practical tips on SMPS design

Is that LTC's an70, the low-noise SMPS app note? BTW, are you sufficiently set-up yet to work on my low-noise power supply?

One would assume it's LTC's app note (the one that has the amusing flowchart in the front).

IMO, LTC has some of the best app notes / documentation in the SMPS arena.

Cheers

PeteS

AN70 is a good application note

martin

Did you make any progress on it?

I've toyed half a day with Tony's two cores idea for the low parasitics, with dual apertures cores that I have at hand (EPCOS B62152-A4-X30, AL=10uH).

primary 2x2T interm. 1T secondary 2x2T

measured leakage ind = 260nH primary/secondary parasitics = 2.5-3.5p

I'll resonate Lleak on the 1 turn link between the 2 cores at 4.5MHz with

Sims are encouraging:

I don't have adequate mosfets and the schotkies to try it fast, but I'm confident it should work nicely. I may have to use it in the coming weeks, so we'll know.

In order to reduce the coupled common mode voltage I also think it'd be nice to center tap the intermediate winding with 2x47p caps or something refered to the primary side GND.

Any other idea?

In message , Fred Bartoli writes

Can you get a fine coax through the cores?

One turn each, join the inners to form the loop. For signal handling connect each screen to it's respective 0v and for power handling have a continuous screen, connected to primary 0v.

My old boss would also earth each core to it's own 0v, usually with a strap of 26-gauge tinned copper over the outside of the core.

"Tony Williams" a écrit dans le message de news: snipped-for-privacy@ledelec.demon.co.uk...

nice

refered

Good idea! The coax has to be pretty fine (hole diameter=3.4) but RG178 or RG179 should fit in together with the dual 2 turns 0.2mm windings. The cross coupling cap should be about 4p for the RG178 and 3p for the RG179, so no pb on this side. The leakage inductance will increase though, but I should be able to tune it out correctly, so probably no pb here.

Better than the dual 47p which was estimated to require a 20mA pk current.

Let pass some time...

Hmmm... thinking a bit more... The screen loop can't be closed: it'll become the winding turn. I now think there's finally no advantage (for a one turn winding) to go the coax route: same voltage induced into the shield than into the center conductor. Even more the parasitic capacitance will probably increase.

If I can find a small piece of these coax (hey the plural patrols, how do you make the plural mark on this one?) without buying an whole roll I'll make a comparison, just to make sure.

I'll probably strap the cores with the sticky copper foil used to solve EMC problems: stick it to the core and solder on ground plane. Better shielding and the cores are guaranteed to not move.