Hal-
Rather than describe it as "one PHY" or "two PHYs" in the traditional
sense,
I think it appropriate to describe what is done in PON/PoC technology
as a
"PHY set"
Definition: In the context on EPON/EPoC (i.e. point-to-multipont)
technology within 802.3, the specification for a master-slave set of
PHYs operating as a communicating pair.
Thus what was specified for the original EPON project (within P802.3ah)
was:
"- A single PHY set"
Then the 10G upgrade project further specified two sets:
10/1
10/10
(admittedly only trivially more complex technically than just one more)
Best regards,
Geoff
On 95//12 9:12 AM, Hal Roberts wrote:
Ed,
I
agree with you that the burst mode PHY in the FDD upstream could be
used in the downstream for TDD; in fact that is an interesting way to
look at it and maybe minimize the number of PHYs.
Geoff
said that that “two PHYs” really are necessary, with one PHY for
upstream and one PHY for downstream.
10GEPON
itself would then have “two PHYs” in that there is 1/10 EPON mode. Is
1/10 EPON considered to have two PHYs using IEEE terminology? (Question
for Geoff I guess).
In
any case, then FDD would need both PHYs (bursty and continuous) and TDD
only one bursty PHY.
Hal
Hi
Jorge,
In
my mind, we will have a burst PHY defined for the upstream operation of
EPoC or EPON. I think that downstream of a half duplex system would be
the same burst PHY. I don’t even see the need for the FDD or TDD
description that Steve proposed. In EPON, the data detector in the PHY
determines when to enable the laser and send data upstream. The half
duplex PHY would enable the same function in the downstream. I agree
that we are defining a PHY for Coax that supports burst mode and
continuous mode of operation. I think that could cover our objectives
for the PHY.
I
think that all of the changes needed for half duplex (TDD) are in the
MAC and MAC Control. This is traditionally where duplex is handled in
802.3.
Hope
that helps,
Ed….
Steve,
Thanks. Yes, that would work for me.
Ed,
Chano,
Would that work for you?
Thanks!
Jorge
Chano
and Jorge,
The terminology we have started to use is a PHY with both FDD and TDD
modes. That conveys the idea that FDD and TDD are very similar but
they may have some differences.
Maybe that wording would work in the objectives.
--------------
Kevin,
Nice job in stimulating email traffic.
J
Will we be discussing the objectives on Friday?
Regards,
Steve
Hi Jorge, thanks for the clarification.
I tend to agree that supporting both FDD and TDD
would be a good justification for having two PHY - although hopefully
both PHYs would have enough commonalities that silicon vendors could
easily implement both in the same chip without much duplication.
Chano
On May 8, 2012, at 3:17 PM, Salinger, Jorge
wrote:
As
one of the advocates for this objective, I'd like to clarify that the
goal we are after is to have TDD system operation for passive coax
plant and FDD for active coax plant. This is what might drive the need
for 2 PHYs, and not the spectrum range.
In other words, we would like to operate a TDD system in passive coax
plant networks above and below 1 GHz, or operate an FDD system in
active coax plant networks above and below 1 GHz.
As for the question of why to have both TDD and FDD systems, the answer
is that TDD systems have the benefit of not fixing the US:DS ratio
(which is desirable flexibility), but active cable systems would not
support a TDD system and would have to operate on a FDD system.
In summary, it is desirable to operate TDD everywhere in the spectrum,
but since that is not possible to do so in active coax plants then the
best solution is to have both TDD and FDD systems. FDD systems can be
used in passive and active coax plants, but the lowest common
denominator of just having an FDD system is undesirable because we are
slowly but surely driving to passive networks everywhere over time.
If having TDD and FDD systems can be achieved with a single PHY, then
that's great. If not, then let's have one PHY for TDD and one for FDD.
Hope this is clear and makes sense.
Thanks!
Jorge
In
that case, “Specify at most two PHYs … “ should be simpler to parse … I
guess nobody will specify 0 PHYs …
I
agree there is no worry about more than two PHYs. So here is a modified
proposal.
Specify at least one PHY and two PHYs
if necessary to support
subscriber access networks using the EPON protocol and operating on
point--‐to--‐multipoint RF distribution plants comprised of
all--‐coaxial cable or hybrid fiber/coaxial media.
If
others think the current wording will not be confusing when going to
IEEE for SG approval then I am not adamant about it. I know the EPoC
group has this understanding, it is outside readers who may find the
current text indicating a preference for more than one PHY.
Never
before was there a problem with such an “understanding”. I would not be
averse to the change you propose, but see it as unnecessary
nonetheless. We have spoken until now about two cable plant types (at
most) so it would be a clear indication for two PHYs (at most)
If
you think “at least one PHY” must remain, then we should have more than
an ‘understanding’ that the industry is driving towards a single PHY
solution. The way it reads now, one might think that if we define one
PHY we meet our objectives and if we define multiple PHYs we exceed our objectives.
The
following wording does not preclude two PHYs but doesn’t require the
reader to have ‘an understanding’ that the industry is driving to a
single PHY solution.
Specify at least one PHY and at most
two PHYs if necessary to support
subscriber access networks using the EPON protocol and operating on
point--‐to--‐multipoint RF distribution plants comprised of
all--‐coaxial cable or hybrid fiber/coaxial media, with a goal to limit
Neither
are my preferred choices of words – I would certainly like to have them
nailed down further.
However,
given the discussion we had at the last meeting, I think leaving the
option of doing two PHYs is OK as long as we understand that the
industry is driving towards a single PHY solution. Put it in other
words. If we write down “one PHY” and a year from now we figure out we
need two, we need at least to modify objectives, which is not as easy
as it seems. If we write down “at least one PHY” right now and one year
from now we figure out we can do everything with a single PHY, we are
good to go.
From
the safety perspective, I think it is prudent to leave it as it is and
work diligently to make sure no more than one PHY is needed.
Define
“minimal augmentation to MPCP” for me please.
There
are plenty of undefined terms in the document, I don’t see why the term
“flexibility” has a higher burden than other terms.
We
will define flexibility more precisely when we know what the channel
models are. If you are still confused what flexibility means it means
variable “QAM constellation density, FEC, guard interval and frame
structure”.
We
should eliminate the opening for multiple PHYs until someone can show
that we can’t do what other standards 802.16e, LTE have done with more
difficult channel model variations.
Define
“enough flexibility” for me, please.
Yes,
I have heard the same and would submit the same reasoning I sent out
before, “If a single PHY can operate over the wildly varying conditions
found in wireless transmission (i.e. LTE or Wi-Fi with huge swings in
signal strength, interference, multipath etc) surely a single PHY
should suffice within the more controlled cable environment”.
Obviously
the PHY parameters would have to be adjustable, i.e. QAM constellation
density, FEC, guard interval and others. This is what is done in LTE
and Wi-Fi.
What
shouldn’t need to change is subcarrier spacing, symbol duration and
other more difficult to change parameters.
As
I sent out before, “The burden of proof should be on demonstrating the
need for two PHYs.”
Having
said all of that, it is much more likely that two PHYs might be
considered to work well with TDD vs FDD.
With
FDD the downstream would be a continuous signal (with idle patterns
when no data is sent) just like DOCSIS which allows easy
synchronization to the downlink and eliminates the need for preambles
at the beginning of a burst (i.e. there are no bursts in the
downstream). Whereas with TDD we have to define downstream frames and
upstream frames along with guard times.
LTE
solves this with two types of frame structures, Type 1 for FDD and Type
2 for TDD (see below).
As
long as we consider two frame structures as defined under one single
PHY then we still need only one PHY.
“The
burden of proof should be on demonstrating the need for two PHYs.”
Therefore
I would suggest a variation on Gomez’ suggestion:
"Specify one PHY
to support subscriber access networks using the EPON protocol
and operating on point-‐to-‐multipoint
RF distribution plants comprised of all-‐coaxial cable or
hybrid fiber/coaxial media, with enough flexibility to operate both
below and above 1GHz and in TDD or FDD duplexing modes."
From
the 5 Criteria discussion, it seems to me the reference to more than 1
PHY has to do with passive versus active coax plants rather than below
versus above 1 GHz.
If
a single PHY can operate over the wildly varying conditions found in
wireless transmission (i.e. LTE or Wi-Fi with huge swings in signal
strength, interference, multipath etc) surely a single PHY should
suffice within the more controlled cable environment.
There
is some concern that operating in the lower bands is significantly
different than operating above 1GHz, thus the PHY might need to be
different. The reason for "at least one" is not to have equal and
competing specifications, but to allow for two implementations that
address multiple sets of coaxial conditions.
If
the same goal can be attained by a single configurable PHY, then we
would probably be okay with that.
I
did not attend the last conference calls, but I have been following
progress through the reflector. I have a question about the wording in
Objective 1, which is now
"Specify at least one PHY to support subscriber access […]"
while
previous proposal was:
"Specify a PHY to support subscriber access
networks […] "
Can
somebody clarify the rationale for the new wording, which seems to
"almost" encourage the group to develop more than one PHY? We have all
seen other IEEE groups that quickly yielded to the temptation of
adopting multiple PHYs as soon as they realized that having the group
agree on a single PHY was hard work. I would hate to see the same thing
happening in EPoC, as this would cause market fragmentation and slow
down adoption.
The
objective should be "one PHY". Only if the group finds at a later stage
a very good reason to develop more than one PHY then that option should
be considered.
Just
to be clear, for me it would be OK if the PHY for downstream and
upstream channels have different parameters. As long as vendors do not
have to face the decision of choosing between multiple and equally
valid PHY options when implementing their products, that would still
qualify as "one PHY" for me.
On
May 8, 2012, at 9:38 AM, Noll, Kevin wrote:
<epoc_objectives_for_may_2012.pdf>
<="" p="">
<="" p="">
<="" p="">
<="" p="">
<="" p="">
<="" p="">
<="" p="">
|