Switcher Layout Any Tips?

[Frenchie]

Hi All,

Before I get too carried away with this board, is someone please able to let me know if I've made any heinous sins in the way this switching pre-regulator section is laid out? Schematic attached for reference.

Capacitor/Inductor values are ballpark for the minute, they'll probably get bumped a bit before I finalise this design. Unlikely they'll be in different package sizes though. I suspect I probably need a 100n ceramic cap across the switcher input as well. The vias on the regulator thermal pad are probably not in their final positions.

When it comes to switchers I've got some vague ideas of how they should be laid out, but really I'm just winging it somewhat  .

Edit: Just spotted the feedback going to the emitter of Q3, it should be going to the collector.

Thanks!

[CarlG]

Hi,

If you want a cleaner outupt, I would suggest the following:

The FB node is within the switching current loop; that should be avoided. Noise on the FB affects the jitter, increasing the inductor peak current.

Also, as a general rule, minimize the switching node (U2-2) and minimize the current loops. For a buck reg, the input cap current loop is more critical (rectangular) than the output current loop (triangular).

- put C2 below L1
- rotate C1 and C2 anti-clockwise and place them so that their GND pads are as close as possible
- connect D1-anode to C1-GND with as short wire as possible
- preferably, connect C1, C2, U2, D1 GND only at C2 (on the "quiet" side). If U2 needs cooling, provide a separate heat spread plane connected to tab
- keep trace C1+ to U2-1 as close to the switch node and D1 as possible (minimize current loop)
- connect R2 at C2+ (the point you want to regulate) (not at L1 where you have added switching noise)

If you have the time, do more than one layout and compare the results. You can learn a lot from that.

BR,
Carl

[free_electron]

The feedback should be taken from the capacitor, not the coil...

Keep in mind that the largest current flows through the coil and the recirculation diode !
There are two current loops : one to charge the coil : so , input, input tank capacitor, chip, coil , output tank capacitor.
One to discharge the coil energy:
Coil , output tank capacitor , through the recirculation diode and back into coil...

Do not share current in via's ! That is a bad idea. A via has an impedance that changes over frequency. It also has a resistance. So , in systems like this with pulsed currents you get pulsed voltages across the via. You dont want crap from one branch injected into other branches.
That isa. Golden rule. Vias are never shared between branches, no matter how close as you cross contaminate signals.

See it this way :
Two currents. One switching from 1 ampere to 2 ampere and back. Another is 10 milliampere.
Lets say the via is 5 milliohm. The first current will create a signal that bounces from 5 millivolt to 10 millivolt across the via. So the second , weak, signal will see this asa voltage source... If that happens to be part of the feedback system you get instant 5 to 10mv of crap into that system...
Since the feedback only uses 1.25 volts...

Give each current its own return path and that problem disappears.

[Psi]

Double check your switching controller footprint is suitable for the actual component dimensions.

It maybe fine, but looks a bit small,  like maybe the metal tab would extend past the copper pad.

You want some extra copper past where the metal tab ends to get a solder joint you can inspect. If the copper ends right where the tab ends hand solder is more difficult.

[Frenchie]

Okay, based on the suggestions so far, I've revised and come up with the following:



I haven't done so yet, but I've left room to extend the thermal pad as per Psi's suggestions. I'll put a top copper fill around it and put thermal stitching vias down the sides.

Can't see how I can keep the feedback signal (and the enable signal) out of the current loop, but I'm assuming it'd be far better to go under the cap than the inductor.

Hand soldering this layout is going to be interesting... but such is life.

Thanks for the help! This is a long way from done so if there are further suggestions please let me know.

[peter.mitchell]

Hand soldering this layout is going to be interesting... but such is life.

it's all about order, id do the inductor first as there is no way to get to pin one with the switcher in place, then the switcher, diode, and bottom components, leaving the caps to last, because they can be soldered pretty messily in the middle, it doesnt matter if you put a massive blob on them  + the diode, in fact, generally, the more the merrier.