Cheung mentions *tan (as in the NP verb tan- "to weave") as the Ir. descendant of *ten. Accordingly, tang is from IE *tenk (I don't know if *tenk itself is an extension of *ten or not). For deriving tār from *tan, the Skt. parallel tantri "string" is helpful. My limited understanding is that the Iranian form should have been like tṇθr- or tanθr-. In any case, the n disappeared, or merged or turned into a. I guess the θ became h in MIr. and disappeared (cf. OIr. puθra "son" > MP/Parth. puhr > NP pūr). I wonder a different route for *-θr was to l (giving tāl "braids").