MTTF of circuit

[Othello]

I have a very simple circuit which switches on an LED when a magnet passes a reed switch. In order to extend the life of the reed switch I kept the switched current to a minimum, about 1mA, by using a transistor to switch the 30mA for the dual color LED.

The manufacturers for the reed switch and for the LED gave me the expected MTTF for their components, but I am at a loss to find the life expectancy for the two resistors and the transistor.
Hopefully someone here has an idea how to find this info.

 

[audioguru]

Name Brand resistors and transistors last "forever".
But cheap no-name-brand junk bought on E-Bay will probably fail tomorrow.

 

[canadaelk]

If you google "free mtbf calculator" you'll find what you need

 

[Othello]

thanks canadaelk, but the calculator you recommended seems to be for windows users only.
Discrimination everywhere...

 

[crutschow]

The life of you circuit will probably be determined by the life of the reed switch, which will likely fail long before any of the electronics.

 

[EternityForest]

I would think at that low of a current the transistor is unnecessary(although probably won't hurt anything)
Number of mechanical cycles and vibration,temp,etc will affect lifetime of the reed switch more than switched current, at least at 30ma. but maybe I'm wrong.

 

[kjennejohn]

I ran the MTBF calculator on this.
Part | Fail Rates per Million Hours
---------------------------------------------------------------
relay, mercury wetted contacts | 6.27
transistor, small signal, silicon | 5.73?!
resistor, ceramic compound | 1.13

I used MIL-217E-1 Reliability Prediction Stress Method, benign ground use, at 25 deg C.
I was amazed at the transistor failure rate! Damn near as bad as the relay! You might as well
save the money and leave the transistor out of the design!
kenjj

 

[dknguyen]

You sure something wasn't missed in the transistor MTBF calculation? Those number are a bit different than what you'd think they would be, particularily the smilar relay and transistor values and the much lower resistor and ceramic compoudn values.

 

[MrAl]

Hi,


I have to agree with dk on this. I've seen many reed switches go bad, stuck closed in many cases, but have seen so few transistors fail with normal operation that i cant even remember them.
Not only that, the failure rate of the transistors changes with applied voltage and temperature too.

 

[Gary B]

Since reed switches tend to stick because they become magnetized over time rather than contact failure, why not use a Hall effect device instead to detect the magnet?

 

[Othello]

Initially I did not use a transistor. I just had the reed switch and two resistors for the two colors of the LED.
The switched current was around 30 mA.
The manufacturer of the reed switch came back to us with a life expectancy of about 30 million cycles, at that current. Introducing the transistor lowered the reed switch current to about 1 mA with greatly increased the life expectancy of the switch.
This little circuit will be used, we will have 12 000 pieces made, in a permanent outdoor art installation, and the switch is wind actuated. The installation only works during darkness, which means the switch only carries current during those hours.
The lower life expectancy at 30 mA yielded a 5+ year useful life, with the lower current the life expectancy of the circuit would go up to 10 to 15 years, a number more acceptable to the architect. The site of the installation is San Francisco.

Our reed switch is of the non-wetted type.
Nobody gave us the optimistic numbers kjennejohn seems to indicate, if I understand his results properly. The reed switch people rate their switch in cycles, with applied current and voltage as parameters. We expect at most one switch cycles per second.

 

[kjennejohn]

You sure something wasn't missed in the transistor MTBF calculation? Those number are a bit different than what you'd think they would be, particularily the smilar relay and transistor values and the much lower resistor and ceramic compoudn values.

I have no idea how the so-called Reliability Prediction Stress Methods are applied or what weights they assign for parts. Otherwise, it's all straightforward. And the transistor and resistor results look bogus.
Your thinkage may vary.
kenjj

 

[kjennejohn]

Othello, at first I didn't think the figures were optimistic at all. Then I did the math and discovered one million hours spans 114-plus years! And I should think that since that is with power applied constantly, and these only run at night, that parts' life spans will double!
kenjj

 

[ronv]

Once upon a time many years ago we built a controller for a mainframe application. It had 4,000 TTL IC's or transistors. We got the failure rate under .3 failures / machine/month or about 1 per 1000 years. I'm sure ic's are better now. They did run in a very nice environment. Even after we had buit 10's of thousands of these we would find design flaws based on failure rate so I have always thought it as important to look at the design as the failure rate.
Having said that if you get 300,000,000 actuations as a life number and you have lets say 10,000 of these little guys going at 1 cycle /second you have about a ten year life /10,000 or .001 years for the whole system. Or you will loose 1 every 8 or 9 hours. Even if you add another 0 to the life (which is a lot for a mechanical device) it's one evey two weeks or so.
If you use the reed here are a couple of things to consider:
Adding the resistor and the transistor have there own failure rate.
If you do use the transistor (probably need to) put the base resistor close to the reed switch to reduce the capacitive inrush current.
Don't bend the leads to close to the glass seal.
Think about cycling them at a few hundred HZ to weed out the early life failures. (Typically 3X the long term failure rate)
If you can do this at elevated temperature all the better. Thermal cycle? Even better.
Have someone inspect the solder joints - they should be perfect.
Any connectors? Batteries?

 

[crutschow]

If you are concerned about reliability, why not use an inexpensive Hall effect device instead of the reed switch, as Gary B suggested, such as **broken link removed** ($0.50). That's probably cheaper than the reed switch.

 

[nsaspook]

I have reed switch position sensors in a device that's been running 7/24 for 20 years (several million cycles switching a few mA). It just started to have sensors errors about a month ago due to contact magnetization (we replace them all) so they are pretty reliable.

 

[RadioRon]

Its all in here:

**broken link removed**

 

[Othello]

If you are concerned about reliability, why not use an inexpensive Hall effect device instead of the reed switch, as Gary B suggested, such as **broken link removed** ($0.50). That's probably cheaper than the reed switch.

This might be worth looking into.
The only reason against it is that I don't know anything about it, but I guess I could search for circuits how these Hall switches are used.

The issue with the reed switch is their odd characteristics, I will have to get the position down rather precisely to give the effect we want. We have a magnet swing by the switch and do not look for a solid on or off for a large arc of the magnet's motion, but we look for a rather brief 'blink'.
With a reed switch that means that the "sensitivity area" needs not to be solidly entered but only lightly crossed, mechanically this is a demanding task.

If the Hall sensors would give me any advantage in this I would wholeheartedly embrace that device. But how is the reliability of those??

 

[Othello]

This might be worth looking into.
The only reason against it is that I don't know anything about it, but I guess I could search for circuits how these Hall switches are used.

The issue with the reed switch is their odd characteristics, I will have to get the position down rather precisely to give the effect we want. We have a magnet swing by the switch and do not look for a solid on or off for a large arc of the magnet's motion, but we look for a rather brief 'blink'.
With a reed switch that means that the "sensitivity area" needs not to be solidly entered but only lightly crossed, mechanically this is a demanding task.

If the Hall sensors would give me any advantage in this I would wholeheartedly embrace that device. But how is the reliability of those??

I just read up on the hall sensors and would would help in our application is a vane interrupter. But while a Hall sensor is small and cheap, I haven't found a similarly inexpensive vane interrupter package.
Maybe someone here has a source??

 

[nsaspook]

I just read up on the hall sensors and would would help in our application is a vane interrupter. But while a Hall sensor is small and cheap, I haven't found a similarly inexpensive vane interrupter package.
Maybe someone here has a source??

Honeywell might have some that will work but maybe not very inexpensive. https://www.electro-tech-online.com/custompdfs/2010/08/1917.pdf