 2005 data but unfortunately reoccurred for runs 9136-9180. For long periods it may be gone again as for the last period with runs 9513-9597.

For 2005 data there were no VLF-breaks since new SWAMPs filtered out the disturbances.

For completness we have made a detailed list with other breaks longer than 5 seconds as was done for 2003 and 2004 data. Those were very few but gaps between runs are not included in this list.

Web-tool for live- and dead-time calculations for the 2005 data set.

Log-files per run for 2005 data.

2006: Investigation of 2006 data (probably complete until 21st of August and run 10010)

Using the nano-DSTs in Madison 2006 data is processed as it comes in.
The early runs are heavily affected by IceCube drilling and calibration and there are highly varying rates affecting the results. Some too bad files have been disregarded but later, and for analysis, we should reject more bad files/runs as for previous years. See plots of raw data rates per minute (pdf, gzipped PostScript)
As pecularities we may mention:
- Day roll-over bugs (note that in raw data MJD=day+53735 is used):

Run	File	Event	Change day
9602	199	8927134	6 to 5
9783	002	134597	45 to 44
9790	156	7442982	52 to 51
9791	157	7455003	53 to 52
9797	181	7551634	59 to 58
9798	184	7732513	60 to 59
9811	020	954943	72 to 71
9815	006	297493	75 to 74
9818	006	295812	78 to 77
9822	145	7028482	83 to 82
9825	155	7215948	86 to 85
9828	011	529662	87 to 86
9848	141	6348815	103 to 102
9851	140	6598097	106 to 105
9894	135	5697257	140 to 139
9897	133	6512169	143 to 142
9905	094	4648845	149 to 148
9918	096	4700374	162 to 161
9930	096	4735908	174 to 173
9940	090	4446279	183 to 182
9942	090	4449513	185 to 184
9951	029	1459837	193 to 192
9985	122	5931568	225 to 224
9994	087	9000369	232 to 231
10062	099	4724614	267 to 266
10064	100	4783838	269 to 268
10095	118	5571732	296 to 295
10125	013	617876	312 to 311
10128	013	616995	315 to 314
10162	113	5405736	324 to 323

A future problem to get the 2006 data base complete from nano-DSTs may be to fill in some missing sequences:
- [OK Apr'06] Between run 9601 file 109 and run 9602 file 124 all files are missing.
- [OK Feb'06] The transfer of nano-DSTs from the South Pole to Madison was stopped due to PBS license reasons on the 17th of January at run 9613 file 110 and some files were lost.
- Run 9800 has two corrupt files (47 and 142) and is missing a lot of files (48-137,140-141).
- [OK May'06] File 150 of run 9880 to file 150 of run 9882 missing due to disk being offline in Madison.
  Resent from South Pole and patched up on 19th of May
- Missing files 56-58, 61-73, 77, 90, 95, and 101-104 from run 10010!
- Missing files 116, 118-132, 134, and 136-end for run 10041 but also files 117 and 133 damaged.
- Missing files 000-001 for run 10042.
- Missing data from file 135 for run 10047 until file 021 for run 10048.
- File 073 corrupt and files 074, 076-105 missing for run 10063 (file 089 exist but corrupt).
- File 111 corrupt and missing files 113-123 and 126 for run 10064.
- Missing files 116-126 and 131-145 for run 10065.
- Missing 007-009 and files 004 and 027 corrupt for run 10072.
- File 026 corrupt for run 10078.
- File 151 corrupt amd most probably missing 152-end.
- Missing files 000-012,015-031 and 014 corrupt for run 10080.
- Missing files 061-065, 067-111, 113-116, 120-135 and files 066, 118 and 155 corrupt for run 10089.
- Missing files 000-011, 034-050 and file 071 corrupt for run 10090.
- Missing files 006-022 for run 10094.
- Missing files 016, 018-036, 038-052, and 054-077; files 017, 037, and 053 corrupt for run 10097.
- Missing files 012 and 015-033; 013 corrupt for run 10098. Possibly also missing files 096-end.
- Missing files 000-004, 006-008, and 010-027; corrupt file 009 for run 10103.
- After run 10227 started 27th of November no files transferred to the North.
- For 2006 data there were no VLF-breaks since new SWAMPs filtered out the disturbances.
  
  For completness we have made a detailed list with other breaks longer than 5 seconds as was done for 2003 and 2004 data. There were very few gaps before run 10103 but after this plenty of them. Gaps between runs are not included in this list.
- COMMENT:In general 2006 data seems quite complete until run 10010 at 21st of August after which quite a lot of files were lost and some were corrupt. After 7th of December 2006 and run 10227 (from 27th of November) no more data have been seen transferred to the North.
- In 2009 Alfio Rizzo found that some 2006 files which turned up in the data stream did not have corresponding live- and dead-time data. It was found that data (in our case nano-DSTs) from tapes brought from the South Pole later had been put on disk in Madison. With some effort we managed to get out old scripts to work and made an update in early June 2009. See special page for some details. About 3.5 days of lifetime corresponding to almost exactly 3 days for dead-time corrected lifetime was added.
Some other features in data :
- Generally there are more features than usually seen in raw rates as shown in plots in pdf- or gzipped PostScript-format.
  Most of those are probably due to IceCube-drilling and calibration.
- For files 108-263 of run 9788 and 000-070 of run 9789 the rate was about 500 Hz and those files were excluded from the analysis. This drastic and yet unexplained feature started at 17:07 on the 18th and ended at 05:37 on the 19th of February (with very sharp edges).
- Run 9821 had rate spikes affecting the analysis for files 85-88, 93-95, and 97-100 which were removed (perhaps a larger continuous interval ought to be disregarded).
- Run 9894 files 105-181 had rates 100-400 Hz and were disregarded.
- Files 126-154 and 162-167 of run 9991 and all runs 9992-10008 disregarded due to rates close to 800 Hz and runs 9998-9999 and 10005-10008 being empty.
- High rates observed also occasionally for runs 10014-10025 and runs 10012-10013 empty.
Web-tool for live- and dead-time calculations for the 2006 data set.
Raw data rates per minute in pdf- or gzipped PostScript-format.
Log-files per run for 2006 data.

For 2000 and 2001 data which have been fully analyzed we have created much smaller data-bases with enough information per file to redo the calculations and to properly sum up total live- and dead-time for a certain run/file-selection. Note that the original data consisting of all individual EM- and TRIG-cards occupies 17 Gb for 2000 data and 32 Gb for 2001 data while the small data-bases are 4 Mb and 5 Mb, respectively. Similar databases are now used by our new web-tools for data from all years i.e. 1999, 2000, 2001, 2002, 2003, 2004, 2005 and 2006. Note that we sofar have kept the data from the selected samples analysed during the data-taking for 2002 and 2003 although those are obsolete.

Christian Walck 2007-03-16