TECHNET Archives

April 2006

TechNet@IPC.ORG

Options: Use Monospaced Font
Show Text Part by Default
Show All Mail Headers

Message: [<< First] [< Prev] [Next >] [Last >>]
Topic: [<< First] [< Prev] [Next >] [Last >>]
Author: [<< First] [< Prev] [Next >] [Last >>]

Print Reply
Subject:
From:
Werner Engelmaier <[log in to unmask]>
Reply To:
TechNet E-Mail Forum <[log in to unmask]>, [log in to unmask]
Date:
Sat, 22 Apr 2006 11:44:17 EDT
Content-Type:
text/plain
Parts/Attachments:
text/plain (1 lines)
Hi Jean-Paul,
Before we agree to diasagree, let us be a little bit more precise. Below is 
the e-mail as is with my comments afterwards.
Werner

In a message dated 4/21/06 18:41:51, [log in to unmask] writes:
> Hello Werner,
> SEE COMMENTS BELOW.
> JEAN-PAUL
> 
> In a message dated 4/21/2006 6:08:01 P.M. Eastern Standard Time,
> [log in to unmask] writes:
> 
> Hi Pratap,
> Side-by-side ACT of SnPb and LF SJs is comparing apples and  oranges, 
> because
> of the significantly different creep rates.
> JUST TO ADD TO YOUR COMMENT, THERE IS ALSO THE GRANNY SMITH AND THE GOLDEN
> APPLES AND MANY MORE.
> 
> THE LF SOLDER JOINT RELIABILITY ISSUE HAS BECOME MORE  COMPLEX  BECAUSE OF
> THE MULTITUDE OF ALLOYS BEING  USED.  E.G.: SAC396/SAC387, THEN SAC305, 
> LATEST
> COMPOSITION I HEARD  BEING USED IS SAC105.  AND DON'T FORGET SN-CU(NI), 
> SN-BI,
> AND MANY OTHERS,  INCLUDING QUARTENARY ALLOYS.
> 
> SO UNLESS THE ALLOY AND THE CONDITIONS ARE BEING SPECIFIED, I HAVE NO IDEA
> WHAT "LF RELIABILITY" REFERS TO.
> 
> TALKING ABOUT SAC396/SAC387 UNDER THERMAL CYCLING CONDITIONS, A LOT OF DATA
> IS AVAILABLE AND THE TREND IS PRETTY CLEAR:
> 
> THE SAC TO SN-PB LIFE RATIO GOES DOWN, AND CAN BE LESS THAN 1, AS  FOLLOWS:
> 
> 1) WHEN THE DELTA T (TEMPERATURE SWING) INCREASES
> 2) WHEN THE DWELL TIMES BECOMES LONGER
> 3) WHEN THE MEAN TEMPERATURE (AVERAGE OF HOT AND COLD TEMPS.)  INCREASES.
> 
> That is why it is impossible to extend IPC-7901 to LF-solders until a 
> generally valid acceleration model for
> LF-solders  exists.
> 
> Werner
> 
> I THINK WE DISCUSSED THIS BEFORE ON THIS FORUM. FOR SAC396/SAC387, OVER A
> DOZEN THERMAL CYCLING RELIABILITY MODELS HAVE BEEN DEVELOPED. I GAVE 
> REFERENCES
> IN MY SMTAI'05 PAPER LAST YEAR AND THERE IS MORE COMING OUT.  I DON'T
> BELIEVE IN 'GENERAL' MODELS (all of them require some kind of component 
> specific
> calibration), SO I USE AT LEAST TWO MODELS WHEN I CAN AND THEN I TRY TO
> UNDERSTAND WHY PREDICTIONS CAN BE DIFFERENT (which is not unusual for SAC 
> and was
> not unusual for SnPb either).
> 
Here are my comments, starting with some defining clarifications.
1) Any fatigue reliability model, not just for solder joints, has 2 parts; 
the first a model describing the fatigue behavior of the material, and the 
second a way of quantifying the loading conditions.
2) The fatigue behavior model is most often some variant of a power law; for 
materials exhibiting both low- and high-cycle fatigue this model may consist 
of two parts, but they can be combined; for materials exhibiting only low-cycle 
fatigue behavior [in the temperature range of most interest, anyway], like 
solders, one model suffices.
3) The second part can be relatively simple, but for creeping metals, such as 
solders, it can get quite complex. Simpler situations can be handled in 
closed form; as complexity rises so will the complexity of the approch needed to 
assess fatigue damage, and some situations can only be addressed by FEA—that is 
the situation we have with SnPb-solders.
4) If we mix the two model parts together, apparent disagreements may arise 
simply because the two sides are not talking about the same thing.
5) I certainly would agree that for each different LF-solder there has to be 
a separate model—presumably they all will be in some form of a power law with 
different parameters and coefficients; but for a given solder alloy there 
should only be one 'general' model describing the creep-fatigue behavior. AND IT 
IS MY OPINION, THAT UNLESS WE HAVE A 'GENERAL' MODEL FOR THE CREEP-FATIGUE 
BEHAVIOR FOR A GIVEN SOLDER ALLOY, WE ESSENTIALLY HAVE NO MODEL AT ALL. 
You state: "THE SAC TO SN-PB LIFE RATIO GOES DOWN, AND CAN BE LESS THAN 1, 
AS FOLLOWS:
     a) WHEN THE DELTA T (TEMPERATURE SWING) INCREASES
     b) WHEN THE DWELL TIMES BECOMES LONGER
     c) WHEN THE MEAN TEMPERATURE (AVERAGE OF HOT AND COLD TEMPS.)INCREASES."
A 'general' model needs to be able to determine the impact of a) the delta-T, 
b) the dwell time, and c) the mean cyclic temperature—for SnPb my model does 
exactly that.
6) Certainly, there may be many different approaches to assess the cyclic 
fatigue damage, including some 'component specific calibrations' as well as FEA 
for more complex situations, e.g. CSP with internal compliancies.

Werner

ATOM RSS1 RSS2