OFDM for the equalization and FEC for the channel errors has been a
successful technique used in residential power-line communications.
However, commercial buildings tend to have long and unpredictable
power delay profiles, requiring long symbol times. In addition, there
are enormous sources of interference (primarily impulse noise) from
things like elevator motors and paper shredders (!), as well as strong
shortwave RF induced signals. Finally, some installations use
multiple transformers between floors, with their own very
unpredictable channel response.
Hey Jonathan,
This was a super informative post, it's really interesting to hear about experience of actually building PLC in practice. I haven't seen very many studies on it, but it seems like different buildings have greatly varying profiles. I didn't think of things like elevator motors in office buildings, but that's got to introduce a lot of interference.
If you ever do this again, I would love to see some of the channels over time. I ordered some basic PLC equipment, but all I really have access to are packets :P
As an alternative, one commercial R&D contract I did a while back was
to use DSSS/CDMA with long codes to reduce the interference by the
coding gain, then use additional delay correlators to receive and
combine echos (a basic RAKE receiver). This worked extremely well for
interference rejection, without FEC at the data level, but at the
expense of a low bit rate and long synchronization times. Since the
communication requirement was a unidirectional broadcast, this worked
out well.
Nice, I guess with unidirectional broadcast that does work out well for you. How do you actually get echoes in a PL? I never thought about this.
The initial implementation was done in GNU Radio with a chipping rate
of 2 Mcps, then ported to the USRP2 FPGA (minus the rake) for chipping
rates up to 25 Mcps.
I think this goes to the point that GNU Radio is an excellent tool for
experimentation, R&D, prototyping, etc., and not so much for chasing
existing standards.
> My main question regarding this topic is to figure out what would be
> necessary to create a test bed to try it out on? Obviously, we aren't going
> to stick the Rx/Tx port of the USRP into a wall socket :)
That company made some generic 50-ohm to wall socket power couplers
for the project, but I don't know if they ever made them commercially.
I still have a pair in my lab to play with :)
It would be really cool to be able to hook up the USRP to a PL and understand various channels in different types of buildings. Totally not conducive to my dissertation, but I find it interesting.
- George