As you might expect, the reliable range of RDS depends on subcarrier injection. In Europe (where RDS was developed) the standard deviation of the main FM carrier by the RDS sub is 2.0 kHz (about 2.7% modulation) and I would guess most US stations also run around 2 to 3 percent. I've found 3% is sufficient to display PS info (callsign or "handle") on signals so noisy that many listeners would probably tune out. If the station leases a portion of its RDS bandwidth to a paging service or another data provider, injection is often increased to 9 - 10 percent for even higher reliability.
The RDS subcarrier is referenced to the third harmonic of the stereo pilot and uses binary phase-shift keying, so it is quite robust. It's unfortunate that American broadcasters and auto manufacturers were so hesitant to adopt this technology and still don't take advantage of all of its features. The automatic retuning (AF) function (Europe's
raison d'être for RDS) could be especially valuable to noncommercial stations with translator and repeater networks, as well as commercial FM simulcast groups in markets with challenging terrain (such as the several operating in the Wilkes-Barre/Scranton area.)
The cost of RDS encoders continues to drop, Czech out this website:
http://pira.cz/rds/index.htm
Last year I ordered some of the PIRA encoder cards and a sample of the MRDS 192 chip (RDS encoder on a PIC for 12 Euro in unit quantities) and tested them on a local station; they do a nice job for the money. The "MiniRDS" postage-stamp encoder is rather cramped, however -- it uses only a 6-bit D to A converter and lacks a pilot sync input, so I'm considering designing a alternate PC layout for the chip that will allow space for an 8-bit converter and pilot filter/amp.
Now, can anyone explain why most HD receivers don't include RDS? Could it be that iBiquity considers this open source, reliable, inexpensive, proven system a threat?