Derating Analysis, how do you do it exactly? Any experiences or stories?

[tj_theEE]

One of the job orders that we have to do to conduct as a reliability assessor is to do derating analysis work on finished products, which is to directly measure the power usage of each components while it is active. Usually my coworker is the one who handles this job because he's much faster. But the products that he needs to analyze gets more dangerous because one of the products that we sell are electric vehicle chargers which involves high current and voltages. Any sort of wrong move could potentially break the product that we're analyzing. This added the fact that we would have to de solder some components to install wires just to get current readings from a component. I personally think that this is a dangerous and unnecessary step and when I asked this on another forum, people told me that this sort of step is non-existent and they would just conduct thermal imaging tests to determine which components are out of spec. What do you guys think? Does anyone here have any experience in doing derating analysis? How do you guys do it in your company?

[tszaboo]

I do it a lot. But wait until you have to do de-rating not only in normal working conditions. I have to do de-rating as well, when another part is broken, depending on the standard, two part broken. So imagine your design has 100 parts. Normal working operations would be 100 measurements. Now imagine that two of the 100 parts might break in several different configurations, like open circuit, short circuit. Suddenly, you have ~100x100x100 measurements to make.

[David Hess]

I have only done derating analysis for my own designs, where I knew what the operating parameters for each part are.  It was very simplified because I already knew which parts are likely to fail due to operating conditions.  I also tested assemblies to destruction in order to identify weaknesses.

Current, including DC current, can be measured without breaking wires using either an inductive probe or an in-circuit current measurement which some multimeters support.

I did not use thermal imaging, but would have.  When necessary, I attached thermocouples.

[CatalinaWOW]

Am I missing something.  Every derating analysis I have ever been in contact with was done analytically.  Perhaps done differently sixty years ago before development of Spice and other simulation tools.  And yes, it does get to be a huge job on large circuits and when multiple failures occur.  But doing by measurement seems virtually impossible.  To do it properly you would have to know where in the tolerance range all of the components were, and assure that you tested samples with components all over their ranges.

What I have seen, and what makes sense to me, is testing to verify the accuracy of the simulation.  Thermography could be useful to reduce cost and increase coverage of this testing.

[bostonman]

Am I missing something.  Every derating analysis I have ever been in contact with was done analytically.

My last job was derating analysis and it was done analytically; so this is a correct statement.

Some was worst case where we built the circuits in MicroCap and ran a Monte Carlo analysis. Most of our analysis seemed theoretical at best and wasn't the ideal job for me due to being repetitive and not hands on.

In most cases we took an agreed upon derating factor for the design based on the customer needs and just created endless monotonous spreadsheets. As an example, all capacitors shall be derated 50% (the agreed percentage based on customer needs), so we'd create a spreadsheet with all the parts, enter the derating factors, the voltage/temp/etc... and see if anything was out of this specification. The final result would be a report along with a dumbed down fancy spreadsheet.

The Monte Carlo analysis would run many loops changing values based on the component tolerances.

Then we also conducted reports based on some list of numbers that were created (nobody in the department knew their origin but some were based off of NASA specs). It would add all the component numbers from the list based on failure rate, blah blah, and tell us the device as a whole would last at minimum the X years the customer wanted the unit to run for (similar to how commercial electronic specifications state they are good for X thousand hours (i.e. MTBF).

My personal input was most of this was all fluff and felt the numbers just appeased the customer. My feeling of projected life expectancy is a calculated risk that a percentage of failures may occur in less than say 100k hours that would outweigh the profits made on units that met the minimum and/or the customer would "upgrade" thus reliving the company of early failures.

Monte Carlo analysis was somewhat fun, but the circuits may have components that were created such as a delayed switch to simulate charging rates. So the "ideal" component was nothing more than a calculated charge/discharge rate.

[David Hess]

Am I missing something.  Every derating analysis I have ever been in contact with was done analytically.  Perhaps done differently sixty years ago before development of Spice and other simulation tools.  And yes, it does get to be a huge job on large circuits and when multiple failures occur.  But doing by measurement seems virtually impossible.  To do it properly you would have to know where in the tolerance range all of the components were, and assure that you tested samples with components all over their ranges.

What I have seen, and what makes sense to me, is testing to verify the accuracy of the simulation.  Thermography could be useful to reduce cost and increase coverage of this testing.

I did everything before we had simulation tools, and never used SPICE for it.  Temperature was always the big unknown.  I could calculate what airflow is required, and plan accordingly, but I would always verify temperatures of finished assemblies.

[tj_theEE]

Current, including DC current, can be measured without breaking wires using either an inductive probe or an in-circuit current measurement which some multimeters support.

You said something about an inductive probe. Is there such thing as an inductive probe for oscilloscopes? How does it work?

Am I missing something.  Every derating analysis I have ever been in contact with was done analytically.  Perhaps done differently sixty years ago before development of Spice and other simulation tools.  And yes, it does get to be a huge job on large circuits and when multiple failures occur.  But doing by measurement seems virtually impossible.  To do it properly you would have to know where in the tolerance range all of the components were, and assure that you tested samples with components all over their ranges.

What I have seen, and what makes sense to me, is testing to verify the accuracy of the simulation.  Thermography could be useful to reduce cost and increase coverage of this testing

Problem is, we're currently working with designs that don't have simulations or models because it wasn't ours to begin with, due to the fact that some of our products are OEM products that we tune specifically for our needs. Do you have any idea on how to mitigate this problem?

[David Hess]

Current, including DC current, can be measured without breaking wires using either an inductive probe or an in-circuit current measurement which some multimeters support.

You said something about an inductive probe. Is there such thing as an inductive probe for oscilloscopes? How does it work?

It is designed for oscilloscopes, but I think it could also be used with a multimeter.  It works by detecting the h-field around a wire, so does not require breaking the circuit.  I think HP had something similar at one time.  Dave did an EEVBlog video about it and shows it also connected to a multimeter:

https://www.aimtti.com/product-category/current-probes/aim-i-prober-520

https://youtu.be/kmCvrGVtC0M?si=qgGHX1bUcGg40Lf0

The other contact option is a multimeter like the Keithley 2001 which can make a DC in-circuit current measurement without breaking a wire or trace.