[time-nuts] ngTADD-1

[Hal Murray]

kevin at rosenberg.net said:
> For < 1ns jitter, do you have a recommendation of  modulation (ASK/PSK?) and
> carrier frequency? 

All the digital communication over fiber that I'm familiar with is simple 
on-off baseband, no carrier.  Digital geeks use encoding rather than 
modulation.

One complication with fibers is that the receiver has to work with a huge 
dynamic range, so they all have an AGC.  I don't know what the time constant 
on the AGC is.  The other complication is that the receiving end has to do 
clock recovery.  That requires transitions so there is a limit on the max 
number of consecutive 0s or 1s.  Modern protocols use 8B/10B or 64B/66B to 
guarantee enough transitions.  (Wiki pages are good.)

The other goal of things like 8B/10B is an equal number of 0s and 1s so you 
AC couple the signal path without introducing pattern sensitivity.  (The 
4B/5B encoding used for FDDI had a couple of nasty cases.)

The AGC makes it hard to send a simple low duty cycle PPS pulse over fiber.

For something like 10 MHz, I'd send alternating 1s and 0s, at either 10 
megabits or 20 megabits.  Somebody has to multiply by two.  I'd do it on the 
easier end.

Suppose you want to send a PPS signal.  It's easy if you have the 
corresponding 10 MHz clock.  I'd use Manchester encoding.  That is send 
alternating 01 at 20 megabits for the non pulse time and 10 when the PPS 
pulse is high.

-------------

Several years ago, there was discussion (or at least mention) of the radio 
astronomy guys down in Chile doing clock synchronization over fibers.  They 
had to correct for temperature and such.





-- 
These are my opinions.  I hate spam.