Question: should the tip/ring polarity on the SDSL connection jack J3 on the
future SDSL DSU boards (OSDCU rev 1 and BlitzDSU) be kept the same as on OSDCU
rev 0, or should it be flipped the other way?

No authoritative answer is available because there exists no standard for
"SDSL" at all, let alone for the connection jacks on the back panels of SDSL
CPE devices.  CR201 and XSB-2000 have 6-position modular jacks; Netopia,
FP/EN/Siemens, NV3133 and 9783-C CPE devices have 8-position ones - yet only
the two central pins are relevant in each case.

Therefore, some experimentation is in order.  In each experiment the plugs
inserted into CPE device jacks were 6-position plugs, whether the device jack
was the 6-position or the 8-position type.

Experiment: connection between two OSDCU rev 0 units.  Results: as expected,
both units showed "correct" T/R polarity when a straight-through cable was used
(one on which the end plugs look identical when held up side by side), and
both units showed "T/R reversed" when a polarity-inverting cable was used.

Experiment: OSDCU rev 0 connected to an XSB-2000 (Lucent DSLPipe).  Observed
the T/R polarity reported by the OSDCU with different cables.  Results: the
OSDCU reported "correct" with a straight-thru cable and "reversed" with a
polarity-reversing cable.  Conclusion: XSB-2000 must be using the same central
pin polarity on its line connection jack as OSDCU rev 0.

Experiment: OSDCU rev 0 connected to a Hack-o-Rocket.  Results: with each cable,
the two ends agree as to whether the T/R polarity is "correct" or "reversed".
However, the SW-reported polarity was "reversed" with the straight-thru cable
and "correct" with the polarity-reversing one.  Conclusion: the jack central
pin wiring polarity is opposite between the CR201s and OSDCU rev 0 boards.

Experiment: OSDCU rev 0, running L2CNF1 configured to emulate a Nokia DSLAM,
connected to Netopia R7200.  Observed results: the jack wiring polarity on the
R7200 appears to be the opposite of that implemented on OSDCU rev 0.

Experiment: OSDCU rev 0, running L2CNF1 configured to emulate a Nokia DSLAM,
connected to Netopia 4652.  Observed results: same as R7200.

Experiment: OSDCU rev 0, running BITDSU configured to emulate a CM DSLAM,
connected to a CM-flavored EN 5851 (3.3V board, model suffix -039).  Observed
results: same as Netopia's SDSL/ATM CPE devices.

Experiment: OSDCU rev 0, running BITDSU configured to emulate a CM DSLAM,
connected to Netopia R7171.  Observed results: same as R7200, N4652 and EN 5851.

Experiment: OSDCU rev 0, running BITDSU configured to emulate a CM DSLAM,
connected to a CupreDSU (Larscom's version of the XSB-2000).  Observed results:
same as with Lucent DSLPipe: the XSB-2000 uses the same jack wiring polarity
as OSDCU rev 0, but it is the opposite of that used by all other CPE devices
for both CM and Nokia flavors, it seems.

Experiment: OSDCU rev 0, running BITDSU configured to emulate a CM DSLAM,
connected to a CR201s running the original fw.  Observed results: same as with
the Hack-o-Rocket: CR201s, despite having a 6-position modular jack, uses the
same T/R polarity as the other "mainstream" CPE devices, NOT the same as
XSB-2000 and OSDCU rev 0.

Experiment: OSDCU rev 0, running L2CNF1 configured to emulate a Nokia DSLAM,
connected to a Siemens 5890 router running in the Nokia CPE mode.  Observed
results: the 5890 uses the same T/R polarity as the 5851 and other mainstream
CPE devices.

Experiment: OSDCU rev 0, running L2CNF1 configured to emulate a Nokia DSLAM,
connected to a 9783-C unit.  Observed results: 9783-C uses the same T/R
polarity as Netopia and FP/EN/Siemens router CPE.

Experiment: a line served from the real D50 DSLAM.  The DSL signal is brought
out from the LCS backplane via what Nokia docs call a sequential cable assembly;
the 50-pin Amphenol connector on the end of the latter mates directly with the
preinstalled connector on the prewired Amphenolized type 66 block I've got from
Fry's; to the best of my inspection ability and understanding all tip/ring
wiring in this setup follows the general telco canons.

The topmost contact on the 66 block is white/blue (white dominant, blue marks -
I assume that's what "white/blue" means in telco-speak) on the prewired Amphenol
rig, and goes to the narrow side of the D shape which, as I understand it, is
called pin 26 in the Amphenol connector pin numbering.  (I have a 25-pair
"telco" patch cable, and the connectors have numbers imprinted in the plastic.
The Amphenol connectors on the end of Nokia's sequential cable have the same
pin numbers on them as well.)  Nokia's manual says that pin 26 on the Amphenol
connector is tip, confirming my understanding that the topmost contact on a
type 66 block is supposed to be tip.  The next contact down on the 66 block is
blue/white, goes to the wide side of the D shape, which appears to be Amphenol
pin 1 which Nokia's manual says is ring - so we are still sane.

I have taken a piece of cross-connect wire pair, punched one end down on the 66
block, and crimped a modular plug (the central pins of a 6-position one) on the
other end.  My cross-connect wire is white/blue and blue/white.  Following what
I believe to be the telco canon in that white/blue should be tip and blue/white
should be ring, I have punched it down on the 66 block side in this order: the
white/blue wire on the very top, on the contact which should be tip, and the
blue/white wire on the contact right below, which should be ring.  On the
modular plug side, I have put the blue/white wire on pin 3 by the RJ-12
numbering, and the white/blue wire on pin 4 by the same numbering, which I
believe are the classic ring & tip positions, respectively, for the primary
pair.

Connected this modular plug (carrying SDSL from the real D50 in what seems to
be the most canonical tip/ring polarity to the best of my understanding) to
OSDCU rev 0: link-up successful, the software on the OSDCU reports the T/R
polarity as "reversed".  Conclusion: this observation strongly supports the
view that the T/R polarity implemented by all other examined SDSL/ATM CPE
devices is right and OSDCU rev 0 is in the wrong.

Experiment: same as the above, but with the CE200 DSLAM instead.  I've unplugged
the Amphenolized type 66 block from the D50 LCS cable stub and connected it to
our CE200 instead.  The Amphenol connectors on the block and on the CE200 are
of the same gender, so I've connected them with the 25-pair telco patch cable
mentioned earlier.  The wire pair running from the 66 block to a modular plug
has remained untouched from the previous D50 experiment - going to the same
OSDCU rev 0, but reconfigured to run BITDSU in the CM CPE mode.  Observed
results: the OSDCU reported the T/R polarity as "reversed", and so did the
CE200.

Conclusion: the T/R polarity implemented on CR201 and on all other examined CPE
devices with the single exception of the XSB-2000 appears to be canonically
correct for Copper Mountain as well.  Taking a 6-position modular plug out of a
Netopia R7100 or out of a CopperRocket and plugging it into an "approved"
CupreDSU would result in "T/R reversed" operation - still perfectly working,
but "officially wrong".  Perhaps the XSB-2000 was "officially" intended to be
used with a polarity-reversing line patch cord.

There seem to be several schools of thought regarding xDSL inside wiring:

1. While not compatible with POTS, SDSL and ISDN-U are POTS-like in that each
   is a single pair, canonically presented on the central pins of the connection
   jack.  While some people seem to frown on it, SDSL/IDSL/ISDN-U devices can
   be connected to their respective wall jacks with ordinary POTS patch cords,
   the same ones you use for your desk phone, analog dial backup modem and fax
   machine.  That happens to be my preferred way of connecting xDSL devices.

2. For the "bigshot" CPE devices that were/are produced in volume and shipped in
   fancy cardboard boxes, it seems to be standard to include an "official" line
   patch cord in that package box.  In the case of CPE devices with 6-position
   jacks (CR201 and XSB-2000), the included "official" cord seems, at least to
   my simple peasant mind and eye, to be no different from the ordinary POTS
   kind.  (The POTS cords usually connect 4 wires, and I don't have an
   "official xDSL" one handy to see if it connects 4 or 6 wires - but in both
   cases, POTS or xDSL, only two wires are actually needed/used.)  In the case
   of CPE devices with 8-position jacks (the apparent majority), the included
   patch cord may be either the same (would fit into a POTS-style wall jack),
   or an RJ45-style cord, requiring the wall jack to be big as well.

3. If both the CPE device jack and the wall jack are of the big kind (RJ-45
   style), it may be very tempting to use one of the ubiquitous Cat 5+ patch
   cords that are used for "Ethernet" - although it is really PseudoEthernet,
   as true Ethernet is a coaxial half-duplex shared medium.  The "official"
   line patch cord shipped with some xDSL CPE devices may be of this kind as
   well.

What does it mean for the issue of tip/ring polarity?  Here are my observations:

* With POTS patch cords it is a toss-up whether tip & ring are connected
  straight through or reversed by the cord.  Both kinds seem to be equally
  common, and as telco-nutty as I am, I still ain't privy to the arcane
  knowledge as to which way is correct in the Kantian categorical sense.

* Ethernet-style Cat3/Cat5/etc 4-pair patch cords with RJ45-style plugs on the
  ends are always straight-through in the sense of T/R polarity: an "Ethernet
  crossover" cable will swap pairs 2 & 3 (pair 1, the only pair that's relevant
  for POTS and xDSL, is irrelevant for "Ethernet" and hence undefined in an
  "Ethernet crossover" cable!), but the T/R polarity of each pair always stays
  intact.

* Some line patch cords that have been shipped with xDSL CPE devices feature
  big RJ45-style plugs on both ends, but aren't Ethernet-suitable: untwisted
  and connect only the central 4 pins.  I have one such cord, labeled
  "ISDN OR ADSL"; the T/R polarity on this cord is straight-through.  I haven't
  seen any cords of this kind that flip the T/R polarity between the ends.

Given a wall jack of the RJ45/RJ48 kind and assuming that the most canonically-
correct in-wall wiring polarity is one that puts ring on pin 4 and tip on pin 5,
two possibilities emerge for the CPE device and the patch cord that goes with
it:

1. If you are designing a CPE device with a 6-position line connection jack,
   with the intent that a POTS-style patch cord will connect it to the wall
   jack, it appears that both of the following options are equally valid:

1a: Use the same T/R polarity on the CPE device as on the wall jack, and make
    the "canonical" patch cord straight-through - the option chosen by Copper
    Mountain for CR201.

1b: Make the CPE device's T/R polarity the opposite of the "canonical" wall
    jack, and make the "canonical" patch cord a polarity-inverting one - the
    option that seems to have been chosen by the makers of the XSB-2000 DSU -
    provided, of course, that it was a deliberate design choice, rather than a
    blunder discovered and made up for after the fact, which we'll probably
    never know.

2. If you are designing a CPE device with an 8-position line connection jack,
   it is probably a safe assumption that all "canonical" patch cords for the
   "8-position mode" are wired straight-through, hence the polarity of the CPE
   device jack should match that of the "canonical" wall jack.

The OSDCU has an 8-position jack for the SDSL connection.  The reasons for it
are historical and no longer matter in the present tense (for the historians: I
once planned to stick the OSDCU board in the plastic enclosure from a gutted
EN 5851, hence the layout of the rear panel connections was loosely copied from
the latter), but now we have our own sheet metal chassis design for our
xDSL-to-EIA530 DSU products, hence the 8-position line connection jack is here
to stay.

Because the option of using device T/R polarity that is the opposite of the wall
jack, with the expectation of using a polarity-reversing patch cord, seems to be
valid only for devices with 6-position jacks, our future DSU boards with
8-position jacks should probably use the T/R polarity used by the "mainstream"
CPE devices with such 8-position jacks, i.e., the opposite of that used on
OSDCU rev 0.

NV3133 update: I have scored one of these new Adtran NetVanta 3133 Nokia SDSL
CPE units from ebay, and subjected it to the same physical exam and T/R polarity
experiment as the earlier CPE devices.  It is a 2-line SDSL CPE device, each
line is connected through its own 8-position modular jack (only the central pins
connected on each jack, no possibility of connecting both lines via the same
jack), and all SDSL interface guts are replicated x2 inside the box.

Single-line experiment: OSDCU rev 0, running L2CNF1 configured to emulate a
Nokia DSLAM, connected to one jack on an NV3133 unit.  Observed results: NV3133
uses the same T/R polarity as Netopia, FP/EN/Siemens and 9783-C CPE devices.

Decision: OSDCU rev 1 will use the opposite T/R polarity from OSDCU rev 0,
aligning our CPE to the de facto standard that has been implemented by everyone
else.
