Bypass your switching-regulator IC supplies, and filter their current-sensing circuits!

It would be nice to add some supply inductance when using some of those stinkin' fast controllers, but alas, they aren't made that way -- if they had separated the switch supply and control supply pins, perhaps.

So without snubbing as an option, you're stuck with heavy bypass. Which is okay for the chip, but can kick out some nasty EMI (examples: John's probable LMxxxx step recovery diode waveform; I've seen sub-5ns edges from LT designs).

Tim

-- Seven Transistor Labs Electrical Engineering Consultation Website:

A part with a fast current sensor inside might be prone to false-alarms if its Vcc is noisy. That might apply to peak-inductor-current sensing buck or flyback switchers. Some of them look for very small trip voltages.

The LM3102 switcher actually worked fine. But nothing else on the board did.

Hah, I've been sticking 10 ohms in my supply rails when things mis-behave, for a while. (I haven't done any SMPS's yet.) (Of course I've probably stuck a few ohms almost everywhere, when it's not working right.)

George H.

The combination of a 150-kHz Simple Switcher, a toroid, and a cap multiplie r is good for a lot of SMPS applications where low-level circuitry is invol ved. It's not brilliant at low input voltages, because the ~2V drop in the Darlington switch will eat your lunch, but at higher voltage, with its pred ictability and nice slow edges, it's the bee's knees.

Cheers

Phil Hobbs

e,

ier is good for a lot of SMPS applications where low-level circuitry is inv olved. It's not brilliant at low input voltages, because the ~2V drop in th e Darlington switch will eat your lunch, but at higher voltage, with its pr edictability and nice slow edges, it's the bee's knees.

Do you put the cap multiplier inside the feed back loop? Or just hang it on the output?

George H.

I tend to use 500kHz+, since it is easier to filter (more attentuation for a given LC or RC filter). Also copper planes work to shield the magnetic fields.

(My stuff is not super low-level like yours though! The actual signal is well below this switching frequency - so filtering works, and the higher switching frequency the better I guess.)

For positive supplies, I usually use split feedback, as you do when driving a heavy C load: the FB pin's voltage divider goes from the CM output to gr ound, and there's a smallish cap from FB to the switch pin for stability. Y ou have to watch how much ripple gets injected into the CM's emitter circui t.

For inverting supplies, where you ground the inductor and let the chip pump its ground pin negative, this doesn't work since the FB pin is referenced to V- instead of ground. An op amp can fix that, or else a TLV431 in the ba se of the PNP CM.

National used to make refs where the FB pin was referred to the cathode, wh ich was nice for negative regulators. Not as cheap as a 431 though.

Cheers

Phil Hobbs

ng a heavy C load: the FB pin's voltage divider goes from the CM output to ground, and there's a smallish cap from FB to the switch pin for stability. You have to watch how much ripple gets injected into the CM's emitter circ uit.

Ahh right. (I'd forgotten.) You've got to split the feedback, that always makes it a little trickier.

George H.

mp its ground pin negative, this doesn't work since the FB pin is reference d to V- instead of ground. An op amp can fix that, or else a TLV431 in the base of the PNP CM.

which was nice for negative regulators. Not as cheap as a 431 though.

With toroids, the main issue IME is capacitive pickup from the winding, so other things being equal, it's better at 150 kHz. Of course the toroid could be smaller at 500 kHz, but copper foil or a Laird shield gets rid of it pretty well either way.

Cheers

Phil Hobbs

Whoops--wrong side of the inductor. The cap goes to the output of the SMPS part, of course, where the voltage divider normally attaches.

Cheers

Phil Hobbs

In my application, generating +/-20V AC pulses with 1A currents, switching noise would seemingly not matter. But not necessarily so, whew, even the SMPS IC that was responsible for all the noise needed some quiet.

Considering the 300kHz switching frequency, and possible 85% duty cycles (that's under 450 ns off time) I chose a 4.7-ohm gate resistor, but that's no doubt too low. A snubber for the 10MHz ringing would make sense as well.