CN3791 demo board

2020-10-30, 02:10:00

I wanted to use a higher-voltage solar panel but the 4056 types only go up to 6V or so.
I've tested some chips from Analog Devices such as LT3652 (more about that in a future article), but, well, some components would fail because of the generated heat. Once, a diode exploded.
I have a 5W panel that can go up 22V and that's a lot.

So I came upon a demo board many sellers (i.e.) have on Aliexpress, which uses the CN3791 ic, and it works.

I want to do my own board with it, so I had a thorough look at it. Since I've bought it, I had found the circuit a little weird.
Before writing about it, I've searched if someone hasn't already wrote about it, and I've found this article.

The circuit isn't very complicated.
A solar panel connects to the + and - (right side in the picture). There's a chip in the middle of the + track and the + hole, which is a P-channel mosfet and a capacitor filter. The mosfet, TPC8107, is connected drain to source, so it always conducts with a small drop over its internal diode. Its gate is controlled by a resistor divider in parallel with a zener diode ('w9', BZ series). Then there's a big 47uF capacitor to store a stable charge for the 3791 IC, the mosfet (4185) required for it to pulsate a current to the battery, signaling leds, two big clamp schottky diodes (SB105 in flat package), a very big inductor (10 uh CDPQ2010) and a very big current sense inductor (0.04) that ends up in a tantalum capacitor (10uF) to provide a stable voltage to the battery.

Now, issues with this circuit:

-the panel input mosfet unwillingly conducts and willingly dissipates some energy over its internal diode and the panel(s) drain the battery at night, which could be prevented;

(5.4V at the gate, mosfet on, source to drain is the conventional direction)

The weird part is that its gate is controlled with a zener controlled resistor divider, which should pull the gate ON upon wrong polarity or charged status, disconnecting the circuit from the solar panel; however, the battery would leak current through the mosfet (drain to source unwillingly) back into the IC, drawing 30uA according to its datasheet,  which would last until the capacitor empties and the gate will reopen... and on overvoltage it would, what, shut the mosfet off... voltage drops, gate is reopened, overvoltage again, off and on and on and off...

(2.88MO from drain to source, panels are disconnected)

But the panels are still under the sunlight and they're going to overheat!
When energy will be required, the panels would underperform.

I understand this is done for the "lead-acid" part of this board, since at least mine is stamped as such, and it does make sense, you'd want to stop the charging when it's full since further charging would permanently damage a lead-acid battery... if someone would actually make 3.7V lead-acid batteries... since the maximum this charger grants it's 4.2V at the output, you'd never charge a 12V lead-acid battery with this IC.

Maybe these boards are used in some cheap solar panel kits that come with lead-acid batteries and it's part of the charade (there is an CN3765 that might output higher voltages). 

-there's absolutely no reason to place that schottky diode near the input mosfet, when it would have served a better purpose for the battery, as indicated by the datasheet;

-there's absolutely no reason to place two clamping zener diodes;

and now for the worst part

-the battery's capacitor is a tantalum capacitor. when subject to reverse polarity, it shorts, it's going to short the battery, and the capacitor is going to burn itself out. I strongly suspect the mosfet won't close itself fast enough, if it does;

-that big inductor and big resistor is the result of bad estimation and design;

While I'm actually impressed at the cautious choice of the inductor, its designed for four times (48Vdc in the datasheet) the rated input voltage (12V) and three times (10A) the current (3A), precisely bad design leads to it: the current sense resistor sets a 3A charging current... if the board was simply rated for a specific amperage, it could have been a lower value and the components would have been different. It's a very expensive inductor for what this board does and it doesn't even make sense to charge a single cell 3.7 battery at 3A;

-the MPPT behind this IC doesn't really work as such and it's a waste of time.

You might want to read this article to find out why. 

The circuit, as it is, uses a voltage divider. The board is "rated" for 12V with a marker (there isn't anything to rate for this board rather than that resistor divider and the other one to shut the gate, since the IC works 4.5-28V). The logic inside the 3791 works in such way that if the voltage on the mppt pin is higher than 1.2V, it starts charging. So you could set that resistor divider for a minimum voltage expected from the panel to deliver high-power, say, for example, the panel is rated for 12V but it's highly-efficient only at about 16V in full sunlight, 12V being an ok-ish not-great power source, but rather the bare minimum to say it's actually delivering useful power (nominal voltage). So you'll set the resistor divider in such a way that when the panel reaches 13V, the resistor shall mark >1.2V and on the charging goes. This is useful to deliver maximum power to the charger and getting the amperage set for the battery. But this is also wrong! The panel is still trying to deliver power at 10-12V, maybe significant enough to heat the panel. In summer, the ambient temperature is already hot enough for the electrons in the panel to start refusing extra movement. Those 1-10-XYwatts cannot dissipate anywhere because the circuit is closed, heating the panel up to 60-80ºC. This charging IC restarts the charging cycle if the voltage on the charged battery drops 5%. How is it going to charge efficiently if the panel is now too hot to deliver power?
Same issue with the p-mosfet disconnecting the panels after charged status.

Leave Comment