RE: Equalization
Gair:
I am a member of Fo2.2.1, and I have all the DMD data since the beginning of
DMD activities several years ago.
The information needed for equalizing the DMD frequenquncy response is not
the static data of the FO2.2.1 optical data. However, those data can help
to generate some specifications for the equalizer design. In another word,
we may be able convert some of those information to electrical circuit
design parameters.
The impulse response is not directly usable for circuit design, which
indicates the bandwidth, but not the whole plot of the frequency response.
As a result, one has to selectively from those Round Robine cables to take
frequency plots. In addition, those frequency plots are very irregular,
which can not be treated straight forward by Bode Plot to design a circuit.
Furthermore, they are all different shapes with multiple corner frequencies.
You have to study many plots to come up some approximation technique to fit
into design equations.
On top, you have to deal with actual system timing parameters. These are
not academic exercises, they have to meet the real system BER of 10^-12
repeatedly for all installed MM fibers.
By restricting ourselves to RML will save us time not to deal with the worst
case headaches, but we still have to go through all characterization process
to DEFINE PROBLEM first, then we can apply tools to cure the problem.
Once we have all the DMD under control, then we will go back to deal with
the TIA with equalization circuit issues. We discussed this issue on
reflector several month ago, and it will not be a easy one at 10 Gbps data
rate to assure BER of 10^-12.
Again, if we have sufficient time, I believe we will come up with some
solution eventually.
Time is the factor.
Regards,
Edward S. Chang
NetWorth Technologies, Inc.
EChang@xxxxxxxxxxxxxxxx
Tel: (610)292-2870
Fax: (610)292-2872
need the system data when DMD is involved
-----Original Message-----
From: Gair Brown [mailto:gdbrown@xxxxxxxxxxxxx]
Sent: Friday, July 21, 2000 4:56 PM
To: Edward Chang
Cc: HSSG
Subject: Re: Equalization
Ed,
You need to talk to some of the participants of the FO-2.2.1 round
robins. I suspect that some of the participants do have data that
contains phase information.
On a different note. I believe that the fiber response is relatively
invariant with respect to time. There is variation between different
fibers, but for fibers selected using the RML requirements the variation
should be manageable.
Gair
Edward Chang wrote:
>
> Gare:
>
> The characterization done by Fo2.2.1 is are from optical property point of
> view, which is very important of understanding the static optical
> parameters.
>
> The equalization need different parameters, time domain frequency response
> with the information of amplitude, phase versus frequency, and the dynamic
> nature of the response and error correction rate relative to fiber
motions.
> Furthermore, one has to define, how far we intend to deal with the DMD
> caused BW deficiency based on the installed MM fiber, then we can setup
the
> target specifications for equalization circuit to be designed.
>
> In all, the timing related parameters, and time domain frequency responses
> will be established.
>
> If we have time, eventually we will get some meaningful result. This is
> reason, I have proposed 5 PMDs to allow each PMD opportunity to be
improved
> later.
>
> Regards,
>
> Edward S. Chang
> NetWorth Technologies, Inc.
> EChang@xxxxxxxxxxxxxxxx
> Tel: (610)292-2870
> Fax: (610)292-2872
>
> -----Original Message-----
> From: Gair Brown [mailto:gdbrown@xxxxxxxxxxxxx]
> Sent: Friday, July 21, 2000 2:15 PM
> To: Edward Chang
> Cc: HSSG
> Subject: Re: Equalization
>
> Ed,
>
> The characterization has already been done by FO-2.2. I believe Mike
> Hackert's working group has a ton of data on the frequency response of
> RML compliant fibers.
>
> Gair
>
> Edward Chang wrote:
> >
> > Rohit:
> >
> > It is interesting. You ,may like to share your results with us?
> >
> > However, fiber dispersion and DMD is not exactly the same thing.
> > The linear dispersion proportional to cable length is straight forward,
> the
> > dispersion related to the non-linear DMD is different. We need
> > characterization.
> >
> > Regards,
> >
> > Edward S. Chang
> > NetWorth Technologies, Inc.
> > EChang@xxxxxxxxxxxxxxxx
> > Tel: (610)292-2870
> > Fax: (610)292-2872
> >
> > -----Original Message-----
> > From: owner-stds-802-3-hssg@xxxxxxxx
> > [mailto:owner-stds-802-3-hssg@xxxxxxxx]On Behalf Of Rohit Sharma
> > Sent: Thursday, July 20, 2000 9:21 PM
> > To: HSSG
> > Subject: RE: Equalization
> >
> > i believe DMD can be (has been) modelled as a Rayleigh multipath fading
> > channel and fast Decision Feedback Equalization (DFE) techniques with
RLS
> > (Recursive Least Squares) or LMS (least mean squares) algorithms applied
> for
> > rapid convergence with a short training sequence (# of bits roughly
equal
> to
> > the # of taps). J. Winters and cohorts in IEEE Transaction (1990 - will
> dig
> > out exact ref. tomorrow) and Stanford ISL with John Cioffi did a bunch
of
> > work in this area in 1991-95 timeframe.
> >
> > part of my grad studies demonstrated an 11 (goes to 11!) tap adaptive
> > optoelectronic filter to equalize dispersion in mmf for 1 and 2.5 Gbps
> IMDD
> > signals. Similar results were shown by other researdch groups using
> > electrical equalization with bipolar taps and DFE structures. some
basic
> > simulation results are available (and were published in Optics Letters)
> and
> > if anyone is interested, send me email and that will prompt me to dig
> > through any relevant results i have...
> >
> > -rohit
> >
> > Rohit Sharma
> > ONI Systems.
> >
> > > -----Original Message-----
> > > From: Jonathan Thatcher
> > > [mailto:Jonathan.Thatcher@xxxxxxxxxxxxxxxxxxxx]
> > > Sent: Thursday, July 20, 2000 4:07 PM
> > > To: 'vipul.bhatt@xxxxxxxxxxx'; HSSG
> > > Subject: RE: Equalization
> > >
> > >
> > >
> > > Vipul,
> > >
> > > You have really caught my attention on this one. What exactly is a "
> > > time-variant impulse?"
> > >
> > > jonathan
> > >
> > > >-----Original Message-----
> > > >From: Vipul Bhatt [mailto:vipul.bhatt@xxxxxxxxxxx]
> > > >Sent: Thursday, July 20, 2000 1:46 PM
> > > >To: HSSG
> > > >Subject: RE: Equalization
> > > >
> > > >
> > > >
> > > >Rich,
> > > >
> > > >I was skeptical, like you, about the ability of equalization to
> > > >overcome DMD. I still am. But what made me more open minded is my
> > > >willingness to examine two key propositions: DMD behavior can be
> > > >approximated as that of a fading multipath channel. And equalization
> > > >can overcome the effect of a fading multipath channel.
> > > >
> > > >Why fading multipath? Consider this. A multimode fiber with DMD is
> > > >seen by an optical signal as a channel that propagates various
> > > >portions of its energy (modes) through a refractive index profile
> > > >that is sharply different (at places) than intended. The propagation
> > > >velocity of a mode depends on refractive index. Over distance, on
> > > >average, some modes will have a cumulative average of a low (lower
> > > >than intended) refractive index path, thereby abnormally decreasing
> > > >their path delay, while others may not. In some bad cases, a bit
> > > >arriving at the receiver is almost split in two or three replicas.
> > > >And the amplitude of each replica is changing dynamically. Guess
> > > >what, that is not far from the behavior of a fading multipath
> > > >channel. Equalizers that deal with this phenomenon are used in TV
> > > >de-ghosting circuits and digital radio.
> > > >
> > > >To an equalization expert, DMD may not look so challenging - it's
> > > >just another channel with randomly time-variant impulse
> > > >response...:-)
> > > >
> > > >But then, we both are putting the cart before the horse. Let's wait
> > > >to hear some presentations from equalization experts. Even if it can
> > > >be done, I will want to know whether it can be done cost-effectively
> > > >and on time. Our discussion should serve as a guide to them about
> > > >what we would like to hear.
> > > >
> > > >Thanks,
> > > >Vipul
> > > >
> > >
>
> --
> Naval Surface Warfare Center
> browngd@xxxxxxxxxxxxx
> Code B35 PH: 540-653-1579
> 17320 Dahlgren Road FAX: 540-653-8673
> Building 1500 Room 110A
> Dahlgren, VA 22448-5100
--
Naval Surface Warfare Center
browngd@xxxxxxxxxxxxx
Code B35 PH: 540-653-1579
17320 Dahlgren Road FAX: 540-653-8673
Building 1500 Room 110A
Dahlgren, VA 22448-5100