Hi all. I hope you can help me think and shed some light on my idea.
The background of this project is a very high current load in the form of a motor. It draws around 400A @ 12V steadily when running. Its runs about 30 seconds per day. The distance between the motor and its supplying battery bank is approx 10 meters round trip, which would call for some really thick cables to avoid an unacceptable voltage drop. Solution: I have installed a buffer battery next to the motor that acts like a big capacitor! All good. Works well. But to feed the charging current into the buffer battery, from the main battery bank, I want to run rather thin cables, because slow charging is acceptable and they are easier to run. This could however be rather dangerous in the eventual case where the buffer battery is flat. Too high current could start flowing and put these thin feed cables on fire. Not good. So I want to reduce the average current to a maximum value of say 50A by simply PWM'ing a high side mosfet at the buffer battery end of these thin cables.
What would be the best way around this?
I could make an Arduino measure the current and drive and adjust the duty cycle of the mosfet accordingly. But I can only measure on "the high side" of the PWM pulse train, when the mosfet is conducting. Will I be able to make current readings fast and accurate enough? At a PWM freq of lets say for example 1kHz, the mosfet is only conducting for 1ms at the time.
The typical hall sensor module can measure DC current everywhere.
You can build a self controlling circuit that stops the pulse when the maximal allowed current is reached, then waits for the required cooling time. Essentially a very slow PWM. Add a coil to produce slowly rising current.
That's pretty-much how a switching power supply works. With a regular (constant voltage) power supply the switched-output is filtered to DC (with an L-C filter). The voltage is fed-back to control the "PWM" duty cycle and hold constant voltage.
A switching constant current supply works the same way except the current is monitored and feed-back. With no load (or a high resistance load) the constant-current supply goes to its maximum voltage as it tries to push current through an infinite resistance. If you short the output, it happily keeps feeding current and the voltage drops to (nearly) zero.
Of course there isn't normally a microcontroller, just a comparator in the feedback loop to switch it on & off, creating an oscillator. That's usually built-into a specialized chip. But 50A is a LOT!!!
If this calculation is correct, 400 Amp * 30 seconds == 400 * 30 /3600 Ah = 3.333 Ah consumed. In order to put that energy back some 140% of the energy needs to be delivered to the final battery.
Let's say 5 Ah delivered during 23 hours 59 minutes and 30 seconds...
5/23.99 = 208 mA of average charge current.....
Use any conventional charger maybe powered by the first, primary battery, or powered from mains.
You're basically asking how to design a 4.8kW DC-DC converter. That is not a project for the faint of heart!
Here's a potential solution: purchase an automotive grade inverter to take your 12V supply and turn it into 120VAC. That's much easier to run 30m from thinner cable. At the motor end, turn that 120VAC back into the DC that your motor needs using a high power (5kW) AC/DC power supply. OR use it to feed a battery charger for your local battery.
In the end, I think you'll find that for such a short distance, it's more cost effective to use a large wire gage. Perhaps multiple cables.
Heat is produced by actual current, not by average. A 50A average at 10% duty cycle will burn thin cables at 500A with heat dissipation blocked by insulation.
Adding an inductor is mandatory for really limited current. Principle of switching power supplies.
Why not use commercial, reliable, equipment?
I did some projecting of just a very similar matter, charge a cold box battery in the car from the car system. Some serial power resistors and a few relays shunting out resistors when charging get closer to done.
Don't forget the large time frame, 23++++ hours to charge. Why bounce on large currents?
But to feed the charging current into the buffer battery, from the main battery bank, I want to run rather thin cables, because slow charging is acceptable and they are easier to run. This could however be rather dangerous in the eventual case where the buffer battery is flat. Too high current could start flowing and put these thin feed cables on fire. 
Not good. 
So I want to reduce the average current to a maximum value of say 50A by simply PWM'ing a high side mosfet at the buffer battery end of these thin cables.
