<div dir="ltr"><br><div class="gmail_extra"><br><br><div class="gmail_quote">On 18 October 2013 13:56, Phil Hodge <span dir="ltr"><<a href="mailto:hodge@stsci.edu" target="_blank">hodge@stsci.edu</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div class="im">On 10/18/2013 08:48 AM, David Berry wrote:<br>
> Since I have come across several SIP headers that contain badly<br>
> inaccurate values for the inverse polynomial, I've followed astropy's<br>
> lead and changed AST so that it *always* calculates it's own inverse,<br>
> ignoring any inverse in the header.<br>
<br>
</div>How do you know it's the inverse transform that's wrong, rather than the<br>
forward transform? </blockquote><div><br></div><div style>That's a good question. I'm basing it on the assumption that in practice people will probably determine the forward transformation directly, and get the inverse from some sort of inversion of the forward transformation. So the forward transformation is likely to be more accurate.</div>
<div> <br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"> It seems to me that if the inverse transform is<br>
specified in the header, that's what should be used<br></blockquote><div><br></div><div>I've come across several cases where the inverse transformation is not a true inverse of the forward transformation. As in the case of Maik's header, if you go from (x,y) to (ra,dec), and then back to (x,y) using the inverse contained in the header, the (x,y) you end up with can be a long way from the one you started with. Clearly such inverses are not very helpful. <br>
</div><div style><br></div><div style>Maybe the thing to do is to do some tests on the supplied inverse, and only replace it if it looks inaccurate.</div><div style><br></div><div style>It may be the case that the inverse in the header is only accurate over some limited domain. </div>
<div><br></div><div style>David</div></div></div></div>