I tried to put this reasoning into some sort of math. This is what I came up with:

X is the fraction of light emitted by bulb that restrikes the bulb for a perfect mirror material(mylar) used for a suboptimal reflector geometry.

Fraction of light emerging out of fixture for Mylar = 1-X

Now consider a diffuse reflector.

If light is incident on a diffuse surface (lets assume perfectly diffuse...meaning equal probability of reflection in all directions in a hemisphere from the surface) then a fraction of the total light will actually restrike and be reflected back into the bulb. IF we call this fraction Y, Y can be expressed as the ratio of the angle subtended by the bulb at the point on the reflector to the total angles it could be reflected at (ie 180 degrees for a hemisphere). Now this ratio Y would change depending on which point of the reflector you are at. Right above the bulb, at the point on the reflector closest to the bulb, this ratio would be the greatest (it is clear in the picture in my previous post), whereas at a point on the left or right portions of the reflector, this fraction Y, is lower. A good guess at what the mean value of Y should be, if you integrate across the whole surface might be Y=0.2-0.3. Lets assume Y=0.25.

This implies that in the case of a perfectly diffuse material you would expect 1-Y=0.75 of the total light being output by the bulb (in the upwards direction) to be available. So roughly you would get around 175% of the light as compared to if you had a perfectly black reflector.

Now if you have a perfect reflector geometry, X can be extremely small for the perfect mirror material. However the efficiency for a perfectly diffusive surface will not improve much at all since the diffuse nature will cause light to scatter in all directions. An interesting corollary is that perfectly diffuse reflectors will be very insensitive to the geometry of the reflector in terms of how efficient they are. The variability in performance of diffuse reflectors for a whole host of geometries should be pretty low.


It is very easy to see that in the case of a perfect mirror surface like Mylar, for a spiral CFL and non optimal reflector geometry, the fraction X could easily exceed this value of 0.25, and in this case a diffuse reflector would be a better option as compared to a perfectly reflecting material.