The updated set of response matrices is produced in answer to various messages
including PR # 80, 82, 84, 89, 90 etc..

As with the previous data sets, these have been produced by SciSIM. These
have not been derived by the hardware teams, and the current SciSIM remains to 
be validated by the ground calibration data, as soon as the latter data set are 
analysed.  

Please, please, please, please see the README notes provided with the Jan99 set. 
Many people ignored them - and would have saved a lot of heartache if they had.
Some of the caveats therein apply also to the current set, but see below:

		"If all else fails RTFM"

However thanks to the efforts of the HEASARC help desk colleagues we have now
some updated responses that solve the problems of:
i) Large and inconsistent data channel sets
ii) Incompatibility with SciSIM v2.0.4 data sets.

These response matrices should be usable for fitting the output of simulations
using the latest SciSIM, WITH IMPORTANT CAVEATS YOU SHOULD READ BELOW !!
Read the documentation provided with the QuickSIM/ProcSIM releases at HEASARC
to find out how to extract XSPEC readable PHA files after running SciSIM.

1)
Some minimal testing was performed with a range of trial spectra. For typical 
power laws and moderate temperatures, the fits returned were reasonable. There
is a tendency for the PN fits to return a slightly low N_H and a slightly harder
spectrum than is fed into the simulation. For MOS the N_H is not returned with
good accuracy, while the spectra tend to be reported slightly softer than the
input values. There seems to be a slight normalisation problem in that the MOS
fits are systematically 10% low.

2) 
The mechanism of generation of these response matrices is undoubtedly the cause 
of the discrepancies and we can identify a number of technical issues that could 
create these problems:
	
	(a) To reduce the size of the matrices, low-level probability channels
	that really do arise have been truncated, but in the SciSIM simulation 
	such events ARE reproduced. Depending on the spectrum and the pile-up you
	simulate this means that EITHER incomplete (false low energy) OR piled-up
	(false high energy) events are NOT represented by the matrices. Of
	course we hope that IN-ORBIT there will be exposure-specific RMF 
	generation tasks in software to circumvent these problems - we are
	still working on this ( (;-)  ). It is not possible to provide a
	universal fix at present to cover all GO simulations !!!!!!

	(b) Depending on the huge number of configuration changes that are
	possible with SciSIM you may get conditions that are not representative
	of those that were used to generate the matrices. Two obvious ones we
	immediately know about are the possible loss of photons through CCD gaps
	(which depending on location within an energy dependent PSF can create
	systematic errors), and the effects of charge transfer losses. A nominal
	amount of CTI is applied in SciSIM, but depending on the location and
	extent of your source the actual charge loss simulated may be different
	than the average loss invoked during the matrices' generation. Thus you
	might see the Iron lines occurring at the "wrong" energy in spectra, for
	example.

	(c) The PN data have to be analysed post-SciSIM to associate split
	events. The algorithm used in PROCSIM and for the response generation
	were undoubtedly different.

3) Two PostScript files are included in the directory which show the typical 
results obtained by Steve Snowden (HEASARC) using the new matrices.

Caveat emptor........