EMI, Pigtails, and Isolators
Good Morning All,
I've been giving some thought to the 10G Serial TDM solution. If you look
at OC-192 Xcvr's, the DFB's are typically biased to 1.5 - 2.0 mW min. facet
power at room temp before adding modulation current. With lower bias
currents, edges go to hell. Add modulation current, double for connector
repeatability issues (assuming the use of a fiber stub) and you're looking
at a drive at low temps of <60 mA....maybe as much as 100 mA, with a rise
time of >50 pS. And this ~an inch from the bulkhead which will be full of
mounting holes. The EMI implications, irrespective of data patterns, will,
I believe, require pigtailing.
Moreover, unless people are willing to convert to APC connectors throughout
the system, the back reflection from PC polished connections means we'll
have to deal with an ORL of -14 dB. The out of phase return signal will act
as a seed oscilation in the lasers cavity, killing your optical SNR (bear
in mind that the grating's wavelength is constantly changing as a function
of junction temperature: single mode is sort of reletive). The answer to
this is to incorporate a Faraday Rotator and polarizer in the optical path.
While not particularly expensive in terms of component cost (~$20),
isolators are notoriously fiddly to set up and don't lend themselves to
automation. We once built a connectorized isolated laser and found that the
ORL varied by 12 dB at a fixed temp, so we abandoned the idea. Again....a
pigtail.
EMI should be sustantially reduced using PAM-5; first because of the
halving of the frequency and second because of the reduction in dI/dT owing
to the reduced current swing per baud. The reduction in frequency means
that biasing can be reduced as the rise time requirements are halved. ORL
may still be a problem and an isolator may (probably will) still be
required.