From: Goran Hosinsky 
To:  atm{at}shore.net
Reply-To: Goran Hosinsky 


Is the distance from focus not an important factor? I have learned that
when using filters you should
place them as near focus as possible, just far enough off that you cannot
see the dust on them.
Goran
La Palma

mdholm{at}telerama.com wrote:

>Certainly it is foolish to become mesmerized by primary mirror quality and
>miror cell induced deformations without also considering the quality of the
>secondary mirror.
>
>The argument that the tolerances on the secondary are much tighter because
>there is a higher photon flux at it's surface doesn't pass muster with known
>optical behavior.  The secondary's error budget can be analysed in essentially
>the same fashion as the primary.  The 45 degree angle makes the geometry a bit
>tougher (I still get hung up on it some times.) but the basics are the same.
>What matters are path length differences (and dark areas such as spider
>vanes).  In a mirror system, path length differences come directly from
surface
>errors.  Path length difference to a good approximation is twice the surface
>height error.
>
>The easiest scheme that has a good chance of being near correct for assessing
>the overall performance of the system is to calculate the RMS of the RMS
values
>for each element. i.e.
>
>   RMS total = Sqrt( (RMS mirror)^2 + (RMS cell)^2 + (RMS secondary) ^2 +
>               (RMS tubecurrents)^2 + (RMS other)^2)
>
>This assumes that each of the error sources is not significantly correlated
>with any of the others.  It is possible, though a little unlikely, for
>correlated errors in different components to add or subtract.  That would make
>the system worse or better than calculated using RMS.
>
>A consequence of the RMS calculation is that, if one component has 1/2 as much
>RMS error as another, the sum of the two will be
> Sqrt(1^2 + 0.5 ^2) = Sqrt(1.25) = 1.118
>
>If the two components have equal RMS error the result is
> Sqrt(1^2 + 1^2) = 1.414
>
>So, if you want your secondary to contribute no more than 12% of the RMS error
>attributable to the primary, it only has to be twice as good as the primary.
>
>If the two have equal (but of course uncorrelated) RMS error, the resulting
>wavefront RMS error will only be 41% higher than the primary alone.
>
>If it is easier to make a good primary than a good secondary, and you use that
>advantage to produce a primary twice as good as the secondary, you will lower
>the summed error from 1.414 to 1.118, about a 15% improvement.  15%
improvement
>in RMS might be worth working for.  The next increment from another factor of
>two improvement in the primary probably isn't.
>
>I have said several times that hunting for cell induced deformation less than
>about 1/250 wave RMS is wasted effort.
>
>Mark Holm
>mdholm{at}telerama.com
>
>
>
>
>

--- BBBS/NT v4.01 Flag-4