SMPS PCB Layout problem

[eem2am]

smps pcb layout

hello,

I wonder if you can help me with a PCB Layout issue for a 10W boost converter LED driver:-

(8 to 17V in and ~28V, 350mA OUT)
Switching Frequency = 510KHz,
Continuous conduction mode.
40% ripple current.
controller = LTC1871

Basically, the PCB is double sided and very small (15cm by 65cm).

Space was tight and i placed the boost inductor, Mosfet, & boost diode on the top layer.
I placed the rectifier loop cap (4.7uF ceramic ) under the boost diode on bottom layer
I placed the switching FET loop cap (4.7uF ceramic) under the boost inductor on bottom layer

There was then not enough room (so i thought at the time) for the pwm controller IC on top of the board.

...So i then placed the controller IC almost underneath the FET on the bottom layer(since i wanted the current sense and gate drive connections to be short)

.......However, this means that the controller IC is basically underneath the FET and the boost inductor.
-In other words , instead of the controller IC having a nice ground plane underneath it, it has the power switch loop current underneath it !

-To make matters worse the LED current feedback track is tracking some 3.5cm along the side of the top of the PCB back to the controller IC. -It has no ground plane directly underneath it (there was no room for ground plane there) , and it traverses alongside the power switching current paths......

-Since this Feedback trace connects to the high impedance feedback input of the controller IC, it is obviously going to pick up a horrendous amount of noise from the switching current loops and feed this noise into the controller's feedback pin.

I am pretty certain that the above means that the PCB that i've just layed out hasn't a cat's chance of ever working.

-Are there any stories out there of rubbish SMPS PCB layouts which strangely ended up working? or should i just trash this PCB and start over again?

 

[frankvh]

smps pcb

I guess I don't need to tell you that page 23 of the datasheet contains some pretty helpful PCB layout info, both with an example layout as well as by showing what the high current paths are.

I've done a number of PCB layouts using LTC switchers (and others too). I certainly sympathise with your problem. The good news for the part you've chosen is that it doesn't have an integrated FET. So the high current paths into the controller are minimal, and the power dissipation of the controller is also reduced. However, keeping the feedback, compensation & other sensitive inputs as nose-free as possible is as important as ever.

From what you're describing, my guess is that your current layout may well "work", but probably work pretty badly. If you haven't physically built a PCB yet, I'd tear it up & try again. Call this one a learning exercise.

I can tell you that on the board I'm currently working on, for the rev 2 I put a *lot* more effort into the switchers layouts. And it really paid off. The switchers on the rev 2 board are much quieter than on my rev 1. Almost identical schematics, just different layouts. It's worth the effort.

My other point is to make sure you're really really clear on where the ground return currents are flowing. When dealing with switchers, it can often be a good idea to put little keepouts in the ground plane to constrain where in the ground plane those currents can flow. If you take a look at any of the LTC demo board layouts, you'll see a lot of that. Little cuts in the ground planes, so that the switcher return currents will flow in that particular portion of the plane and nowhere else. This can really reduce the coupling of switcher noise to other regions of the board.

Frank
https://www.frankvh.com