A16.6.1 Introduction

First of all, a black mark to all of us for not having noticed what is to follow earlier than we did; we've all been trusting that the computer was getting it right without making even simple cross checks. As a result of this we have been working for 4 1/2 years with incorrect energy bands in the track files.

Some little while back, George and I were looking at some tracks flux spectra and decided that we needed to modify the track definitions to eliminate the pile up of crud in (typically) energy bin 12 due to the turnaround of the tracks as particles penetrate the C-detector. When looking at some very long-accumulation PHA matrices to decide where best to trim the tracks, it became clear that at the higher energies many of the tracks were a poor fit to the data. Therefore the obvious thing to do was to attempt to define a new set of tracks.

So far so good: having generated what seemed to me a suitable track file, I sent it to Carol for comments. Then the problems began! Carol noted that the energy bin numbers in the new track file were consistently 1 less than those in previous track definition files. After a number of cross checks we came to the conclusion that the new bin numbering is correct and that we have been using incorrect energy bin numbers ever since we started using track data.

A16.6.2 The New Track Files

The initial "new" track file (hereafter lantracks008) and its corresponding centrelines file centerlines8 were defined by fitting by eye to a PHA matrix for the first 400 days of the mission. This used a spline fit to user-selected nodes, with the option to try again if the fit was unsatisfactory. Having thus defined the centreline, a width in D energy was selected to define the track. Determining the energy bins is done simply by calculating the energy/nucleon for the species under consideration, using the C and D energy relations (for an unrolled matrix):

EC(eV) = 10^{0.015625(C-14)+5} for C<45 
EC(eV) = 10^{0.0122(C-4)+5} for 45 <= C<88 
EC(eV) = 10^{0.015625(C-23)+5} for C>= 88

ED(eV) = 10^{0.0117647(D-52)+5}

and then establishing in which energy bin the total energy lies.

Having defined this raw tracks definition a number of refinements were then made:

1. A triangular region in the low-D high-C corner of the matrix was eliminated to ensure a uniform elimination of the penetrator turnaround from all species.

2. All bins containing less than 10 C-energy D-energy pixels were eliminated as these are bins in which there is just a tiny corner of the track, for which it would be difficult to obtain a reliable flux.

(Figure A16-13 - not available)

That these tracks fit the actual tracks in the matrix can be seen by examining histograms of the counts summed along the track in the C direction as a function of displacement in D from the centreline. This is illustrated in Figure A16-14.

Figure A16-14 Histograms for each track of the number of counts as a function of displacement in D from the centreline. The track file is lantracks008, the data are for the first 400 days of the mission. The dashed line indicates the extent in D of the track.

Unfortunately, when lantracks008 is used for data from more recent intervals, the tracks on the matrix clearly lie below the tracks as defined by lantracks008. For the 1993-1994 interval, it appears that this shift is very close to 1 pixel in the D direction. It seems that the best solution to this is to generate a new track definition lantracks009 which has all the D locations shifted downward by 1. In generating lantracks009 it has been assumed that the cause of the shift is a reduction in the actual energy deposited in D, rather than a change in the translation of D-pulse to energy. As the shift is only 1 pixel, the difference is probably not significant, but it does affect the energy/nucleon and hence the energy bin (in essence it will shift the bin boundaries by 1 pixel).

A16.6.3 Software Changes Needed

As it now seems that we need two lantracks files, we need to consider the implications for the software displaying tracks data.

The main problem is in the count-rates to flux conversion in the extremal bins of the track. At present this issue of partial bins is handled in two different ways:

comp: For those bins where only a small fraction of the bin is occupied by the track there are a set of "hard-wired" energy-range corrections.

idl_hs: The problem is ignored.

A more general solution seems to be to use a file trkbins00n.dat which contains the effective upper and lower energies of each track. (This file is an extension of the trkbins00n.dat file already used without the number extension in idl_hs to indicate the highest and lowest bins in each track.)

The only difficulty with this is defining the effective energy limits of the tracks. Generally the upper limit is relatively simple, and most reasonable methods give very similar answers because the energy is dominated by the C-pulse and so the energy contours are nearly parallel to the D axis. However, for the lower limits the energy contours are very oblique, often almost parallel to the C axis, which means that simple solutions like the energy of the low end of the centreline or the lowest energy in the track are not good measures. The solution I have adopted as a preliminary attempt is to use the mean of those points with the extremal C-pulse which lie in the extremal bin (see Figure A16-15). This seems to work reasonably well with the possible exception of the magnesium track.

Figure A16-15 Schematic illustration of the determination of the effective minimum energy of a track.

From the user's standpoint this means specifying a PHA map number to comp and to idl_hs when requesting track data be plotted.

Status of software: I have made the necessary modifications to the PHA software, and the resulting PHA version 4.02 is now believed to be functioning correctly. The spectra generated by comp are now much more believable than anything I have previously seen from it. This version also restores a number of fixes that Carol and I had made to PHA version 3 which had gone missing in the original version 4. This will be made available shortly. (I need to sort out the VMS version and to check that the relabelling of He3 as H3 is complete.)

I have a functioning development version if idl_hs (V3.08) which incorporates these changes, and need to check it for proper interlocking of settings before releasing it. At least both packages agree about the values of the track fluxes, and I believe that the idl_hs conversion is correct in that when I convert rates by hand I get the same answers.

A16.6.4 Implications

What are the implications of this for previous work?

It seems to me that this can be divided into three categories:

Minor changes: Many of the papers published using track data use mainly composition ratios between different species at the same energies. These are only affected insofar as the quoted energy bands are incorrect; e.g., a ratio of (say) 0.5-1.0 MeV/nuc He/O will now be 0.354-0.707 MeV/nuc. In most cases this will not affect the science. (There is one exception to this: for any work involving high-energy heavies-especially Sulphur and Iron-the actual rates and hence fluxes are in error due to the tracks being in the wrong place.)

Significant changes: Any work using actual flux levels and/or spectra will be rather more seriously affected, because not only will the energy range be wrong as a label but the wrong range will have been used in calculating the flux. Work looking at temporal changes in fluxes should be salvageable provided it doesn't rely on absolute values of flux or energy.

Invalidated: Anything involving the identification of features in the spectrum above around 3 MeV/nuc is now very shaky as the old band 11 (now reclassified as band 10) had not only the wrong absolute energy but also the wrong fractional width which will introduce a spurious feature into the spectrum. Also anything above about 7 MeV/nuc (formerly listed as about 14 MeV) (for mediums; less for lights and more for heavies) is dominated by penetrating particles. Any work relying on fluxes or spectra at high energies needs to be redone.

Actions: In view of the potential seriousness of this error there are two actions that should be taken:

1. All team members should check work that they have taken a major part in to see how the conclusions are affected.

2. The persons responsible for installing software should install PHA 4.02 and IDL_HS 3.08 as soon as possible after release. In the meantime, put any current work using track rates/fluxes on ice. 

Return to HISCALE List of Appendices

Return to Ulysses HISCALE Data Analysis Handbook Table of Contents