Switcher post regulator

[Alex Eisenhut]

Those caps are awfully small in relation to the inductors.  Are you sure you'll have enough value there -- at rated voltage and temperature?  This module likely isn't going to stay cool, even with an efficient chip!

Inductor size, I guess, seems reasonable given the gross ratings I'd expect from the form factor, or you can shrink both by choosing an even higher frequency chip (should be able to get some in the >MHz range?).

Also, there's no input filtering (beyond the bypass caps).  Is there other analog stuff this might back-feed into -- is that a concern?

Remember, the input side of a buck is the noisier side!

Tim

Hey, I said be harsh to my *schematic*! I expect nothing but praise for my layout. 

The big cap on the input, top left, is a tantalum that is removed now and I'll put more ceramic caps, with the resistance of the PTC fuse that should damp out a bit the oscillations. Output current is more on the order of half an amp for the application, not the ~3 amps from the form factor.

As for the feeding back, I was wondering about that, the linear transformer feeding the circuit has two secondaries and about 4 inches of loose wiring from the transformer to the bridge and main input caps. Inter-winding capacitance is not known (to me) but for sure it'll be the right value to let that 1MHz signal jump all over the place.

But yeah, coil and capacitor selection continues. I'm also considering the .75 inch "top hat PCB" to put the switcher on there and filtering on the base.

[Alex Eisenhut]

For example, L3 is far too much of a BMFL and there are ceramic inductors that are far smaller in size, which allows space for BMFC C9. Etc..

I'm starting from the TI datasheet and porting over the equations one by one. Their stupid software is too clumsy for me.

https://docs.google.com/spreadsheets/d/1v11noxAnF4kvpm88LzwxX3Mkwh_AWj5LsXuw3eF2t6A/edit#gid=0

The datasheet kinda sucks though.

[T3sl4co1l]

Don't know where those numbers are coming from...

I always go back to basics.  V = L * dI/dt.  This does it all for a few inputs,
http://schmidt-walter-schaltnetzteile.de/smps_e/abw_smps_e.html
e.g.
Vin min = 5
Vin max = 20
Vin calc = 10
Vout = 5
Iout = 0.5
f = 1000
L = 5E-6

The "proposal" is automatically generated (solves L for dI = 20% or something like that), which is rather conservative, more suitable for old powdered iron chokes than compact ferrites or modern powder composites.  So don't be afraid to overwrite that figure and put in something a bit more aggressive, like I've done here.

Note it says "Tip: The lowest value of L is achieved if ?IL=2·Iin for Vin_min."  You usually want to be around BCM (the boundary between CCM and DCM, where inductor current is just barely returning to zero each cycle) to maximize density.  Assuming that you can, since such operation does increase inductor losses.  (Nothing a modern SMT can't handle.)

5uH (say, 4.7 to 10uH real value in circuit?) seems pretty reasonable to me.  That'll be much smaller, which will allow you to put much bigger capacitors around.  Remember those have to be >>5V parts (X7R or better) to maintain value.  Probably 16V at the output, 25V or more at the input.  A single 1206 part should be more than enough (for each cap), plus additional filtering as needed/desired.

The layout itself -- don't get me wrong, it looks pretty good.  I never do anything that dense, though I always have the area to expand into, as well...  I probably wouldn't do much better myself.  And it's a neat constraint to work against, as long as you have the time to optimize it of course.

Tim