| Thread Links | Date Links | ||||
|---|---|---|---|---|---|
| Thread Prev | Thread Next | Thread Index | Date Prev | Date Next | Date Index |
a few comments:
RJ-45 pins will almost always not mate within the same few microsecond period, so there is always
some differential mode, some pair to pair mode, and some common mode currents generated by cable discharge.
Yes, the PHY must be protected, internally, or otherwise, since at ESD frequencies, the magnetics pass most of the
differential mode energy, and some common mode energy.
Do we really need the MOV to protect the PSE output when it will probably have a large output cap, and smaller caps for EMC?
Maybe, but that is an implementation detail.
As far as protecting silicon on the wire, we need to know what type of silicon process we are talking about.
A 60V power process can obviously withstand a lot more abuse on the wire than a low voltage, high speed PHY process.
One of those the devil is in the details types of problems.
- Rick
-----Original Message-----
From: Miller, Larry [SC7:322:EXCH]
Sent: Monday, September 18, 2000 12:08 PM
To: Dieter Knollman
Cc: R karam; stds-802-3-pwrviamdi@ieee.org
Subject: RE: Static Discharge
These pulses are over with in just a few microseconds, though the current can be in the many-amperes. However, the i-squared-t product is low enough that it does not melt anything other than perhaps a bit of connector plating.
I wish one or more PHY people would jump in here.....
Larry
-----Original Message-----
From: Dieter Knollman [SMTP:djhk@lucent.com]
Sent: Monday, September 18, 2000 11:49 AM
To: Miller, Larry [SC7:322:EXCH]
Cc: R karam; stds-802-3-pwrviamdi@ieee.org
Subject: Re: Static Discharge
Larry,
I was not suggesting the use of an MOV to protect the PHY. I agree it won't work.
The MOV can clamp the voltage between the power pairs and protect the silicon on the wire.
The MOV's can be low cost if you can tolerate hand insertion.
The LAN would not see the MOV. It is simple more power supply capacitance.
What's troublesome to me is that a clamp between the pairs may increase the cable discharge current.
Do the magnetics limit these surges via saturation?
Dieter
Larry Miller wrote:
The clamperator networks we use operate just a bit above normal signal voltages.
That means that the voltage is constrained to be within the IC ratings at all times. (Most 10/100 PHY cable inputs and outputs can stand 10 volts or so for short intervals.)
Obviously, this means that by the time you get up near PHY-damaging voltages the clamp has to be stealing ALL of the current. (Things like MOVs do not have nearly sharp enough clamping characteristic curves, Dieter.)
In view of the rise time of the discharge, the amount of energy available in the pulse, and the fact that the clamping network must not upset 100 MHz 100TX lines that are bothered by just a few pF of shunt capacitance under non-clamp conditions, this is a pretty tall order.
Add to that the fact that you do not want to be spending $$ per port for protection.
I have not found any pre-packaged single gizmo that meets all of the requirements.
PHY vendors have been working on this by doing interesting things on the internal pin connections, etc.
Larry
-----Original Message-----
From: Dieter Knollman [SMTP:djhk@lucent.com]
Sent: Monday, September 18, 2000 10:42 AM
To: R karam
Cc: stds-802-3-pwrviamdi@ieee.org
Subject: Re: Static Discharge
R karam wrote:
> Hi
>
> I ditto Larry on this with a small addition,
> when I showed up to see someone with the 750v number, the next question was
> this sounds about enough to power the whole switch? (laughs)
>
> so we are really talking about a charge that we measured in "volts"....
> add to that what Larry is saying down here, when the Moon aligns with the sun,
> and humidity is just right, and installers are new to the game....
> this can be an issue....
>
> to model this to a first order, Power design has done well at using a charged
> cap
> discharging into the RJ45....
>
> but at the end, Larry/Rick has shown the EFTB test on his detection scheme, we
> need to
> do the same for the silicon on the wire, this is where spikes are coupled onto
> a cable
> and fed into the RJ45 as detection is happening in real time.
[Dieter]
The silicon on the wire can be protected via an MOV varistor across the PSE output.
This will clamp the voltage between the pairs. Unfortunately it does not limit the
current created by the cable discharge.
The PHY will need the solution Larry mentioned.
>
>
> roger
>
> >
> > Unfortunately, that does not work.
> >
> > The problem is not that of a connected cable charging up, but that of having
> > a pre-charged cable (with no built-in resistors) suddenly being plugged into
> > the RJ-45 connector.
> >
> > What you get is a sub-nanosecond rise time pulse with several hundred
> > millijoules of stored energy. This pulse is coming from a resonant cable so
> > what you see is an exponentially decaying damped oscillatory discharge. 5 kV
> > on a 25-foot cable, 700 to 900 V on a 100-m cable.
> >
> > If you are lucky, both connector pins mate at the same time and you get the
> > common-mode rejection of the magnetics helping you by not passing some of
> > this through to the PHYs. A worse case is if only one pin of a pair mates,
> > and you get differential transfer through the magnetics.
> >
> > We (and PHY vendors here) have seen as much as 60 volts appear on the PHY
> > pins.
> >
> > In any case a bleeder resistor that would absorb this would have to be such a
> > low value that it would interfere with normal signals. We have had good luck
> > with diode clamping networks that have a dynamic (clamping) impedance on the
> > order of 1 ohm.
> >
> > This is, fortunately, a comparatively rare phenomenon that usually only
> > occurs when a cable is being plugged in in a new installation. Careful
> > network installers would make sure that they discharged all cables BEFORE
> > plugging them in, dont'cha fellas?
> >
> > We have had a few Environments from Hell where a dry climate, combined with
> > moving belts on assembly lines (opportunistic Van der Graaf generators!),
> > have caused constant, repetitive failures. You know who you are out
> > there....(grin)
> >
> > Larry
> > -----Original Message----- From: Yair Darshan [SMTP:YairD@powerdsine.com]
> > Sent: Saturday, September 16, 2000 2:41 PM To: 'Dieter Knollman';
> > stds-802-3-pwrviamdi@ieee.org Subject: RE: Static Discharge
> >
> > The 2kv cap is not suppose to leak.
> >
> > Assuming that the energy level develops on low capacitance value (few
> > nano-farads), than a "bleed" resistor located between the positive and the
> > negative feeding lines of 10 to 100K can discharge it (lower value - better
> > results).
> >
> > Better way is to clamp the voltage developed from excessive charges by using
> > low impedance path that activated at a certain voltage that fits the system
> > requirements at normal working conditions.
> >
> > Yair.
> >
> >
> >
> > Darshan Yair Chief R&D Engineer > PowerDsine Ltd. - Powering Converged
> > Networks > 1 Hanagar St., P.O. Box 7220 > Neve Ne'eman Industrial Zone >
> > Hod Hasharon 45421, Israel Tel: +972-9-775-5100, Cell: +972-54-893019 Fax:
> > +972-9-775-5111 > E-mail:
> > << < <mailto:yaird@powerdsine.com>>> < <mailto:yaird@powerdsine.com>>>. >
> > < <http://www.powerdsine.com>> >
> >
> > > -----Original Message----- > From: Dieter Knollman [SMTP:djhk@lucent.com]
> > > Sent: å, ñôèîáø 15, 2000 6:53 PM > To: stds-802-3-pwrviamdi@ieee.org >
> > Subject: Static Discharge > > > Hi, > > My background is telephony.
> > I'm totally new to 802.3, so please excuse > my ignorance. > One thing that
> > appears strange is the lack of an intentional discharge > path for the
> > potential on the cable. > The only means that I have found is breakdown of
> > the common mode > termination capacitors. > Are these caps with 2 kV rating
> > intended to be leaky? > > On POTS line interfaces the Tip lead is typically
> > biased around ground > and serves as a discharge path to earth ground for
> > Analog Sets. Do the > IP Phones require a discharge path via the LAN? > >
> > Dieter Knollman > DMTS > Lucent > djhk@lucent.com >