Repeated overload capacity of 5 watt WW resistor

[Circlotron]

I have an application where a lead acid battery has to be automatically load tested once every 24 hours by applying a load for 5 seconds and ensuring the voltage doesn't fall below a certain value. Four 5 watt wire wound resistors switched by a MOSFET package nicely. The actual dissipation works out to 20 watts per resistor (4x nominal rating) and for 5 seconds this looks to be okay. What I would like to know, is there some kind of long term degradation possible using the resistor this way? Some datasheets say the resistor is okay for up to 10x rating for 5 seconds but they don't tell you if this is a once off or as many times as you like, provided things cool down in between. Any ideas?

[T3sl4co1l]

Probably, due to thermal cycling.  I've never seen a MTBF rating though.

Tim

[Tomorokoshi]

One thing I have to look at in some circuits is what happens with a software failure or certain hardware failures. For instance, if the controller got stuck with the load being left on because of a software failure. Then do the analysis with the resistors left on all the time. How hot will they get? Will they fail open? Will it cause a fire? Etc.

[Circlotron]

Good point. Maybe I'll conduct a few tests with a PTC  series and in thermal contact so if they do get too hot it will back off to a reasonable temperature.

[MK14]

The following application note, about pulse overloading their wire wound resistors, may give you some help.

A quick glance at it, seems to say that you MUST keep the resistors wire temperature below the maximum allowed, during the overload.

http://www.ttelectronicsresistors.com/pdf/application_notes/Pulse-Overload_AN.pdf

[SeanB]

130C 10A non resettable thermal fuse sandwiched between 2 of the resistors covers the long term case, but you would be better off using 8 resistors and stacking them in 2 layers. The long term reliability is not going to be good with more than double the dissipation, as the internal element will undergo a massive physical shock on each stress cycle, and will eventually go open circuit or high resistance on the end caps.

The 10x rating for 5 seconds is a once off rating, based on fusing limits of the internal element wire or foil, it is guaranteed to survive 2 cycles of this, one applied during manufacture as a test of the unit ( often a sample test on a few per batch) and one on acceptance test. In use not recommended, it is a massive stress. Going 2 times the dissipation is a much more survivable event.

[oldway]

In fact, you measure the internal resistance of the battery.

The test you use is fairly rudimentary, if not barbaric.

It is also dangerous in case of failure (short circuit) of the mosfet or of his driver. (risk of fire)

Another option is to measure the internal resistance by injecting a 10 khz current in the battery (same principle as the measurement of ESR of the capacitors).

It may be necessary to provide an inductor in series with the load to make the measurement independent of the load, especially if there are capacitors in the circuit.

Fluke uses this principle in its battery tester 500 series.

[Siwastaja]

In fact, you measure the internal resistance of the battery.

The test you use is fairly rudimentary, if not barbaric.

It's exactly the correct way to measure the DC resistance of the battery.

It is also dangerous in case of failure (short circuit) of the mosfet or of his driver. (risk of fire)

There is nothing dangerous designing power dissipating electronics when designed properly. Proper design includes means such as:
- slow blow fuse to protect from fire if something gets shorted (properly dimensioned, can protect from software lockup too)
- using so-called fusible resistors
- using non-resettable thermal fuse

These measures protect from fire just fine. Similar measures (such as a fuse) are needed for lower power analysis as well, as a short circuit can appear in a lower power device as well, so you have just showed us a typical case of false sense of security.

Another option is to measure the internal resistance by injecting a 10 khz current in the battery (same principle as the measurement of ESR of the capacitors).

This is completely wrong, as it ignores or misinterprets the ionic resistance of the battery, which, unlike in the capacitors, is very significant.

The result is 10kHz AC resistance, which was used in battery manufacturing quality control in the past before EIS analysis; but this AC thing still remains a nice forum myth that needs to be busted all the time. It has very limited significance, and no one has been able to show any practical use for the AC number for a battery system or device designer.

The fact that battery datasheets sometimes provide the 1kHz or 10kHz AC resistance number feeds the myth. But people who actually have to read the battery datasheets every day, quickly notice that lack of proper information is what we have to live with, and is not by any means limited to the lack of DC resistance quantification.

In batteries, there is surprisingly poor correlation between the AC and DC resistance. Two different batteries that have exactly the same voltage sag under the actual load may have the AC resistance differ by the factor of 10.

The actual application most likely needs the quantification of the DC parameters, because the actual load is behind wiring inductance and local elcap bypassing, causing any ESR measured at 1kHz or 10kHz to be irrelevant.

Hence, the 5 second load test is most likely just the right tool for the job.

Sure, the AC impedance test might be good enough for the application, and it may predict the End-Of-Life problem. But it's certainly not the easy (or more "correct" or "finer") solution, because not only it's more complex to implement, interpreting the results requires more knowledge about the battery chemistry, whereas the DC load directly models how the voltage sags under the actual load.

[Circlotron]

whereas the DC load directly models how the voltage sags under the actual load.

And that degree of voltage sag is exactly what I want to know.