What is generally not well understood is that there are really eight diffraction spikes from a normal spider, not four. Each vane causes two spikes, perpendicular to the vane:
Add another vane opposite the first, and we get two more spikes, though it will still look like two since they fall on top on the first ones (if the vanes are parallel, that is). But even though it still looks like two, each spike will be brighter because of the added diffraction.
So, what does this have to do with the price of coffee in Miami? Nothing. What does this have to do with curved spiders? Everything!
Straight-vaned spiders reinforce the diffraction spikes making them brighter. A properly designed curved-vane spider will spread the diffracted light throughout the field, eliminating the spikes.
First off, a curved vane will not make a pair of spikes, it will make a pair of fans:
Like the pattern produced by a searchlight.
The arc of each diffraction fan is equal to the arc of the vane in the light path. For example, if the vane below has a 90° arc in the light path, it will form two 90° fans:
So you want to select the number of vanes and the arc of each vane such that the fans form a complete 360° circle.
Some combinations that will work, from a purely optical point of view, are:
1 vane with a 180°
2 vanes with 90° arcs, set 90° apart. Note that if the vanes are set opposite one another the fans will reinforce and form a brighter searchlight pattern.
3 vanes with 60° arcs, set 120° apart.
3 vanes with 120° arcs, set 120° apart.
4 vanes with 45° arcs, set at 45°, 90°, 180°, and 315°.
5 vanes with 36° arcs, set 72° apart.
Other combinations are possible, but all designs follow the rule that no vane has a vane opposite it.
Note that it seems to make little difference which way the vanes are curved. One design that takes advantage of this is two 90° vanes set 90° apart, but curved in opposite directions. This allows you to use a single 180° piece with ample mechanical strength.
Fellow ATM Gary Seronik,
who uses this design, reports: I've found through expermentation
that there is considerable lee-way in the designs. It seems so long as
some kind of curve is present the diffraction effects are erased. At
first I went to great pains to assure that the bends were precisely
90 degrees and so on, but later found out that it made little
With most designs, the total amount of diffraction will actually be slightly more than with an equivalent straight spider for two reasons:
This brings up one additional item: fellow ATM Richard Schwartz points out that in certain circumstances the noticable spikes may be prefereable. For instance, if you are looking for the faint companion of a bright double star, the circular diffraction pattern from a curved spider could bury it, but with bright spikes you can place the star where the spikes aren't.