Surge Pulse Clamping with Ceramic Capacitors

On Jun 1, 4:41=A0pm, Joerg <inva...@invalid.invalid> wrote:
> Klaus Kragelund wrote:
> > On Friday, May 31, 2013 7:54:16 PM UTC+2, Joerg wrote:
> >> Klaus Kragelund wrote:
>
> >>> On Friday, May 31, 2013 1:16:42 AM UTC+2, Joerg wrote:
> >>>> Klaus Kragelund wrote:
> >>>>> On Thursday, May 30, 2013 9:32:19 PM UTC+2, Joerg wrote:
> >>>>>> Klaus Kragelund wrote:
> >>>>>>> On Thursday, May 30, 2013 4:43:00 PM UTC+2, Joerg wrote:
> >>>>>>>> Klaus Kragelund wrote:
> >>>>>>>>> On Thursday, May 30, 2013 1:32:10 AM UTC+2, Joerg
> >>>>>>>>> wrote:
> >>>>>>>> [...]
> >>>>>>>>>> But isn't that bleed-through only an issue if the
> >>>>>>>>>> ground of the LTC2862 has come off? Otherwise this
> >>>>>>>>>> would really be a problem because you'd just have
> >>>>>>>>>> moved the dissipation from one place to another. I
> >>>>>>>>>> think the only real protection for 30V continuously
> >>>>>>>>>> is some sort of cut-off, whether inside or outside
> >>>>>>>>>> a chip. At least a partial one where the current
> >>>>>>>>>> becomes very small.
> >>>>>>>>> When the device is in transmit mode, it has no way of
> >>>>>>>>> =A0knowing if an applied voltage is just a load or a
> >>>>>>>>> fault. So it has current limit on the outputs and if
> >>>>>>>>> sourcing into the VDD is detected, the driver is shut
> >>>>>>>>> off. If 5V is supplied to the device externally, only
> >>>>>>>>> the current limit is active and the current is
> >>>>>>>>> steered to the GND or VDD rail.
> >>>>>>>> They probably should have built in an internal VDD
> >>>>>>>> run-up trigger that then limits the current to zero. If
> >>>>>>>> it doesn't let go of the bus at that point that could
> >>>>>>>> present a problem.
> >>>>>>> It actually has that function, AFAIR the trigger level is
> >>>>>>> =A01-200 mV above rail
> >>>>>> Interesting. Then you shouldn't have an issue with the rail
> >>>>>> =A0floating up, or you only need a very light load on the
> >>>>>> rail.
> >>>>> The maximum rail current is 80mA as defines in the datasheet,
> >>>>> so the shunt regulator on the VDD rail must be able to sink
> >>>>> this current
> >>>> Just took a look at the datasheet again. I can't see that 200mV
> >>>> =A0above-rail cut off anywhere. Just an entry on page 3 that the
> >>>> thing can sink up to 250mA into its supply. Because it says
> >>>> -60V to 60V in the conditions box. Then a graph on page 7 that
> >>>> says something different. I'd really contact LTC app support
> >>>> before using it, this datasheet looks too ambiguous to me.
> >>> The 200mV number is a number from the engineers from Linear
> >>> Technology, I contacted them since I saw the same thing you did.
> >>> From LTC: Qoute The answers to the questions regarding LT2862 is
> >>> the following. In receive and shutdown, the lines are high ohmic,
> >>> 112k, meaning no problem if connected forever to +/-60V. In
> >>> transmit mode there is the graph showing pin current for output
> >>> low and output high as function of voltage, assume worst case
> >>> 1.66 x nominal. The 80mA worst case current going to Vcc is
> >>> happening at the narrow spike (output high, a few volts positive,
> >>> peaking at about 60mA typ) in diagram below, at higher voltages
> >>> the current goes to GND. Unqoute
>
> >> Assuming they mean the graph on page 7, middle right:
>
> >>http://www.linear.com/docs/40761
>
> >> That shows that if someone would apply +30V it'll still draw 50mA
> >> times
>
> >> 1.66 for worst case. That's 2.5 watts worst case ... phsssst ...
> >> *POOF*
>
> >> ... I'd say then it's not safe.
>
> > That's the correct figure, but only valid if the device is in
> > transmit mode, during receive mode the currents are zero.
>
> > The only way the device can be in transmit mode is if a telegram is
> > send to it and that's impossible if 30V DC is applied to the line. So
> > no phssssst...... *POOF* :-)
>
> Sure, if there is this sort of "natural" protection via protocol and
> your system will never be sending unsolicited half-hour sermons (or be
> the communications initiator) then you should be safe.
>

you could put some form of inhibit on tx_enable when you see the
fault condition


-Lasse