CRIS Level 2 Data Release Notes
=========================================================================
January 23, 2013 Kelly Lave
REANALYSIS OF THE FULL CRIS DATASET
The CRIS Level 2 data product has been re-evaluated following changes made to
the data cuts and the calculation of the geometry factors and SOFT hodoscope
efficiencies. Bugs in the positioning and thicknesses of the silicon detector
stacks and the calculation of the depth of incident particles have been
corrected. This resulted in a small change in the geometry factors (less than
~1%). Note that the geometry factors tabulated for the user are slightly
different for each species in a given energy level. This is expected since we
first calculate the differential geometry factor per unit energy and then
integrate over the energies of the particles stopping in the given energy
level. The isotopic composition used for this calculation comes from the
values posted in the contributed CRIS level 3 data product
(www.srl.caltech.edu/ACE/ASC/DATA/level3/cris/isotopic_composition.html). The
tabulated elemental geometry factors are a weighted average of the isotopic
geometry factors.
The corrections described above also change the number of events accepted in
each energy level. For abundant species (i.e. carbon, oxygen, silicon, and
iron) the number of events change by less than ~1%; rarer species (i.e.
phosphorus, chlorine, and scandium) changed by less than ~3%; cobalt, which
has limited statistics, changed by less than ~5%.
The calculation of the SOFT efficiencies now implements the work of de Nolfo,
et. al (2006). These efficiencies are parameterized by the energy loss in
silicon at the top of the instrument for particles stopping in each of the
four stacks of silicon detectors; they have been reproduced in Lave (2012a,
2012b). The boron efficiencies change by up to 1.5%, while heavier species
change by less than 0.7%; the heaviest species, such as iron, change by less
than 0.2%.
The available documentation
(www.srl.caltech.edu/ACE/ASC/DATA/level2/cris/cris_factors.txt) now includes
additional information previously unavailable. In addition to the energy
bands, recommended central energies, and geometry factors, we now include the
SOFT efficiencies, spallation correction factors, and the uncertainties on the
spallation factors. THIS NEW FACTORS FILE SHOULD BE USED WITH THE NEW LEVEL 2
DATA! The uncertainties on the spallation correction factors are determined
by assuming a 10% uncertainty on the total interaction cross sections. Note
that the spallation factors assume that all particles interacting in the CRIS
instrument are identified and removed from the analysis; although there is
reasonable certainty that the efficiency of rejection is high, this has not
yet been confirmed.
The appropriate systematic uncertainties on the CRIS intensities should be
calculated for each charge and energy bin. Combine in quadrature the
contributions due to the geometry factors (2%), the SOFT efficiencies (2%),
and the spallation (given by the absolute uncertainties in the data file
discussed above).
For supplemental information regarding the calculation of the CRIS
intensities, please refer to George, et al. (2009) and Lave (2012a, 2012b).
References:
G. A. de Nolfo, et. al. Adv. Sp. Res., Vol. 38, pg 1558, 2006.
J. S. George, et al. ApJ 698, pg 1666, 2009.
K. A. Lave. "The Interstellar Transport of Galactic Cosmic
Rays", Ph. D. Thesis, Washington University in St. Louis,
http://openscholarship.wustl.edu/etd/707, 2012a.
K. A. Lave. Thesis Erratum,
www.srl.caltech.edu/ACE/ASC/DATA/level3/cris/Erratum_11-20-122.pdf, 2012b.
- Apr 2, 2010 Rick Leske
NEW CRIS Level 2 Data
The CRIS Team has re-evaluated the production of CRIS Level 2 data,
implementing several changes that were incorporated in the analysis
presented in George et al., ApJ 698, 1666 (2009). These include
revisions to the hodoscope grammage, interaction probabilities, and
data cuts and geometry factors for events stopping deep in the
instrument. In addition, a bug in the maps of the SOFT hodoscope has
been repaired, slightly changing the calculation of trajectories, and
the parsing of the data telemetry stream has been made more robust,
which results in the recovery of a small number (~0.1%) of additional
events. The overall effect of these changes is typically to reduce the
absolute intensities by a small amount, by up to a few percent. Together
these changes have some effect on the width and centers of the CRIS
energy bins; PLEASE USE THE NEW ENERGY VALUES (available at
http://www.srl.caltech.edu/ACE/ASC/DATA/level2/cris/cris_energy_bands.txt)
WITH THESE NEW DATA!
***CAUTION*** CRIS DATA ARE OUT OF CALIBRATION BETWEEN 28 AUGUST 1997
AND 4 DECEMBER 1997. Early in the mission the camera discriminator
in the SOFT hodoscope was adjusted several times. Between 28 Aug and
4 Dec 1997 the discriminator settings were not optimal, resulting in
a bias against heavily-ionizing particles in the hodoscope. For species
below about Si (Z=14) we do not notice any problems, but the intensities
reported in the CRIS level 2 data (both in this new release and in all
previous versions) are too low during this period for heavier species,
by ~10% for Fe (Z=26) to as much as ~30% for Ni (Z=28). This fact was
overlooked in the George et al. paper cited above and will be addressed
in an upcoming erratum. We are working to determine the appropriate
correction factors and will implement them in a future level 2 release
when they are available.
=========================================================================
- Mar 3, 2010 Rick Leske
Unintended Data Change
The CRIS Team has re-evaluated the production of CRIS Level 2 data,
implementing several changes that were incorporated in the analysis
presented in George et al., ApJ 698, 1666 (2009). These include
revisions to the hodoscope grammage, interaction probabilities, and
data cuts and geometry factors for events stopping deep in the
instrument. In addition, a bug in the maps of the SOFT hodoscope has
been repaired, slightly changing the calculation of trajectories, and
the parsing of the data telemetry stream has been made more robust,
which results in the recovery of a small number (~0.1%) of additional
events. The overall effect of these changes is typically to reduce the
absolute intensities by a small amount, by up to a few percent. At
present we are evaluating the effects of these changes on our energy
bins (which are expected to be at the few percent level), and once
this is completed in a few weeks we will release the new data.
Unfortunately, test files using this new analysis were inadvertently
placed on the CRIS level 2 site on Feb 24, 2010, 2:40pm PST, replacing
the older files there; also data from Bartels Rotations 2407 and 2408
were generated using the new analysis. CRIS DATA FILES DOWNLOADED
BETWEEN FEB 24 and MAR 3 2010, OR ANYTHING FROM BR 2407 DOWNLOADED
PRIOR TO MAR 3, SHOULD NOT BE USED WITHOUT THE APPROPRIATE CHANGES TO
THE ENERGY BINS, WHICH WE ARE STILL WORKING ON. For the time being, we
are restoring the old files to the site until we have finished
evaluating the new ones.
We thank the users who called this issue to our attention, and we
apologize for any inconvenience this may have caused.
=========================================================================
- Nov 7, 2003 Rick Leske
Energy Band Information Update
Up until now the energy band info included only the minimum energy and
maximum energy for each element for each detector range. These values
are for the characteristic energy band which is defined as the minimum
and maximum energy required to stop in each range at the median zenith
angle for that range. The incorporation of this characteristic energy
band into the flux and geometry factor calculations is described in more
detail below in the entry for Level 2 Software, Version 1.0 July 26,
1999. None of the geometry factors, energy intervals, or fluxes have
changed. However, we now add to the energy band documentation file a
recommended central energy point, which is defined as the arithmetic
average of the (isotope-weighted) absolute min and max calculated energies
for each range. Note that this is NOT the average of the upper and lower
values of the characteristic energy band.
=========================================================================
- Aug 16, 2002, Jeff George
Quiet Time Criteria Update
The CRIS quiet time criteria defined below for Level 2 data was found
to allow periods of solar activity in which SOFT is partially
saturated with triggers from low energy particles. The criteria
removed periods in which the SOFT trigger rate was essentially zero
but did not remove the transitions from normal operation to the fully
saturated state. This became very apparent during the July 22, 2002
SEP event in which the particle fluxes kept the SOFT trigger rate in
this intermediate state for several days. These periods need to be
removed as the instrument is not operating in a nominal state.
We have corrected the problem by adding one additional test to the
quiet time criteria for basic instrument operation. Now the average
of the four E1 detector rates is required to be below 300 Hz. This
has no effect on normal operation when the rate is typically 10-20
Hz. The 300 Hz level corresponds well to the point at which the SOFT (0 AND
1) trigger rate begins to saturate. The new criterion rejects periods
with normal SOFT rates but high E1 rates, exactly the transition
conditions that were a problem. The new condition removes a total of
5.8 days (0.3%) from the full mission to date. Nearly 2 days
corresponds to the July 22, 2002 event alone. All of the other
rejected periods come from short transitions of a few hours each
scattered throughout the mission. There is no effect on the data for
periods outside these transitions during SEP events.
All CRIS Level 2 data available from the ACE Science Center was
reprocessed using this additional quiet time criterion on Aug 16,
2002. Data is currently reported through Bartels Rotation 2306.
=========================================================================
- Level 2 Software, Version 2.0
(updated Mar 20, 2002 by Jeff George)
(all CRIS Level 2 data available from the ACE Science Center was
reprocessed using this new code on Mar 21, 2002)
This new version of the Level 2 processing software is a complete
rewrite of Nathan Yanasak's original code to make the procedure much
more modular, efficient, and readable. It corrects several minor
errors in matching event and rate data. We also took advantage of the
need for reprocessing the data to implement a new, objective criterion
for selecting periods during which the instrument is operating at full
performance and there is little possibility that the galactic cosmic
ray spectra are contaminated by solar energetic particle events.
The new software package processes a single Bartels rotation at a
time, improving the speed and reducing the memory requirements. A Perl
script automatically extracts event, rate, and housekeeping data. All
CRIS data are reported to the ground in 256 second instrument cycles
stamped with a spacecraft clock time for each cycle. In keeping with
this format, the new code works with an efficient mask for selecting
valid data periods on a cycle by cycle basis. Any mask index which
cannot be addressed by a valid clock time in the rate data is declared
missing. Because rate and event data are physically reported in
different cycles in the spacecraft telemetry, the index immediately
following a missing cycle is also rejected. Instrument performance
and solar activity criteria flags (described below) are also applied
directly to the mask array. A single loop through the mask array
indices generates fluxes only for known valid periods and applies -999
fill values for the remainder. This procedure makes it extremely
difficult to erroneously report fluxes for invalid periods.
Data provided to the public will continue to be averaged on hourly,
daily, and 27-day timescales. In addition, the factors used to
calculate the fluxes from the instrument counts remain identical to
those used in the previous Yanasak code and described in previous
releases.
Instrument Performance and Quiet Time Criteria:
Data from any given instrument cycle must pass a set of minimum
criteria to be considered "valid":
1) Spacecraft telemetry frames for both the event and rate data
must exist and have been received without error on the ground.
2) The SOFT multi-channel plate (MCP) high voltage must on and at
normal values. During high rate periods (i.e., during solar
energetic particle events) the current limits on the SOFT MCP
trip off and the instrument must be manually commanded back on
when the event is over.
3) The SOFT trigger rate (the logical AND of the two trigger
planes) must not exceed 5kHz. The SOFT trigger is
edge-detected and at high rates there is so much ambient light
in the fibers that the trigger level generally remains high
and few if any event triggers are generated.
4) The fractional livetime for Z>2 particles must be above 60%.
CRIS livetime is typically over 80% and the instrument is
capable of performing even when the available livetime is
low. Such periods introduce a large correction to the flux
calculation, however, so we prefer to simply reject them.
These criteria define the minimum requirements for valid data and
remove only ~7% of all time since launch. The purpose of the Level 2
data, however, is to provide a clean sample of galactic cosmic-ray
data. As such, we impose an additional tighter restriction on the
SOFT trigger rate to avoid even the possibility of contamination by
solar energetic particles.
5) The SOFT trigger rate (logical AND of the two trigger planes)
must not exceed 500Hz.
Each of these criteria are applied individually to each instrument
cycle and with 5), remove about 13% of the time in the full
mission-to-date. By contrast, the previous code used a very
conservative and somewhat subjective criterion that rejected entire
days. That code rejected ~43% of the full mission data up to this point.
Event and Rate matching errors corrected in this release
Rate data for a given instrument cycle is physically reported in the
telemetry for the next instrument cycle. Because of this, the
previous Level 2 code version matched events from one spacecraft clock
timestamp with the fractional livetime reported with the next
timestamp. We recently realized that this correction is already being
made in the Level 1.1 processing where the timestamps for the rate
data are corrected to match those of the cycle which contains the
events for which the rates are being reported. This also affects
which cycles are rejected due to missing rates. CRIS data are
normally quite stable and the rates do not typically vary quickly so
this is expected to have little impact on the previously reported
fluxes. Nevertheless, the error is corrected in this release.
=========================================================================
- Level 2 Software, Version 1.3 (updated February 1, 2001 by Nathan Yanasak)
No major structural changes were made in the V1.2 programs listed above.
However, two errors in the data processing were discovered. The first error
occured as a result of switching from ce1 to ce2 data (see V1.2 discussion).
A general cut applied to data from CRIS eliminates particles which pass
close to the outside edges of the active detector region. The intent of this
cut is to eliminate particles which exit the detector through the sides.
During the switch from ce1 to ce2 data, the routine in the data extraction
software which handles this cut was altered, decreasing the size of the cut.
The calculated geometry factor assumed the older cut from the V1.0 and V1.1
data set, which was inappropriate for the new cut in V1.2. After correcting
the geometry factor so that it corresponded to a larger active volume in the
detector, the flux for level 2 data from V1.3 software decreased by
approximately 4-5%.
The second error is associated with the calculation of the event median angle.
Comparison of the median angle calculated in the V1.2 factors.pro program with
angles calculated by another independent program revealed an error in the V1.2
code responsible for the angle calculation. With the error corrected, the new
values of the median theta are given below:
instrument range | median theta
=================================
2 | 20.1
3 | 19.9
4 | 19.7
5 | 19.3
6 | 18.9
7 | 18.9
8 | 18.6
The increase in the values of median theta give slightly increased average
values for the corrected energy bands. This increase depends on the depth of
the reported energy band in the detector, ranging from ~0% for range 2 to
~1.5% for range 8.
The systematic uncertainties reported above at the end of the section entitled
"General Information"--15% for low-Z elements and 10% for higher-Z elements--
need to be re-evaluated in light of the changes which took place in V1.1, V1.2,
and V1.3 of the level 2 code. The primary sources of uncertainty in the flux
calculation for level 2 data are as follows:
1) statistical
2) uncertainty in average energy of a band -- flux data are reported
for an energy band, and the spectra depend on which energy the data
are plotted.
3) geometry factor uncertainties
4) SOFT efficiency
5) correction for fragmentation in the detector -- factors.pro uses the
expression of Westfall et al. (PRC, V19, p1309, 1979) instead of a
more complex, energy-dependent cross-section code.
The uncertainty for item #2 was estimated in the following manner. An average
spectral shape (f=log(flux)~A*log(E)^2+Blog(E)+C) can be found for all species
by plotting flux data at energies in the center of the energy bands. Using
this shape, one can examine how much the flux should change over the range of
energies in a given band. Assuming ad hoc that the reported flux value given
for an energy band Elo < E < Ehi can lie anywhere between Elo+0.25*delE and
Elo+0.75*dele (here, delE=Ehi-Elo), the largest variation in flux in any given
energy band is 5%.
The geometry factor has been verified by three independent codes to a level of
5%. The precision of the factor calculated in factors.pro is statistically
significant to the level of 1.5%, but 5% should be used as an upper limit at
the time of this version release.
The improved SOFT efficiencies for V1.2 and later versions are precise to 3%
for Z=4 and 2% for other elements.
Differences between the total fragmentation cross sections from Westfall et al.
(1979) (used by factors.txt) and measured total cross sections have been
estimated using the formulae of Tripathi et al. (NASA/TP-1999-209726, 1999).
The formulae from Tripathi et al. are fine-tuned to match an extensive set of
cross-section data, and they include an energy dependence as well. The amount
of fragmentation depends on the depth at which the cosmic rays stop in CRIS,
and the depth is also a function of the particle energy. So, the amount of
variation between the formulae of Westfall et al. and Tripathi et al. is a
function of the particle species and energy. The low-Z and Fe,Ni-group nuclei
show the largest difference between these formulae (ranging from 0% for
stopping depths near the top of the CRIS SI detector stacks to 5% at the
bottom of the stack), and species around Z=11-15 show the least amount of
difference (<2% over the full range). To be conservative, we suggest an
uncertainty of 5% in the fragmentation correction.
Adding all of the systematic uncertainties listed above in quadrature, we find
a global uncertainty for level 2 data of ~9% for all elements.
Data from CRIS during periods of high solar activity are "flagged" with a
value of -999.9 because the instrument was not designed for optimal
operation during these condition
(see http://www.srl.caltech.edu/ACE/ASC/level2/cris_l2desc.html on the ACE
Science Center website for a full description of the CRIS level 2 data). These
periods are the most conservative estimates of days during which CRIS
performance may be affected. Currently, the criteria for determining periods
of high activity are being re-evaluated so that more days can be included in
the level 2 data. These criteria will be applied at a future data release and
documented in the release notes.
=========================================================================
- Level 2 Software, Version 1.2 (updated May 22, 2000 by Nathan Yanasak)
No major structural changes were made in the four V1.1 programs listed above.
However, three minor changes were made.
Change #1: Discrepancies between the two different methods for calculating the
average geometry factor (mentioned above) have been resolved, and the average
geometry factor calculated by factors.pro has changed to reflect this
agreement. The maximum amount of discrepancy between the geometry factor in
older versions of factors.pro and version 1.2 is a 7.5% increase in the
corrected geometry factor, and the average amount of discrepancy is
approximately 6.7%. Calculation of the median theta for the zenith angle
described above is dependent on the geometry factor code, and the new values of
median theta for version 1.2 are presented below:
instrument range | median theta
=================================
2 | 18.4
3 | 18.4
4 | 17.5
5 | 16.7
6 | 15.9
7 | 15.4
8 | 14.6
Change #2: Version 1.2 uses ce2 data as input instead of the ce1 data used as
input for previous versions. ce2 data is slightly different from ce1, using a
different algorithm to determine SOFT trajectories. The improved algorithm
results in a higher SOFT efficiency on average for any GCR species
(~ +1.5% average increase from ce1 data), with much greater efficiency at
lower Z (~ +18% for Be). SOFT efficiencies calculated by the version 1.2
factors.pro are appropriate level 2 data processed from ce2 data.
Change #3: In our calculation of particle energies, we have switched from an
older range-energy IDL procedure to a newer one based on Anderson and Ziegler.
Particle energy values decrease by 1-2% for particles stopping near the top of
range 2 using the newer method, and the energy values experience a negligible
decrease for particles stopping deeper within the stack.
Finally, the IDL procedure eleplot.pro has been replaced by another procedure
which has two versions with differing names: frsteleplot.pro and
itereleplot.pro. The first version, frsteleplot.pro, is identical to
eleplot.pro and should be used if one wants to reprocess level 2 data from
the beginning of the mission. The second version, itereleplot.pro, is similar
to eleplot.pro, but it uses output from previous runs of frsteleplot.pro or
itereleplot.pro instead of reprocessing data from the beginning of the mission.
=========================================================================
- Level 2 Software, Version 1.1 (updated October 26, 1999 by Nathan Yanasak)
No significant changes were made in the four V1.0 programs listed above;
however, a different version of code was used to extract the raw ce1 data which
acts as an input to the statflux.pro code. A slight discrepancy concerning the
coordinates of the detector telescope elements was discovered. The pre-level
2 processing of raw event files makes certain geometric cuts appropriate to
the geometry factor used for level 2, which serve to eliminate data which
traverse SOFT near its edge or pass very close to the guard rings in the
detector telescopes. Some of these geometric cuts were affected by this
discrepancy, and a set of positional offsets was applied to correct these in
pre-level 2 processing. Consequently, the number of raw events which are
being processed by statflux.pro has increased by approximately 1.3%, which
increases the particle fluxes found using V1.0 code by some fraction of this
1.3%. The input section of statflux.pro was altered accordingly to reflect
this increase.
=========================================================================
Level 2 Software, Version 1.0 July 26, 1999
N. Yanasak
Programs:
1) statflux.pro
Language: IDL
General description: This program applies the data cuts to raw
ce1 data.
2) eleplot.pro
Language: IDL
General description: This program associates the data to the
appropriate values of the instrument livetime.
3) lvl2gen.pro
Language: IDL
General description: This program is responsible for formatting the
data into the level 2 format and placing it into the
appropriate files.
4) factors.pro
Language: IDL
General description: This program calculates the necessary factors
for converting the event rates given in the output files of
lvl2gen.pro into particle differential fluxes (in units of
# of events/(MeV/nuc)/seconds/(cm^2 sr) ).
General Information:
Software used for the generation of level 2 CRIS data outputs
data grouped into two categories: more abundance elemental species
(C,N,O,Ne,Mg,Si,Fe), and species of lesser abundance (all other elements not
listed, between Z=5-28). Data for the more abundant species are
reported for each 256-second window in time, and the less abundant species
are reported for 1024-second windows. There are two types of output data
files, with each type corresponding to one of the two categories described
above. Each file holds data reported during one Bartles rotation.
Data are reported as the rate of particles observed to stop in each of
the 7 instrumental ranges. These ranges correspond to different vertical
stopping depths in the instrument. Because the particle zenith angle for this
reported data lies between 0 and 30 degrees, a unique vertical depth will give
a spread in particle stopping energies. This means that the energy bands which
characterize particles of a given species stopping in each instrument range
will overlap in particle energy. To define characteristic energy bands for
each instrument range, we have used the minimum and maximum energy required to
stop in that range at the median zenith angle. The different median angles for
each range are as follows:
instrument range | median theta
=================================
2 | 18.6
3 | 18.6
4 | 17.9
5 | 17.2
6 | 16.4
7 | 15.9
8 | 15.2
The differential flux of cosmic rays reported at the level 2 site is inversely
proportional to the product of the energy band for a given instrumental range
with the geometry factor appropriate to that band. A proper calculation of the
flux requires using the true energy band (with overlapping boundaries) instead
of the characteristic band for a given instrumental range. To account for this
difference, an adjustment was made to the geometry factor for each instrumental
range to keep the product of the energy band and geometry factor the same.
Data are required to pass a very loose charge consistency cut to
eliminate possible contamination of less abundant species by abundant species
which fragment in the instrument and are misidentified. Charge consistency
is defined as the number of standard deviations off of the mean value of
the charge ratios ZEST_LAST/ZEST and ZEST/ZEST_E1 for a particle stopping at a
particular vertical depth. Here, ZEST, ZEST_LAST, and ZEST_E1 are event
charges calculated using different detector combinations to compute dE, E'
(for a description of particle identification using the dE/dx vs. total E
method, see Stone, et al, 1998, Space Sci. Rev., 98, 285-356):
ZEST -- E'=stopping detector, dE=all previous detectors
ZEST_E1 -- E'=second and all subsequent detectors, dE=first detector
ZEST_LAST -- E'=stopping detector, dE=next to last detector
We applied a charge consistency cut on the data to include events with less
than 10 standard deviations off of the mean charge ratio value. Because of
the presence of dead layers in the instrument telescope elements, events
stopping fairly close to the top and bottom surfaces of individual detector
elements will record substantially lesser values of E' for ZEST and ZEST_LAST
than is actually deposited. This will cause the mean charge consistency for
particles stopping near dead layers to change quite rapidly as a function of
vertical stopping depth. To simplify the difficulty in identifying particles
stopping near dead layers, a cut on particles stopping 160 microns from the
dead layer surfaces was implemented.
To convert the particle rates to fluxes, the level 2 site divides rates
by the geometry factor and correction factors associated with the efficiency
of the SOFT hodoscope and the expected amount of loss from fragmentation in
the stack to yield particle fluxes. Flux uncertainties reported in the level
2 site are based on the number of counts reported for a particular window of
time.
Additional systematic uncertainties related to these factors must
also be considered by the level 2 data user who uses version 1.0 data. The
average geometry factor for CRIS has been calculated by two independent
methods, and there is a discrepancy of ~5% between the two for a maximum
zenith angle of 30 degrees. In addition, the trigger efficiency of the SOFT
hodoscope was calculated using two different methods. The difference in the
results between the methods is ~5% for elements with higher Z (>10), and for
elements of lesser charge where the light output of the SOFT varies more
significantly with dE/dx, the discrepancy is somewhat larger (~10%).
Considering these systematic discrepancies, we recommend adding an additional
10% to the statistical uncertainties of higher Z elements and 15%
for lower Z elements to reflect the geometry factor and SOFT
efficiency indeterminancy more accurately. These systematics will be
improved in future versions.