Finding sufficiently crappy inductors

I use 35v tants on the outputs of 317 and 1117 type regs, 3.3 and 5 volts, sometimes 12, and they seem fine. The ESR is just right for those regs, even with a lot of ceramic bypasses all over the place.

Yeah, that's why I changed it, to putting a transformer in series with the undamped inductor. Unlike a pi filter with a resistor, there's no DC drop to contend with, which could be important if the necessary resistance is high.

Divided? I thought you /wanted/ low Q!

Tim

yeah, but not zero :)

NT

Q goes to 1 at high frequencies, because skin effect is diffusive. :)

Tim

add a network at the PS and loop output but before the filter..

the network consists of R in series with C to ground. this will D Q the load with no impact on the post filter. maybe you need R C __and L__in series.

series resonate at the freq it wants to oscillate adjust the R as required.

Hi,

What would your switchers do if you use an IC to create a spread spectrum clock from the 2.15MHz? That could have the effect of cancelling a lot of the output filter feedback resonance maybe.

cheers, Jamie

Hi,

Another idea, what about reducing the feedback loop frequency response of the switchers and then adding electrolytic capacitors after the PI filter, near the loads. If the loads current ripple isn't too high it shouldn't need to be compensated for by the switcher anyway, and the PI can manage it.

cheers, Jamie

I misunderstood your second post. That might well work, if I can (a) find such a part, and (b) fit it into the design.

Thanks

Phil Hobbs

I do. But shorted-turn effects greatly reduce inductance, which I don't want.

Cheers

Phil Hobbs

Hi,

I think using SEPIC converters in place of the buck and boost converters could also solve the issue, as they could push/pull their output to keep it regulated.

cheers, Jamie

I must have hopelessly misunderstood what you want, it sounds as if you wan t cores that work great at some unknown hf, perhaps 2MHz, but lose all q at 150kHz.

Wire rope, even if each strand were insulated, which stainless steel is, wo uld surely have no prospect of doing that. But I can think of something tha t just might.

NT

Putting an L-R network in parallel with the L is good for damping input fil ters. I don't see why it wouldn't work for a buck output stage as well. A nd the damping L is smaller than the original L, as opposed to using a damp ing C larger than the original C.

I always reference this paper for picking the value of R:

Well, it's best to start from what you really want and work down to what you can get. ;) Whit's transformer-coupled resistor would be a good fit if I could afford the space. (The usual thing.)

Cheers

Phil Hobbs

Thanks. In this case the output filter resonance is destabilizing the FB loop--there's no significant frequency component of the load current that's up above 1 kHz. I'll have a squint at the paper though.

Cheers

Phil Hobbs

of

e.

't

want cores that work great at some unknown hf, perhaps 2MHz, but lose all q at 150kHz.

, would surely have no prospect of doing that. But I can think of something that just might.

A wire with ferrite beads on behaves as you want at dc & 150kHz. A small ca p across it could let hf through.

NT

Ferrite is pretty good at 150 kHz. Where would the loss come from?

Cheers

Phil Hobbs

yes :) I meant whatever inductor suits your 150kHz purpose.

NT

That's the one I'm trying to find. Works great at 2 MHz. ;)

Cheers

Phil Hobbs