Solar DC circuit: Preventing Buck Converter from Dropping out Due to Motor switching on/off (in Parallel)

Question

I have an off-grid orchard which has a well pump running directly off of a 24v PV array. (27-32v when at full power) Parallel to this direct circuit, I have a buck converter to power a Raspberry pi (via GPIO), which has a switching solid-state relay and a camera to watch the pump and its gauge.

When the Pi triggers the SSR, the voltage coming into the buck drops from 27v to 24v on my multimeter (I don't have an oscilloscope, so perhaps its more of a severe drop), and this causes the output voltage to briefly go to zero, causing the Pi to reboot.

While it's entirely possible this is the fault of the cheap 12v/24v to 5v buck I bought on Amazon, I'm wondering if this is expected, and if so, can I do something like add a capacitor, either on the input or output to buffer this, and keep the pi running.

Answer 1

To keep the input voltage to the buck converter from dropping down to some really low value during the inrush current period of the 150 watt motor, use a diode and capacitor. This will "hold-up" the voltage at the input to the buck circuit despite it dropping down across the motor. The diode gets reverse biased during this short period and therefore the motor load doesn't discharge the capacitor.

But, you have to choose a capacitor that is sufficient. For instance, if the buck regulator will work down to (say) 8 volts without a problem, then you can start to calculate the hold-up capacitor. Based on the pi taking (or needing) say 1 amp we can say: -

$$I = C\dfrac{dv}{dt}$$

Where dv is the permissible drop from 24 volts to 8 volts i.e. 16 volts and dt is the the inrush time for the motor (say) 2 seconds hence: -

$$1\text{ amp} = C\dfrac{16 \text{ volts}}{2\text{ seconds}}$$

C therefore equals 125,000 uF or above.

The numbers used above are my guesstimates and may not bear much resemblance to the real values needed by the buck regulator or the pi current.

Answer 2

If the power converters & supplies you have are well-made, they should already have capacitors in parallel between the +5V and ground lines to smooth any jumps in voltage. However, you can definitely add a cap as close to the RPi as possible to smooth out the voltage even more. The problem is also coming from the RPi triggering a power hungry device, so you may be able to add an inductor in series with the device to slow down power surges.

An alternate (and pricier) option is to use an un-interrupted power supply for the RPi power, or connect a battery pack that is both continuously 'charging' the pi, and being charged by the buck you already have. The battery method isn't recommended due to the current draw from the RPi but is a possible short-term fix depending on the type of battery.

Answer 3

The buck converter apparently can't keep up with the abrupt change in input voltage. A cap probably won't be enough to smooth the voltage drop, but if you are willing to build a circuit, you could put in a capacitance multiplier to smooth it much more effectively. This solution could work if the voltage drop is not in fact much more severe than a few Volts.

This circuit is a common ripple rejection circuit. Q1 is a Darlington transistor. R1 and R2 form a low-pass filter with C1, which prevents abrupt changes at the collector of Q1 from passing through to the emitter. If the PV array output steps abruptly by 3V, it should take 10 seconds or so for the Q1 emitter output to make that transition with the component values shown below. That should give the buck converter plenty of time to adjust.

The RPi and camera probably draw somewhere in the ballpark of 200 mA at the input of the buck. With these component values there is about five Volts across the Q1 collector-emitter terminals, so Q1 will dissipate about a Watt. It should be able to handle without a heat sink, but it might be a good idea to put one on in case the Pi & camera draws more than that.

The disadvantage of this approach, apart from the complexity, is that you give up a Watt or so to the ripple rejection circuit.

^{simulate this circuit – Schematic created using CircuitLab}