[time-nuts] Anybody want a Thunderbolt power supply?

jimlux jimlux at earthlink.net
Wed Aug 31 17:00:33 EDT 2016


On 8/31/16 1:52 PM, Adrian Godwin wrote:
> I'm being Devil's Advocate here because I certainly realise switchers do
> generate high frequency noise. But wasn't the intention of them to make
> filtering easier, with smaller filter components, exactly because they
> operate at higher frequencies ?
>
> So why do they fail ? Is it cost-cutting to make them only just good enough
> for typical uses ? Shouldn't it be possible to make the best possible
> supply from a switcher, if only cost and weight weren't the first
> considerations ?
>

Sure.. you can do a bunch of LC low pass sections and knock it down, but 
building a *wideband* low pass filter with good ultimate rejection is 
also challenging.

In a system I'm actually sitting next to, we have a switcher followed by 
a 60dB rejection lumped LC followed by the power distribution, followed 
by a 60 dB rejection lumped LC followed by a linear regulator.

When you're at this kind of level, layout and mechanical arrangement is 
important, because the switching noise (which is also on the input) can 
couple "around" your other circuits.

The other challenge is that those filters have resistive loss - if you 
want good regulation, how do you get your "sense" voltage back to the 
regulator without that path being the path for the noise - so you'd need 
to put LP filters on the sense line.

And of course, the L's need to be shielded.
And you put it in nesting boxes made of steel

A linear regulator is a LOT easier.

If you aren't power dissipation constrained, then it's often the easiest 
way to get to where you want to be.


More information about the time-nuts mailing list