ATM Re: short f# spheres

[epabcc@epa.ericsson.se (Bratislav Curcic)]

Hi Mark,

> I'm in a niggly mood so I'll disagree with the above ;)

_You_ are in a niggly mood! I thought that after the latest spell of
absolutely awful weather we have in Melbourne I had an exclusive right to
bad moods ...

> I haven't  seen any evidence that faster f# are easier to polish spherical. 
> My evidence for this is via interferemtric testing, during polishing. If
> you were performing say a ronchi or foucault test on them you may be led
> to beleive mistakenly they are good spheres because the sensitivity of
> ronchi and foucault is dramatically decreased with short focus mirrors. I
> have seen `fast' mirrors that looked good spheres with a casual inspection
> with a ronchi but proved pretty horrible by interferometer. As a 
> dramatic example a mirror with an f# of 2 at the Center of Curvature  
> will have a knife edge shift of 1/36 that of say an amateurs F6 telescope 
> primary tested at its COC, for any given amount of wave abberration...
>    

Hm. It seems you caught me on the thin ice here. I haven't thought
about dramatical reduction of Ronchi/Foucault tests at fast f/ratios.
Maybe we've found the source for a common myth that secondary
"magnifies" primary's errors ? If primary wasn't properly tested in the
first place, errors will be much easier to see at higher f/ratio.

> Have you any mechanical explanation why a steep curve should polish more
> spherical? 

Well, sort of. Low prices of instruments based on a fast spherical
primary (SCTs) may be one indication. From my limited experience it
looked that was much easier to get to a good sphere whith short
f/ratios. But that doesn't hold much water because of what you just
said, and I never used these as spheres in final instrument anyway.
Friend's Maksutov/Casegrain had to be refigured by definition (residual
spherical was too large at f/16), so it is difficult to say how much of
that refiguring was to cover for (still detectable at foucault!) errors 
on the primary.
If we say that we can detect 1/100 wave defects with Foucault used as a
null test at "normal" f/ratios (not that impossible), at say f/2  
detectable error represents ~1/11 wave ... So seems that we _can_
still figure good f/2 sphere, but it has to look absolutely flawless
with Foucault. And that _isn't_ automatically acheivable even at f/2, I
admit ...

> I find steep curves harder to polish out, whereas I can hold a
> flat within a fringe of absolute flat for hours with the right stroke, and
> mantain an even polish all over. I agree that it is easier to hold a steep
> curve to a given radius of curvature tolerance, as it is harder to change
> the radius, but this is not the same thing.

You can change radii of deep spheres quite rapidly in one direction - make
them longer, by reversing tool and mirror. Shortening goes much slower.

Bratislav