MTBF calculation for power supply (DC-DC converter)

[pjshah]

Hi all,
I know that most of the power supply manufacturers, give data on room temperature (25C). Every (95%) power supply data sheet has MTBF at 25C. Is there a way we can calculate this number at higher temperature?? I would like to calculate this at 75C.
I am not looking exact number (in terms of years) but at-least rough estimate.

Any help would be greatly appreciated.

Thank you,
pj

 

[andre_luis]

There are some standards which aims to predict the life cycle of each kind of electronic devices subjected to many environmental conditions. One of the most consistent in my oppinion is the MIL-HDBK-217F, which gives parametrically the value in years in function of the temperature inserted on the formulas.

A practical procedure that could be used to estimate the expected life of a complete product in a systematic way would be putting these values in a proper field on the library schematic of the components placed at a design, so that once the BOM list is generated, a simple application that reads text of this file is able to assign to each component the value calculated for certain temperature, therefore by employng some statistical analysis, could get the life cycle of the whole product.

For the specific case above, you can try to identify on this board the most critical item which accounts the lowest MTBF at current temperature , and then calculate at the best case, what would be the time decreasing due solely to this component.

 

[chuckey]

75C will not give you any headroom for semiconductors to dissipate their heat. Likewise the "best" quality high temperature electrolytic capacitors are only rated at 85C. I think you should redesign your power supply for high temperature working rather then trying to guess if its life will be 1000 or 5000 hours.
Frank

 

[KlausST]

Hi,

to your live expectations:

i think it is not predictable. It depends on all used components.
For some components one expects (experimental?) half the livetime for every 10°C temperature increase. (unconfirmed)
This would mean: 25°C --> 75°C is 5 x 10°C increase. makes 0.5^5 = 3% of the livetime.

Other devices may not be that sensitive on temperature rise, others may be even more sensitive.

*********

Likewise the "best" quality high temperature electrolytic capacitors are only rated at 85C

85°C is standard temperature for leaded electrolytic capacitors.

at farnell 6 out of 10 are rated with 105°C
others are 125, 135,150 °C rated.


Klaus

 

[ravindragudi]

I agree with Chuckey.

MTBF will not be given at elevated temperatures by any OEM, unless it is asked for. May be you can check with your supplier for this directly.

Other point is, what is the need to calculate the MTBF for continuous operation at 75C. Thats too much of a stress for the unit. Obviously it will not be any good number. Instead of that try considering the working temp in periods - like 25C for 40%, 40C for 25%, 50C for 20%, 75C for 15% of life time (which may be 5 years/ 10 years or in terms of k hours). With this you can approximate to one average temp of constant operation for the device. Assume this avg temp comes out to be ~40C. Then you have to get the MTBF from supplier @ 40C, if you don't get it, you can as well experiment it by going for HALT (Highly Accelerated Life Test). For this also you may need some assistance for information on product from the supplier.

 

[KlausST]

Hi,

what is the need to calculate the MTBF for continuous operation at 75C. Thats too much of a stress for the unit.

True for commercial products.

But it is quite commmon for industrial equipment. We are a developing company for industrial devices.
In one application we need to implement a sensor into a pipe with hot fluid. So it is normal that the ambient temperature is 75°C and even more.

In another application we build a sensor mounted on the outside of a chemical reactor. Inside the reactor there is about 1100°C, and even at the cooled outer wall there is more than 75°C.

We have to give our customers the guarantee that the devices work for at least 5 years under that conditions. 24/7 working time.
If a device fails it is our problem. If "some" devices fail, then the customer will not order again. Then we have a big problem.

But to be honest, we "try" to design it to withstand this, but i am not able to calculate/predict the lifetime.
We made our experience and therfore we use some techniques and try to avoid to use some devices (like electrolytic capacitors) to extend lifetime

Klaus

 

[andre_luis]

I worked on a project that we had to literally change the entire BOM in order to turn the product complying to industrial specifications. Except the microcontroller, we were able to find options rated to extended temperature for all parts, but we did not make any tests to measure quantitatively the gain in terms of the additional lifetime achieved. Anyway, on such cases it is not only needed to work in a more efficient coller system, but also to take a more sophisticated approach, performing analysis on the scope of the geometry of the components, its placement on the board, and also on the heat flow.

 

[schmitt trigger]

In addition to what has been observed by the other posters, for power supplies there is a key word: derating.

For instance, if your load requires 5 volt and 10 amps @ 75C, you may want to use a supply that it is rated at 20 amps @ 25C.

And please don't rely solely on fans to cool you down. Fans, unless they are expensive (with ball bearings and hi-temp wire) are low MTBF items. And the power supply itself must have a fan controller to lower its speed and/or turn it off completely when it is not required.