Q about (ab)using a PC's parallel port

[Norm Dresner]

Background: A co-worker came to me about a month ago and asked,
"I have this small, embedded PC with only three slots which are already
full but I need an interrupt from an external signal. What can I do?"
"No problem," said I, "you can use the printer port's interrupt."
"But I don't know how to do that. It's not DOS, it's running Windows
NT."
"No problem," said I again, "I'll just write a device driver for you."
And so I did. And it tested perfectly on my systems here and on the
embedded system using an interrupt-simulation box I built.

Then, the other day, I got a phone call while I was working on one of our
airplanes.
"We measured the PTTI signal and it's 0-10v. Will that be a problem
with the printer port?"

God bless them for actually measuring rather than just assuming
compatibility and connecting it up.

Anyway, a simple pair of 1K resistors solved the problem, but I got to
wondering. Exactly how much damage would it likely have done to get a
~15-30 microsecond wide, 10V pulse into the parallel port (pin 10 IIRC) at a
1 Hz rate? I realize that if the port were actually designed "properly"
with reversed-bias diodes to ground and +5v that it would probably have
worked for years without any real damage. BUT ... how many PC ports are
actually designed with that kind of safeguard?

Does anyone have any explicit knowledge or anecdotal evidence one way or the
other?

TIA
Norm

 

[John Miles]

Anyway, a simple pair of 1K resistors solved the problem, but I got to
wondering. Exactly how much damage would it likely have done to get a
~15-30 microsecond wide, 10V pulse into the parallel port (pin 10 IIRC) at a
1 Hz rate? I realize that if the port were actually designed "properly"
with reversed-bias diodes to ground and +5v that it would probably have
worked for years without any real damage. BUT ... how many PC ports are
actually designed with that kind of safeguard?

Does anyone have any explicit knowledge or anecdotal evidence one way or the
other?

PC printer ports are tough as nails. I doubt it would have hurt
anything.

(I knew a kid in high school who pulse-dialed a telephone line with a
PIA output from his Commodore 64. It actually worked. I would have
been less surprised to see it leaking holy water.)

-- jm

 

[Alexander Grigoriev]

LPT ports used to be easily fried, if connected/disconnected while a printer
or PC are powered.

 

[daytripper]

Background: A co-worker came to me about a month ago and asked,
"I have this small, embedded PC with only three slots which are already
full but I need an interrupt from an external signal. What can I do?"
"No problem," said I, "you can use the printer port's interrupt."
"But I don't know how to do that. It's not DOS, it's running Windows
NT."
"No problem," said I again, "I'll just write a device driver for you."
And so I did. And it tested perfectly on my systems here and on the
embedded system using an interrupt-simulation box I built.

Then, the other day, I got a phone call while I was working on one of our
airplanes.
"We measured the PTTI signal and it's 0-10v. Will that be a problem
with the printer port?"

God bless them for actually measuring rather than just assuming
compatibility and connecting it up.

Anyway, a simple pair of 1K resistors solved the problem, but I got to
wondering. Exactly how much damage would it likely have done to get a
~15-30 microsecond wide, 10V pulse into the parallel port (pin 10 IIRC) at a
1 Hz rate? I realize that if the port were actually designed "properly"
with reversed-bias diodes to ground and +5v that it would probably have
worked for years without any real damage. BUT ... how many PC ports are
actually designed with that kind of safeguard?

Does anyone have any explicit knowledge or anecdotal evidence one way or the
other?

pin 10 would be the -ACK pin, a good choice as an external interrupt.
A Nat Semi PC87413 super io chip has a "ttl compatible" input with its well
firmly nailed to vdd (5V). They spec the max dc input voltage of 5.5v (@vdd =
5). Not much wiggle room on paper, they don't list exceptions for this type
io cell.

There are practical allowances for over/undershoot for pretty much every
technology I've ever used, but the time domain was usually in the five's of
nanoseconds, not tens of microseconds, and the voltage excursions were
measured in 10ths of volts, not 5 volts of overshoot.

In this case they'd have been whistling past the graveyard, I think ;-)

/daytripper

 

[Ben Bradley]

Background: A co-worker came to me about a month ago and asked,
"I have this small, embedded PC with only three slots which are already
full but I need an interrupt from an external signal. What can I do?"
"No problem," said I, "you can use the printer port's interrupt."
"But I don't know how to do that. It's not DOS, it's running Windows
NT."
"No problem," said I again, "I'll just write a device driver for you."
And so I did. And it tested perfectly on my systems here and on the
embedded system using an interrupt-simulation box I built.

Then, the other day, I got a phone call while I was working on one of our
airplanes.
"We measured the PTTI signal and it's 0-10v. Will that be a problem
with the printer port?"

God bless them for actually measuring rather than just assuming
compatibility and connecting it up.

Anyway, a simple pair of 1K resistors solved the problem, but I got to
wondering. Exactly how much damage would it likely have done to get a
~15-30 microsecond wide, 10V pulse into the parallel port (pin 10 IIRC) at a
1 Hz rate? I realize that if the port were actually designed "properly"
with reversed-bias diodes to ground and +5v that it would probably have
worked for years without any real damage. BUT ... how many PC ports are
actually designed with that kind of safeguard?

All the parallel-port expansion cards I've ever seen have about ten
capacitors, one between every line (I forget which direction, but
I'll guess input)and ground (or that's what the traces looked like).
These might be high enough value (like .001 or .01 uF) that a 15 uS
pulse through a 1k resistor might not even reach the proper level at
the pin.
Like others said, you're not going to hurt it. The problem is its
robust design may keep it from doing what you think it should do.

Does anyone have any explicit knowledge or anecdotal evidence one way or the
other?

There are surely schematics of parallel interfaces online...

 

[Paul Burke]

LPT ports used to be easily fried, if connected/disconnected while a printer
or PC are powered.

That was what all the manuals said, but I think it was just arse
protection, unless the mains wiring was so primitive that earth levels
on adjacent sockets were wildly different.

Paul Burke

 

[Frank Bemelman]

That was what all the manuals said, but I think it was just arse
protection, unless the mains wiring was so primitive that earth levels
on adjacent sockets were wildly different.

I ruined an on-board lpt1 port once, plugging in a printer. Perhaps
the printer wasn't grounded at all, and static discharge being the
culpritt. Bought a LPT isa card to 'fix' it.

 

[Stefan Heinzmann]

pin 10 would be the -ACK pin, a good choice as an external interrupt.
A Nat Semi PC87413 super io chip has a "ttl compatible" input with its well
firmly nailed to vdd (5V). They spec the max dc input voltage of 5.5v (@vdd =
5). Not much wiggle room on paper, they don't list exceptions for this type
io cell.

Such chips usually have protection diodes to Vdd built into the inputs,
hence the 5.5V limit (the diode starts to conduct there). The more
interesting spec is the maximum current through this diode, but that
info is not always given. It is not unreasonable to assume that they
will survive 10-20mA. Hence you can protect the input from overvoltage
by using an appropriate series resistor.

 

[Jan Panteltje]

That was what all the manuals said, but I think it was just arse
protection, unless the mains wiring was so primitive that earth levels
on adjacent sockets were wildly different.

Paul Burke

Look, it says max 5V, so make it that.
Ports can be fried, these days it goes directly to the chipset.
Look up dataheet of chipset (if you can get it) and look at specs of input,
how they measured it.
Replacing mobo is expensive.
More expensive then designing in a good interface.
JP

 

[budgie]

Such chips usually have protection diodes to Vdd built into the inputs,
hence the 5.5V limit (the diode starts to conduct there). The more
interesting spec is the maximum current through this diode, but that
info is not always given. It is not unreasonable to assume that they
will survive 10-20mA. Hence you can protect the input from overvoltage
by using an appropriate series resistor.

and in the absence of knowing for sure that there is an internal clamp diode,
I'd stick on an external schottky clamp.

 

[Rob Stow]

I ruined an on-board lpt1 port once, plugging in a printer. Perhaps
the printer wasn't grounded at all, and static discharge being the
culpritt. Bought a LPT isa card to 'fix' it.

Reminds me of something that happened to some computers I was
in charge of in the mid 1990's. I had an Alpha-based server
running NT 3.51 - and no LPT on the server. When a techie
from Digital wanted to copy some stuff off a zip disk and onto
the server, he hooked up a parallel port Zip drive to a PC,
intending to copy stuff from the Zip onto one of the server's
shares.

He toasted three parallel ports on three different PCs before
he twigged to what he was doing. And no, he wasn't doing
hot connects - the PCs were spare network clients he pulled off
of my shelves.

Digital was extremely embarrassed and promptly replaced all
three PCs with significantly better ones. I was ready to fight
them if they tried to fob me off with some LPT cards, but
they didn't even try to do that.

 

[Stefan Heinzmann]

and in the absence of knowing for sure that there is an internal clamp diode,
I'd stick on an external schottky clamp.

You can't without modifying the PC hardware since the 5V supply is not
available on the connector.

 

[Spehro Pefhany]

You can't without modifying the PC hardware since the 5V supply is not
available on the connector.

Yes, and the Schottky could form a path to put 10V on Vdd for the
whole computer if the series resistor was omitted or got shorted.

I think I'd prefer to see a BJT pulling down one of the lines with an
internal 4K7 pullup..

Best regards,
Spehro Pefhany

 

[daytripper]

Such chips usually have protection diodes to Vdd built into the inputs,
hence the 5.5V limit (the diode starts to conduct there). The more
interesting spec is the maximum current through this diode, but that
info is not always given. It is not unreasonable to assume that they
will survive 10-20mA. Hence you can protect the input from overvoltage
by using an appropriate series resistor.

You're certainly free to exceed the vendor's specified Absolute Maximum Rating
for Vih. In the absence of any vendor information regarding the existence of
input clamp diodes, I'd definitely put external protection on those nets
- or just avoid exceeding the spec ;-)

/daytripper (fortunately, this specific subject is going the way of the dodo)

 

[Ian Stirling]

You can't without modifying the PC hardware since the 5V supply is not
available on the connector.

5V zener.

 

[Rich Grise]

That was what all the manuals said, but I think it was just arse
protection, unless the mains wiring was so primitive that earth levels
on adjacent sockets were wildly different.

Paul Burke

I went through 2 or 3 combination boards (2 serial, 1 LPT) I think by
accidentally hot-plugging, or maybe by looking at them wrong. I worked
around it by building a little daughter board that replaced the
Chinese chip with the circuit right out of the IBM tech. Ref.
It worked for the rest of the life of the computer.

I also had to put an oscillator on the daughterboard, because the
Chinese lame chip had the clock for the UARTs on it. That was
kinda cool - a Pierce (or is it Clapp?) osc, with a hand-wound
collector choke.

Cheers!
Rich

 

[larrymoencurly]

Background: A co-worker came to me about a month ago and asked,

"We measured the PTTI signal and it's 0-10v. Will that
be a problem with the printer port?"

Several years ago, I was able to buy a used FIC PA-2007 mobo cheap
because its serial ports didn't work. I found that a couple of
surface mount capacitors in the circuit that converted between TTL and
serial port voltages were missing, and installing new capacitors fixed
them. But a few months later a friend of mine plugged or unplugged a
parallel printer while everything was turned on, and again the serial
port went bad, and those capacitors disappeared. The south bridge
handled both the parallel and serial ports, and new capacitors fixed
everything again.

Would it hurt to put a low-value resistor in series with the signal
line and a zener and a reverse diode for protection?

 

[budgie]

You can't without modifying the PC hardware since the 5V supply is not
available on the connector.

I do so much port interfacing here that all the PC's have the +5 and +12 supply
rails extended to support external devices. I have taken it for granted, and
keep forgetting that this is not everyone's MO.

So it's a 5v1 zener clamp and a series resistor, OR tie the clamp diode to an
external +5 rail if there is one in the connected hardware. After all, it
doesn't have to be the internal 5V rail to clamp excursions, as long as they
share a common reference rail eg 0V.

 

[Franc Zabkar]

Background: A co-worker came to me about a month ago and asked,
"I have this small, embedded PC with only three slots which are already
full but I need an interrupt from an external signal. What can I do?"
"No problem," said I, "you can use the printer port's interrupt."
"But I don't know how to do that. It's not DOS, it's running Windows
NT."
"No problem," said I again, "I'll just write a device driver for you."
And so I did. And it tested perfectly on my systems here and on the
embedded system using an interrupt-simulation box I built.

Then, the other day, I got a phone call while I was working on one of our
airplanes.
"We measured the PTTI signal and it's 0-10v. Will that be a problem
with the printer port?"

God bless them for actually measuring rather than just assuming
compatibility and connecting it up.

Anyway, a simple pair of 1K resistors solved the problem, but I got to
wondering. Exactly how much damage would it likely have done to get a
~15-30 microsecond wide, 10V pulse into the parallel port (pin 10 IIRC) at a
1 Hz rate? I realize that if the port were actually designed "properly"
with reversed-bias diodes to ground and +5v that it would probably have
worked for years without any real damage. BUT ... how many PC ports are
actually designed with that kind of safeguard?

Does anyone have any explicit knowledge or anecdotal evidence one way or the
other?

TIA
Norm

The original IBM AT serial/parallel adapter has an internal pullup
resistor on the -ACK input. Measuring the voltage on pin 10 should
confirm if it is tied to +5V or if it is floating. I can't imagine it
would be floating, as this could generate spurious interrupts (?).
Assuming pin 10 is pulled up to +5V, I would then use an optocoupler
(eg 4N2x) to provide isolation between your PC and your signal source,
for total peace of mind, and to avoid earth loops, etc.


- Franc Zabkar

 

[Stefan Heinzmann]

I do so much port interfacing here that all the PC's have the +5 and +12 supply
rails extended to support external devices. I have taken it for granted, and
keep forgetting that this is not everyone's MO.

So it's a 5v1 zener clamp and a series resistor, OR tie the clamp diode to an
external +5 rail if there is one in the connected hardware. After all, it
doesn't have to be the internal 5V rail to clamp excursions, as long as they
share a common reference rail eg 0V.

A 5V1 zener clamp is not equivalent to clamping the signal to the PC's
5V rail with a diode. Just think about what happens when the PC is
switched off with the external device still on. The original IBM
implementation of the parallel port used TTL logic chips (LS series).
These do clamp the input signals to the rails, so feeding in 5V with the
PC turned off could harm the PC. Contemporary interface chips may be
implemented differently, they might be able to tolerate 5V regardless of
whether they are being powered or not.

If you're not depending on maximum speed you may resort to a solution
others have proposed, namely to rely on the ACK pin being pulled up in
the PC, and using an open collector driver (or just a transistor) to
drive it.