RE: XAUI skew
Tom,
Thanks for the reference to the FC document. This also seems to be where
the numbers for XAUI skew limits came from. To use the same measurement
method for XAUI, AC-coupled 50 ohm loads would be used. Swapping the
cables at the measuring point for calibration is of course a good
practice.
Dawson,
If EMI control is the purpose of the skew spec, only one out of several
factors impacting common mode noise is accounted for. Just as you say,
an AC common mode voltage limit would cover more of them. However, I
think it must be up to the implementer to make the trade-off between
circuit performance and system shielding properties. EMI/EMC are
important properties of any implementation but assumed to be
insignificant from an inter-operability point of view in the XAUI
specification. I think skew numbers belong to a category of properties
that the standard otherwise leaves to the implementers choices for good
designs.
To take skew measurements with the FC method that are repeatable with an
accuracy of 10% of the limit value of 15ps, you need be able to swap
connections at the test point with a physical positioning accuracy of
200um.
If Fibre Channel specifies skew for reasons other than EMI, it could be
a good idea to find out what calculations/trade-offs led to the limits
of 15ps and 75ps. The far end signal from a typical differential pair
with good coupling generally has less skew than the input signal. Poor
trace design with branch length differences can increase the skew, but
that, I assume is also the implementer's domain.
My discomfort with the skew spec as it stands is mainly because there is
no definition, method, purpose, or reasoning behind the numbers.
Ali,
Per resolution of your comment #1067 to draft 2.0, differential skew at
the receiver was changed to include ISI. The connection between skew and
ISI is not obvious to me. I can't find any reference to ISI in the Fibre
Channel spec on which the XAUI skew numbers seem to be based. What kind
of test set-up do you have in mind for generating a skewed waveform
differential signal that includes the effects of ISI to test a XAUI
receiver?
Best regards Tord.
-----Original Message-----
From: Kesling, Dawson W [mailto:dawson.w.kesling@intel.com]
Sent: Friday, February 23, 2001 01:10
To: Serial PMD reflector (E-mail)
Subject: RE: XAUI skew
Tord,
First for some historical background. Differential skew and has been
included since early XAUI proposals. There was a proposal to replace it
with an AC common mode voltage limit in Tampa last November. It was
pointed out then that the skew measurement definition is complex for the
XAUI waveform (particulalry with pre-emphasis), that an AC common mode
voltage specification would better serve the purpose for EMI control and
that jitter is already spec'ed anyhow. The group decided at that time
that no action was required on differential skew. Two comments during
the last balloting cycle specifically asked that skew information be
make MORE explicit. (These were both approved: #1067 and #965.) Finally,
one new participant raised a concern with measuring skew last month in
Irvine, but no specific proposal was made at that time so no action
could be taken.
So where do we go from here? I can see several options:
1. Do nothing and keep the text as is. People apply reasonable
engineering practices to define their own test methods to verify the
skew of the driver and skew tolerance of the receiver. The reasoning
here is that the standard does not have to define test methods (although
it does so in cases such as jitter where considerable confusion might
otherwise exist). This is the default decision unless another option
gets a 75% vote.
2. Define a detailed skew measurement method and document it in section
47.4 (Electrical measurement requirements). Tom's text from FC looks
like a good start and you point out several other details that would
need to be considered. If anyone is interested in doing this, please
submit a comment by tomorrow or at least let me know you'd like to make
a proposal at the Plenary. I suspect this would be supported by a 75%
majority, but someone has to sign up to do it.
3. Remove all references to differential skew as you suggest. I doubt
this will gain the 75% majority needed, but there is only one way to
find out.
-Dawson
-----Original Message-----
From: Tord Haulin [mailto:Tord.Haulin@optillion.com]
Sent: Wednesday, February 21, 2001 2:38 PM
To: Kesling, Dawson W; Serial PMD reflector (E-mail)
Subject: XAUI skew
Dawson,
Clause 47 has specifications on differential skew both for the
transmitter and receiver. There is however no definition for
differential skew, or method given for measuring it. In paragraph
47.3.4.1 the differential skew is said to include the effects of
intersymbol interference, ISI. I would think the main effect of ISI is
deterministic jitter, which is part of the jitter specifications. To
define and measure skew, we would probably have to relate it to rise and
fall times and define a common mode load and load balance with
tolerances. The transmitter skew and possible crosstalk along the line
is filtered by the channel common mode transmission magnitude and group
delay, which falls outside the channel model work done to date.
In the receiver compliance test some additional sinusoidal jitter is
applied to ensure some margin for effects otherwise not accounted for.
How about lumping skew into this, to avoid going into specifying enough
parameters and conditions to take measurements on skew and related
parameters? That would mean the skew limits could be removed from the
spec. Comments?
Tord.