Incidentally, for those of you who are curious why SSE is a problem for digital RPTVs (LCD, DLP, LCoS) but not for CRT RPTVs, you might want to take a look at Goldenberg J, Huang Q, and Shimizu J, Rear projection screens for light-valve projection systems, Proceedings of SPIE -- Volume 3013 Projection Displays III, Ming H. Wu, Editor, May 1997, pp. 49-59. They explain that speckle (what we call SSE) arises from interference between scattering centers on RPTV screens. To get interference, the light needs to be coherent. Over sufficiently small distances on the screen, even light from an incoherent source can behave as if it were coherent, if it has been subjected to sufficient spatial filtering (though it is of course not fully coherent, as light from a laser source would be). How small that distance needs to be (what they call the "coherence length") is inversely related to the size of the projecting element, because it is this size that determines the degree of spatial filtering. In CRT RPTVs, the projecting element is relatively large (typically 5"), leading to a small coherence length (7 microns). By contrast, in digital TVs the projecting element is much smaller (~1"), leading to greater spatial filtering and thus a larger coherence length (75 microns) (the distance from element to screen also factors into this). The typical spacing between scattering centers ("feature spacing") on an RPTV screen (at the time this article was written) was 4 microns. [I.e., these screens are not perfectly smooth and homogeneous; instead, they have some surface roughness, and this roughness has a typical length scale of 4 microns. The article has a nice scanning electron microscope image of an RPTV holographic diffusing screen that shows this clearly.] So with a CRT RPTV, the coherence length is too close to the feature spacing to cause significant interference. But with a DLP RPTV, the coherence length is more than an order of magnitude greater than the feature spacing, ensuring that the there will be significant interference. And this interference -- which creates a random variation of light and dark patches across the screen corresponding to the randomly distributed features present on the screen -- is what causes speckling (aka SSE) (at least according the authors of this article).
Here's my (hopefully not overly naive) interpretation/restatement of the effect the authors are observing: you can make light more coherent if you spatially filter it. The smaller the filter, the more coherence you get (full coherence results from a point source, if the light is also monochromatic). Digital projection chips act as smaller filters than CRT guns, resulting in greater partial coherence. This in turn means more interference from closely-spaced features, resulting in greater speckling.
The authors (all of whom were then working at Philips USA) recognized that speckling is undesirable. They measured it for both a 52" CRT RPTV and a 52" LCD RPTV. In arbitrary units, the CRT and LCD had speckling values of 2 and 12, respectively. In their viewing tests, speckle was "disturbing" at values above 6. They indicate it can be reduced using diffusers, which break up the coherence, but at the expense of reduced resolution.
And of course, all this leaves unanswered why these screens have the surface roughness that causes the speckling in the first place! I assume this is there for some other functional reason (perhaps having to do with viewing angle or brightness uniformity), but I don't know what that is. I read in a thread somewhere (sorry, no longer recall the details) that some screen manufacturer has produced an RPTV screen that minimizes SSE, but at the expense of light output (fine with me!). It's going to be available on some company's (Pioneer ??) pro units next year, but not their consumer models. Or something like that.
This is all outside my area of expertise, so if anyone would care to correct/elucidate/expand, knock yourself out.