synchronous nastiness

Aren't they all ?:-)

Kill the "Q" of the inductor?

I'm very leery of these modern switcher architectures, playing insane games to reduce external component sizes.

...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 http://www.analog-innovations.com |    1962     |
             
 I love to cook with wine     Sometimes I even put it in the food

If the switcher is operating at only a few MHz (my brain cramps when I say "only a few MHz" about a switcher, and I'm not all _that_ old) then it shouldn't take much to do this.

--

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

Do you need to implement control loops in software?
"Applied Control Theory for Embedded Systems" gives you just what it says.
See details at http://www.wescottdesign.com/actfes/actfes.html

Back when I was designing off-line switchers for GenRad, I rarely operated above 25KHz ;-)

Back then we were more interested in heat inside the portable equipment rather than size or weight (*)

(*) Most hilarious moment for my technicians was when I grabbed the flag on a TO-220 transistor to see how "hot" it was. Since I had heavily snubbed everything, I didn't even need heatsinks. However I had forgotten about the 340V P-P on the flag... that'll really ring your chime ;-)

...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 http://www.analog-innovations.com |    1962     |
             
 I love to cook with wine     Sometimes I even put it in the food

The inductor (15 uH) isn't the bad guy... not at 150 MHz! And the rising edge + ring is so brutal that it laughs at/fries RC dampers.

My first switcher used a 709 opamp as a combined error amp/pwm generator, and a 2N2905+2N3055 pseudo-PNP as the switch. The 3055 didn't quite saturate, so was fast enough. It ran at 23 KHz, since I could hear 22K in those days.

John

parasitic noise, use some low value R's between the stages. Lowering the Q should take care of that. that's just an opinion of course. :)

"

How's the pcb layout? I've found that pcb layout is pretty important around the switchers I've done and seen. Loop area is really important with these high freq switchers. This is one area where isolation of ground is important to keep the nasty currents from spreading into other circuitry. I also like placing a ferrite bead on the output of the power supply before it fans out to additional circuitry if its feeding analog stuff (feedback connection is on the power supply side of the bead).

Mark

Try to put a small resistor in series with the boost capacitor. Doesn't make sense but it fixed an EMC problem for me.

--
Programmeren in Almere?
E-mail naar nico@nctdevpuntnl (punt=.)

An Amobead will probably fix it:

Look at some of the pdf's at the bottom of the page for examples of what an Amobead will do for you.

1. The parasitic inductance in the loop just turned off ( mostly trace from PGND to Cout) is getting nasty and oscillating with the node capacity at the SW output.

2, First try a 100nF cap from the high side of Cout to the PGND pin. This loop must be as short as possible. Now take a new picture of ringing to see where we are, a snubber from SW to Cout will be added next. Cheers,

Harry

It's the upper fet turn on that has the nasties; the falling edge is well-behaved. At the speed of the rise and ring, the 15 uH output inductor is an open circuit, so there's no dramatic current flowing into Cout.

I'm figuring there's 10's of amps flowing somewhere, in the ballpark of 1e10 amps/second.

We tried snubbing. The spike is so big, and the equivalent node impedance so low, that any snubber that even partially damps the ringing fries its resistor.

That's not the answer.

John

Ok, ok, I see your bottom side FET turned off for about 15nS before the top side turned on and started to pour current into that node charging up the parasitic capacity and supplying inductor current. When the inductor current is satisfied, the topside parasitic inductance gets nasty and keeps supplying current, overcharging the node capacity. The parasitic inductance is in the top side because the topside parasitic diode never turns on, it is isolated from the input cap by this inductance.

Answer, slow down that topside turn on to > 0.35/150MHz => 2.4nS

Bypass the BJT inputs of the opamps. A few ten picofarads from IN+ to IN-, SMT bead right in front of each. If you want to be extra good use one of those Murata T-filters. Small, SMT, cheap. If it's really, really nasty you'll need a piece of copper tape over each opamp with a short tie to the ground plane.

BTDT. In our case cell phones were causing the grief, especially the GSM kind. Try it out, ask around who has an AT&T GSM phone (not iPhone) or T-Mobile. Then turn it on while holding it right next to the box. Blackberries also pack quite a punch.

Using CMOS opamps helps a lot but sometimes there justa ren't any that fit the bill.

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

Two points:

1. It's inside the IC, so I can't slow it down.

1. The fet isn't turning on fast enough to make that rise.

John

How about a different inductor using a much lower Q core at the present switching freq?

Also, one could implement a notch filter via a series LC at that point tailored for the ringing freq.

Another thought, placing a series RC in parallel with the inductor to off shift the ringing freq to something much lower. The R would lessen the Q on the over all. That's just a thought.

I've dealt with this issues before, inductor design has a lot to do with it.

We have 100 KC hot cathode oscillator tubes systems which drive a rectifier stack in a SF6 gas pressured vessel that houses a rectifier multiplier stack to give us ~ 1.5 M volts of DC at the other end. The inductor on the plate of this OSC tube is a multi wound segmented type. This is done to prevent unwanted artifacts from developing due to parasitic LC in the coil construction of a large body type. So this unit has multiples of smaller inductors in series to form a single unit.

"

This is what's going on:

ftp://66.117.156.8/SwitcherRise.JPG

For the first 15 ns, the lower fet of the synchronous switcher is on. The output voltage is about -0.5 volts, roughly, which is Rds-on of the lower fet times the load current. In preparation for turning on the upper fet, the chip switches the lower fet off for the next 20 ns. Now the load current is flowing through a pn junction diode internal to the chip, essentially the substrate diode of the lower fet. That has a drop of about 1 volt. The 20 ns of both-fets-off time is intended to make sure they don't generate a huge shoot-thru current.

But when the upper fet turns on, at 35 ns, the substrate diode reverse-conducts, and a huge current flows through the upper fet through the diode to ground. After about 5 ns of reverse conduction, the diode snaps off, and the output zooms up. It's acting like a step-recovery diode, or maybe even a Grekhov DSRD. This huge current winds up the chip's wirebond and lead inductance, which is what's ringing. The waveform is classic.

Two fixes:

Add an external schottky diode to clamp the output at -0.5 or so, so the slow PN diode doesn't turn on.

and/or

Replace the bipolar opamps with fet amps, which are less eager to rectify RF.

National should have done something about this. Used a schottky if possible, or a soft-recovery diode, or warned people to add an external schottky.

John

It doesn't take much C to resonate at 150 MHz (0.075 pF). If the energy is that high, you'd see high currents circulating (and heating things up) between the inductor and the mystery C..... ...unless the C is in the inductor itself. Parasitic capacitance between the windings perhaps?

--
Paul Hovnanian     mailto:Paul@Hovnanian.com
------------------------------------------------------------------
The opinions stated herein are the sole property of the author. All
rights
reserved. Void where prohibited. For external use only. Standard
disclaimers 
apply. If irritation, rash or swelling occurs, discontinue use
immediately
and consult a physician.

If this is what I think it is, the amo beads would most likely fix it, but you don't have direct access to the FET drain or source leads so I guess that ain't gonna work...

Another trick is to slow down the turn on of that top FET. You don't have direct access to its gate, but you sort of do by way of the bottstrap capacitor, Cbst.

Try adding a series resistor to Cbst of a few Ohms to sort of accomplish the same thing since it's the only available access point.

I saw this in some application note or something somewhere so it isn't something I just thought of.

boB

If they did that then they couldn't get you to pay them for tech support!

This is sorta 'related' to your problem...

Page 5 shows the fix I believe, which, with that part would have to be implemented with a resistor in series with the bootstrap cap. (and the diode too maybe)

boB