LM2676S load response failure

Hi Falco,

Question 1:  Are you applying the load resistor while the Razberry PI is powered.  This would add the load currents together and you might hit the current limit.

2) With your Vin caps far from Pin 2 and the Anode of the recirculating diode you are adding a lot of ground bounce to the circuit.  This might be causeing the part to falsely hit current limit early.  Try Addin a capacitor close to pin 2 and the gorund of the diode to lower noise

3) Your power traces are long and thin.  Each square is 0.5 mOhms for 1 oz copper.  You have hundreds of squares in the traces I see.  this would be several hundreds of mohms and while this wouldn't cause the issue you are seeing it isn't good practice because of the large voltage drops.

Regards,

Marc

Hi Marc,

Thanks for your reply! No, while testing with the load resistor the Raspberry Pi was not connected, just the bare board.

About 2), I hope that moving the Vin caps will solve the problem, and I guess adding the omitted Vin caps will hopefully solve some problems too. Also, 5/3,1=1,6A is pretty far from the limit, or can this really cause that big kind of spikes? And when it 'crashes' to 2,2V the current it draws is a measured ~0,7A, shouldn't it just recover back to 5V then?

By squares I think you mean the grid? That isn't a static factor, since I can change the grid size.. Anyways, the Vin trace is 0,9mm wide, all the traces related to the LM2676 are 0,6mm and all other signal traces 0,4mm. I'm not sure about what planes to use though. On my current PCB only the top layer is a ground layer, the bottom is unused. But because I have 2 voltages (3,3V and 5V) I don't know which one to make a plane. Or could I make 2 planes since the 5V pretty much only needs to go to an USB connector and a voltage reference chip? Or even 3 planes, making the 3V3 and 5V plane more like a really wide trace surrounded by ground? It is a 2-layer PCB.

With kind regards,

Falco

Edit:

Would this be a better PCB design?

Counting squares is just a way to estimate the resistance of a trace.

The bottom line is that your traces, in the power paths, are much too thin. 

You can use planes or just much wider traces to connect the inductor, diode, and input capacitor.

Once the device goes into current limit, it requires a much smaller load to come out again.

FD

Uh, maybe I'm missing the point of what you are trying to tell, but the inbuilt trace width calculator in Designspark PCB says that with a trace thickness of 70µm, ambient temp 20 degrees celcius and an allowable temperature rise of 10 degrees a trace width of 0,4mm would already suffice for the maximum load of 2A..

Also, I can't really fit any wider traces to the pins right? 1mm really is the max then.. And when using planes, is it okay for it to also be underneath the inductor taking EMI into account? Or should that be a ground plane then?

Edit: ah, after googling the counting squares method I realize what you mean. So that's a yes for using planes for both 5V and 3V3 on the bottom layer next to each other I guess.