LM3578AN boost converter, 2.4V to 5V problems

[darko31]

Hi guys, I'm having some trouble trying to design boost converter based on LM3578AN, 2.4V to 5V, 500mA output current. The 2.4V is coming from two Ni-MH batteries. The circuit works as it should to about 200mA, after that Vout goes down rapidly to about 1.5V, and batteries (Vin) falls to about 2V, and dummy load starts pulling 800mA all of a sudden (CC dummy load, like the one from Dave's video). I've first breadboarded the circuit and it worked (I'm aware that DC-DC converters won't behave nicely when using breadboard), then I've transferred it to stripboard making sure that feedback is as short as possible. The circuit behaved pretty much the same as on the breadboard. The inductor I'm using is this

http://store.comet.rs/Catalogue/Product/27661/

it has the current capability. I've used with MC34063 buck regulator for the same current. How higher than calculated inductor affects the circuit? I suppose just more room for stored energy isn't a bad thing.

The values I'm using are

C1 = 1 nF
C2 = 1000 ?F
C3 = 20 pF
C4 =  20 uF
fosc = 100 kHz
R1 = 39 k?
R2 = 10 k?
R3 = just copper wire
R4 = 220 k?
L1 = 47 ?H
D1 = 1N5818

Any tips are appreciated.

[jeremy]

Sorry, I've not had any experience with the LM3578, but you did say "any tips" 

Have you oscilloscoped your circuit? Any weird waves on the switching output (pin 6) under a light load? what about as you transition to a much heavier one? How does the wave change?

Are you using caps with a low ESR?

How much current is coming from the batteries during good/bad states (how is the efficiency changing?)

Also: I'm also not quite sure how a constant current load can change the current it is drawing? In fact, I'm pretty sure that by definition it can't, or it ain't CC 

[Skimask]

As I've figured out the hard way, if you don't have a big ground plane, you pretty much don't have anything.
A stripboard isn't much better than a breadboard in that respect...still not much of a ground plane.

https://www.eevblog.com/forum/beginners/still-trying-(and-failing)-with-the-lm2588-5v-gt-12v/45/
Look about half way down on this page.  I started off with a breadboard, moved to a protostack board (didn't get any better).  Only when I did the circuit on a chunk of single sided PCB did it start to work right.

[darko31]

@jeremy

Yeah, I don't have scope at hand to check out what's going on (which is pretty lame). Output cap is low ESR. I haven't measured the input current yet. And, yeah, that is pretty weird, the current simply ramps up, when moving the fine pot just a little bit, from 200mA to 800mA with CC dummy load, and voltage goes down rapidly, something odd is going on.

@Skimask

Ok, big ground plane it is, I'll etch some PCBs, try out different layouts, and report back results.

[Skimask]

As you can see in the pics, I didn't even etch a PCB.  Just drilled a handful of holes for ground connections, and let the parts hang out in the wind on the other side.

[jeremy]

If I was you, I'd be focussing on understanding the behaviour of CC load before you try to fix the PSU. You may have two broken devices instead of one, and trying to test a broken project with another broken project will just lead to headaches

[darko31]

I'm not sure that dummy load is broken, I've tested the converter with 10 Ohm resistor connected at the output, the same voltage drop happens.

[jeremy]

Right, but then if your CC load behaves exactly like a fixed-resistive load, it isn't really a CC load? A CC load would change its resistance based on the changing voltage so as to only ever draw 200mA (or whatever you select). If it can't stay at 200mA and it is CC, then it is either broken or being put out of spec. As far as I know that is the definition of a CC load