THERMAL RESPONSE PLOTS, Chapter 4
Many of the Thermal Response testers sold by FEC include software which (among other things) produces two types of plots that we call "Cooling Plot" and "Heating Plot." This paper describes these two plots and their intended applications.
This plots THERMAL RESPONSE vs reading delay TMD. The TMD normally used for a single reading is typically in the 50 to 100 microsecond range. The plots however start at a short time such as 10m S and extend out to tens of milliseconds. The high and low ends of the plots have different intended uses.
The low end of the plot allows you to evaluate the performance of the tester, the test fixture and device under test to find the lowest safe value of TMD to use for your screening tests.
It is desirable to use as low a TMD as possible since the junction is cooling significantly during this delay. We would use zero delay if we could but practical considerations related to the tester and to the device under test make this impossible.
When you look at a Cooling Plot you will see the Thermal Response increasing smoothly as the TMD decreases until at some point there is a marked discontinuity in the curve. You can easily see the lowest TMD that is on the smooth predictable part of the curve. Of course if you use a lower TMD, your test results will be unreliable.
The high end of the plot has a use similar to the Heating Plot so much of that will be covered in the section on Heating Plots. It is useful to note here though, that the Cooling Plot is somewhat safer to the device under test as well as faster to produce. This is so because it uses a fixed heating pulse width thereby reducing the possibility of overheating the part. The cooling time between points on the plot can be relatively short and constant over the plot. This reduces the total plotting time.
You will note when you compare the two plots that they are very nearly mirror images of one another.
The above might lead you to think there is no need for heating plots, but this is not so. Heating plots have much better resolution at the high end of the time range. When the TMD for a cooling plot is very long the junction has cooled to such a degree that DVF is very small and hard to measure.
Heating plots cause the junction to get very hot at the high end of the time scale. This results in a large easy to read DVF. If DVF gets too large, we will suspend the plot and mark the remaining programmed points as "Invalid". This protects both the device under test and the tester.
PLOT COOLING FIRST
Why? When you do a heating plot you must choose a value for TMD. The cooling plot allows you to find a low but safe value so that your heating plot will give reliable results.
WHAT ARE WE LOOKING FOR ANYWAY?
The most important use of the heating plot or the long end of the cooling plot is to find the most effective pulse width and test limit for detecting bad bonds.
Many registered specifications set the test conditions and limits that you must pass and of course you must do so, but passing the registered specification is no guarantee that the parts have good bonds.
For a given part number and a given manufacturing process, the difference between a good bond and an unreliable bond may be fairly small. This is particularly true if the pulse width is not optimum. Each manufacturer’s process yields a unique profile of junction temperature vs heating time. Usually all of the processes result in good parts providing that the process is in good control.
An industry wide fixed Thermal Response test must necessarily be written to accept the product of most quality manufacturers. This type of conservative specification will usually fail really bad parts but will probably not be optimum for your particular process.
We suggest that you make heating plots for a reasonably large sample of your parts and examine the plots looking for a few parts whose curves show a noticeable increase in slope at some moderate pulse width. This small deviation usually occurs at about the time that the heat flow from the junction first reaches the solder bond.
We at FEC have had the opportunity over the years to work with several manufacturers who suspected that they had bonding problems. Here is what usually happens.
First we make a lot of heating plots and look for parts with a noticeable difference in the rate of increase of temperature. Some of the good and some of the suspected bad parts are cut open to confirm the correlation between the curves and the bonds.
We then setup a test with optimum pulse width and test limits to reject the bad parts.
That is good but the next part is even better. Now that we have a reliable method of evaluating the quality of the bond, the process can be adjusted to give better results. This usually results in parts that are typically better than even the formerly "good" parts.
Please contact Frothingham Electronics Corp. if you have any questions or need assistance with your Thermal Response testing.