synchronous nastiness

At the end of the day the $64000 question will be: Does it pass EMC?

Drill a hole and put a drop of muriatic acid in it :-)

That ought to slow it down ...

To the uninitiated: DON'T DO THAT! Muriatic acid is about the worst stuff, it can eat a whole into concrete.

Yep, it could do that, or maybe eat a HOLE into it ;-)

...Jim Thompson

-- | James E.Thompson, P.E. | mens | | Analog Innovations, Inc. | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | Phoenix, Arizona 85048 Skype: Contacts Only | | | Voice:(480)460-2350 Fax: Available upon request | Brass Rat | | E-mail Icon at

is

Will it eat the hole IC too?

Oops ...

Well, I talked German most of the morning. Out of roughly 130 people at church service there were ten (!) Teutons. Quite unusual for this area.

is

Ya sure, ya betcha, da hool IC :-)

message

keeps

inductance

is

Probably ;-)

...Jim Thompson

Will it eat the IC's electrons after it's done with the holes?

Tim

That's what I get for not looking at the LM3102 data sheet. I just figured it used an external FET. :-( ..

You misspelled "welectrons".

--=20 Keith

I don't think so? Google only gives 207 hits for "welectrons", most of them academic, curiously (I suppose I could be missing a subscript in there). I've not encountered the term before.

Tim

The addition of external diodes may be ineffective, as diverting current flow from one loop to another, under the influence of only a few 100mV of diode forward voltage differential, will be impractical in the time frame available.

Reducing turn-on speed of the upper fet is effected through an increase in gate drive source impedance, by placing impedances in the drive source's supply - the boost cap - as is suggested elsewhere.

RL

I recently had the same problem, only the ringing frequency was 200 MHz. The topside FET oscillates well into it's conduction cycle. These newfangled efficient, fast FETs and N-channel charge pump controller drivers are the culprit. The waveform you captured looks familiar.

It helps to look at the switcher like a Colpitts oscillator. The bottom FET, as it's turning off and losing charge looks like a capacitor. In addition to the parasitics between the switch node and ground, and in concert with the source-to-drain capacitance of the topside FET, you get the classic capacitor divider of the Colpitts. The inductor is formed in the few nH of inductance between the gate of the topside FET and the controller output. The successful solution involved one or more of three methods (We did 1 and 3)

Frank

any exposed aluminium will gone in a flash, it won't eat the gold, copper(much) or exposed silicon.

it takes nitric acid to disolve copper, aqua regia to dissolve gold, and some really nasty chemicals to disolve silicon.

the acid only attacks the cement in concrete the agregate (sand and gravel) will be left behind.

Bye. Jasen

The mother of acids (sulfuric) will dissolve copper quite fine. Hydrochloric not so much, but it will.

If you want fast rates, flourboric acid (BF3) buffered with hydroflouric acid does quite nice.

Actually it will attack most sedimentary rocks, some igneous rock and will weather metamorphic rock. It won't bother sand much.

I came to this thread late, but

This is a very common problem in synchro switchers - Don't let that slow body diode turn on! Paralleling with a lower Vf schottky fixes the problem. There are some IC's designed for external FET's that claim adaptive shoot through protection that may time the fets well enough that this is not a problem. I think there are also MOSFETs available that have a built in schottky diode, however, I have not used them, and wouldn't help with your problem, since the fet is internal to the IC. Most of the switchers I design start at 50+ watts, and I always include the schottky diode.

Your right - there should be a built in schottky (can they be produced in a standard cmos process???), or a least a warning to use a external schottky, but that would mean a higher parts count for a simple switcher, which would make it less attractive to small = better and cost conscious buyers.

It's also not the first chip to have very serious issues from major vendors. Stuff like code not executing as documented in some processors, low voltage / low power quad OP amps getting strange offsets below a certain temperature threshold (like -20 deg C) (vendor admitted problem!), 30 db variance in sensitivity of RF FSK demodulator, etc

This can fix the problem, but will drop the efficiency of the power supply. The efficiency drop might be significant at those frequencies. Not likely a problem in John's case, since the unit is bench powered, unless the power dissipation in the chip becomes too much.

I have found that this can also ~increase~ the efficiency of the power supply by lowering the losses from this particular mechanism problem.

boB