boost-converter trouble

If the output voltage quality isn't critical, a simple on-off regulation with a fixed PWM duty cycle can work well. Used it

Best regards, Piotr

For this power level you don't even need an ADC, a comparator will do. ADC-based PID control is an unnecessary luxury and a waste of CPU cycles.

Best regards, Piotr

Tim, we are speaking about a 350mW power level and powering a LED. I bet that even an unregulated boost with a fixed PWM ratio would do the job.

Best regards, Piotr

Yes, make the job simple and try gating the PWM signal with a built-in comparator. Just try.

This is not a mains-powered 500W switcher where it pays off to use a more advanced control algorithm, just a sub-0.5W LED power supply.

Which is unpredictable, as this thread witnesses.

Best regards, Piotr

Going into the CCM land for the required 350mW would be an interesting adventure on its own. Accompanied with a synchronous rectifier, perhaps? ;-)

Best regards, Piotr

the usual objection is what if it crashes

NT

The usual solution is to use an MCU with sufficiently sophisticated hardware to cram your switcher into that. Even a tiny PIC16F1709 has all the required circuitry to take care of 2 current-controlled SEPIC converters. I've done that, also for the purpose of LED powering (14W in each channel). The software part is there just to configure the hardware blocks and run the slow voltage loop (if needed at all). And for digital brightness controll + other bells and whistles.

With a PSoC you can even move your PID into hardware, but I'm afraid the cost would not be 20 cents, even if blackmailing the CEO.

Best regards, Piotr

Tell me about the hardware dead-time generators.

That's interesting. Set a PWM duty cycle a bit higher than needed, and just do bang-bang control whenever you get around to it.

Well, don't do that.

I did note discontinuous mode, which someone snipped. But it wouldn't be hard to write code for continuous-mode control. I've done CCM switchers with comparators or schmitt trigger gates as the control elements; just limit the max duty cycle and it usually works.

You can always buy a smaller inductor.

It does use CCM, especially when starting from zero, and there needs to be an input current limit that's not far above the operating input current. This drive up the parts count and area for a comparator or uP solution.

The uP version could ramp up pulse width and check that the output voltage is increasing properly. Supervised soft-start.

The LEDs won't pull much current until the voltage gets high, so it should be easy to test the basic boost mechanism at low duty cycles.

You can go a step further and require the bang-bang event to happen at least once per some predefined interval. If it doesn't, you have a short circuit or a feedback network failure. You shut down the system in either case and report malfunction/perform delayed restart, whatever fits you.

Back then I had believed that *this* is the way all the switchers work, have learned the "proper" PID control about 7 years later.

Best regards, Piotr

I've been suggesting these, here, for a while. They're really slick parts.

The PICs will do PID.

That's a heap of stuff for 20 cents. The 8051 opcodes are morally repugnant, but if they are hidden under a c compiler, they are almost tolerable.

Doesn't seem to be in this exact branch of this thread.

CCM with duty limiting only works as long as you consider the load as part of the control system and not an adversarial unknown. Or for certain values of "usually" and "works".

Tim

I've got to admit that power supply design is more difficult when the load hates you.

This works fine

But a Cuk is a little quieter.