[8023-POEP] 100BASE-T normative specification of 350uH inductance
All,
On a number of occasions it has been mentioned that the normative
requirement in 100BASE-T for a 350uH inductance, combined with the
increase in DC current imbalance that results from the increase PoE+
current, will result in a increase the cost and physical size of PoE+
capable magnetics compared to POE capable magnetics. It has also been
suggested that there are now other ways to address the problem that the
inductance specification is there to address, baseline wander.
A while back I had drafted a maintenance request to change the IEEE 802.3
standard to allow an implementation to either meet the 350uH inductance
requirement or, alternatively, not meet an inductance requirement if the
PHY can still achieve the target BER under worse case conditions through
some other implementation. I have attached the draft below. I suggested
this approach as it is the BER that is really what is of concern, not the
way it is achieved. Havening the specified inductance is indeed one
approach but other implementation approaches should be allowed by the
standard so long as they achieve the same BER.
I think that due to the issues with current imbalance IEEE P802.3at should
consider including the proposed change below as part of its draft. I will
request presentation time at next weeks IEEE P802.3at meeting to allow
consideration of this proposal.
Regards,
David
Rational for revision
=====================
IEEE Std 802.3-2005 mandates, through a reference to ANSI X3.263-1995
(TP-PMD), a requirement to meet a specific measurable inductance on
100BASE-T ports. The reference to ANSI X3.263-1995 (TP-PMD) is contained
in subclause 25.2, the text in ANSI X3.263-1995 (TP-PMD) can be found in
subclause 9.1.7 'Worst case droop of transformer'. This text reads:
---
Baseline Wander tracking by the receiver is dependent on the worst case
droop that can be produced by a transmitter. Droop is directly related to
the Open Circuit Inductance (OCL) which varies with temperature,
manufacturing tolerance, and bias current.
Worst case Baseline Wander Frames vary the transformer bias which causes
the droop to change with data content. This variation must be accounted
for by the receiver to track the Baseline Wander over long frames.
Variation in inductance caused by bias of the transformer can be on the
order of 2:1.
The minimum inductance measured at the transmit pins of the AOI shall be
greater than or equal to 350 uH with any DC bias current between 0 mA and
+8 mA injected as shown in figure 13.
---
I believe that this requirement for a specific inductance in IEEE Std
802.3-2005, through it reference to the TP-PMD standard, is an over
constraint. The fundamental requirement is for the link to operate at an
acceptable Bit Error Ratio (BER) under worse case conditions, including
worse case baseline wander. I believe that the inductance requirement
found in TP-PMD subclause 9.1.7 was added to specifically overcome the
problem of baseline wander but since the development of the TP-PMD
specification other techniques can be successfully and economically be
used to overcome the baseline wander.
The requirement is therefore an unnecessary additional mandatory
requirement for many equipment vendors that has to be met simply to
conform to IEEE 802.3. In these cases the magnetics are physically larger
than would be otherwise required, making the construction of todays
smaller form factor equipment more difficult, as well as being more
costly.
To address this an additional exception should be included in Clause 25
that allow implementers to either meet the inductance defined in TP-PMD
subclause 9.1.7 or to meet a BER requirement. Unfortunately neither the
TP-PMD standard nor Clause 25 provide a BER target for 100BASE-TX however
10BASE-T has a BER target of 10^8 and 1000BASE-T has a BER target of 10^10
therefore a target of 10^9 for 100BASE-T seems reasonable.
Proposed change
===============
Insert a new subclause 25.4.4 as follows, renumber the following subclause
as required.
25.4.4 Change to 9.1.7 "Worst case droop of transformer"
The Worst case droop of transformer specified does not apply if the
following specification is met.
While receiving data from the transmitter specified in subclause 7.1 of
TP-PMD through a link segment specified in 25.4.6, a receiver shall send
the proper PMA_UNITDATA.indication message to the PMA at a symbol error
ratio of less than one part in 10^10 when the receiver is connected to a
noise source as described in 25.4.6.3.
Impact on existing networks
===========================
None. It is additional option to conform to the BER requirement rather
than the inductance requirement. Implementations that in the past have met
the inductance requirement will still comply with the standard.
Implementations that choose to meet the inductance requirement in the
future will still comply with the standard.