Hello again,
I have figured out the answers to some of my questions regarding DC on
xDSL by talking to a support engineer at Midcom, one of the vendors for
the custom transformer parts used for xDSL. (Trying to get support from
the SDSL transceiver chip vendor would have been hopeless as it's
Mindspeed, but I hadn't previously thought of trying the transformer
vendor instead.) Here is what I've got:
* The most pimped-out design for an ISDN/IDSL/SDSL/SHDSL subscriber
terminal (the kind that would be made by someone who likes to cover
every possible corner case and has the parts and PCB real estate to
spare) uses a fancy DC termination scheme. DC through the line
transformer's winding is blocked with a cap, but another device is
connected in parallel across the line which provides the fancy DC
termination.
Lucent's LH1465 ISDN DC termination IC is the example I am familiar
with. The DC termination is fancy in that the device presents a high
DC resistance until the voltage reaches a certain threshold; then it
"switches on" and the DC resistance becomes quite low; it then stays
"on" for as long as the current flowing through it remains above
another threshold. There is yet another threshold for the maximum
allowed current; when that one is reached, the device increases its DC
resistance to limit that current. Finally, the presence or absence of
DC current is indicated to the ISDN device's microprocessor via an
optocoupler. Fancy indeed.
The stuff above is what I already knew prior to my original inquiry.
* The subject of my inquiry was devices made for SDSL and IDSL (rather
than ISDN or SHDSL) which do away with the expensive fanciness
described above on the presumption/realisation that it would be a
waste of effort and parts because the SDSL/IDSL lines in North America
don't have intentional DC on them.
In the absence of intentional DC on the line one does away with the
LH1465 or equivalent, but there remains the question of what to do
with the DC blocking capacitor: leave it in or take it out? The
choice made in the design of dominant SDSL CPE devices from Copper
Mountain, Netopia and Inefficient Networks has been to leave the cap
in, and that is what I have reproduced on my OSDCU.
The Midcom support engineer didn't have a clear answer to this
question either except to confirm that their HDSL/SDSL transformers
have been designed with the provision for DC flowing through them and
that they are guaranteed to perform within specs with up to 160 mA
flowing through the primary winding. (160 mA is a lot, all the specs
I've seen for over-the-line powering of remote terminals and repeaters
say the limit is 60 mA.)
I have decided to leave the DC blocking cap on the OSDCU board in
place in any future revisions because:
a) In the CPE application I simply "do like the others" and that's
what I'll tell to anyone who asks :-)
b) In the CO application with DC power feeding we do need an open
rather a short in there as explained below.
* Now moving on to the question of interest to me: how would one provide
DC power on the line if such a thing was needed. I've asked the
Midcom support engineer the same question I had asked on this list: is
it better to apply the DC power to the centre split or outside the
transformer? The Midcom guy agreed with my reasoning that either way
ought to work, and the discussion then turned to the question of why
they have designed their transformers to work with up to 160 mA of DC
when the requirement could have been avoided if one simply opted to
apply the DC feed in parallel with the DC-blocked winding.
The Midcom guy explained to me that the core needs to be gapped in
order to make the transformer DC-tolerant, but since they also need to
gap the core in order to provide the required primary inductance (the
SDSL transceiver chip requires the custom transformer to dual-function
as a high-pass filter with a precisely specified primary inductance),
it appears that making the transformers DC-tolerant didn't cost them
much extra.
While the only authoritative answer was that Midcom had designed their
transformers with the specs that they have (including the 160 mA DC
current) because that's what the SDSL chip vendor had asked them to
make, by analysing the situation rationally we came to the conclusion
that applying the DC source or load to the centre split and letting
the feeding current flow through the winding results in a smaller part
count. If one were to apply DC in parallel with the DC-blocked
winding, the DC power supply would at the minimum need a big inductor
at its output in order to provide a high impedance at the data signal
frequencies, whereas if one applies DC to the centre split, the
transformer itself takes care of that.
* The conclusion I have reached for any SDSL designs that we may create
in the open source community is as follows:
+ On the OSDCU we can keep things simple by leaving the DC blocking
cap where it is and not providing any extra taps beyond the main
SDSL jack. The vast majority of users would use it as a CPE device
or perhaps a non-power-feeding CO terminal, and it would simply work
as is. Those very special users who would like to use it as a CO
terminal with DC power feeding would be able to connect an external
power supply (including the big inductors or whatever is needed for
high impedance at the data signal frequencies) in parallel to the
SDSL pair coming out of the OSDCU.
Leaving the DC blocking cap inside the OSDCU as it is would be quite
important for this application scenario, as otherwise any attached
DC supply would be shorted locally!
+ If we ever design an SDSL repeater or some other device that needs
to draw power from the line as an inherent feature, we can make use
of the high DC current capability of Midcom's transformers and
connect the DC load to the centre split without much extra effort.
* Regarding future revisions of the OSDCU board, there is no need to
bring out taps for access to the line transformer's centre split.
I would however like to do something better with the transient voltage
suppressor's earth ground pin than what I'm doing currently.
On the current board this pin is wired to a via that is supposed to
accommodate a single header post, but I now realise that what I have
is not terribly useful. I am reluctant to connect it to the chassis
ground, and I think that the principle of "first, do no harm" would be
better fulfilled by bringing this pin out to some kind of screw
terminal post on the rear panel. The user can then either leave it
unconnected (resulting in a situation identical to the mainstream SDSL
CPE devices which use 2-terminal TVSs with no Earth ground connection)
or connect it to a really good Earth ground with a dedicated green
wire.
What I need to find then is a single-circuit screw terminal block that
would solder onto the PCB and protrude on the edge like a connector,
and hopefully be as compact as possible so that it can be squeezed
into the rear panel of the OSDCU. Any suggestions?
MS
_______________________________________________
geda-user mailing list
geda-user@moria.seul.org
http://www.seul.org/cgi-bin/mailman/listinfo/geda-user
