blown serial ports

It's probably failing due to induced _currents_ and no limit (like your 12V case).

Can you add R's in series close to the chip (or make a plug adapter that contains the R's)? ...Jim Thompson

On a sunny day (Sat, 30 Jan 2010 10:56:46 -0800) it happened John Larkin wrote in :

If speed is not an issue, then optos I have seen used (say 1200Bd or use fast optos). The way this is done is like the diagram below. The communication software should set DTR to + and RTS to - For the reverse direction the same mirrored, that is TXDATA can directly drive an opto.

PC port ------------------------ collector

----| |-------------------- RXD data in ------------------------ emitter | R1 | -- RTS (-12V)

the limiting factor is the cable capacitance.

Please also not that you SHOULD NEVER connect a hard +12V to a RS232 *OUTPUT*. That will kill the chip. And in case of connection to inputs, use a current limiting resistor, just in case.

optos).

an opto.

PC port

case.

The Maxim/TI part is rated for +-25 applied to an input. Apparently they are rated for +-13.2 at a transmitter output.

I always dislike situations where silicon is wired directly to connector pins.

John

On a sunny day (Sat, 30 Jan 2010 11:24:20 -0800) it happened John Larkin wrote in :

optos).

an opto.

PC port

case.

The extra limiting resistor will protect the chip if accidently connected to output, and also what makes contact first? Ground? D-25 or D-9 have no leading ground pin.

Some PC software wants to see a DSR (data set ready) signal, normally, for example in case of a nul modem cable, this is done by locally connecting DSR to DTR (dat terminal ready), not by stuffing 12V into the PC DSR from the device.

It occurs to me that chips of this sort use a disconnect scheme, much like tri-state, to claim "voltage tolerance". So any inductive source spells death :-( ...Jim Thompson

Transzorbs have high capacitance, could be an issue if you use fast data rates and the drivers might run on the toasty side.

optos).

drive an opto.

PC port

case.

The only solution I found that realy works is to have diodes to the rails, a small resistor to the respective IC pin _and_ a rail bleeder circuit.

The resistor just makes sure that the substrate diodes don't take a large share of the hit but the new protector diodes do. The rail bleeder makes sure the induced surge doesn't drive the rail up to unsafe levels for the chips. Something around the TL431 is what I usually do, the hard stuff is taken care of by the bypass caps. This has dropped field failure to almost zero in many cases.

Mind PE ground difference situations especially in older buildings like yours. If the gear is on circuits coming off different subpanels you really need isolated RS232. Can cost a couple hundred but compared to the downtime and repair hassle that's often peanuts. I've seen situations where two chassis connections lit a car light bulb. Then some big load came on ... phut ... blew the filament.

=A0 =A0...Jim Thompson

=A0 =A0 ...Jim Thompson

=A0 =A0| =A0 =A0mens =A0 =A0 |

=A0 | =A0 =A0 et =A0 =A0 =A0|

=A0|

=A0 =A0 =A0 |

I can tell you that at Maxim, the ability to short output or input pins to rails was not done at a tri-state condition. Apps would test with both the charge pumps providing the voltage and even with hard supplies on the rails that were normally outputs of the pumps. The only weasel word in the whole scheme was the short condition would be indefinite.

Not totally relevant, but the latchup testing is done with both the pumps floating and with hard supplies at the pump outputs. The only weasel in the scheme would be latch-up testing with the pumps actually running. This is an ugly test since forward biasing parasitic diodes can cause some analog hickups. Later generation parts were designed to be as "digital" as possible so that carrier injection due to foward biasing parasitic diodes wouldn't be an issue.

Now when I say not totally relevant, in some ways applying external voltages is relevant during start up. At some point in a 232's life, it has to be powered up while an external voltage is applied since the other guy in the network could be on. Apps took this a step further and would test the parts with shorts to ground or hard supplies when starting up.

If John is really hard core, I'd suggest checking if HP will allow the system to be use in failure analysis. If all went to plan, HP would pull the chip, send it off to TI, then get a report back as to why it failed. This is how parts or systems get more robust. Of course, you would have to insist on getting the system back sans 232, and you could be out a computer for a month or two.

"John Larkin" wrote in message news: snipped-for-privacy@4ax.com...

Maybe an isolated converter would help. 40' of cable tells me you may have a ground potential problem. Especially when things go "POOMP".

These are neat :

Pricey tho.

I always do isolated RS232 on the PIC stuff I do, just a few extra components.

And I've found the USB to 232 dongles to be very tolerant in our house test sets with out isolation.

Cheers

A crude look at the specs for transorbs seem to show that they are about as bad as MOVs or SiC protectors. Use reasonably fast lo C clamp diodes to the +/- 15 supplies and a goodly cap right at where the diodes connect to supplies (caps to ground). Velly short leads on all that, signal feed via resistor to clamps, another resistor from there to RS232 IC; repeat for each and every signal lead from/to outer world. I think that it is Computer Micro Semi tat makes some rather decent diode arrays for ESD protection (if i have the name right).

optos).

drive an opto.

PC port

case.

Well, the RS-232 spec requires that 25V rating, but every maker of RS-232 chips does it differently and usually does it wrong and in some cases fail to meet (or even get close to) spec. Push comes to shove, set-up a chip on a test board and check out the V/I curves (use current limiting!). You may be greatly surprised.

...and a wiped HD at best!