<p>It's an old thread, but I might as well put the data here as this comes up with Google.<br>
<br />I measured (or tried to) the pupil ratio of Canon EF 100/2.8 L at five magnification settings. It seemes that 0.94 -0.61 * m is a reasonable approximation. The "accurate" curve levels at ends and is a bit steeper in the middle, but not enough to matter for DoF calculation. I wouldn't mind if somebody checked the end values. If they are about right, the others likely are as well.</p>