Electronics > Projects, Designs, and Technical Stuff
LED Boost converter efficiency improvement
TmaxElectronics:
So i have just built a constant current led driver for 30-60V LEDs that can drive them at up to 3.5A from a ~20V source. The chip i use is the LT3795.
The circuit is identical to the one at the top of the datasheet (obviously with adjusted component values), which should have efficiencies of >90%, but at only 400mA output current.
So far the design is working well, but i am only getting ~80% efficiency. One thing that i found was that there is a very odd gate drive waveform of the Low side mosfet (see below), if i understand correctly the biggest losses occur while the mosfet is changing from on to off, which is taking longer than it should (i think). I have no idea where it is coming from; i have checked FB, IntVcc and LED Current Sense for a possible fault conditions, that would cause the chip to force the Gate low, but have not found anything.
So now i am wondering if that reduction in efficiency is even because of that switching weirdness of if i just got pulled down the rabbit hole. Also i am not sure if the efficiency at this point is just a fundamental limit, after all this is a 200W boost converter.
EDIT: i have just checked and the chip also momentarily disables the PFet at the same time as the glitch on the NFet (i have no picture of that though), no idea what to make of that though :-//
T3sl4co1l:
...Circuit and layout?
Tim
TmaxElectronics:
--- Quote ---...Circuit and layout?
--- End quote ---
right... kinda forgot about that ::)
Here it is (down below), please be easy on me about the design, i haven't been doing this for too long ;).
EDIT: as a temporary mod i have also added a 10n capacitor across R4, as the noise caused the chip to go into output voltage limit...
T3sl4co1l:
Huh, why use Hall effect sensors for monitoring, when the controller has them built in (ISMON, IVINCOMP)?
Trace lengths to the gates are pretty long, and unsupported by ground pour. Switching loop could be tighter. Could definitely afford to be on 4-layer build, would give better thermals too.
I might guess the hiccup on the edge is measurement error, and not necessarily in the actual circuit. That's not exactly a good thing, as it means you've got at least as much common mode noise dropped across ground. (I hope this thing is going in a filtered metal box?)
Tim
TmaxElectronics:
--- Quote ---why use Hall effect sensors for monitoring
--- End quote ---
that is just a carry over from an older design. i just copied over the temperature and current sensing parts to this. But using the internal ones is certainly a better idea.
Four layer is a good idea. This PCB has only two, because i wanted a quick and cheap-ish prove of concept. The issue is that i can't make it any wider because of the enclosure i have for it.
--- Quote ---I might guess the hiccup on the edge is measurement error, and not necessarily in the actual circuit
--- End quote ---
i measured that edge on the pins of the mosfet, with a ground spring if thats what you mean :-//
The thing will go into an aluminium pipe :D
Thanks for the quick reply :)
Navigation
[0] Message Index
[#] Next page
Go to full version