BC857 Destroyed

[Kasper]

You're 100% confident no diodes have gone open circuit ?

99.9 but i will happily test again tommorow

Rapzak

 

[Eeyore]

If they have found the source to the problem, this could be the same
problem, so i don't see that as stupid.

Well, Phil Allison reckon it means C-E punchthrough but I don't see how that
helps much.

Your approach to the finding of the fault is not very scientific. Humans once
examined the entrails of dead animal to foretell the future. Examining dead
transistors is little better.

This is also because that i believe it is not a common way the transistoer
is destroyed

You don't ?

do to the functionality it has afterwards.

Uh ? The functionality it has afterwards is NONE.

Graham

 

[Kasper]

Well, Phil Allison reckon it means C-E punchthrough but I don't see how

that
helps much.

Your approach to the finding of the fault is not very scientific. Humans
once
examined the entrails of dead animal to foretell the future. Examining
dead
transistors is little better.

http://rapzak.mine.nu/tr8_1.pdf
Just not easy to see anything on the equipment i have availible for this
point.

Uh ? The functionality it has afterwards is NONE.

If you have read what i have writtenm you have nottiched the layers not
seems destroyd due to even when the impedance between C and E is arround
250ohm, it will be lower when there is a base/emitter current. And also both
base/emitter and base/collector "diodes" are present.

Rapzak

 

[Spehro Pefhany]

No 18V 1276 Ohm for 1 and the 2 other is 1620 Ohm.
Also the software locks for not activating all 3 releays at once.
Then we could discuss just when the uC starts up, all 3 + boost is
activated... hmm...



The allowed peak current is 200mA.

Do you have a datasheet for that part? *With* an SOA curve?

Under some conditions, the PNP is exposed to as much (or more) stress
during switching as the NPN (NPN and PNP switched off simultaneously,
for example).

I would *not* use such wimpy parts as those in this application. Put
MMBT4403 or MMBT2907 parts in there (same footprint, maybe same pinout
(?)) and I'll bet your problems go away, if they are not diode
related. There is negligible price difference. Might be a good idea to
increase the base current on the PNP a bit too, like up to 5mA or
10mA. It's only on for a brief period.

Best regards,
Spehro Pefhany

 

[Kasper]

I would *not* use such wimpy parts as those in this application. Put
MMBT4403 or MMBT2907 parts in there (same footprint, maybe same pinout
(?)) and I'll bet your problems go away, if they are not diode
related. There is negligible price difference. Might be a good idea to
increase the base current on the PNP a bit too, like up to 5mA or
10mA. It's only on for a brief period.

Please take a look:

http://rapzak.mine.nu/tr8_1.pdf

Rapzak

 

[Kasper]

BTW, what prevents the relays and C10 from ringing? Did you try a zener to
ground so at least it can't swing above safe limits for TR8 or below
ground? It might be worthwhile to hang a fast digital scope on TR8's
collector and trigger that with a relay command. It could yield an ugly
surprise.

Yes, this point also came to my mind after i have used a litle time with
spice.

So that i defentlig will try tommorow to measure

(and yes, Kasper is my real name )

 

[Joerg]

Kasper wrote:


Now, is this your real name?

No 18V 1276 Ohm for 1 and the 2 other is 1620 Ohm.
Also the software locks for not activating all 3 releays at once.
Then we could discuss just when the uC starts up, all 3 + boost is
activated... hmm...





The allowed peak current is 200mA.

Be careful with that assumption:
http://www.onsemi.com/pub/Collateral/BC856ALT1-D.PDF

This has a nice thermal graph for pulse loads in there:
http://www.infineon.com/dgdl/bc856s...037a5&fileId=db3a304412b407950112b431bd0b5559

Can't you just use a stronger transistor for a while and see if those
fail as well?

The biggest problem with that are we have seen problems back to last year,
so the problem is made over several production lots.
But we have ofcause checked the diodes on several bad units, and they
measured correctly.

BTW, what prevents the relays and C10 from ringing? Did you try a zener
to ground so at least it can't swing above safe limits for TR8 or below
ground? It might be worthwhile to hang a fast digital scope on TR8's
collector and trigger that with a relay command. It could yield an ugly
surprise.

 

[Joerg]

http://rapzak.mine.nu/tr8_1.pdf

That report says BC856, not BC857. Hmm...

Just not easy to see anything on the equipment i have availible for this
point.

Take a DSO. It doesn't have to be faster than 100MSPS for this job. Look
at the collector node while relays are switching (trigger on that). Try
with all relays since you wrote that there are at least two types used.
If you guys don't have a DSO then borrow it, and buy one as soon as
possible. I don't know which country you are in but it looks like
Europe. In that case Farnell has some pretty good deals, I even bought
one there here from the US.

 

[Eeyore]

If you have read what i have writtenm you have nottiched the layers not
seems destroyd due to even when the impedance between C and E is arround
250ohm, it will be lower when there is a base/emitter current. And also both
base/emitter and base/collector "diodes" are present.

Once it's BROKEN it's BROKEN.

Crying over the dead body of the silicon chip serves NO purpose.

What's important is to stop it happening. There WILL BE some design deficiency
that probably hasn't previously shown up most likely because the parts
previously being used were better than the data sheet suggested.

Graham

 

[Spehro Pefhany]

Please take a look:

http://rapzak.mine.nu/tr8_1.pdf

Rapzak

Okay, here's the data sheet:
http://www.nxp.com/acrobat_download/datasheets/BC856W_BC857W_BC858W_4.pdf

The part appears to be a BC856BW.

An MMBT2907A is actually a bit cheaper, BTW, probably because it's
more common and will replace the wimpier parts in general purpose
switching applications.

Best regards,
Spehro Pefhany

 

[Kasper]

That report says BC856, not BC857. Hmm...

That is an mistake, it is BC556, sorry for that.

Take a DSO. It doesn't have to be faster than 100MSPS for this job. Look
at the collector node while relays are switching (trigger on that). Try
with all relays since you wrote that there are at least two types used. If
you guys don't have a DSO then borrow it, and buy one as soon as possible.
I don't know which country you are in but it looks like Europe. In that
case Farnell has some pretty good deals, I even bought one there here from
the US.

Yes we have several, and it has been done already with no really bad things
to see.

 

[Eeyore]

No it ISN'T ! The rated continuous Ic is only 100mA.

Do you have a datasheet for that part? *With* an SOA curve?

The data sheet shows a PEAK Ic of 200mA with unspecified duration. 100mA
continuous.

The saturation voltage climbs rapidly past 100mA. At 200mA it's off the graph.
Even at 150mA Vce(sat) is around 1V. Over-dissipation failure looks likely at
200mA.

Graham

 

[Eeyore]

I would *not* use such wimpy parts as those in this application. Put
MMBT4403 or MMBT2907 parts in there

Agreed.

Graham

 

[Eeyore]

That report says BC856, not BC857. Hmm...

Danger Danger !

Infineon and Philips disagree over the power rating of the -6 version.

Graham

 

[Eeyore]

That is an mistake, it is BC556, sorry for that.

A BC556 in TO-92 can safely dissipate much more power than a BC856 in SOT-23.

Graham

 

[Joerg]

Yes, this point also came to my mind after i have used a litle time with
spice.

There is some stuff SPICE won't show because it doesn't know about it.
For example ringing of the relay coil itself. Also, the inrush current
of a relay coil can be quite substantial. You can measure that by
placing a 10ohm resistor in the emitter to ground node of the switch
transistors. If you want to measure current spikes at nodes that aren't
at ground level you can make yourself a little current transformer. But
never, ever, omit the burden resistors when you do that. It's a very
handy addition for your toolbox.

So that i defentlig will try tommorow to measure

(and yes, Kasper is my real name )

Sounds German ;-)

 

[Tam/WB2TT]

Schematic:

http://rapzak.mine.nu/tr8.pdf

Offhand, it does not appear that there is anything to keep the collector of
TR8 from going to an arbitrarily large negative voltage.

Tam

 

[James Arthur]

Offhand, it does not appear that there is anything to keep the collector of
TR8 from going to an arbitrarily large negative voltage.

Tam

D1 prevents it.

Cheers,
James Arthur

 

[Eeyore]

Offhand, it does not appear that there is anything to keep the collector of
TR8 from going to an arbitrarily large negative voltage.

Yes.

None of the relays have the classic inverse parallel diode at all.

Whoever drew that schematic ought to have been shot too. I'm hard pressed to
tell what are connections and what aren't.

Graham

 

[James Arthur]

Please take a look:

http://rapzak.mine.nu/tr8_1.pdf

Rapzak

I don't detect anything wrong with the circuit offhand.

I'd double-check the layout to be sure the clamp diodes exist and are
wired as drawn on the schematic.

I'm with Spehro in thinking a bigger, stronger transistor will fix
your problem, but I wouldn't be happy. AFAICT the BC556 shouldn't
break, and you really need to find what's breaking it, and make sure
you've fixed _that_.

I'd try to break more transistors--e.g. beadboard a version where you
o parallel two relays for the load, and / or
o cycle it very rapidly
o increase the supply voltage
etc.

Find out how much TR8 can really take, and learn thereby its mode of
destruction & margin of safety.

Fixing it will be easy once you know what's breaking it.

Cheers,
James Arthur