Regulator Life

It would help to know the conditions that the LP2985IM5 was operating under when the Device pin 5 temperature was measured.

The usual stuff: Vin, Vout, Iout, and ambient air temperature. At a minimum, the Iout value would be helpful to create the operating box.

" ... is it best to reduce the input voltage by 1V or reduce the junction temperature by 1°C ..."

The short answer is that by reducing the input voltage by 1V you will be reducing the junction temperature. By how much? Hard to say without more information.

       Tj = (((Vin - Vout) x Iout) x RΘja) + Ta  

or

       Tj = (((Vin - Vout) x Iout) x Ψjb) + Tpcb

Either way, Vin is a factor.

http://www.ti.com/lit/an/spra953b/spra953b.pdf

I could guess, but I would prefer to give as accurate of an answer as possible.

The LP2985IM5-x.x (SOT-23 / DBV / 5-pins) has a RΘja thermal rating of 175.7°C/W (this is for the standard JEDEC 4-layer High-K PCB), and a Ψjb thermal rating of 30.3°C/W.

Since we know that the Tpcb is about " ... 115 degrees Celsius (maximum) ... " and the Ψjb thermal rating is 30.3°C/W

So, here's where I start guessing ...

I will presume that the ambient temperature is 25°C.

I will presume that maximum pin 5 temperature of 115°C will be at Vin=19.4V and Vout=3.30V.

I will presume that you are right at, but not exceeding, the maximum recommended operating junction temperature of 125°C.

       Tj = (((Vin - Vout) x Iout) x Ψjb) + Tpcb

        Iout = ((Tj - Tpcb) / Ψjb) / (Vin - Vout)

        Iout= ((125°C - 115°C) / 30.3°C/W) / (19.4V - 3.3V) 

        Iout = 20.7 mA

Then Pdiss = ((19.4V- 3.3V) x 20.7mA) = 333 mW

If Pdiss = 333 mW, Ta= 25°C, and Tj =125°C

then RΘja = ((Tj - Ta) / Pdiss) = (100°C / 0.333W) = 300°C/W

Now, 300°C/W is a pretty poor thermal rating, but this is where guessing get you sometimes.

Anyway, working backwards ...

If you reduce Vin by 1V, and keep everything else the same ...

   Tj = (((Vin - Vout) x Iout) x RΘja) + Ta  = (((18.4V- 3.3V) x 20.7mA) x 300°C/W) + 25C = 118.7°C

So a reduction of Vin by 1V ~might~ yield a 6°C drop in the junction temperature. It all depends.

So what would this mean for device ~lifetime~?

TI has tools for ~estimating~ device lifetime vs temperature. It's a bit of an awkward two-step process, but it's do-able.

You need to go the the LP2985-N product folder

http://www.ti.com/product/LP2985-N?keyMatch=lp2985-n&tisearch=Search-EN-Everything

Near the top of the page there are eight tabs, click on the 'Quality and packaging' tab. This will take you to the LP2985 'Quality & environmental data' page.

The far left column is 'Part#'s', the far right column has links that take you to the 'DPPM/MTBF/FIT Rate' page.

Go down the far left colum and find your part#, then go across to the far right column and click on 'View'

This will take you to the 'Reliability Data: Reliability Estimator Results' page specific to your part #.

Near the top of the page you will see two gray bars. One bar says "Download Derating Tool', this is the tool that does the calculations, it's pre-loaded with data from some part that we don't care about. The other says 'Download Speadsheet', this holds reliability data specific to the part# that you selected.

There is one row of data in the 'Download Spreadsheet' (Row 8, I think) that needs to be copied and pasted into the 'Derating Tool' at Row 11. The thing to be careful about is that both of the downloaded files will be named 'reliability_estimator_results.xls', so if you want to save them just pick useful file names. If you select the entire row to copy and paste you might get some error message about '... not the same size ...' or something when you paste. Just ignore it, I do.

In the 'Derating Tool' spreadsheet you can tweak the 'New Confidence Level' value, and/or the 'New Useage Temp' value, and see the new estimated MTBF and FIT values. This is probably what you want.

Don't mess with any of the values that you pasted into Row 11, that's the base data needed for the 'New' calculations.

You will notice that the lower portion of 'Derating Tool' spreadsheet is loaded with disclaimers of all sorts.

Thanks for the detailed response.

I noticed that today's specification for 'RΘja' is 175.7°C/W. However, the National specification that I have (which is dated July 1st 2007) quotes 'Θja' as 220°C/W. I could email this older specification to you if you prefer. This reduction in thermal resistance is quite an improvement. Why did this value change?

Regarding the 'Download Derating Tool' spreadsheet, what does 'New Useage Temp' mean? Is this the regulator's junction temperature or is this the ambient temperature?


Thanks,
Kevin

"... I could email this older specification to you if you prefer ..."

No need. I have a copy in my archives.

" ... Why did this value change? ..."

Beause National Semiconductor was bought by Texas Instruments. For the past year, or more, we have been in the process of reformatting the National Semiconductor datasheets to the TI format. Part of that reformatting requires that we replace any NSC package thermal values with values that come from the TI package thermal database to ensure consistant thermal information across the entire TI product line. The methods and benchmarks used by NSC were not always consistant. 

The old NSC datasheet says "... the SOT-23 package is 220˚C/W in a typical PC board mounting ..." , and that one word 'typical' can mean almost anything. While the TI values are all referenced back to the JEDEC standards (JESD51-x) via the app note SPRA953.

"... what does 'New Useage Temp' mean? ..."

I have always presumed that it indicated Junction Temperature since using Ambient Temperature is a bit pointless. I honestly don't know.

 

Ok, I'll assume that 'New Useage Temp' means junction temperature.

Regarding voltage regulator LP2985IM5-5.0/NOPB in our application, we have the following two scenarios:

Scenario 1 (the most conservative end of our application)
Tj = {[(Vin - Vout) x Iout] x RΘja} + Ta
Tj = {[(10.7V - 4.96V) x 0.0651A] x 175.7°C/W} + 50°C
Tj = 116°C

Scenario 2 (the most aggressive end of our application)
Tj = {[(Vin - Vout) x Iout] x RΘja} + Ta
Tj = {[(18.65V - 4.07V) x 0.0651A] x 175.7°C/W} + 50°C
Tj = 217°C

Using 116°C as the regulator junction temperature in Texas Instruments calculator for the LP2985IM5-5.0/NOPB device, I get an FIT (failure in time) of 281 failures per 10^9 device-hours. If I have a population of 100,000 LP2985IM5-5.0/NOPB voltage regulators, does this mean that I can expect 0.0281 failures / hour (281 x 100,000 / 10^9)?

Can you please clarify what 'device-hours' means as my interpretation may be wrong in the example above.

In the definition of FIT (failures in time), are you assuming that the voltage regulator is on 100% of the time? Or is there an assumed duty cycle, meaning that the voltage regulator is on only (for example) 80% of the time and off 20% of the time?

Also, in the definition of FIT (failures in time), what is the criteria for defining a failure? Unstable voltage fluctuation on the output side of the regulator? Melting of the silicone core?


Thanks,
Kevin

Your questions are outside my area of expertise, so I have forwarded them to the SVA Reliability/Quality group.

I am awaiting a reply from them.

Hi Donald,

Have you received any responses from the SVA Reliability/Quality group?


Thanks,
Kevin

Unfortunately, nothing that answers your questions.

They did point out two issues:

1) Exeeding Tj of 150C violates the AbsMax rating

2) Tj = 217°C will not happen since thermal shudown protection will prevent it

 

Yes, it is unfortunate that TI cannot provide a definition for the 'failures in time' metric that is provided in the TI calculator.

As I stated above, the most aggressive end of our application is:
Tj = {[(Vin - Vout) x Iout] x RΘja} + Ta
Tj = {[(18.65V - 4.07V) x 0.0651A] x 175.7°C/W} + 50°C
Tj = 217°C

I have test data that shows the LP2985IM5 device does not go into thermal shutdown in this worst-case condition. We did notice however that, in this worst-case condition, the LP2985IM5 output voltage drops off significantly (but does not shutdown).

We also tested the TPS709 voltage regulator in this worst-case condition and it did go into thermal shutdown. So it seems that the thermal shutdown characteristic of the LP2985IM5 and TPS709 devices behaves in a different manner.

I would like to share with you the test data that I have from the LP2985IM5 and TPS709 devices in our application. Could you please contact me outside this forum (ksaid@husky.ca) as I would like to understand why there is a difference in the thermal shutdown characteristic of the LP2985IM5 and TPS709 regulators. I would also like to get a response on the definition of 'failures in time' as per my question in my previous post.

I have a management meeting on Wednesday this week so I would appreciate some answers before then.


Thanks,
Kevin

I sent your questions out to several other people that are more knowledgeable about this subject than I am. I got one reply back this morning (23-June):

---

" Here are some answers to your questions.

FIT is the expected statistically derived (or measured) constant failure rate of a given device over time.  Most times it statistically based in that when an op life test is run for say 1000 hrs , there usually are no fails.  So no fails , 0 FIT?  Realistically there will be failures with a large enough sample size so a few assumptions are made. 

First we assume that the distribution of the population of the devices we have can be reflected by a chi square distribution with 2 deg of freedom. 

Then we apply a confidence factor indicating how conservative we want to be when predicting the failure rate – typically 60% conf is what is used but some use 90% for a more conservative estimate (this will give you a higher failure rate with all else being the same).

Then you have to determine the predominant failure mode of the device and look up the activation energy of that failure mechanism.  The lower the number the less the acceleration of the test – typically we use 0.7eV which is an average of failures typical for semiconductors. 

Next in the calculation you determine (or assume) the use temperature – typically 55deg C – and as long as you know the stress temp during op life and the number of devices stressed you can calculate the lifetime. Fits is the inverse of MTBF  x10^9  - your calculation below is correct. 

But think of the many assumptions made to get there, use temp, activation energy, confidence factor of 60%, since stress is done at max operating temp we assume device in application is using max operating voltages  and at 100% duty cycle.

If the failure rate is too high the given application, the designer must build in redundancy."

---

As for the TPS709 voltage regulator, I have never worked with this device, it's outside my circle of responsibility, so I can't speak to the performance. I can say that's it is a totally different design, so differences should be expected.