secondary turn-on

Don't think construction of choke is an issue, depending on how much leakage is intentionally present.

The voltage waveform will differ slightly on leading and lagging phase, dependant on PWdelay and load.

If your measurements show zero current in the lower fet before the second conduction event, then there's no issue there, either.

RL

with

harris

Thanks to all who helped with this- raising the high side gate resistors to

51R solved the problem- the peak turning on the low fet was current thru the CM choke from the opposite bridge??? I know it seems strange but there you go. this has helped the design no end, THD has plummeted to 0.01% @10KHz -10dB not bad for battery backed voice amplifier!!

On a different note- does anyone have any sneaky circuits for stopping the error amp output exceeding the ramp- Zeners are slow and act two early- I have used a complimentary pair of transistors with the bases dc biased to the ramp peaks less 200mV and a pair of 1n4148 to remove current from the error amplifier- this is fast and scales well with supply rails- but can't help thinking there is a more elegant way.

Thanks again Anthony

To the contrary, it makes good sense. In fact I wrote you a message from home this morning pointing out the issue, and suggesting you move the CM-choke to _after_ the normal-mode filter (which should be balanced and differential), or eliminating it entirely.

However before I could send the note, my computer crashed.

In fact it crashed so thoroughly a little Chinese voice came out of my speaker, "system failure, CPU... system failure, CPU... system failure, CPU... system failure, CPU..." Now the computer fails reboot attempts, or anything. I'll probably be out another $300 for a new Pentium 4 chip. It's OK, I'll send you the bill... I mean it's only fair, your note broke it.

Tell us more about that there error amplifier. What is it?

--
 Thanks,
    - Win

I see, you're trying to convert the floating differential H-bridge output to a grounded single-ended output at high frequencies, without using a transformer, etc., what a potential mess...

Your balanced filter should come first, and probably it should be in two stages, with a MHz input filter (low-capacitance inductors), followed by your audio-frequency filter. Then follow with your CM filter, which must have very good high-frequency performance, and finally, only if you insist, a pair of small caps to ground. Ouch.

I'm using my backup computer, a nearly empty linux machine.

Sorry, I just can't picture what you've got. Like to help, but...

--
 Thanks,
    - Win

Not in my book. :>) Stored-reference differencing is unreliable because voltage-probed high-frequency details can change as you move the probe from point to point. To measure current spikes I use Tektronix current probes, and I have six to choose from, from 500A at one extreme to 1GHz at the other. To measure differential voltages, such as the floating Vgs for the high-side FETs, or the H-bridge output, I use differential probes; either handmade, or a LeCroy / Preamble amplified wideband probe. With a 500MHz digital scope. That's what it takes to get the T-shirt.

1R series resistor?

--
 Thanks,
    - Win

Well, maybe I should say matched pair with common ground. This is a poor man's probe that works surprisingly well in difficult scenes.

.. scope RG-174 50-ohm coax .. ____________________/ .. (O)____________________)-950R--->

.. w/ 50-ohm |__________ local gnd. .. term ____________________| .. (O)____________________)-950R--->

Both scope channels have internal wideband 50-ohm terminations. So the end of the coax looks like a 50-ohm load to the matched 950-ohm axial-lead resistors, resulting in a pair of matched 20:1 attenuators. Or use 50-ohm resistors for 10:1 attenuation. The coax transmission line serves to isolate the local probed ground from the scope ground. And most power and high-frequency circuitry doesn't mind a 500 or 1k probing load.

But I still prefer my Tektronix, HP and LeCroy differential probes.

Get a better scope (mine doesn't do that). Or use an external fast clamp circuit to protect the scope from far-out-of-range signals.

Yes. Jim shows my resistor-coax-probe trick, page 80.

--
 Thanks,
    - Win

Removing the CM choke increases the current consumption as the filters charge/discharge- wasting valuable battery time, using the CM choke means the filters only work when audio is not at 0V (which it is most of the time! The output filter first section is balanced, second section the caps are ground referanced hence the charging- i needed to do this to help with getting the noise down on previous boards to pass emc- might try re-arranging the filter this weekend!

the H bridge output is RC filtered and "de-balanced" before passing into an inverting op amp node- input is cap coupled 3k3 resistor, feedback is 10k (from debalancer to node), node is compensated with 4k7/47p in series- op amp is TS462 from SGS and operates single supply 12V (swings0.3V to

11.6V) but the ramp only has 10V swing- derived from a xtal and passed thru integrator- gives an excellently symetrical triangle- most important for low THD

regards Anthony

Hi Win,

Care to elaborate on the home-made differential probes?

Also: trying to measure small Vds_on voltages when Vds_off is very large doestn work well - when the scope is set to small V/div it gets grossly over-driven by Vds_off. This causes thermal tails, which can take a long time to settle out, and can do a great job of corrupting the small-amplitude measurement just attempted. IIRC linear tech AN47 covers this in some detail, along with a great tute on scope probing.

Cheers Terry

a lovely way to make a broadband probe that nobody cares about when you melt it with a heat gun. Never thought of a pair thoough - thanks, I'll pinch that idea.

*sigh* one day when I'm rich and famous...

As does howard johnson in "High Speed Digital Design" although he (and I) use 1k resistors for divide-by-21.

Cheers Terry