RE: [802.3ae_Serial] another question on pattertn defined in the current draft:
Juergen,
The loopback should be done as low in the WIS as practical so it tests most
of the WIS circuitry. Therefore, the WIS framer is involved in doing the
loopback. That is why the decision was made to send a repeating 16-bit
output.
Traditionally, an Ethernet node in loopback has not driven its output
because an Ethernet device doesn't transmit if it can't receive. This of
course was critical when operating in a shared CSMA/CD network but it has
been maintained in full duplex Ethernet.
In the latest draft we have departed from that for the PMA and PMD
loopbacks. The statement about what the transmit output should be during
loopback has been deleted. Since there is not statement under loopback about
how the transmitter behaves (for the PMA/PMD) then one interpretation is
that it behaves as usual and the transmit input appears on the transmit
output as well as being looped back.
For the higher layers we are still sending a static output during loopback
such as the 0x00FF in WIS.
Transmitting the same signal that is being looped-back would be consistant
with what we are now doing at the PMA and PMD sublayers. Having a field set
that keeps the receiver at the other end from thinking the link is up would
be even better.
Pat
-----Original Message-----
From: Rahn, Juergen (Juergen) [mailto:krahn@xxxxxxxxxx]
Sent: Thursday, June 28, 2001 1:17 AM
To: stds-802-3-hssg-serialpmd@xxxxxxxx; 'DAWE,PIERS (A-England,ex1)'
Subject: AW: [802.3ae_Serial] another question on pattertn defined in
the current draft:
Hi,
A Sonet receiver will generate LOF but no real LOS as power is there. This
will be interpreted as invalid input signal so may be mis connection. There
is not any standardized interpretation for this as the Primary alarm will be
displayed. So the interpretation of this is open. In this Sonet element AIS
will be inserted that will prevent further alarming in the Sonet world. The
10GE receiver will receive the AIS.
In case of legacy WDM or OTN where you have non-intrusive monitors you will
have this alarms through al Translators and other NE up to the point where
this may hit a Sonet NE or the 10GE receiver. This LOF notifications again
may be interpreted as signal type mismatch so in particular if several
operator domains are passed may be interpreted as mis - connection.
OTN also ( expecting a Sonet signal but receives something different will
interpret this as signal type mismatch and insert Unframed AIS which will
than be also delivered to the 10 GE receiver. (this will lead to LOF in this
receiver).
An issue with this pattern may be that the LOF process is slower than the K-
bytes interpretation for protection switching and such fixed pattern may be
interpreted as stable K-bytes and a protection action may be partly
initiated. (By the way, this is a known issue and was also considered when
defining the Unframed AIS in OTN)
Regards Juergen
> ----------
> Von: DAWE,PIERS (A-England,ex1)[SMTP:piers_dawe@xxxxxxxxxxx]
> Gesendet: Mittwoch, 27. Juni 2001 18:19
> An: Rahn, Juergen (Juergen); stds-802-3-hssg-serialpmd@xxxxxxxx
> Betreff: RE: [802.3ae_Serial] another question on pattertn defined in
> the current draft:
>
>
> Juergen,
>
> Draft 3.0 had said "In this mode, ... the WIS shall transmit a continuous
> stream of all-zero data words to the PMA sublayer, and shall ignore all
> data
> presented to it by the PMA sublayer." This meant zeroes on the line:
> zeroes
> where the header would be and zeroes where the payload would be. I
> thought
> that this could lead to optical power glitches and/or chattering optics
> and
> even worse problems than you describe, so I suggested changing it to any
> balanced pattern. It turns out that the pattern chosen by the group is
> being proposed for the square wave test pattern for WAN PHY.
>
> That's the history: I'll let the SONET experts debate the merit of AIS,
> which I presume is balanced, or nearly so.
>
> Piers
>
> > -----Original Message-----
> > From: Rahn, Juergen (Juergen) [mailto:krahn@xxxxxxxxxx]
> > Sent: 27 June 2001 10:28
> > To: stds-802-3-hssg-serialpmd@xxxxxxxx
> > Subject: AW: [802.3ae_Serial] another question on pattertn defined in
> > the current draft:
> >
> >
> > Hi,
> > Can somebody help me to understand why the pattern in case of
> > Loop back is set to 00-FF?
> > My question is was there a strong reason for this that I do
> > not know. What I know is that there are some
> > ugly effects possible with such pattern when the interface is
> > connected to a Sonet or other type transport network.
> > I understand that at the last meeting some work has been done
> > on this loop back facilities. Now I read in
> > 50.3.9 Loopback:
> > ..............................................................
> > ..............
> > ...........................the WIS shall transmit a constant pattern
> > to the PMA sublayer, and shall ignore all data presented to
> > it by the PMA sublayer. The pattern output to
> > the PMA transmit path at this time shall consist of a
> > sequence of 8 logic zero bits and 8 logic one bits, form-ing
> > the 16-bit word 00-FF hexadecimal. No SONET overhead or fixed
> > stuff shall be output to the PMA at this time.
> >
> >
> > While agreeing that the data incoming to the PMA sublayer
> > should not be
> > sent further such constant pattern is likely to generate some ugly
> > additional
> > effects . So maybe this would lead to undefined and miss
> > leading alarms
> > or( or even protection actions) in a possible transport network. This
> > transport
> > network that can either be a plane Sonet transport network, but also a
> > traditional
> > WDM network with Sonet non-intrusive monitoring or transport
> > via an OTN. In all those
> > cases the consequence on such patterns are not defined and
> > the alarms that will
> > be generated are likely to be miss interpreted. For instance
> > it can be
> > interpreted that a wrong signal is connected. In case a proper AIS is
> > inserted however the transport network will react in the standardized
> > way and no alarms (or even protection switches as worst case
> > scenario)
> > will be activated (which should not be done when a client
> > equipment is in test mode).
> > So can somebody help me with the reason for this pattern ( In
> > contrast to a normal AIS
> > that would not generate such effects). If there is no
> > particular overriding reason for this
> > I would strongly suggest to take the AIS signal instead.
> > Regards Juergen Rahn
> >
> >
>