Telescope Reviews: How many points for a primary cell ?

Yes, that's the way it's implemented. Changes in radius of the support circles and angles are all that's allowed. I ran some comparisons between about 5000 iterations and 500,000 and there was little variation in maximum error.

There's nothing to prevent you SFAIK from setting up a ring for every support point except the time it would take to define the cell.

Regarding the use of (solid) pads, no it doesn't affect the support much at all so long as they're centered. Pads will have to conform to the glass though. You can get an idea of this by running the 3D PLOP option after you do a cell, it models the glass more accurately and generally returns lower overall errors by about 50%.

Hmm, I got 14.6nm RMS surface. Same specifications for size, modeling an f/100 plate in PLOP. 3D PLOP shows 11nm RMS surface (see below), and it's nearly identical to Mauro's FEM result for the meniscus! Something is not correct somewhere in this line of thought then.

Unless (of course) there is in reality negligible surface change due to the shape itself - which was my original hypothesis and one that I still mark as the winner for actual mirror support.

This is why I was going to deploy a 27-point cell on the 1/3" 14.7" mirror in my first design (well more than good enough, 1.6nm RMS surface at least). But to test Mel's observation I'd put it on a simple fixed-angle 9-point first, planning to upgrade to 27-point as a further experiment, which would bring me back to my original design. The cost of doing the experiment is pretty small, and I'll know the cell isn't doing any harm by the end.

Also note that the error on the 9-point should be easily visible in star testing, if it performs as expected. And by building two interchangeable cells (they just bolt to the back of the collimation supports I can field test one against the other as needed.

So a tentative conclusion here is that Mel's reasoning is incorrect somehow - either the observation, the actual effect of cell support on thin mirrors (how many people TRY a 3-point cell on a 13.2x1"?), or the conclusion based on equivalent thickness.

But since the primary error afflicting thin mirrors on multi-point supports is sag not shear my thinking was that (on the small scale between support points) the overall curve of the glass can play only a small role. And that's why I modeled the f/100 mirrors in the first place, and am prepared to build 81-point supports (or beyond) for much larger versions.

Uhm, one final point. Tapered mirrors (Mauro's suggestion) have been around forever. I've seen some horrible FEM modeling on such, but in practice they seem to work well enough. Which could just mean many of the considerations involved in support and thermal analysis are a level beyond real world performance - or that real world "performance" tends to be a level beneath what is theoretically possible. Not really wanting to go there on those.

Best,
Mark

Edited by mark cowan (06/28/12 04:58 PM)