Weird transistor failures

[Spehro Pefhany]

On Fri, 08 Apr 2005 15:00:01 +0200, the renowned René
<rjz~REMOVE~@xs4all.nl> wrote:

On Fri, 8 Apr 2005 11:43:42 -0000, "Tom R" <TomReer@hotmail.com
wrote:

I have had a number of failures of simple transistor LED drives which I
really can't figure out why:

I am using a standard bc182 driven from 1 k directly from a PIC. The emitter
is grounded, the collector has two white leds (connected in parallel) , a
red LED in series and a 240 Ohm resistor. The drawn collecter current is
around 25mA.
The voltage is 12v dc and everything works perfectly for months
until...eventually (and not always) the transistor begins to leak such that
the LED's become just visible in the dark. The transistor will still turn on
fully but not fully off.

Anybody ?

Thanks
Tom

FWIW - I have seen this behaviour in cheap transistors used in bulk
(e.g. C945, general purpose npn) - in any application it was necessary
to provide a good ground potential to the base to prevent it from
leaking. (it is amazing to see how little current is required to make
a LED glow visibly)
Sure the pic emits a solid "0V" in off position? Resistor directly
from base to ground (~3K3) added to series resistor is std issue with
these part ever since in my book.

Unless you do something silly in the programming, the PIC output is
push-pull, with a fairly low resistance MOSFET to ground when "off",
so effectively he has a ~1K resistor from emitter to base when off.


Best regards,
Spehro Pefhany
--
"it's the network..." "The Journey is the reward"
speff@interlog.com Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog Info for designers: http://www.speff.com

 

[Tony Williams]

In article <d35sd1$uba$2@news6.svr.pol.co.uk>,
Tom R <TomReer@hotmail.com> wrote:

Thanks Win, the PIC is powered from the same supply (via
regulator) as the LED drive. I have deliberately stopped the pic
to simulate a crash and the transistor is quite happy. I have
used drive resistors from 1K thro to 10K and still have the
effect. I have used 2n3704's but still... Not had any complete
failures, just not quite shutting off failures. Have not measured
collector voltages as there appears no heat buildup.....

Do you have any appreciable wire lengths between
the transistor and the LEDs... ie, inductance,
with maybe enough flyback energy to wound the
transistor. It will be the total wire length
from the nearest 12v decoupling capacitor.

.............................and sadly I
cannot get the things to fail in the lab.

Long wires in the field and short wires in the
lab could be a confirmation of the inductance
suggestion.

--
Tony Williams.

 

[Tom R]

Do you have any appreciable wire lengths between
the transistor and the LEDs... ie, inductance,
with maybe enough flyback energy to wound the
transistor. It will be the total wire length
from the nearest 12v decoupling capacitor.

.............................and sadly I
cannot get the things to fail in the lab.

Long wires in the field and short wires in the
lab could be a confirmation of the inductance
suggestion.

--
Tony Williams.

Thanks Tony, the LED's are at the end of quite a long run (about 12") with
the transistor, resistor, caps and PIC all within about an inch of each
other.
Tom

 

[John Woodgate]

I read in sci.electronics.design that Tom R <TomReer@hotmail.com> wrote
(in <d361a7$ndr$3@newsg3.svr.pol.co.uk&gt about 'Weird transistor
failures', on Fri, 8 Apr 2005:

Thanks Tony, the LED's are at the end of quite a long run (about 12")
with the transistor, resistor, caps and PIC all within about an inch
of each
other.

Lecher lines! Betcha they are oscillating!
--
Regards, John Woodgate, OOO - Own Opinions Only.
There are two sides to every question, except
'What is a Moebius strip?'
http://www.jmwa.demon.co.uk Also see http://www.isce.org.uk

 

[Tom R]

Thanks Tony, the LED's are at the end of quite a long run (about 12")
with the transistor, resistor, caps and PIC all within about an inch
of each
other.

Lecher lines! Betcha they are oscillating!
--
Regards, John Woodgate, OOO - Own Opinions Only.
There are two sides to every question, except
'What is a Moebius strip?'
http://www.jmwa.demon.co.uk Also see http://www.isce.org.uk

Thanks for that John. My only question is how is oscillation on the
collector (I presume) erradicated by a cap across the base/emitter ?
Tom

 

[John Woodgate]

I read in sci.electronics.design that Tom R <TomReer@hotmail.com> wrote
(in <d36689$10s$1@newsg3.svr.pol.co.uk&gt about 'Weird transistor
failures', on Fri, 8 Apr 2005:

Thanks Tony, the LED's are at the end of quite a long run (about 12")
with the transistor, resistor, caps and PIC all within about an inch
of each
other.

Lecher lines! Betcha they are oscillating!
--
Regards, John Woodgate, OOO - Own Opinions Only.
There are two sides to every question, except
'What is a Moebius strip?'
http://www.jmwa.demon.co.uk Also see http://www.isce.org.uk

Thanks for that John. My only question is how is oscillation on the
collector (I presume) erradicated by a cap across the base/emitter ?

You don't get oscillation on one electrode; the whole device is
involved. The point here is that you potentially have a VHF tuned
circuit in the collector, formed by the long wires, and that, with the
internal capacitances of the transistor, can make a quite efficient
oscillator.
--
Regards, John Woodgate, OOO - Own Opinions Only.
There are two sides to every question, except
'What is a Moebius strip?'
http://www.jmwa.demon.co.uk Also see http://www.isce.org.uk

 

[Bob Eldred]

"Tom R" <TomReer@hotmail.com> wrote in message
news:d35n56$1jj$1@newsg1.svr.pol.co.uk...

I have had a number of failures of simple transistor LED drives which I
really can't figure out why:

I am using a standard bc182 driven from 1 k directly from a PIC. The
emitter
is grounded, the collector has two white leds (connected in parallel) , a
red LED in series and a 240 Ohm resistor. The drawn collecter current is
around 25mA.
The voltage is 12v dc and everything works perfectly for months
until...eventually (and not always) the transistor begins to leak such
that
the LED's become just visible in the dark. The transistor will still turn
on
fully but not fully off.

Anybody ?

Thanks
Tom

Do you at any time TRIS the port(s) to high impedance in the code? How many
ports are involved, are other bits of the ports inputs at any time, possibly
causing an inadvertent unintended tristate on the active port bit. Check the
code carefully, it may be doing things you don't intend especially if you
write to the port in the wrong bank. When a transistor shows the mentioned
leakage, what is the output voltage on the respective port bit? Is the leaky
transistor actually bad? It's also possible that the PIC under certain
conditions is driving the transistors with short unintended pulses and not
really off. Again check the code. Interestingly the PIC's can putout 25mA so
you may not even need the transistor(s). You might consider FETs with a
2-3volt threshold instead of transistors. Do the PICs ever fail?

You should probably regulate the LED voltage to 8 volts or so and use 20Volt
transorbs on the 12 Volt rail as has been mentioned. It's hard to believe
that transistors may be oscillating as mentioned if they saturate properly.
I guess it could happen but they would never oscillate if you wanted them to
in that kind of a circuit.

Another possiblity is ESD damaging the transistors. They could have been
damaged long ago and failures just showing now. How about heat? What is the
maximum temp this circuit is subjected to? Just some ideas but I'm sure
you've thought of them.
Bob

 

[Pooh Bear]

Bob Eldred wrote:

"Tom R" <TomReer@hotmail.com> wrote in message
news:d35n56$1jj$1@newsg1.svr.pol.co.uk...
I have had a number of failures of simple transistor LED drives which I
really can't figure out why:

I am using a standard bc182 driven from 1 k directly from a PIC. The
emitter
is grounded, the collector has two white leds (connected in parallel) , a
red LED in series and a 240 Ohm resistor. The drawn collecter current is
around 25mA.
The voltage is 12v dc and everything works perfectly for months
until...eventually (and not always) the transistor begins to leak such
that
the LED's become just visible in the dark. The transistor will still turn
on
fully but not fully off.

Anybody ?

Thanks
Tom

Do you at any time TRIS the port(s) to high impedance in the code? How many
ports are involved, are other bits of the ports inputs at any time, possibly
causing an inadvertent unintended tristate on the active port bit. Check the
code carefully, it may be doing things you don't intend especially if you
write to the port in the wrong bank. When a transistor shows the mentioned
leakage, what is the output voltage on the respective port bit? Is the leaky
transistor actually bad? It's also possible that the PIC under certain
conditions is driving the transistors with short unintended pulses and not
really off. Again check the code. Interestingly the PIC's can putout 25mA so
you may not even need the transistor(s). You might consider FETs with a
2-3volt threshold instead of transistors. Do the PICs ever fail?

You should probably regulate the LED voltage to 8 volts or so and use 20Volt
transorbs on the 12 Volt rail as has been mentioned. It's hard to believe
that transistors may be oscillating as mentioned if they saturate properly.
I guess it could happen but they would never oscillate if you wanted them to
in that kind of a circuit.

Another possiblity is ESD damaging the transistors. They could have been
damaged long ago and failures just showing now. How about heat? What is the
maximum temp this circuit is subjected to? Just some ideas but I'm sure
you've thought of them.

Are you a University lecturer by any chance ? ;-)

Graham

 

[Tom R]

You don't get oscillation on one electrode; the whole device is
involved. The point here is that you potentially have a VHF tuned
circuit in the collector, formed by the long wires, and that, with the
internal capacitances of the transistor, can make a quite efficient
oscillator.
--
Regards, John Woodgate, OOO - Own Opinions Only.
There are two sides to every question, except
'What is a Moebius strip?'
http://www.jmwa.demon.co.uk Also see http://www.isce.org.uk

Thanks for your time John, looks like I will belt and braces with a cap and
transient supression.
Tom

 

[Tom R]

Dirty fix- a zener or unipolar TVS across C-E on the transistor could
do the trick (dropping excess voltage across the resistors and LEDs).
You should be able to find an upper limit on typical transients from
various sources (load dump etc, both time and voltage) by searching on
the net, then calculate what kind of part you need. Gut feel-- a 3
cent 1W 30V zener (eg. 1N4751) will be more than good enough, but
that's worth about what you paid for it.. TVS parts can take a lot
bigger surges than regular zeners.

Now, that might start killing the LEDs (or resistors). A better
solution would replace the resistor with a surge-resistant type (if
it's not already), and connect a TVS from the resistor to ground. That
would kill transients without putting current through the LEDs. BTW,
I'm using 30V for the zener/TVS to deal with the situation where
people series two batteries to start the vehicle, so it might see
double the nominal input voltage for a significant period of time. I
hope you're using a regulator that is rated for automotive
applications, or otherwise protecting a more fragile regulator.

Don't lower the base drive, it's pretty good. You might consider using
a more rugged transistor such as a 2N4401, which probably wouldn't
cost any more, and will be even further from coming out of hard
saturation, even under transient conditions.


Best regards,
Spehro Pefhany
--
"it's the network..." "The Journey is the reward"
speff@interlog.com Info for manufacturers:
http://www.trexon.com
Embedded software/hardware/analog Info for designers:
http://www.speff.com

Thanks for your time Spehro, will try your suggestions and report back.
Tom

 

[Tom R]

Do you at any time TRIS the port(s) to high impedance in the code? How
many
ports are involved, are other bits of the ports inputs at any time,
possibly
causing an inadvertent unintended tristate on the active port bit. Check
the
code carefully, it may be doing things you don't intend especially if you
write to the port in the wrong bank. When a transistor shows the mentioned
leakage, what is the output voltage on the respective port bit? Is the
leaky
transistor actually bad? It's also possible that the PIC under certain
conditions is driving the transistors with short unintended pulses and not
really off. Again check the code. Interestingly the PIC's can putout 25mA
so
you may not even need the transistor(s). You might consider FETs with a
2-3volt threshold instead of transistors. Do the PICs ever fail?

You should probably regulate the LED voltage to 8 volts or so and use
20Volt
transorbs on the 12 Volt rail as has been mentioned. It's hard to believe
that transistors may be oscillating as mentioned if they saturate
properly.
I guess it could happen but they would never oscillate if you wanted them
to
in that kind of a circuit.

Another possiblity is ESD damaging the transistors. They could have been
damaged long ago and failures just showing now. How about heat? What is
the
maximum temp this circuit is subjected to? Just some ideas but I'm sure
you've thought of them.
Bob

Thanks for your time Bob, I'm happy there are no firmware probs. The PIC
shows, as near as dammit, 0v when I have investigated the problems and the
tr's are u/s. PIC's have never failed and am currently considering a rare
FET alternative (pin out issues - ain't many DGS or SGD fets about !) .
Temperature is not a problem. My frustration is being unable to catch them
failing. I am considering subjecting new tr's to some big spikes and 'iffy'
base voltages to see if they damage in the same way i.e. partial failure
rather than terminal.
Tom

 

[Rich Grise]

On Fri, 08 Apr 2005 07:01:44 -0700, Mark wrote:

might be load dump and esd

For ESD, I'd put a 0.1 uF cap directly from the from the base to the
emitter and another 0.1 uF cap from the collector to the emitter

and figure that the "+12V" can go to +80 volts during load dumps and
surges and either choose a suitably rated transistor or put some other
form of protection. I like a 100 Ohm 5 Watt resistor in sereis with the
+12 volt feed to a 22 Volt zener and 100 uF cap to ground.

If at all possible, track down a TransZorb to use instead of a Zener.
A transzorb is really just a zener, but it's designed to take huge
transient hits and dissipate them safely. And they're fast - I don't
know how "ordinary" zeners compare for speed, but you won't get one
that's as good as a transzorb for transients - that's where the name
came from.

http://www.google.com/search?q=transzorb

And, of course, as others have mentioned, some ferret[sic] beads. ;-)

Cheers!
Rich

 

[Tom R]

If at all possible, track down a TransZorb to use instead of a Zener.
A transzorb is really just a zener, but it's designed to take huge
transient hits and dissipate them safely. And they're fast - I don't
know how "ordinary" zeners compare for speed, but you won't get one
that's as good as a transzorb for transients - that's where the name
came from.

http://www.google.com/search?q=transzorb

And, of course, as others have mentioned, some ferret[sic] beads. ;-)

Cheers!
Rich

Thanks for this Rich, will give them a try.
Tom