[MidAtlantian]

I'm sure that "how components fail" is a huge topic, but right now I am focused on how a 555 chip, especially the cmos version, is likely to fail.

There are two main conditions for failure: when it "shouldn't have failed", and when it is in some way overloaded: either with voltage or current. What are the chances that the output (pin 3) of the dead chip would end up being high in either of these failures? And perhaps most importantly: how, apart from experience, would one find out this sort of information?

Also, do transistors ever fail in a way that they continue to conduct across the emitter-collector; that is, a short across the base region?

Thank you.

 

[Harald Kapp]

What are the chances that the output (pin 3) of the dead chip would end up being high in either of these failures?

50/50, depending how the output transistor was destroyed.

Also, do transistors ever fail in a way that they continue to conduct across the emitter-collector; that is, a short across the base region?

That's a common type of failure for transistors.

 

[davenn]

I'm sure that "how components fail" is a huge topic, but right now I am focused on how a 555 chip, especially the cmos version, is likely to fail.

There are two main conditions for failure: when it "shouldn't have failed", and when it is in some way overloaded: either with voltage or current.

and a 3rd one since you are worried about CMOS chips
Static discharge is a big killer of CMOS chips

Dave

 

[MidAtlantian]

Hi Harald,

I have a circuit that is ultimately switching on a 30 amp heater for between 5 and 10 seconds. If it stayed on much longer, things would not go well for my three 16 amp, but insufficiently heat-sinked, triacs. I could put in another 555 with a slightly longer interval in series as insurance, or use a a temperature probe as a cut off. The mains circuit is fused and on an RCD, but if the triacs get too hot, I am not sure what happens when they fail dramatically. The circuit as it is cannot be retriggered for 2 minutes, so there is already some protection in the design.

Harvey

**************************************

Hi Dave,

I guess the question is: how much protection do I design in for ANY sort of solid state failure, so it is a bit like a life-insurance policy: you don't have to worry about knowing exactly what will trigger the payout.

Harvey

 

[Gryd3]

Hi Harald,

I have a circuit that is ultimately switching on a 30 amp heater for between 5 and 10 seconds. If it stayed on much longer, things would not go well for my three 16 amp, but insufficiently heat-sinked, triacs. I could put in another 555 with a slightly longer interval in series as insurance, or use a a temperature probe as a cut off. The mains circuit is fused and on an RCD, but if the triacs get too hot, I am not sure what happens when they fail dramatically. The circuit as it is cannot be retriggered for 2 minutes, so there is already some protection in the design.

Harvey

**************************************

Hi Dave,

I guess the question is: how much protection do I design in for ANY sort of solid state failure, so it is a bit like a life-insurance policy: you don't have to worry about knowing exactly what will trigger the payout.

Harvey

Here's what I would do, but I'm not a professional...

-Tie the case to earth ground. (Ensure components inside the case are isolated)
-Design similar to a car audio amplifier, and line up the triacs to use a single heat-sink. Place a thermal fuse on the same heat-sink.
-Place a fuse on the input side of the device.
-Depending on how 'fail-safe' you want this. Design a 'watch-dog' to monitor and time the output. If the output does not behave as expected disable the device. If the watch-dog does not behave, disable the device. (Street lights do this... this is why you see red flashing lights in a 4-way. It's a fail safe and is part of a secondary control board in case the primary control board does something unexpected... ie, two green lights in both directions.)

 

[MidAtlantian]

Hi Gryd3,

Thanks for your suggestions. I can see that a "backup plan" can end up dwarfing the original design, but I like your approach. The idea of a thermal cutoff fuse is an excellent suggestion that I had not considered. I would assume that I could use, say 4 or even 5 ten amp fuses, since they are triggered by temperature, and if they began to go, they would cascade to a full cut-off quickly enough. 30amp fuses may be available, but the 10a are easy to source, and inexpensive. Since the fuses would be anyway a catestrophic shutdown, I would guess that I could get 192C fuses, but I think you might make a much more realistic "guess". So: suggestions please!

Yes, I had got to "one watchdog", and I see that one could go into infinite regress, but I think that the thermal fuses and one watchdog should be enough. And I have a fuse on the DC side of things. If the thermal fuses ever go, then I will have to see what happened, and do a bit of a redesign. But that begins to sound safe enough.

Again, thank you for your help.

 

[Gryd3]

Hi Gryd3,

Thanks for your suggestions. I can see that a "backup plan" can end up dwarfing the original design, but I like your approach. The idea of a thermal cutoff fuse is an excellent suggestion that I had not considered. I would assume that I could use, say 4 or even 5 ten amp fuses, since they are triggered by temperature, and if they began to go, they would cascade to a full cut-off quickly enough. 30amp fuses may be available, but the 10a are easy to source, and inexpensive. Since the fuses would be anyway a catestrophic shutdown, I would guess that I could get 192C fuses, but I think you might make a much more realistic "guess". So: suggestions please!

Yes, I had got to "one watchdog", and I see that one could go into infinite regress, but I think that the thermal fuses and one watchdog should be enough. And I have a fuse on the DC side of things. If the thermal fuses ever go, then I will have to see what happened, and do a bit of a redesign. But that begins to sound safe enough.

Again, thank you for your help.

Sorry, but I can't guess a temperature for you.
If I were you, Id pick a temperature based on the data sheet for your triacs. As long as the thermal fuse and triac share the same heat-sink it should be fine.

 

[Harald Kapp]

If it stayed on much longer, things would not go well for my three 16 amp, but insufficiently heat-sinked, triacs.

The best cure is to use suitably rated triacs with an accordingly large heat sink.

I would assume that I could use, say 4 or even 5 ten amp fuses,

Never use fuses in parallel. That is a sure way to blow them reapidly. Once the first fuse has blown due to whatever reason, the load is distributed along n-1 fuses and these will blow even faster. Avoid overheating in the first place, see above.

Obviously, the 555 and a failure of that chip is not your main problem. Your triac circuit neds to be designed properly.

 

[MidAtlantian]

Thanks Harald.

The reason I do not want "adequate" heat sinks is that it would mean a huge increase in size - not to mention ventilation - for a circuit that should not be active for more than say, 7 seconds, 3 or 4 times a day. Of course, I understand that the fuses would cascade, but since I am relying on them being primarily temperature triggered, and I "overrate" by a factor of even two, I assumed that I would have the protection I want.

And it is not obvious to me that the 555 is not the main issue, since if that works reliably, then the triacs will never be on for more than 7 seconds, and never on again for several minutes. I understand that if the triacs do fail, they do create another hazard, but it seems to me that the timer is the weakest link given its capacitor dependence and cmos design. And that even if the triacs fail, they are not likely to create a serious hazard in 7 seconds.

And even if I am disagreeing, I am finding this discussion really helpful in broadening my way to look at these issues.

Hi Gryd3,

Quite right that I needed to look at the specs. I was not far of, but I really don't have any experience gauging the time it takes for heat to get though to a fuse. The triacs have a working temperature of 125C, or about 250F. But in a failure the temperature of the triacs will be moving up quickly and I do not really have a sense of how fast the fuses respond: how fast heat conducts to the actual sensor, so I need to leave an adequate "head room". That is where thought you might have a better idea than I would.

 

[Gryd3]

Thanks Harald.

The reason I do not want "adequate" heat sinks is that it would mean a huge increase in size - not to mention ventilation - for a circuit that should not be active for more than say, 7 seconds, 3 or 4 times a day. Of course, I understand that the fuses would cascade, but since I am relying on them being primarily temperature triggered, and I "overrate" by a factor of even two, I assumed that I would have the protection I want.

And it is not obvious to me that the 555 is not the main issue, since if that works reliably, then the triacs will never be on for more than 7 seconds, and never on again for several minutes. I understand that if the triacs do fail, they do create another hazard, but it seems to me that the timer is the weakest link given its capacitor dependence and cmos design. And that even if the triacs fail, they are not likely to create a serious hazard in 7 seconds.

And even if I am disagreeing, I am finding this discussion really helpful in broadening my way to look at these issues.

Hi Gryd3,

Quite right that I needed to look at the specs. I was not far of, but I really don't have any experience gauging the time it takes for heat to get though to a fuse. The triacs have a working temperature of 125C, or about 250F. But in a failure the temperature of the triacs will be moving up quickly and I do not really have a sense of how fast the fuses respond: how fast heat conducts to the actual sensor, so I need to leave an adequate "head room". That is where thought you might have a better idea than I would.

Unfortunately, the heating is a little beyond my skill set.
The triac will heat up. The rate will depend on the thermal mass of the heatsink, as well as the rate at which the heatsink can dissipate the heat. If you have some way to measure, Run it for 7 seconds and see how hot it gets.

And as far as the fuses being placed in parallel is concerned. You will have more problems than anything else... just get a bigger fuse. This can be an acceptable practice with mosfets or triacs in parallel if it's engineered properly, but never with fuses.
All it takes is for one of the fuses to get a little hotter than the rest and it will pop, then the rest will cascade. You dont even need to draw too much current for this to happen, it could simply be running normally.

 

[Harald Kapp]

Heating of the triacs is not the only issue when operating them at higher curents than they are rated for. The silicon structures and the bonding wires are not meant to be operated above the rated limits. They can easily be destructed by overcurrent without the triac getting noticeably warm.
With your setup you really risk damage from a blown triac much more than from a defect 555.

 

[MidAtlantian]

Hi Harald,

I am way under the rating of the triacs. I am under heat-sinked IF the triacs are on for very long. If they are only on for the programmed period, they should not get very warm at all. That is why I am focused on the timer circuit.

Also, there have been several comments about parallel fuses, which I can understand in terms of current triggered fuses. But I would have though that temperature triggered fuses would be different. Imagining a case of 100 x 10 amp fuses rated at 100C, and a current of 50 amps. If any of the fuses got warmer than the others from the current flowing though them, the current would shift (slightly) to the others, and it would be very unlikely that any of them would heat up enough from the current flow itself, to trigger their own temperature-triggered fuse - which is usually supposed to be sensing the temperature of something else.

 

[Harald Kapp]

A standard fuse is thermally triggered, that's right, but the thermal energy to trigger the fuse comes from the self-heating by the current flowing through the fuse. Therefore once one of a bunch of paralleled fuses blows, the current shifts to the other fuses, incerasing the thermal load on them and they will trigger onae after the other - unless they are heavily overrated which in turn makes the use of a fuse moot.

One way out of the dilemma could be a self-resetting polyfuse (which is sensitive to external heat much more than a standard glas-tube fuse) mounted next to the triacs onto the heatsink. This way the polyfuse would in a way "sense" the temperature of the assembly and turn it off in case of overheating. In this case, too, only a single fuse should be used.

 

[MidAtlantian]

Hi Harald,

Yes, of course. A standard fuse relies on the heat generated by the current flowing though a metal bridge of a controlled size. But I was not planning to use a standard fuse but one that did not heat up in a way to overload a filament, but that senses an external temperature and blows based on that. I have not found a source for larger current temperature triggered fuses.