For buck, or any converter really, many products will specify a total accuracy specification. Depending on how it is written in the datasheet, it is usually the DC errors + AC errors where:

DC errors are voltage deviations over Line, Load, Temperature
AC errors are ripple and over/undershoot

Let's take a converter with a 5.0V dead nuts output voltage. You apply a load step which leads to voltage undershoot, droop during the DC load portion , and then overshoot during the load release. This transient pulse width is not critical - I am only interested in the load step and load release behavior.

How I would measure undershoot: From Vout_nom = 5.0V before the step is my reference and down to the peak of the undershoot. This is undershoot in mVp.
Debate on measuring overshoot: The output has slightly drooped during the plateau of the current load, so let's say it is 4.95V. Then the load release happens and the peak of the overshoot is 5.10V. What is the overshoot in mVp?
Would it be from the actual output voltage differential of 4.95V to 5.10V?
Or would you measure it from Vout_nom which is 5.0V?

 

Would it be from the actual output voltage differential of 4.95V to 5.10V?

Yes.

 

Why?

 

Why?

Why not?
You want to measure the actual value of the undershoot/overshoot, not some theoretical value based on the designed output voltage.

 

Because if you measure from the output droop level to peak of the overshoot that provides the absolute value. Taking that value as a percentage of VOUT_NOM that would be potentially higher than a customer specification.
Measuring pk-pk would give you the entire tolerance window (line, load, noise, ripple, and transient) IMO.

 

Because if you measure from the output droop level to peak of the overshoot that provides the absolute value. Taking that value as a percentage of VOUT_NOM that would be potentially higher than a customer specification.

Then you need to define exactly what the customer specification is.
Fudging on measurements is not the way to have a happy customer.

 

An application that is powered by a "(nominally) 5.0 vol"t supply will normally be able to survive some percentage of variability in the supply voltage. That is often because much tighter power voltage regulation costs a lot more.
And if a customer has some specifications as to minimum and maximum then there will be no question, while if it is couched in terms of proportion or total deviation, interpretation of the actual values is more challenging.
And I will not discuss theoretical issues, if there is an actual requirement then state it. And we will try to interpret the actual numbers relative to what can simply be achieved and what is theoretically possible.

 

The point of the question is to understand if there was a generally accepted methodology. I was hoping someone would reply with some insight if that is the case - IEEE, JEDEC, IPC etc.

At one time, long ago, I saved a copy of the IEEE Standard Dictionary of Electrical and Electronic terms from the bin. And I wanted to look up transients.

transient overshoot: An excursion beyond the steady-state value of the output as the results of a step input change. Note: It is usually referred to as the first such excursion; expressed as a percent of the steady state output step.

Which yet another way to reference it. I really wished they had a diagram.

And, mind you, this isn't "theory to be discussed." This is a lab practical question - to determine if the data is in spec, the method; i.e. the reference of the peak to what (droop or nominal).

Given that no one has provided an authoritative reference, I walk away with "It depends."

 

"It depends" is the correct answer, and always will be.
.
.
.

 

Actually thinking this through, if one were only interested in the magnitude of the undershoot and overshoot then it would be measured from the nominal output voltage.

It makes zero sense to measure undershoot from the nominal output to the trough of the undershoot, yet measure overshoot from the output voltage droop level to the peak of the overshoot. In the instance of overshoot, one would include the DC error from the droop.

You're all welcome.

 

Measuring a turn on "overshoot" transient of a power supply was often done with a storage type triggered sweep oscilloscope Presently there are digital instruments made for that exact test, Check "Rhode and Schwartz" company for a high accuracy version of the tester.