Hi guys
I am trying to design a DC-DC converter to power a Raspberry Pi 3 SBC. The spec calls for 5 volts +/- .1 volts. I am using 12 volt battery (12 V - 14.2 V). The Pi, plus some associated hardware, (GPS Unit, motor controller etc) causes a current drain of between 500 Ma and 3.5 Amps.
My initial design was based on a 7805 voltage regulator which worked fine, but was not very efficient. To improve efficiency, I have now upgraded to a buck converter based on a LM 2678T-5 switcher. The circuit is attached. It almost exactly follows the application note on the TI website (http://www.ti.com/lit/ds/symlink/lm2678.pdf). However, I find that the voltage swings between 5.65 and 4.8 volts, depending on the load.
I have damaged one Pi, and I find that even when the Pi works, it reboots every five or ten minutes, presumably a hot boot caused by fluctuating voltage.
Can anyone suggest a way to make the output more stable, or an alternative circuit.
I take the point I have seen on other posts that its easier, quicker and cheaper to buy a module to do the job, but I want to be able to put the entire project on a single board. Anyway, it surely must be possible to derive a stable 5 volt supply. Anyone got any ideas?

 

You are likely seeing a voltage drop due to resistance in the output and ground connections.
The regulator itself likely maintains the voltage within 100mV over load between its feedback and ground pins.

How did you construct it?
Proper layout with a solid single-point ground (preferably a copper ground plane) are required for proper operation of a switching regulator at those currents..
The feedback connection should be directly to output pin as close to the load as possible, and the load ground should be connected directly to the LM 2678T-5 ground pin.

Look in section 10 of the data sheet for layout and PCB guidelines.

 

I would very carefully test your DC-DC converter before connecting it to something precious like a Raspberry Pi!
That would include testing the output voltage at no load, various amounts of load and if possible looking at ripple and spikes on the output with a good oscilloscope.

 

You are likely seeing a voltage drop due to resistance in the output and ground connections.
The regulator itself likely maintains the voltage within 100mV over load between its feedback and ground pins.

How did you construct it?
Proper layout with a solid single-point ground (preferably a copper ground plane) are required for proper operation of a switching regulator at those currents..
The feedback connection should be directly to output pin as close to the load as possible, and the load ground should be connected directly to the LM 2678T-5 ground pin.

Look in section 10 of the data sheet for layout and PCB guidelines.

You are likely seeing a voltage drop due to resistance in the output and ground connections.
The regulator itself likely maintains the voltage within 100mV over load between its feedback and ground pins.

The unit is constructed on a prototyping board with a bunch of parallel tracks. I have mounted all the components, but especially the bootstrap capacitor as close as physically possible to the pins on the IC. I noticed there was a source of heat around the connector on the ground connection, so I have reworked that and I am now getting 5.05 volts on the DMM between input and output. Thanks for your help.