100W 12V to 4.2V buck converter / voltage regulator ?

[ponas]

Hi there,

 So I want to run a small, purely solar powered, linux "server" based on Olimex A64-OLinuXino-2Ge8G-IND.
Olimex board has a battery connector (JST) which can be used to power the board, as well as getting battery voltage, discharge rate, etc all from within the linux running on the board (no external voltage/current sensors required, which simplifies whole setup).

For the solar panel I've chosen 100W SPR-E-FLEX-100.

Now the battery. Olimex battery connector supports single 18650 or multiple connected in parallel, so I plan to make a 20P (or so) battery pack (and use protected cells!).

The challenge I have is how to charge the battery pack as fast as possible, because solar panel location will be suboptimal. So basically I'm looking for a 12V to 4.2V buck converter, which can handle 100W or so.

I found couple of cheapish Chinese buck converters ala.

https://www.prodctodc.com/10a-dc-buck-converter-10v35v-to-42v-stepdown-power-supply-volt-regulator-p-375.html

but I'm not sure if I can trust them running 24x7 all year around, so I'm looking for a high quality and efficient alternative.  Best, of course, would be a solar charge controller which would output 4.2V, but I have not seen such controllers as they usually support 12V or 24V batteries only... Any ideas ?

[fourtytwo42]

I found couple of cheapish Chinese buck converters ala.

https://www.prodctodc.com/10a-dc-buck-converter-10v35v-to-42v-stepdown-power-supply-volt-regulator-p-375.html

but I'm not sure if I can trust them running 24x7 all year around, so I'm looking for a high quality and efficient alternative.  Best, of course, would be a solar charge controller which would output 4.2V, but I have not seen such controllers as they usually support 12V or 24V batteries only... Any ideas ?

Hoverboards come to mind (the exploding catching fire variety)! I don't think Lithium chargers are as simple as that.

[george80]

If you are using 18670's you will need a proper charger.
They are cheap as chips on ebay. you would be very unwise just to hook them to a power supply especially one that does 4.2V which is right at their upper, life shortening limit.

Just look through Ebay, aliexpress, bangood... whatever.
 Hardly a rare thing to get.

[NiHaoMike]

Connecting all the cells in parallel avoids the usual balancing problem, at the cost of high currents. You'll be looking at up to 30A or so, enough to make the converter design still fairly easy but not trivial. Limit the voltage to 4.1V/cell or less and the charge current to 1C or less for good service life.

100W of solar sounds like a lot to run just one single board computer - I have personally used a single 100W panel to power 4 cheap smartphones, 2 old tablets, two Raspberry Pi 3s, and a small mining ASIC with some power left over. (All of those were part of my cryptocurrency mining setup, the Raspberry Pis used as management controllers and the rest calculating hashes.) I used a 4S LiFePO4 pack (12.8V, 82Ah) with a balancing BMS, then used buck converters to step that down to 5V to actually run all the equipment.

[Seekonk]

Running a solar panel into a buck converter can put it into a death spiral reducing maximum power.  The panel can be kept at power point  with a simple circuit that fakes the feedback pin of the converter into thinking there is an over voltage on the output.

[NiHaoMike]

Running a solar panel into a buck converter can put it into a death spiral reducing maximum power.  The panel can be kept at power point  with a simple circuit that fakes the feedback pin of the converter into thinking there is an over voltage on the output.

There are buck controllers with MPPT that will work properly with a solar source. I don't have any experience with them other than a cheap MPPT board that never performed well, failed after only a few hours of use, and was promptly returned for a refund. Apparently, since the transient response requirements are very modest, any microcontroller that has an ADC and enough PWM outputs (likely 2 or 3 phases for 100W at 3-4V) can work well as a MPPT controller. Add in hardware overvoltage protection and peak current limiting, then there's really not much to go wrong experimenting with algorithms other than that they might not work well.

Spoofing the feedback signal of a standard buck converter can work, but there's really no good reason to introduce a bunch of complexity between the MPPT circuit and PWM modulator. (Makes it harder to stabilize the control loop.) Ideally, the PWM from the microcontroller should directly drive the gate driver except if there's an overvoltage or overcurrent condition forcing the MOSFETs off.

[Seekonk]

There isn't much reason to have real MPPT.  If TI can do it, so can you.

[NiHaoMike]

There isn't much reason to have real MPPT.  If TI can do it, so can you.

If the battery voltage range is very far from the MPP range of the solar panel and requires a DC/DC converter to work well, a MPPT converter pretty much comes at minimal extra cost. But it usually makes more sense in small systems to match the battery and panel voltages so that MPPT would only give a small further improvement and then just use a voltage cutout charge controller to keep things simple.

I have done the "microcontroller adjusting a PWM controller" approach for my senior design project (although it was for a stationary bicycle), and my findings are that two nested control loops aren't exactly the easiest to make stable. Besides, the microcontroller already has PWM outputs and some even have built in comparators that can disable the outputs in hardware.

[Seekonk]

Anytime I see 100W solar panel, I know it is a post not worth responding to. Those types of questions will not be from someone that can design a SMPS.  But, modifying a buck converter only takes about a half dozen parts which should be doable, you'd think.  Experience has shown that is an impossibility for 99.999% of people in solar.

This charges my battery for 60V array.  For less than $10 it is a super low cost system.  No one should be buying battery voltage panels with grid tie so cheap.  This system is also heating water with another module.