Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

Re: Equalization


First, we should actually observe, and test variety of DMD fibers to have 
appreciation of fiber DMD bandwidth deficiency characteristics.  Since it is 
caused by the combination of the imperfection of the refractive index curve, 
launch condition, and light traveling path, the resultant frequency response 
curve is totally unpredictable from fiber to fiber and from transceiver to 
transceiver.  You can see those plotted DMD BW response curve, which is the 
combination of multiple non-deterministic, irregular response curves 
superimposed together.

To generate a counter response curve for equalization, we must have the 
on-fly response plot capability, and generating the equalization circuit of 
the counter response constantly.

Of course, we can use the brute-force to approximate to whatever the extent 
we can get, but what will be the result?  Useful?  No one knows, till we try 

How much equalization will help DMD is unknown to everyone at this moment.  
We are based on the simple linear bandwidth deficiency model to determine the 
unpredictable DMD bandwidth deficiency in an actual application.  Of course, 
if we have time we can improve BW to whatever the extent the rules of physics 
will allow us to obtain; nevertheless, we do not have any idea what the 
result will be?

If we vote 5-PMDs in, in next several years, someone will come up a great 
solution to compensate DMD to certain extent.  However, we will never quite 
cure it, because DMD by nature is a defect parts.  The extent of defect by 
DMD among the installed fiber are all different, which can vary from 0% 
defect to over 80% defect.  I believe, it is unwise to cure those severely 
defected DMD fiber, because the RETURN-OF-INVESTMENT may not be worthy.  

I am very much interested in DMD problem.  If I have time, I may do something 
about it.  I would recommend you to try it to have the actual feeling of what 
we are discussing here.


Edward Chang
NetWorth Technologies, Inc. 

 If I didn't know better, I would have thought you are punishing me
 for starting this thread!
 Actually, this one is easy, almost elementary. (I suspect you read
 "impulse response" as "impulse", and  the 7th cup of coffee made you
 quickly pull the trigger...)
 Still, here goes. Simple models of communications links assume the
 channel to be linear, time-invariant. "Channel" is copper cable,
 fiber, air, whatever. In time domain, we describe its behavior with
 an impulse response. If this impulse response, the h(t) curve, is
 the same no matter when you look, we call it a time-invariant
 response. The impulse response of a singlemode fiber, for example,
 can be called time-invariant. If not, it's time-variant. (Purists,
 please cut me some slack; I know, nothing is truly time-invariant.)
 If it is time-variant, we have to ask if it varies randomly or
 My contention was that a multimode fiber with DMD can be viewed as a
 channel whose impulse response is changing with time in a random
 fashion. Hence the phrase "randomly time-variant impulse response".