Synchronous Switchers

Do you use snubbers on the inductors? Your waveform is too compressed to see what's going on but it looks like you have ringing on the edges. Snubbers might help. How much improvement do you need?

Too bad they're not isolated.

Wouldn't it be better to use better isolation on the analog supply to the D/A? The on-chip stuff is garbage at those frequencies, they make a big deal out of

The on-chip stuff is garbage at those frequencies, they make a big deal out of

The DACs and opamps all have ferrite bead+bypass cap filters on all their supply pins. The switching noise is ripping through ground planes or power pours or something. If I short the post-DAC LC lowpass filter midway through, the spurs change but don't go away. I think the best fix is to use quiet-er switchers.

Here's the board. It's, like, 1/3 power supply, and it's not big enough to get things very far apart.

D/A? The on-chip stuff is garbage at those frequencies, they make a big deal out of 40dB attenuation which is useless in RF applications requiring spectral purity.

Which may be resonant at one of the harmonics. Ferrite beads may be lossy at sufficiently high frequencies but I've had to add resistance in series with a ferrite bead to kill a resonance.

You have to characterise that sort of by-passing fairly carefully if you need heavy attenuation. And you may need to by-pass the regular stacked layer ceramic capacitors with a 1nF percelain microwave capacitor (if they still make them).

Nope-- not needed. You have to know what you're doing, though -- which the integrated switch devices rarely do.

I won't post here, but I'd consider doing a reference design for you on contract ;-)

Tim

Under 10MHz? Nah. If anything, skip the big ceramics. Tantalum or aluminum (not polymer) is better. I'll leave it as an exercise for the student to discover why. :-)

Tim

There has to be some non-overlap in the fet totem pole, or you'll get huge shoot-through spikes. So there needs to be a catch diode to keep things happy during the small non-overlap interval. Doesn't there?

This is a great shot:

That's the LM3102. You can see

Lower fet on

Lower fet off, catch diode conducting

Upper fet turns on, catch diode refuses to turn off

Catch diode runs out of carriers, snaps off. SRD! The rise is 15 volts (and three bizillion amps) in a couple ns.

The snap would freak out opamps clear across the board. Adding an external schottky, in parallel with the internal diode, helped some but not enough.

Switchers with external catch diodes, especially slow stuff like older Simple Switchers, don't do this. The LTM8023 is very quiet.

Obviously. Stacked layer ceramics don't look resistive under a 100MHz. John Larkin has been known to work with higher frequencies.

Tantalums typically have a high enough ESR to damp the kind of resonances I was worrying about. Aluminium electrolytics can be less resistive, but they have higher ESR's than ceramics.

D/A? The on-chip stuff is garbage at those frequencies, they make a big deal out of 40dB attenuation which is useless in RF applications requiring spectral purity.

Is a linear supply out of the question?

Are you dyslexic ?

Jamie

Did you try a B-field probe? An imperfectly shielded inductor can easily couple big local currents into a ground plane.

Hz.

he

Stupid question - which is probably a redundant observation, given the source of the question.

Yep. Another Larkin sycophant. ...Jim Thompson

Apparently not when I do it :)

Tim

Now you're getting Thompsonish on us. Tell us how much you know but won't reveal.

You're just jealous that *you* don't have any sycophants. Well, there is JF, and you are welcome to that one.

Split ground plane! :)

Like James Arthur said, could well be magnetic noise from the not-so-well shielded switcher inductor.

I made a b-field probe with just a 1cm square loop of wire on the end of a BNC. Or you can probably use a smt inductor (I never thought of that at the time). You can put it on the end of another q-tip stick.