RE: [10GBASE-T] RE: EMI Discussion
Hello 10GBASE-T SG Members,
Following up on my earlier post, I spent some time with Larry Cohen, a
senior scientist with SolarFlare and we discussed the measurements used to
generate the EMI compliance slide used in the CFI last November.
Having seen the methods used, I found them to be consistent with methods I
would have used. I have no problem with the data generated by SolarFlare for
that presentation.
That said, I also have some recommendations which I shared with Larry and
believe I should share with the SG.
My primary concern is with the breadth of the measurements. When confronting
a challenge to apply new technology to an installed base, as we did with
10BASE-T or 100BASE-T or 1000BASE-T, a broad study of the installed base is
essential to ensure compatibility. In the case of 10BASE-T, where we were
applying high speed (10MHz?) signals to CAT3 wiring which had been installed
for phone support, a *huge* quantity of testing was done to verify signal
integrity, EMI compatibility, and noise immunity.
In the case of 1000BASE-T, less was required because the signalling operated
within CAT5 test limits. Only MDNEXT, ELFEXT, and RL required
characterization and ultimately specification. For 10GBASE-T, we are talking
about applying signals as much as 4X the design (and test) frequency of the
cabling. I believe this puts the quantity of work to be done on cable
characterization somewhere between that done on 10BASE-T and 1000BASE-T with
emphasis on EMI compliance. The reason I emphasize EMI compliance is that
you can always claim to operate on shorter cable lengths, but the FCC
doesn't care if its 1 meter if you spray EMI all over the place. Balance and
CM_ReturnLoss are areas that give me concern as they may impact EMI
compliance.
I have recommended to Larry that the SG should obtain market-share data on
various cable plant components to understand what patch panels, wall jacks,
patch cords, etc.. hold the majority of market share, then perform similar
tests to the ones he has already done to develop confidence that EMI
compliance in the broad market will be be consistent with his existing data.
We saw some EMI performance variation as a function of patch-cord length
which appears to show an increase in emissions as cable length decreases. We
should characterize that to understand what limitations that poses for
installations.
I believe a certain amount of that work should be done (and hopefully
presented) before the SG completes its work, but more work in this area will
be required once the TF is formed. Once I wrap up my CX4 work, I would like
to help in this area.
Best Regards,
Dan Dove
> -----Original Message-----
> From: DOVE,DANIEL J (HP-Roseville,ex1) [mailto:dan.dove@hp.com]
> Sent: Thursday, July 31, 2003 3:20 PM
> To: stds-802-3-10gbt@ieee.org
> Subject: [10GBASE-T] RE: EMI Discussion
>
>
>
> Hello all,
>
> Since my earlier post with attachment could not be forwarded,
> I have posted
> the attachment onto the CX4 site at the following URL;
>
> http://www.ieee802.org/3/ak/MLT3_2ndOrder.jpg
>
> It shows the PSD reduction that is likely to exist on a
> CAT5(e) or CAT6
> cable plant due to natural filtering in the magnetics, PCB
> and device pads.
> Since the energy above 100MHz is really not necessary for
> data recovery, it
> could be much lower.
>
> My point was that 10GBASE-T (or 2.5GBASE-T for that matter!)
> operating on
> UTP will have to address the possible imbalance and other anomalous
> impairments that could impact FCC/EN compliance specs at
> frequencies between
> 100MHz and 450MHz because vendors of infrastructure equipment
> may not have
> tested in that region. George Z has pointed me to their
> presentation which
> indicates acceptable EMI performance and I am waiting for one of his
> engineers to provide more detail with regard to the breadth
> and depth of
> testing that was done to confirm that presentation.
>
> Best Regards,
>
> Dan Dove
> HP ProCurve Networking
> > -----Original Message-----
> > From: DOVE,DANIEL J (HP-Roseville,ex1) [mailto:dan.dove@hp.com]
> > Sent: Wednesday, July 30, 2003 9:29 AM
> > To: 'Joseph Babanezhad'; stds-802-3-10gbt@ieee.org
> > Subject: RE: [10GBASE-T] PAR and 5 critters
> >
> >
> >
> > Hi Joseph,
> >
> > Thank you for the clarification. I mis-understood your
> > presentation in that
> > I thought you were comparing the PSD of 100MB MLT-3 with that
> > of 1000MB
> > PAM-5, and then 10G (others) to show similarity of the PSD
> > energy above
> > 100MHz. That explains to me why the MLT-3 shape appeared the
> > way it did. My
> > calc's showed a different shape, but similar in slope if the
> > MLT-3(100) was
> > unfiltered.
> >
> > As my attachment showed, MLT-3/100 and PAM5/1000 are likely to have
> > substantially lower energy above 100MHz which means that EMI
> > performance of
> > CAT5(e) or CAT7 cable plants may not have really been
> stressed to meet
> > FCC/EN compliance requirements as they would to perform with higher
> > differential PSDs that would exist using 10G line codes.
> >
> > I was wondering if someone has done a study of installed base
> > components to
> > confirm that we are not completely prohibited by that
> > installed base for
> > using the frequencies and levels that are being proposed?
> > Unless we want to
> > rule UTP out completely, I think such a study is necessary to
> > ensure that
> > the PSD under discussion is going to meet FCC requirements.
> >
> > I remember that 1000BASE-T had a similar hurdle to keep its
> > spectrum ~under
> > 100BASE-T. I believe we can lower that hurdle, but really
> > need to understand
> > just how much.
> >
> > Regarding my attachment, maybe Brad can post to to the 10G
> > website as a
> > contribution? It showed your PSD graphs, and then an MLT3 PSD
> > with/without
> > 2nd order filtering which demonstrates the large reduction in
> > energy above
> > 100MHz that is going to exist in a normal implementation due
> > to parasitic
> > package elements and magnetics.
> >
> > Dan
> >
> > > -----Original Message-----
> > > From: Joseph Babanezhad [mailto:jobaba@platolabs.com]
> > > Sent: Tuesday, July 29, 2003 2:18 PM
> > > To: stds-802-3-10gbt@ieee.org
> > > Subject: RE: [10GBASE-T] PAR and 5 critters
> > >
> > >
> > >
> > > Dan,
> > >
> > > My January 2003 presentation to IEEE 10GBASE-T SG
> > >
> > > http://www.ieee802.org/3/10GBT/public/jan03/babanezhad_1_0103.pdf
> > >
> > > assumes that all the line-signals considered (MLT3,
> PAM5-to-PAM33)
> > > have the same minimum pulsewidth of 800ps. That is why
> all of them
> > > have notches at 1.25GHz and 2.5GHz.
> > >
> > > If you were to review my March 2003 presentation
> > >
> > > http://www.ieee802.org/3/10GBT/public/mar03/babanezhad_1_0303.pdf
> > >
> > > here various line signals considered have different minimum
> > > pulsewidth in order for all of them to achieve 2.5 Gb/s
> data rate.
> > > The minimum pulsewidth for both MLT3 and PAM5 is 800ps. For MLT3
> > > and PAM5 their notches are still at 1.25GHz and 2.5GHz while for
> > > PAM9-to-33 they are at different frequencies. When comparing
> > > MLT3 and PAM5's frequency spectrums it becomes clear that MLT3
> > > shows considerable peaking around pass-band and notch frequencies.
> > >
> > > There was no additional filtering, such as TX pulse
> > shaping, included
> > > in deriving these frequency spectrums.
> > >
> > > Regards,
> > >
> > > Joseph N. Babanezhad
> > > Plato Labs.
> > >
> > > -----Original Message-----
> > > From: "Booth, Bradley" <bradley.booth@intel.com>
> > > To: <stds-802-3-10gbt@ieee.org>
> > > Date: Tue, 29 Jul 2003 11:54:22 -0700
> > > Subject: RE: [10GBASE-T] PAR and 5 critters
> > >
> > > >
> > > > Forwarded for Dan Dove without attachment... Dan, please
> > > give the URL
> > > > for the presentation as our website doesn't accept attachments.
> > > >
> > > > Thanks,
> > > > Brad
> > > >
> > > > -----Original Message-----
> > > > From: "DOVE,DANIEL J (HP-Roseville,ex1)" <dan.dove@hp.com>
> > > > To: "'Sterling Vaden'" <sterlingv@bellsouth.net>
> > > > Cc: stds-802-3-10gbt@ieee.org
> > > > Subject: RE: [10GBASE-T] PAR and 5 critters
> > > > Date: Tue, 29 Jul 2003 11:17:53 -0700
> > > >
> > > > Hi Sterling,
> > > >
> > > > I have been an outsider to the study group work, so please
> > > excuse me if
> > > > I am
> > > > bringing up an issue that has already been dealt with, but as I
> > > > understand
> > > > it the current proposals for coding rely upon a spectrum of
> > > > 400MHz.
> > > > It
> > > > is
> > > > possible that companies manufacturing cable system
> infrastructure
> > > > equipment
> > > > like patch panels, wall jacks, and even the cable itself
> > > have qualified
> > > > their products for FCC and EN compliance based upon
> > > measurements done
> > > > with
> > > > equipment that only uses <100MHz spectrum like 100BASE-T or
> > > 1000BASE-T.
> > > > That
> > > > said, the balance requirements for such equipment may never
> > > have been
> > > > stressed at frequencies above 100MHz for longitudinal
> > > balance, and so
> > > > the
> > > > installed base of cable systems may have issues with EMI
> > > compliance if
> > > > someone were to begin running equipment with a PSD that
> > > produces large
> > > > energy above 100MHz on those systems.
> > > >
> > > > Has any work been done to characterize this issue? I found a
> > > > presentation by
> > > > Joseph Babanezhad but don't think he applied a full
> > analysis in the
> > > > sense
> > > > that it appears he did not allow for natural filtering of
> > the output
> > > > PSD
> > > > by
> > > > device capacitance and magnetics on the MLT-3 PSD. Check out the
> > > > attachment.
> > > > Clearly MLT-3 and PAM-5 (100 and 1000 speeds) can
> operate without
> > > > energy
> > > > above 125MHz and most implementations limit energy above that
> > > > frequency.
> > > > I
> > > > believe the 10G designs will require substantially higher
> > > energy than
> > > > those
> > > > technologies in the region 100MHz < f > 400MHz.
> > > >
> > > > Again, please excuse me if this has been addressed. Just
> > point me to
> > > > the
> > > > data.
> > > >
> > > > Regards,
> > > >
> > > > Dan Dove
> > > > 802.3 member
> > > >
> > > > -----Original Message-----
> > > > From: Sterling Vaden [mailto:sterlingv@bellsouth.net]
> > > > Sent: Tuesday, July 29, 2003 6:43 AM
> > > > To: stds-802-3-10gbt@ieee.org
> > > > Cc: stds-802-3-10gbt@ieee.org
> > > > Subject: Re: [10GBASE-T] PAR and 5 critters
> > > >
> > > >
> > > > OK, I'll bite,
> > > >
> > > > I think it has been "shown" that for the purposes of
> the PAR and 5
> > > > critters,
> > > > 10G will work over the following:
> > > >
> > > > ISO Class F without ammendment.
> > > > ISO Class E (or Cat6) screened or SSTP (with extended
> > > limits to 500 or
> > > > 625
> > > > MHz) extended limits TBD by ISO or TIA
> > > > (note that Class D and Class E screened (overall
> shield) and SSTP
> > > > (individually shielded pairs) are commonly installed in
> > > Europe, so this
> > > > is
> > > > not a "fantasy cable". Screened cabling is also
> specified by TIA)
> > > > ISO Class E (or Cat6) UTP to 50 meters (with extended
> > > limits to 500 or
> > > > 625
> > > > MHz) extended limits TBD by ISO or TIA
> > > >
> > > > ISO Class D (Cat5e) Screened? Perhaps up to 80 meters, but
> > > this is grey
> > > > area.
> > > > ISO Class D (Cat5e) UTP? Perhaps up to 40 meters,
> > > >
> > > > For ISO Class D (Cat5e) there is a basic problem in that
> > the cabling
> > > > standards groups are unwilling to standardize (create
> > limits) beyond
> > > > 100
> > > > MHz, or expend further work on the cabling besides
> > measure it. This
> > > > poses a
> > > > very real difficulty in specifying a protocol that relies upon
> > > > performance
> > > > beyond that specified by the cabling standard. (ask Geoff
> > Thompson)
> > > > Cabling
> > > > manufacturers may decide to forego warranting their cabling
> > > systems for
> > > > such
> > > > a protocol.
> > > >
> > > > Also, for the time being, lets pretend that alien
> crosstalk field
> > > > testing
> > > > does not exist (it doesn't). Also lets pretend that alien
> > crosstalk
> > > > mitigation techniques (all, retrofit and new cable
> > designs) for the
> > > > moment
> > > > do not exist. These are considerations for the task group.
> > > >
> > > > Some may also contend that the protocol will run on longer
> > > lengths of
> > > > Cat5e
> > > > and Cat6 UTP. If so, that is fine, but it is a matter of
> > > dispute, and
> > > > therefore cannot be considered. This is also a
> > consideration for the
> > > > task
> > > > group.
> > > >
> > > > At the Plenary, on the last day we heard the PHY vendors
> > > backpedaling
> > > > on
> > > > their previously stated opiniion that it would run on Class
> > > F. If that
> > > > is
> > > > the case, and they insist on this position, then the
> > > project is dead.
> > > > Therefore I propose that there must be an agreement that
> > > the protocol
> > > > will
> > > > run on at least Class F cabling to 100 meters, or we
> need to start
> > > > over.
> > > > If
> > > > we can agree on that, then we can move forward to
> > consider the other
> > > > cabling
> > > > classes.
> > > >
> > > > Sterling Vaden
> > >
> >
>