PWM dimming is ok when used with a SEPIC converter?

[treez]

SEPIC LED drivers are tough to design with high bandwidth.
Therefore, we make do with a low bandwidth.
However, will the low bandwidth mean that we see repetitive led current overshoot during the OFF-to-ON transition of the PWM dimming operation?

That is, the SEPIC is known for its difficulty when used with high crossover frequency. As such, are there specific problems encountered when PWM dimming a SEPIC converter with the LM3421 LED driver IC (and using a dimming FET in series with the LEDs)

Can we be sure that the voltage on the COMP pin will not rise when the LM3421 is on the OFF time of PWM dimming?......since we don't want the led current to overshoot when the dimming fet is turned back on during PWM dimming cycles.

LM3421 DATASHEET:
https://www.ti.com/lit/ds/symlink/lm3421.pdf

Also...
Taking this PWM dimmed Constant off time controlled sepic converter attached .....it is so easy to make stable...what is it about 'constant off time' that makes feedback compensation so much easier for it?...even in the SEPIC topology...("PWM dimmed SEPIC" schematic and ltspice simulation attached)

 

[rashid0214]

Yes....：微笑：

 

[treez]

"Yes" the 'comp' pin voltage will be frozen during pwm dimming off cycles?

 

[mtwieg]

If the LED PWM frequency is much higher than the controller crossover frequency, then it should not be a problem. From the control loop's perspective you will just be changing the resistance of the load.

 

[treez]

The thing with PWM dimming LEDs in say a sepic, where an in-line dimming fet is used, is that even with PWM dimming frequency being at the crossover frequency ,it shouldn't matter if the controller is advanced enough....
...If every time the dimming fet turns off, the "comp" pin voltage is preserved at the voltage it stood at when the dimming fet first turned off, then when the dimming fet turns back on again, then even though its say 10ms later, its as if time had frozen, and no conditions in the control loop changed, and so it just carrys on driving the leds almost as if the leds had never actually been turned off at all.
I am wondering if the LM3421 has this facility? (preservation of 'comp' pin voltage when dimming fet is off)....some of the linear.com controllers have this, but I am wondering if the LM3421 does this?

 

[mtwieg]

That's another method, though a rather convoluted one. It looks like the LM3421 does have the capability, but be aware that even if the comp pin voltage is preserved perfectly, that only directly controls the peak current through the input inductor, not the LED current. The actual LED current will likely show ringing and overshoot characteristic to the SEPIC topology. You should definitely simulate it thoroughly with an accurate controller model. If an accurate model isn't available, I wouldn't use that controller (or at least that feature).

 

[treez]

The idea of the in-line dimming fet is that it minimises this ringing and overshoot, because the output voltage is also preserved due to the dimming fet literally making the led load suddenly go to zero current when the dimming fet is turned off.....thus the output voltage doesn't discharge...and when the dimming fet is switched back on, the led current is virtually immediately restored...as I mentioned, its as if time was frozen during the pwm dimming off-time.
We wish to use the sepic because the dimming fet can be low-side, as opposed to a buck boost where its high-side, with high-side drive.
I have searched the web for documents saying that "PWM dimming is not possible with sepic converters" and can't find anything.

You'd think at least one led driver app note would advise not doing pwm dimming with sepic if it wasn't adviseable?.....I cant find anybody advising the pwm dimming is bad with sepic.

 

[mtwieg]

The idea of the in-line dimming fet is that it minimises this ringing and overshoot, because the output voltage is also preserved due to the dimming fet literally making the led load suddenly go to zero current when the dimming fet is turned off.....thus the output voltage doesn't discharge...and when the dimming fet is switched back on, the led current is virtually immediately restored...as I mentioned, its as if time was frozen during the pwm dimming off-time.

What you're missing is what happens to the energy stored in the sepic converter itself (that is, the two inductors and DC blocking cap) when it is switched on and off. Those will not "freeze," they will energize, de-energize, and reach a new steady state every time you change the PWM state. Your own simulation shows this clearly. This isn't automatically a terrible thing, but you need to make sure that process is well controlled and doesn't cause any SOA to be exceeded.

I have searched the web for documents saying that "PWM dimming is not possible with sepic converters" and can't find anything.

You'd think at least one led driver app note would advise not doing pwm dimming with sepic if it wasn't adviseable?.....I cant find anybody advising the pwm dimming is bad with sepic.

Of course no one would say it's impossible, that would be silly.

Personally my favored method of doing this would be by modulating the reference voltage for the control loop, but most monolithic converters don't allow this.

 

[treez]

modulating the reference voltage for the control loop

...oh I see, you are referring to what they call "analog dimming".
I favour that too....though people say that the led changes hue as its dimmed.....I have never seen this demonstrated...but maybe its true?....I looked for a youtube video showing this change of color, but couldn't find one, and when I do it in the lab, it just looks the same color to me.

 

[mtwieg]

That's actually a good point. If a LED's color does change vs current then that would manifest in an analog dimming scheme, but not a PWM scheme (assuming the PWM is ideal).

The main limitation of analog dimming is usually the limited dynamic range due to limits on the duty cycle of the converter.