SHDSL aka G.shdsl or formally ITU-T Recommendation G.991.2 is the new
symmetric DSL standard designed to replace the utter mess of
SDSL/2B1Q (a sea of proprietary flavors).
The name SHDSL comes from it being somewhat of a compromise between the
previously divergent standards
of SDSL-as-in-Internet and
HDSL-as-in-T1/E1-transport.
SHDSL features a frame structure that looks almost identical to that of the good old HDSL (the overhead bits have different meanings but occupy exactly the same positions in the frame), but the standard is worded in a way that recognises applications other than T1/E1 transport. The modulation has also been changed from 2B1Q to TC-PAM, which is said to give SHDSL better speed/distance performance compared to SDSL/2B1Q. Finally, repeaters are allowed because the framing and pre-activation mechanisms have been designed with those in mind.
The ETSI version of the standard defines transport protocol-specific
transmission convergence (TPS-TC) for a number of different Layer 2
transport applications which among others include unformatted sync serial
bit stream (clear channel) and ATM.
Although it doesn't say so anywhere in the standard (at least the free ETSI
version), there is an unwritten agreement among DSLAM vendors that all SHDSL
lines served from DSLAMs use ATM.
On the other hand, the bit stream / clear channel mode is predictably very
popular among those who do roll your own DSL
without DSLAMs and
don't care for ATM.
I used to say that I would eat my shoe if I ever saw a SHDSL line in the USA-occupied territories: there is so much already deployed SDSL/2B1Q infrastructure that it would probably take something on the order of another civil war to take it all out of service and convert to SHDSL.
However, the main factor working against the introduction of SHDSL is the
sad fact that symmetric business DSL services in general regardless of
technical details are rapidly falling out of fashion along with the rest of
the philosophy of institutional peering and net neutrality.
They are being replaced with a brave new world paradigm of content providers and
consumers, the connection technologies being divided into multi-gigabit links
for the former and ADSL/cable/FiOS for the latter.
Technologies like SDSL and T1 which are intended for those who refuse to
participate in the provider/consumer paradigm are receiving the
ugly duckling
treatment from everyone ranging from ISPs to local
geek groups.
I can't write more on this topic as it's too painful.
As explained in more detail on our chips page, there are unfortunately no transceiver chips for SHDSL that feature the same degree of openness as can be found in the RS8973 bitpump for SDSL/2B1Q or in the MC145572 ISDN U interface transceiver for IDSL. Of course with SHDSL the signal format is a bona fide published standard, so one is perfectly welcome to build a free and open implementation from discrete components. Needless to say though, our little group has worthier things to do in life.
We are considering building our open connectivity solution for SHDSL based on Mindspeed's M289xx chipset. While on the one hand we have plenty of beef with Mindspeed and from that standpoint I feel like going with Infineon instead just in order to not reward Mindspeed with chip sales, on the practical side I have to also acknowledge our goal of supporting as many SDSL flavors as possible with as few different hardware designs as possible, and M289xx has a clear advantage on that front: it supports G.shdsl, many flavors of SDSL/2B1Q and IDSL with the same hardware.
Our LNX8945 page has more musings on this topic.