Stabilizing the regulation loop

AD tech support contacted me outside this forum and this was the exchange which took place:

AD:

For LED drivers, we dont have a detailed mathematical equation to help calculate the compensation components. 

I would recommend you have place holder for the components in your board. 

You can also find the application schematics shown in datasheet closest to your application and start with those compensation components. 

Other option as a starting point is to use LTSPICE FRA tool to verify your design loop response in LTSPICE. 

We do provide an example on how to setup for LED driver in examples folder. I am attaching that for reference in email here. 

This example is for a different LED driver, but similar FRA can be performed on LT8376. 

With regards to analog vs PWM dimming, analog dimming LT8376 regulates LED current by changing the current regulation threshold. 

As a result if compensation is marginal, it could result in stability issues as you change control voltage. 

For PWM dimming LED current amplitude remains unchanged, converter regulates the same LED current whenever PWM signal is higher. 

Me:

I, with the help of Google’s AI, managed to find my own answer to this question along with an answer to a previous unanswered AD tech support question.

The details are here in this other EngineerZone topic on this component:

  https://ez.analog.com/design-tools-and-calculators/ltspice/f/q-a/602285/ad-supplied-model-oscillates-when-more-realistic-components-used

 It turns out that there are, to some degree, equations which can be used in the solution to this. The LTspice model, which I’m already familiar with (at least the transient response; thanks for the FRA example), can indeed be of some help, but with caveats.  The model that uses ideal components is mostly worthless in this regard because it shows the circuit being stable even if “Rc” is omitted.  The same model, but using real-world representations of components, suffers a bad oscillation if “Rc” is omitted and it can be brought under control if acceptable values of “Rc” and “Cc” are selected/used.  I am indeed using only analog dimming in my application.

 Note that the LT8376 datasheet does its user base a disservice by saying “Stabilizing the regulation loop typically requires only a capacitor Cc connected from the VC pin to GND.”

 An examination of the “Typical Applications” shows only one out of six that have “only” the Cc capacitor so it is definitely NOT “typical”.  Someone at AD needs to reword this datasheet/sentence to something like the following:  “In some cases, where the output capacitance, Cout, is less than 10uF, the Rc resistor may not be required (i.e., 0 ohms) for loop stabilization. Typically however, Rc is required, with larger values of Rc and Cc providing more stable operation and slower response to the control input. The frequency of the compensation loop is given by the following equation  and this needs to be less than the loop crossover frequency for the system (less easily determined).

The AI suggested that the loop crossover frequency is determined in the following manner, though I let it do the calculations and suggest appropriate values for Rc and Cc, which I subsequently tested using the LTspice simulator and more real-world models for the components (see other discussion for that tale).