Re: ATM aargh! maybe

["Nils Olof Carlin" <nilsolof.carlin@telia.com>]

Mark Holm

> >> If you were already much better than 1/4 wave why did you keep
going? Isnt 1/4  the theoretical best?
> >
> > The 1/4 wave value comes from a statement made in the 19th century
by an English  physicist, Lord Raleigh.  He said that if the path
lengths
of the waves arriving  at the final focus differ by less than 1/4
wave, that
one will not be able to tell the difference between that image and a
perfect
one..........................
> > 2. Since Lord Raleigh, both quantitative theoretical work and
careful observation by a number of well qualified astronomers, both
amateur and professional, and opticians, both amateur and professional
has shown that Lord Raleigh was not quite correct.

While I haven't seen it first-hand, I recall seeing a quotation from
lord R. that [regarding low order spherical aberration] 1/4 wave P-V
wavefront error was when the performance was getting "decidedly
prejudicial" - so it was not *his* idea that 1/4 wave of sph abb would
pass for perfect, and he shouldn't be blamed for later
misunderstandings of what he really claimed. Considering what he said
(as I know it), I'd claim he *was* correct.

> >
> >
> > 1. The statement refers to the final image, since light traverses
the path to  the mirror twice, once before reflection and once after,
defects in the mirror surface are multiplied by 2.  Thus the 1/4 wave
rule becomes, for mirrors, the 1/8 wave rule, but everbody still calls
it the 1/4 wave rule.

Let me once again plug for the concept of RMS error - any useful data
reduction program does calculate it. It is by no means less real than
the P-V, and it is a far better predictor of performance. Also, why
not give it in nanometers instead of the convoluted fractional
wavefront (wavelength not specified, but 550-560 nm usually) error?
It shouldn't be difficult to memorize - the old 1/4 wave criterion (of
low-order sph abb!!) corresponds closely to 20 nm RMS error - a
simple, round number. 400 square nanpometers of mean square error
would be even better, but that can wait ;-)

> > 3. The Foucault test and the data reduction program are not
perfect.  It is not possible to perfectly characterize the mirror
surface
using, in my case, only 8 data points.  To be safe, I set my goal a
bit
under the 1/8 wave standard so that errors remaining undetected
would be likely to be less than 1/8 wave at the image.  Hence 1/10
wave at the image, or 1/20 wave at the mirror.

The testing technique is fraught with imperfections - hardly
surprising, what *is* surprising is that it is possible at all to
measure the surface to a fraction of a wavelength of light, and fairly
reproducibly and reliably.
> >
> >  I can tell you from experience that a lot of pits remaining
> > from an incomplete polish, or trouble in fine grinding, will
degrade the image more than 1/4 wave figuring errors.  The
degradation is different. Pits scatter light all over the image,
figuring errors scatter it nearby. With pits the whole image
seems hazy and the background not as dark as it could
be, with figuring errors, bright stars don't look as sharp as they
could, and fine lunar and planetary detail is degraded somewhat

Seems to me that if you have the choice between pits and figuring
errors, you can choose a good, dark background with poor planetary
detail, or  a slightly brightened background with contrasty details.
I'd prefer the latter, even if you don't. What do others think?

Nils Olof