1.0 GENERAL
2.0 PROCEDURE
3.0 DESCRIPTION of each flag
4.0 FLAG MATRIX
5.0 SPECIFICATION FOR SMOOTHED ARRAYS
6.0 RECOVERY OF MATCHED SOURCE LIST
7.0 DERIVATION OF BACKGROUND
8.0 SPECIFICATION OF 'ASSOCIATED WITH NEAREST NEIGHBOUR'

1.0 GENERAL

This is meant to expound on the outlined methods of setting flags which were devised at the MPE meeting in September. We have added the source 'a' flag after the MPE meeting in Feb 95.

There are 3 logical steps to setting flags:

1 - Before visual inspection. this is done so as to minimize the work required in step 2. SASS       parameters are used to set some flags.

2 - During visual inspection (VI)

3 - After VI - since some flags will spawn others.

Obviously the step 3 operations can be incorporated into the step 2 process so long as they are 'invisible' to the operator.

Reminder: all flags are F or T, true when something is suspect.

2.0 PROCEDURE

2.1 Step 1
              
All of these are straightforward database operations once the conditions have been decided.

Field Flags:

B            background

              The algorithm for deciding if the BG is high is given in section 7. We generate               a histogram of absolute background levels and then see if the average               exceeds some arbitrary threshold, currently set at level '8'.

D            deferred

               if there are no sources (e.g. in 12" detect), set the Field D flag, deferred.

S            SOURCE LIST MISSING

               check for empty or missing source file

Source Flags:

u            UN-UNIQUE if detection matches detection with smaller cell, set u.

n            NOT CHECKED
              if S/N is < (tbd) 4 sigma, set n

NB: THE HRI/IDL SOFTWARE HAS NOT  IMPLEMENTED THE FOLLOWING; SO WE SUGGEST THAT IT REMAINS 'INOPERATIVE' AND INSTEAD WE DEAL WITH THIS IN THE DOCUMENTATION: therefore, 'n'  means ONLY that s/n <4....*IT DOES NOT MEAN THAT THE DETECTION HAS NOT  BEEN CHECKED VISUALLY!

What we had originally indended was:

              [if S/N is > 4; AND detection is 'matched' with one or more smaller cell detection(s),               all of which has s/n > 4 (i.e. the detection is 'not significant'); then set n=t.]

a            ASSOCIATED WITH (NEAREST) NEIGHBOUR a test involving radius of extended circle and PRF radius if this test is met, set a, and i.

d            DETECTOR STRUCTURE if EPI is > 0.01, set d, i, p and v.

e            EXTENDED if SIZCOR is > 1.2, set e, i, and p

v            VARIABILITY SUSPECT perform the two var tests described in section 3 for the v flag.

Step 1 will also require other preparatory tasks. Among these will be the creation of a blocked array which will then be smoothed with a Gaussian (details given below). Another task will be the recovery of the matched source table (section 6) and the creation of 'new columns' for each source. These new columns will contain the source numbers from the matched source table for each detect cell size. Although this is a bit wasteful of space, it will make retrieval tests significantly easier.

2.2 Step 2 and 3                           Visual Inspection

We suspect it will be most efficient to combine steps 2 and 3 so that the operator will see the ramifications of each decision. This means that flags spawned by other flags should be changed immediately and so displayed.

FIELD: Based on our V&V experience, it is useful to first view the whole field. This may be done with a blocked photon display, although smoothing (see section 5) or expansion may be useful. In V&V, we used xexamine in PROS which allowed zooming in on a region of interest. Current plans are to have a 512 blocked photon display for the whole field AND a 512 smoothed array of the inner portion of the HRI field of view.

The operator should answer the following:

D               DEFERRED are there problems I don't want to deal with now?

A               ASPECT SUSPECT is there evidence that the aspect has problems

E               EXTENDED EMISSION IN THE FIELD Is there substantial extended emission?

M               MISSED SOURCE(S) Did one or more sources get missed? (overlay ALL unique                   sources)

D query should be pretty obvious and quick. The A question is usually manifest by all sources showing the same elliptical structure or by all sources being double. E should be fairly obvious. M is the last of the field queries as it requires the same display as the beginning of the source flags...the overlay of detect positions. The search for missed sources will be done on the two arrays described in section 5.

SOURCE: The primary source flags to consider for n=f, u=f, (but see section 5.4 for precise selection) are:

c                 CONFUSED DETECT ALGORITHM Is there one detection for two close sources?                    Or one (large) box containing complex emission or more than one source

f                  FALSE DETECTION Is this detection not a real source?

m                MULTIPLE DETECTIONS FOR A SINGLE SOURCE Were there (spurious) multiple detections for only one source?

w                WITHIN EXTENDED EMISSION
                   Is the source within extended emission?

If the operator sees a particular problem, the secondary flags i, n, and p may be set. The operator can focus on deciding if the source is real, if it is confused, if there are multiple detections, and if the source is within extended emission. Note that many 'm' sources will also be 'f' sources; i.e. one of the reasons a detection may be false is that it is a spurious detection in the wings of a strong source. Since we now let m spawn f=t, when you flag the real ('mother') source, you should 'undo' the spawned false flag so that it becomes f=f.

3.0 Description of the FIELD FLAGS
----------------------------

A                asp_susp                 Step 2 only                               aspect is suspect

For the HRI, this condition can only be easily spotted if there is a sufficiently strong source which is seen to be elliptical whereas it is known (apriori) that it is unresolved. OR If there are two or more sources which have the same apparent ellipticity. Extremely pathological cases of double images etc will also require setting this flag, but these are expected to be rare.

Determination of bad aspect by VI is perhaps the most difficult process in flag setting. We do not wish to flag aspect when there is a small residual ellipticity caused by wobble (usually < 5"); therefore, if ellipticity is found, the magnitude should be estimated as >5" before setting the A flag.

There is also a potential ambiguity between the field "suspect-aspect" flag A and the source "confused" flag c - what's the difference between a confused source (i.e. two real sources 3 to 10" apart) and a source that appears as a close double image due to bad aspect? If there is only once source strong enough to see elliptical or double structure, there is no obvious method of determining the actual situation.

Finally, if there are no strong sources near the field center, it will be difficult to see any evidence of bad aspect.

Subsequent Action: Whenever this flag is set, all source flags i,p,v are automatically set. ---------------------------------------------------------------------

B              bkg_susp               Step 1: test of background level
                                                 Step 3: if E (extended emission) flag is set.

In step 1, the algorithm (section 7) will automatically set B. If the extended emission flag (field flag E) is set by the operator, B gets set automatically.

Subsequent Action: none. ---------------------------------------------------------------------

D            deferred                   Step 1: if SASS failed running detect
                                                Step 2: the usual method

If, for any reason, the operator believes that a proper treatment would take too much time, or new software is required etc, the field can be deferred.

Subsequent Action: Whenever this flag is set, all source flags n are automatically set. ---------------------------------------------------------------------

E            ext_emis                   Step 2: substantial extended emission

This is subjective, and will normally be set for clusters of galaxies, supernova remnants, etc. It requires enough extended emission to affect source measurements and background calculations. As a guideline, we recommend that the size of the extended emission exceeds the size of the largest detect cell (2') by a factor of two or so, i.e. the emission patch be of order 5' or greater.

Subsequent Action: Whenever this flag is set, field flag B is automatically set. ---------------------------------------------------------------------

F           fals_det                    Step 3: f source flag set for >= 1 source.

Whenever a source is "deleted" with the false detection flag, this field flag is set as a general field indicator. Subsequent Action: none ---------------------------------------------------------------------

M          src_mis                    Step 2: There are one or more missed sources

In order to avoid numerous fake (marginal) sources, this flag should only be used when it is very obvious that a source is missed by detect. There will always be sophisticated algorithms that can find probable sources not detected by SASS, but that is not what we are concerned with here. A moderately blocked field will probably be the best method for this test. The operator will click the mouse on these missed sources and thereby record the position and some measure of the counts.

Subsequent Action: none ---------------------------------------------------------------------

S          slist_mis                   Step 1: check for missing or empty source file

Subsequent Action: Normally (tbd) the D field flag should be set. ----------------------------------------------------------------------

3.2 SOURCE FLAGS
----------------

a            assoc_nn               Step 1: Source associated with a close neighbour

This flag was invented for the PSPC during construction of the original MPE CAT. For the PSPC, it involved a test to see if two sources were within a certain distance of each other. This distance depended on the PRF (function of hardness ratio and off-axis distance), the extent parameter, and a fudge factor determined by trial and error. Details for the HRI are described in section 8.

Subsequent Action: set the i flag. ---------------------------------------------------------------------

c          confused                  Step 2: detect confounded by close sources

This is evident only by visual inspection, as e.g. two close sources with only one detection between them. Or if a large box contains more than one source, or complex emission.

Subsequent Action: Whenever this flag is set, also set i, p, and v. ---------------------------------------------------------------------

d         det_struct                 Step 1: set if EPI is > 0.01

EPI is the edge priximity indicator. It is zero if the source is clear of the edge or detector hotspots. APR95, set threshold to be 0.01 instead of zero. Values of 10^-4 were found, and 0.01 was deemed a reasonable value for d and spawning i, p, and v.

Subsequent Action: Whenever this flag is set, automatically set i, p, and v. ---------------------------------------------------------------------

e         extended                  Step 1: set if SIZCOR > 1.2

Note that we have changed the original threshold (>0) since on inspection, SIZCOR is often just above one, but not significantly so.

Subsequent Action: Whenever this flag is set, automatically set i, and p. If the VARIABILITY parameter is "POSSIBLE", then the v flag should be set. ---------------------------------------------------------------------

f          false                         Step 2: source reality is doubtful Step 3: if the 'm' flag is set; set f=t

'This can be determined only by visual inspection.' was the original specification. However, for consistency with PSPC practices and as an efficiency improvement for the operator, f=t is set whenever the operator sets m=t.

Subsequent Action: Whenever this flag is set, set the Field F flag. ---------------------------------------------------------------------

i          inten_susp              Step 1: if Source a or d is set
                                            Step 2: possible to set by visual inspection
                                            Step 3: if Field A is set; if Src c, e or w is set

The normal mode will be the automatic operations of steps 1 and 3; however there may be other conditions.

Subsequent Action: none ---------------------------------------------------------------------

m         mult_det                 Step 2: multiple detections occurred

Multiple detections occurred for a single source. For the HRI, this condition will normally be understood to mean 'for the same cell size'. Occasionally however, a larger cell detection is far enough removed in position so that it is not matched with the small cell detection even though by VI, it is obviously the same source. In such a case, both detections get the m flag, but the erroneous position detection also gets the false flag set. The VI operator should pay attention to the position flag whenver he sets m. As a convenience, the IDL implementation automatically sets f=t whenever m=t. Therefore, for the detection which corresponds to a real source, the operator must 'undo' the f=t flag.

Subsequent Action: set the false flag (this is for operational convenience only) ---------------------------------------------------------------------

n         not_checked           Step 1: If s/n of detection is < 4 sigma.

                                             Step 3: If Field is Deferred

The question of s/n threshold has not been addressed definitivly. We can reduce the amount of VI operator effort by not examining the large number of detections near threshold; these detections are difficult to evaluate and would probably be the cause of significant dispersion in the results since subjective judgement would be required. For the HRI, the SAO group agrees that the S/N threshold should be set at 5 sigma, but this may be revised downward. * During testing in early April 95, we decided to try with s/n = 4 for threshold. Strickly speaking, n means ONLY S/N < 4; rather than "not visually checked".

Subsequent Action: none. ---------------------------------------------------------------------

p         pos_susp               Step 1: if source flag d is set
                                           Step 2: may be set by VI
                                           Step 3: if Field A, Source c, d, or e is set.

For the most part p will be set by other flags: field A, and source flags c, d, and e. It may also be set manually during VI.

Subsequent Action: set the 'i' source flag. ---------------------------------------------------------------------

u          un_uniq                Step 1: based on SASS detection matching

We will recover the matching of numbered detections with the different cell sizes, which produces letter designated sources. Then, whenever a detection is matched with a detection made with a smaller detect cell, the U flag will be set. This is the only method of recovering a unique source list (i.e. in the database retrieval, do not take detections with the U flag set to True).

There is, however, a complication when a S/N threshold is set for the N flag. For sources that are extended or off-axis, a S/N < threshold may be obtained for the 12" detection whereas the 24" (or larger) detections will often have S/N > threshold. Therefore, the U flag by itself can not be the sole criterion for desirability.

Subsequent Action: none ---------------------------------------------------------------------

v         varflg_susp            Step 1: compare VAR for source and BG; if Field A or Source d is set
                                           Step 3: if source flag c is set

One needs to make a few tests of the VAR parameter which has the possible values: POSSIBLE               probability that source is constant < 0.01
NEGATIVE                                                                            >= 0.01
TOO_FEW               less than 50 counts
NEAR_EDGE          OFFAX > allowed max (probably 12')
Note that VAR is generated only for detections with the 12" and 24" detect cells.

a) If the e source flag is set, and VAR=POS, then set the v flag.
b) If VAR(src)=POS *AND* VAR(bkg)=POS, then set the v flag.

Subsequent Action: none. ---------------------------------------------------------------------

w            within_ext             Step 2: VI

This flag describes a discrete source embedded in extended emission, e.g. a pulsar within a SNR or an AGN in a cluster. Note that while w implies the presence of extended emission, it may not be so strong or extended as to warrant the setting of the field E flag. It is also possible to have a strong unresolved source in weak extended emission so that w is set, but not the source flag e.

Subsequent Action: Whenever this flag is set, the 'i' source flag should be set. ---------------------------------------------------------------------

4.0 FLAG MATRIX

The field flags are upper case, source flages lower case. An "x" in a given position means the flag for that column should be automatically set whenever the flag for that row is set, i.e. the flags along the left border spawn the flags in the indicated columns.

                                                      THEN....------->

                           A    B    D    E    F    M             S    a    c    d    e    f    i    m    n    p    u    v    w .........................................................................................................................................................
IF...             A                                                                                         x                 x           x
                   B
                   D                                                                                                      x
.........................................................................................................................................................
                   E           x
                   F
                  M
.........................................................................................................................................................

                  S
                  a                                                                                          x
.........................................................................................................................................................
                  c                                                                                          x                  x            x
                  d                                                                                          x                  x            x
                  e                                                                                          x                  x            x?
.........................................................................................................................................................
                  f                                x
                  i
                 m                                                                                   x
.........................................................................................................................................................
                 n
                 p                                                                                          x
                 u
.........................................................................................................................................................
                 v
                 w                                                                                          x

The x? above means set the v flag if variability was labelled as "possible" in the SASS output. m spawns f is purely for operational convenience

5.0 SPECIFICATION FOR DISPLAYS

For the visual inspection (VI), a smoothed array is required because with only photons, weak sources are difficult to evaluate. Since we do not want to loose much resolution (to sort out close sources) we believe that two 512 arrays are required: one for the central part of the field with a resolution close to the instrumental, and the other of the whole field, with a lower resolution which will still be adequate for the outer parts of the field. To improve the S/N ratio, we choose PHA channels 1-8 since 9-15 is mostly noise.

5.1 High Resolution Map

We choose a blocking of 4, which results in 2" pixels, and a FOV=17'. To get 3 pixels per beam, we choose a Gaussain with sigma=1.27 pixels (2.55") which corresponds to FWHM= 6.0". However, during testing, we found that this function was too small....the results were very similar to the blocked photon array. Therefore we changed the recommendation for Gauss: sigma = 1.7 pixels (3.4"), or FWHM=8.0". With a blocking factor of 4, the original size is reduced to a 2048, and to get 512x512 around the center, we choose 769:1280 (for the IRAF designation of section).

.......following paragraph to be deleted once testing verifies above.....
For rh150004, Cen A, even when the 90 detections were reduced to 12 unique sources with S/N > 5, the jumble in the center precluded evaluation with the standard blocked image/overlay. Only when a smoothed image and contour diagram was constructed (block=3, 1.5" pixels, Sigma Gaussian = 2 pixels, FWHM circa 7") which was a 512x512, were we able to evaluate the features in the galaxy and jet, and find a missed source. We suggest that such a smoothed array be made for each field to be used in source evaluation during VI. Which of these two choices is tbd.
..........................................................................

5.2 Whole Field

Experimenting with fields rolled close to 45 degrees, we found that a blocking factor of 10 (i.e. 5" pixels) was satisfactory. Current thinking is to have this be a photon array, i.e. not smoothed. However, if smoothing is to be implemented, similar considerations as in section 5.1 lead to Gauss sigma = 1.27 pixels (6.37"), for a FWHM=15" (3 pix). The section specification is [155:666].

5.3 Graphic Plane overlay

We require a somewhat complex coding for the source position overlays. First, we want to identify the unique sources which have s/n > 4. These should be marked in a prominent colour (e.g. cyan) with either the smallest detect cell box with s/n > 4 OR (tbd) all detect cells with s/n > 4. Next, we need to mark those sources with s/n < 4 with a less prominent colour (e.g. blue). Although operators will not normally examine these sources, they need to be clearly indicated so that they will not be considered to be 'missed sources'. Until evaluation is performed, we expect that these overlays will be useful on both displays, and should be suppressable by the operator.

5.4 A note on selecting sources

For both eventual default database retrieval and for choosing which sources to mark for the operator, we will need a combination of the source match table, n and u flags. In discussions with MC, we assume that the source match table (retrieved as per section 6) will be incorporated into the source table for easy conditional tests. This means that there will be 5 additional columns for each source, with entries being source numbers. The 'best' source list will consist of s/n > 4, which, for us means n=false (thereby also leaving out sources from deferred fields) AND u=false. However, if u=true, it is imperative to check (test) to see if the smaller detect cell has n=f. If it does not, then we will accept the source in question even though u=true.

6.0 RECOVERY OF MATCHED SOURCE LIST
filename: matchsrc.doc
Date: 18 nov 94
originat: deh - from JC email
tlm:
Subject: INSTRUCTIONS FOR RETRIEVING MATCHED SOURCE LIST FOR HRI

6.1.0 Get the ...trl file from the FITS

6.2.0 Run the AWK script:

To use this task, just type the command: "matchsrc input_filename". If the input_filename is missing, the program is smart enough to ask you to re-try it. The input file should be same as OUTPUT.trl in ROSAT HRI PRT FITS file, but you don't have to name it as "OUTPUT.trl".

Two output files are created :
           matchpg.txt - the pages of the match source summary
           matchnum.txt- the list of the match source numbers except the first left one. It is                                      in a numerical order with one column in each row, as you                                      requested.

Also, each output file contains "Sequence Number", so the users know where the data came from originally.

For the RRA, one can either parse the ..pg.txt file if you need complete information (which sources go with which), or if you just want to set the "un-unique" flag, you can use the list of detection numbers in the ...num.txt file.

6.3 The AWK script

# ******************************************************************************************************
# Module: matchsrc
# Project: Rosat Results Archive - SAO RSDC
# Purpose: Capture the pages of the match source summary and
# the source numbers for which a detection has been
# made with a smaller detection cell.
# Author: JCC at SAO - Nov. 1994
# Updated: {n} <who> -- <does what> -- <when>
# ******************************************************************************************************
rm -f ttmpdmp*
nawk '
         BEGIN { if (ARGC != 2)
                  {print("*** Please use the command: *** ")
                   print("*** matchsrc input_filename *** ")
                   exit}
        }' $*
# capture the pages of the match source summary
rm -f tmp
echo $1 >> tmp
root=`cat tmp`
echo $root |
           awk '
           {  {if (NR==1) { {if ($1 ~ /^$/) exit }
              ii = 0
              tii = 1
              print ("*****************************")
              print ("Possible Source Match Summary")
              print ("*****************************\n")}}
          {if ($1 ~/Sequence/)
              print (substr($0,9,length($0)), "\n\n")}
         {if ($1 ~/Possible/ )
              {ii = NR
               tii = NR}}
        {if ((ii>= tii) && ($1 !~ /composite/))
              { print substr($0,11,length($0))
                ++ii}}
       {if ($1 ~ /composite/)
             { print substr($0,11,length($0)) exit}}
       }' $root > matchpg.txt

rm -f tmp
# read the numbers of match sources and remove the dashes
awk '
        {if ($1 ~ /Sequence/) print }
        {if ($1 ~ /^[A-Za-z]$|^[A-Za-z][A-Za-z]$/)
             { $1=""
                 {for (jj=2; jj<=NF;++jj)
                         {if ($jj ~ /^----$/)
                              {$jj = "" }}
                         print $0 } }
       }' matchpg.txt > ttmpdmp
# write the numbers of match sources
awk '
        {if ($1 ~ /Sequence/) print }
        {if ((NF >= 2) || ($1 !~ /Sequence/))
           {for (kk=NF; kk>=2; kk=kk-1)
            {kkk = kk - 1
              {if ($kkk ~ /^[0-9]+$/)
               {print $kk} } } }
        }'       ttmpdmp > ttmpdmp2

# remove duplicates and write it in a
# numerical order
awk '
       {if  ($1 ~ /Sequence/)
          {print}
       else {
         {source[NR] = $1
          {for (jk=NR-1; jk>=1; jk=jk-1)
           {if (source[NR] == source[jk])
             {source[NR] = ""
               break}}}
           {if (source[NR] !~ /^$/)
                 {print source[NR]}}
      }}}'     ttmpdmp2 | sort +0n -1n > ttmpdmp3

# output the results to matchnum.txt
awk '{
       if ($1 ~ /Sequence/) {
         print ("\n### ",$0,"\n")
         print("***************************************************************")
         print("SOURCES for which a detection has been made with a smaller cell")          print("***************************************************************\n")
         }
        else
           print
        }' ttmpdmp3 > matchnum.txt

rm -f ttmpdmp*

7.0 DERIVATION OF BACKGROUND

The HRI background, when high, is suppressed by SASS. However, it may be useful to have a user warning flag which will alert them to total high rates.... be it from anamalous background or estended emission from SNR or clusters. We do not have the experience to decide exactly what threshold to use, but suggest a value of 8 (units are 10^-7 cts/pix/sec). This quantity exists in the .TSI record.

A possible alternative is to make some test on the histogram of background. However, if we use just the time averaged value, there are two ways to get it. One is to recover from the ASCII printout/copy. It comes from SASS output in slightly different units...cts/sec/sqarcmin. The other is to use the script written by Primini.

filename: instr
Date: 13 dec 94
originat: fap/deh
tlm:
Subject: INSTRUCTIONS FOR FINDING THE AVERAGE BACKGROUND FOR HRI
From fap@cfa239.harvard.edu Mon Dec 12 14:39:25 1994

A shell script called hiback.sh which runs fdump and awk to determine the average background level in an HRI observation from the TSI records in the _bas.fits file. Eventual use should probably redirect output to a file or...??

If you don't have ftools, you can probably find another way to get the appropriate columns from the TSI list.

a command such as:

   hiback.sh /proj/rev0/mo/rfits_3_0_new/rh110267/rh110267n00_bas.fits

will produce a histogram showing the time at each level. 1 is low, 14 or so is high. Units are cts/sec/pixel

2 557.998
3 740
4 430
5 120
6 150
7 110
8 10
1 30
Average Background Level: 3.46183

Our initial guess is that if <level> is GT 8, it should be flagged. However, it might be that a smaller value is warranted, or one might like to look at the behaviour of the histogram. We do not have a good sense of how these look for many sequences.

Here is the script 'hiback.sh'

fdump "$1[4]" STDOUT "TIME,FAILED,HIBACK" "-" prhead=no page=no | awk -f hiback.awk

Here is the AWK, 'hiback.awk'

# awk program to compile histogram of high background levels in TSI
# file.
#
# 11/21/94 fap
#
BEGIN{
             TIME = 0.0;
             FAILED= 1;
            HIBACK= -1;
            TOTAL = 0.0;
}
NR>3{
            deltaT = $2 - TIME;
            if(FAILED==0){
                     bkhist[HIBACK] += deltaT;
                    TOTAL += deltaT;
                     }
                    TIME = $2;
                    FAILED = $3;
                    HIBACK = $4;
}
END{
             for (bklev in bkhist) {
                    print bklev, bkhist[bklev];
                    avbk += bklev*bkhist[bklev]/TOTAL;
}
             print "Average Background Level: ",avbk
}

8.0 SPECIFICATION OF 'ASSOCIATED WITH NEAREST NEIGHBOUR'

The rationale for the 'a' source flag comes from the MPE experience with the PSPC. We envisage two close detections (with the same size detect cell) which, although fulfilling all SASS tests to qualify as separate sources, may, in fact, be close enough to affect the determination of intensity of either or both. Therefore, we require an algorithm which will identify 'associated' sources. Although it is conceivable that one of a pair of associated sources may be flagged (by the operator) as a false detection, we choose to leave the automatically generated 'a' flag. Hence the definition to the user should specify that "two detections were found....", (not, "two sources were found...") and in the spirit of "intensity suspect", the user may demonstrate that the SASS intensity is, after all, reliable in such a case.

8.1 Input parameters

The basic parameters for the HRI will be the 50% power radius of the PRF(theta) and the radius of the extended circle. We realize that the latter quantity is not a reliable measure of the extent of a source. However, it appears to be the only available measure of source extent.

To parametrize the PRF we have two options. The first would be to solve for the 50% power radius as given in section 3.2 of the HRI CAL report. We could double this number to get a radius that was closer to 90%. Alternately, we could use the SASS parameter, CRAD, which, according to the RDF/Rev2 output, purports to be the 90% encircled energy radius (the DPG defines CRAD as 3.0 * sigma of PRF). Because of the uncertainty as to the genesis of CRAD, and because it appears to be too large near the field center (20"), we opt for 2*50% radius.

The HRI report gives:

r(50)=(2.35/2)SQRT{sig(hri)^2+sig(asp)^2+[sig(mir)+a*theta^b]2} arcsec

with             sig(hri)= 0.74 arcsec
                    sig(asp)= 1.0
                    sig(mir)= 1.3
                              a = 0.0205
                              b = 2.349
                    theta is the off-axis angle in arcmin

Figure 9 of the HRI report plots this function, showing that it goes from 2.1" (center) to approx 25" at the edge of the detector. Since we recommend twice this value, the leading coefficient simply becomes 2.35

For the extent parameter, we will use the SASS parameter, REC, radius of the extended circle (which encloses all contiguous detection cells above threshold). Unlike the PSPC, this is a raw observable, not a deconvolved estimate of true source extent.

8.2 Derived Parameter

Therefore, in order to determine if a 'nearest neighbor' is likely to disrupt the intensity measurement of a source (counts within CRAD), we define ARAD (association radius) as:

ARAD = 2 * r(50)

If SIZCOR < 1.2 (i.e. no evidence of extent), this will be the only new parameter for this source. However, if SIZCOR >= 1.2, then we define a new circle with center XEC(L2) and YEC(L2) and radius REC. The reason we must maintain REC with its appropriate center position can be seen by considering a point source near the edge of a large cluster of galaxies. For a discrete (external) source which lies outside the cluster, but not very far from the embedded point source, we do not want to erroneously assign 'a' to the external source.

8.3 Procedure for setting the 'a' flag

There may be better methods of achieving this result, but the following 'brute force' approach has the advantage of simple coding and few conditionals.

1) For each source, generate the new parameter ARAD, as specified above.

2) For each source (s), perform the following test with every other source (i) detected with the same detect cell size.

a) If 'a(s)' flag is true, (exit).
b) Calculate the distance, d, between s and i (convenient to use L2 pixels).
c) If d < ARAD(s) + ARAD(i), then set both s and i aflag to true. (exit)
d) If SIZCOR(i) >= 1.2, AND distance between s and XEC(i), YEC(i) < REC(i) + ARAD(s), set      aflag(s) = true; (exit)
e) exit

(exit) means go on to the next source.
....................................................................... end ..........................................................................