You can't have a clear understanding of rendaku unless a major breakthrough comes in linguistics. Besides, even native speakers can get confused when dealing with proper nouns and words from dialects. Ask a random Japanese guy on the street if 茨城 is いばらき or いばらぎ. Chances are he doesn't know.
As for voiced obstruents, you can think of them as those moras usually written by kana with dakuten, though technically they only refer to the consonant part of them. So as yudantaiteki says, if you type a kana with d, g, z, b, or j plus either a, i, u, e, or o, you get a mora with a voiced obstruent.
A tricky thing is that their unvoiced versions may be different depending on whether you're talking about grammar or phonetics. For example, t, k, and s are the unvoiced versions of d, g, and z in both phonetic and grammatical senses. But while the unvoiced version of consonant b is p in the phonetic sense, the grammatically unvoiced version of b is h in Japanese. This is because the sound of Japanese has changed rather drastically while renduku grammar etc. haven't changed to the same degree.
It doesn't seem like the Wikipedia article explains this, but the sound change of renduku occurs only when the first consonant of the latter word is either t, k, s, or h. These four guys and their grammatical voiced versions d, g, z, and b also appear other sound change rules like saku+kyoku = sakkyoku. So my advice is you learn these four guys are kind of special and move on.
If you want to learn renduku a little deeper than Wikipedia, here's a blog about rendaku. All posts are about the phenomenon and written in Japanese.
http://rendaku.exblog.jp/
If you really really want to learn it deeply, there are a lot of research results out there. They require native level understanding of the Japanese language though. e.g., you should be able to intuitively tell how a native speaker's mind works differently when he pronounces やまがわ and やまかわ, both of which are 山川.
Last edited by magamo (2010 June 28, 11:19 pm)
