On 14 Jan 2006 17:27:08 -0800, "w_tom" wrote:

No one can accurately answer your question without numbers. Up
front, numbers such as current and voltage should have been provided.
Is this a 3 volt system or a 300,000 volt system? Also the environment
should be considered.

Which means only a generic answer can be provided. Any single point
failure has a protective backup. For example, transistor switch
current limited by an emitter resistor in series with a fuse,
polyswitch, or overvoltage crowbar. But again, we don't even know what
the danger is - with numbers. Therefore a useful answer is not
possible.

I know what you are driving at, but how exactly will you apply the
numbers? There are plenty of regulations covering increased hazards
for, say high voltage or explosive atmosphere. I did not mention these
- obviously my question implied they do not apply. Hence I made it
explicit that I am interested in generic standards and best practice.

You should also notice that I explicitly said "small transistor" thus
ruling out 300KV and 10KA systems. I would be most interested to learn
how different voltage and current levels would affect your analysis,
given that whatever they are, they need to fit the handling capacity
of a "small transistor". How is 12V 120mA going to be any different
from, say 5V 1mA or even (pushing the "small transistor" term to its
limits) 48V, 1A?

I also suggested the Uncle Fred might fiddle with the external wiring
thus creating a hazard. Since high voltage and high current cabling is
obviously not accessible, this scenario implies that the power levels
are small - just enough to blow a "small transitor" but not enough to
warrent physically protected cables.

As for the danger, you do know what it is. I explicitly said that the
failure scenario is overload of a semiconductor - the small transistor
mentioned later - with the possible end result of ignition of the
part.

As suggested by the term "Uncle Fred", the application is domestic
consumer. However, I do not have any data on Uncle Freds so I cannot
provide numbers for you calculations.

FWIIW ,the small transistor circuit in question is a BS160 FET driving
a 12V load at 120mA but subject to possible short circuits as said.
The system fuse is 1A but fuses do not blow instantly so, with the
fairly high "on" resistance of the FET (rising as it heats up) there
is the distinct possibility of the TO92 device dissipating many watts
before failing. The electronic protection comprises a foldback
regulator and is perfectly adequate unless, of course it fails first,
leaving the circuit unprotected without any indication of the latent
problem.


Derek Potter wrote:
I need some information about the current approach to safety of new
equipment with regard to fire hazards created by a fault. I appreciate
there are generic standards covering the ejection of molten metal and
so on, but I am wondering about the application of the "single
component failure" concept in situations where a failure could
overload a semiconductor with the possible, though unlikely, result
that it ignites or ignites an adjacent part. I'm not asking about
*techniques* to avoid hazard, I'm asking about what is legally
required. Trick question - I'm not asking for legal advice, just
information concerning best current practice. I'm posting from the UK
but I suspect the regs will be substantially the same in all of Europe
and the US.

My particular concern is a small transistor driving an external alarm.
The power supply has a fuse but as it feeds several circuits, it
doesn't provide much protection for the external alarm circuit. I am
considering an active current limit in the supply but I still have
some reservations as to whether this meets the letter of the law. For
example, one scenario involves two events as follows:

1 The current limit fails spontaneously, but as this is not
monitored, the defect remains undetected, waiting for the second event
to happen...
2 Someone fiddles with the external wiring and causes a short.

The result is that the driver overheats, catches fire and there is
hell to pay. Now, it is perfectly true that this involves two
independent "failures" so at first sight would meet the "single
component failure" criterion. However, I suspect that a fault that is
never detected (and just lies there waiting for a chance to create a
hazard) may not count. Likewise, a fault that could be caused by Uncle
Fred with his screwdriver is hardly a spontaneous component failure.
So overall, would such a system meet the "due care" criterion?
...