RE: [802.3ae_Serial] Tx setup window with OMA
Mike,
As you say, one cannot actually set up at the minimum allowable extinction
ratio. As you say, one of the problems with the old specification method
was the narrow, 1 dB setup window in OMA. Let's say 2 dB is reasonable and
you set up 1 dB of OMA away from the edge, close to your example 5.25 dB
extinction ratio. Then, the actual extinction ratio will vary with
temperature, ageing etc. Unfortunately, the mean power or OMA will too.
For any particular part and mechanism of variation and ageing, these things
will be correlated. (The degree and even the sign of correlation may vary
from design to design and part to part). So we are more likely to run into
the corner of the box, which is truncated by the blue line, than if these
things were independent random variables. Therefore it is wise to be
pessimistic and look for a setup window at a lower extinction ratio than set
up (say 4.75 dB) giving about 4.5 dB of useful setup window. But in doing
this you forgo the opportunity to set up at, or even vary into, a really low
extinction ratio.
You say that a 4.5dB Tx output power window is easier than asking for the
extra dynamic range in the receiver. The Tx output power window is mostly
about mechanical stability and precision, test equipment accuracy, operating
temperature range for the customer, etc.; generally, things which cost
dollars on an ongoing per-part basis. While receiver dynamic range is an
issue within an IC; one pays for it as a once-off (whether as design and NRE
or as volume-dependent pricing), the ongoing cost is a tiny increase in
receiver current consumption.
Comparing the PMDs; for the SR/SW case the multimode connectors may have a
smaller variability in loss hence a smaller setup window may be achieved for
the same stability and precision within a transceiver. And as you say,
other things limit each end meaning that paying for tight tolerance is a
good trade.
For the ER/EW case, doesn't Tx setup window trade off dB for dB with channel
attenuation window (presently 6 dB wide)?
Piers
> -----Original Message-----
> From: Mike Dudek [mailto:mdudek@xxxxxxxxxxxx]
> Sent: 03 August 2001 03:15
> To: DAWE,PIERS (A-England,ex1)
> Cc: '802.3ae Serial'
> Subject: Re: [802.3ae_Serial] Tx setup window with OMA
>
> I have a different way of looking at this problem. For the LR/LW 1310nm
serial
> PMD the minimum extinction ratio allowed is 4dB. The actual extinction
ratio
> will vary somewhat with temperature, aging etc. Therefore setting the
> extinction ratio to exactly 4dB is not possible. Let's assume instead
that the
> extinction ratio is set to 5.25dB (which is still significantly lower than
any
> of the SONET minimum specs). This allows the extinction ratio to vary a
> reasonable amount and leaves a 5dB allowable power set window which is
more
> than adequate. I still think that the maximum output power should be
reduced
> to 0dBm as I think a 4.5dB Tx output power window is easier than asking
for the
> extra dynamic range in the receiver.
>
> Note that the attachment showing the 1GBE LX clearly shows one of the
problems
> with the old specification method. If you hold the laser at constant set
power
> using Automatic Power Control and have changes in laser slope efficiency
due to
> termperature and aging it is very easy to fall out of the Extinction Ratio
> specification (ie at constant power there is only 1dB variation allowed in
> OMA.).
>
> The same general arguement applies for the ER/EW and SR/SW. The intent of
> setting the minimum extinction ratio was not to have implementers set
their
> parts to this it was at least for 1300nm and 1550nm to bound the
interferometic
> noise, and to not include in the standard areas of the plot that clearly
no
> implementer would want to use.
>
> For the SR/SW case the maximum power is being limited by the laser safety
> standards. If we want a wider Tx output power setting range we would have
to
> make the receiver more sensitive. I think this would be the wrong thing
to
> do. At reasonable extinction ratios (say 6dB) the setting window is 4.5dB
for
> the same point in the triple trade off curve you used. Again I think this
is a
> good compromise.
>
> For the ER/EW case I think that increasing the maximum output power is
probably
> a good idea. (2-3dB would seem to me to be the right amount based on the
> arguements above.) Here laser safety is not an issue, and as we are
already
> using attenuators we would not have to increase the receiver's dynamic
range.
>
>
>
> "DAWE,PIERS (A-England,ex1)" wrote:
>
> > As I wrote two days ago, the setup window for the LR/LW, 1310 nm serial,
> > PMD, at low extinction ratio is 4 dB when 5 dB was intended.
> >
> > For the ER/EW, 1550 nm serial, PMD, the window is around 3.5 dB if TPD
> > (transmitter and dispersion penalty) is small and extinction ratio is
its
> > minimum 3 dB, or 0.4 dB (!) at maximum TPD, minimum extinction ratio.
> >
> > For the SR/SW, 850 nm serial, PMD, the window at 3 dB extinction ratio
is
> > very roughly 2 dB; it depends strongly on the triple trade off.
> >
> > The attachment Pave_OMA_LX.pdf is an example of an old style (mean power
> > based) standard that works: the 1 Gigabit Ethernet LX standard. The
setup
> > window is 8 dB wide. Because both the upper and lower limits are
measured
> > on the same basis, the window is 8 dB wide for any extinction ratio.
(P1
> > means the power in the ones.)
> >
> > At 10 Gbit/s we don't want to set a very high receiver overload value,
we
> > want to enable lower extinction ratios, and to enable cost-effective
> > transmitters we wish to specify them on an OMA basis. The attachment
> > Pave_OMA_LRW.pdf shows where the LR/LW, 1310 nm serial, PMD is at in
draft
> > 3.1. The setup window is 4 dB wide at 4 dB extinction ratio.
> >
> > How can we fix this?
> >
> > Options are:
> > 1. Reduce the Tx OMA spec. Would need better receivers and/or
reduced
> > link attenuation. Not attractive.
> > 2. Raise the Tx mean power spec back to 1 dBm. This works.
> > 3. Specify the Tx maximum power on a maximum-in-in-the-ones basis
at
> > e.g. +3 dBm. This seems to work too. It's a reasonable compromise
between
> > the transmitter's primary metric (OMA) and the receiver's overload
concern
> > (power in the ones? mean power? OMA? it may depend on receiver
> > implementation).
> > 4. Specify Tx maximum in OMA, at say +1 dBm. This makes the setup
> > window independent of extinction ratio again but it gives the receiver a
> > hard time.
> > 5. A combination of 2 and 3 above, as illustrated in purple. Not
sure
> > that the benefit over the better of options 2, 3 outweighs the cost of
the
> > extra test involved.
> >
> > Option 2 [should have said 3: PD] isn't an original idea. It was
proposed for 1550 nm in
> > http://www.ieee802.org/3/ae/public/jan01/frojdh_2_0101.pdf and shelved
> > because it wasn't essential to fix anything broken and we wanted to know
> > more about how 1550 nm receivers overload before taking it up.
> >
> > Piers
> >
> >
> ------------------------------------------------------------------------
> > Name: Pave_OMA_LX.pdf
> > Pave_OMA_LX.pdf Type: Portable Document Format (application/pdf)
> > Encoding: quoted-printable
> >
> > Name: Pave_OMA_LRW.pdf
> > Pave_OMA_LRW.pdf Type: Portable Document Format (application/pdf)
> > Encoding: base64
>