Pulse devices time estimate

[Mike Belanger]

It is well documented that semiconductor lasers are characterized by the lognormal distribution (ref. Telcordia GR-468). I have seen several recent examples that use the chi-sq approach but I think there is enough data to say that this is not correct.

My question is to do with time estimation. It is not correct to use total device-hours as the device is pulsed, so it is only active based on duty cycle and the harder you drive the device the more aging you do to the device.

It there a method to descibe the active time of a device based on duty cycle? and how do I use this in standard distribution equations

[Pantelis]

Mike,

With respect to the Chi-Square approach, I assume you are referring to the use of the exponential distribution and total operating time approach (i.e. 2*(Ta)/X2..). That approach (as with any approach that utilizes an exponential distribution) makes the underlying assumption of a constant failure rate, which is - more often than not - an incorrect assumption - as you point out.

Now as to your specific question let me first point out that I am not an expert in semiconductor lasers so I am going to try to answer it using a generic devise (and you will have to decide what applies to your specific situation).

To do that lets take a device that is cycled on/off. The device can age (and be more prone to failure) while operating. Additionally (and if you want to add more complexity), the device reliability may also be affected by the number of On/off cycles, and perhaps even the time the device is sitting inactive.

If we assume the simplest case (i.e. that the device’s reliability is only function of operating time and not how many times it was turned on and off, etc.) then you need to look at (convert) the data in terms of operating time (i.e., actual time the device was on). Such data (time-to-failure or suspension for each specimen on test) can then be used to obtain a life distribution. Now, when attempting to fit a life distribution, other than the exponential, you will be assuming a non-constant failure rate, thus your data will need to contain at least one failure time in order to fit the model. Once the model is fitted, inferences can then be easily made as to the reliability of the device after any time of continuous operation.

Now in the case that on/off cycling and/or dormant storage also factor in the reliability then more complex modeling is needed. In these cases one needs to look at these factors as stresses and utilize Accelerated Testing type of models. If the on/off cycling is the only stress then a method similar to the example given at http://www.reliasoft.com/alta6/a6ex5/index.htm can be utilized (ignoring the other stresses). If additional factors are present (i.e. dormant time) then more complex modeling is required, i.e. use of time varying stress and the cumulative damage model.

I hope this answers your question.

[Mike Belanger]

Thanks,
That does help a little. I think I might be getting into an area too specific for this forum.

I have discovered in archive notes at my place of work an equation in an excel file that was used to estimate MTTF of pulsed high power semiconductor lasers (I don't think I can write the equation here in text format). The origin of this equation is unknown, the author is not an employee (or former). I think it comes from some work that a customer did for a specific program and the author of the equation may be a consultant.
I have not been able to find any references in any technical journals or texts.
I have analyzed this equation using different parameters and conditions. The equation seems to make some technical sence but after working the equation several times by changing conditions it failed to make hold water and therefore appears to be flawed.
Any help that anyone can offer to find any technical reference that might explain where this equation came from or might help to estimate MTTF of high rep rate pulse lasers would be great. Thanks

[Pantelis]

This is hard to answer without the equation? Can you put type it in using an excel type format?

[Mike Belanger]

I'm going to try to type out the equation, hopefully it will be readable.

MTTF= 4.5 x (Po/L)^-6 x tw^-2 x F^-1 x f(T)

where 4.5 = constant
Po= output power (W)
L = length of the cavity stripe (um)
tw= pulse width (ns)
F = rep rate (kHz)
f(T) = is a function of temperature for T=20C f(T)=1.18

As I mentioned previously there is some theoretical sense to this equation but it breaks down when put the test. I have never seen this or any similar equation in any literature. It would be interesting if you or anyone else has ever seen something like this. I would be interested in reading any paper on this subject.
thanks

[Pantelis]

New to me...

It is empirical, and may have worked for a specific design at one time.
I fail to see why it should work across designs ...