Half-bridge MOSFET drivers for class D audio?

[Tony]

Hi,

I'm heading down the class D power amp road, and I intended to use IR's dual
half bridge drivers. But the critical dead-time tradeoff (distortion vs
destruction) depends on device selection, temperature, load current, etc, and
therefore seems to require either:
- turn-on only after the opposite side drive has dropped to <1V
(complex driver cct and still slightly more than optimum dead time), or
- Crown's BCA technology which isolates the two MOSFETs
(patented, more expensive output filter).

Now LT's LT1336 driver seems to have the driver interlocking internal and fast,
but it only works to 60V. Does anyone know of a similar part (even if just in
the works) that works to, say, 200V?

Tony (remove the "_" to reply by email)

 

[Winfield Hill]

Tony wrote...

I'm heading down the class D power amp road, and I intended to use IR's dual
half bridge drivers. But the critical dead-time tradeoff (distortion vs
destruction) depends on device selection, temperature, load current, etc, and
therefore seems to require either:
- turn-on only after the opposite side drive has dropped to <1V
(complex driver cct and still slightly more than optimum dead time), or
- Crown's BCA technology which isolates the two MOSFETs
(patented, more expensive output filter).

Now LT's LT1336 driver seems to have the driver interlocking internal and fast,
but it only works to 60V. Does anyone know of a similar part (even if just in
the works) that works to, say, 200V?

My favorite part for this application is Intersil's HIP4081A, because it's
faster than the alternate parts (much faster than the LT1336). Yes, it's
limited to a 80V supply, but in the usual H-bridge configuration that gives
you 160V peak-to-peak, which should be more than enough! 500W rms, Sheesh!

Thanks,
- Win

whill_at_picovolt-dot-com

 

[R.Legg]

Hi,

I'm heading down the class D power amp road, and I intended to use IR's dual
half bridge drivers. But the critical dead-time tradeoff (distortion vs
destruction) depends on device selection, temperature, load current, etc, and
therefore seems to require either:
- turn-on only after the opposite side drive has dropped to <1V
(complex driver cct and still slightly more than optimum dead time), or
- Crown's BCA technology which isolates the two MOSFETs
(patented, more expensive output filter).

Now LT's LT1336 driver seems to have the driver interlocking internal and fast,
but it only works to 60V. Does anyone know of a similar part (even if just in
the works) that works to, say, 200V?

You seem to be confusing drive method with topology and modulation
method.

The BCA topology claims a reduction in delay-induced distortion due to
the fact that equal opposing delays are cancelled, as is ripple
current under certain conditions, due to the bipolar supply and
grounded load.

The same modulation scheme can be applied to four active switches,
though there is no natural guarantee that opposing delays will be
matched in either arrangement. It is potentially achievable in the
BCA, with great care, at zero load current.

The cancellation reduces feedback requirements and simplifies them; it
does not eliminate them. The BCA modulator is a ternary modulator, and
is not restricted to use in the BCA arrangement.

The asymmetric switch can reduce delay and is obviously simpler to
drive, but it does not eliminate the delay or reverse recovery
problems that you've mentioned.

RL

 

[Tony]

You seem to be confusing drive method with topology and modulation
method.

The BCA topology claims a reduction in delay-induced distortion due to
the fact that equal opposing delays are cancelled, as is ripple
current under certain conditions, due to the bipolar supply and
grounded load.

The same modulation scheme can be applied to four active switches,
though there is no natural guarantee that opposing delays will be
matched in either arrangement. It is potentially achievable in the
BCA, with great care, at zero load current.

The cancellation reduces feedback requirements and simplifies them; it
does not eliminate them. The BCA modulator is a ternary modulator, and
is not restricted to use in the BCA arrangement.

The asymmetric switch can reduce delay and is obviously simpler to
drive, but it does not eliminate the delay or reverse recovery
problems that you've mentioned.

RL

Sorry if my post appeared confusing (or confused). I understand the various
BCA-related issues that you mentioned (better now, since your explanation), but
my post really was only concerned with the drive issues. It mentioned BCA only
because its output inductors inherently eliminated shoot-through the two totem
pole MOSFETs by separating them (as does the Crest design, apparently, but in a
much less dramatic way). I considered that 30ns or so of diode reverse recovery
time could be easily handled with any practical MOSFET turn-on time. I haven't
decided on a modulation scheme, but at the moment a 2nd order delta modulator
interests me, as its inherent "immediate feedback" is far better able to
compensate for variations in delays, supply voltage and switch resistance.

Tony (remove the "_" to reply by email)

 

[Tony]

Tony wrote...

My favorite part for this application is Intersil's HIP4081A, because it's
faster than the alternate parts (much faster than the LT1336). Yes, it's
limited to a 80V supply, but in the usual H-bridge configuration that gives
you 160V peak-to-peak, which should be more than enough! 500W rms, Sheesh!

Thanks,
- Win

whill_at_picovolt-dot-com

Sorry to hear the LT is so slow. Actually I need around a 120V nominal supply to
properly drive the intended loads (pro audio speakers). I understood that it
would be unwise to operate the HIP4080A or HIP4081A even near their 80V rating,
not to mention power supply droop and tolerances (perhaps that warning was due
to spikes on bad PCB layouts?).

Tony (remove the "_" to reply by email)

 

[Winfield Hill]

Tony wrote...

... I understood that it would be unwise to operate the HIP4080A or
HIP4081A even near their 80V rating, not to mention power supply
droop and tolerances (perhaps that warning was due to spikes on bad
PCB layouts?).

You should use a well-regulated supply, which reduces distortion.
Spikes can be controlled, get with the program! Bottom line, you
can use the 4081 at 80V, or even higher (!) for personal equipment.
But of course it's not advisable in production.

WRT your 120V spec, do you need to drive your speakers to +/-120V?
For an 8-ohm load that's 1800 peak watts without clipping, more
with some distortion. Seems a bit much.

Thanks,
- Win

whill_at_picovolt-dot-com

 

[Winfield Hill]

Tony wrote...

I considered that 30ns or so of diode reverse recovery time
could be easily handled with any practical MOSFET turn-on time.

Another issue to watch out for is MOSFET turn-off-time delay.

Thanks,
- Win

whill_at_picovolt-dot-com

 

[Tam/WB2TT]

Hi,

I'm heading down the class D power amp road, and I intended to use IR's dual
half bridge drivers. But the critical dead-time tradeoff (distortion vs
destruction) depends on device selection, temperature, load current, etc, and
therefore seems to require either:
- turn-on only after the opposite side drive has dropped to <1V
(complex driver cct and still slightly more than optimum dead time), or
- Crown's BCA technology which isolates the two MOSFETs
(patented, more expensive output filter).

Now LT's LT1336 driver seems to have the driver interlocking internal and fast,
but it only works to 60V. Does anyone know of a similar part (even if just in
the works) that works to, say, 200V?

Tony (remove the "_" to reply by email)

Can't you transformer couple the gate drive to the FETs, the way you would
with an offline switcher? Then the max voltage is determined by the
breakdown voltage of the transformers, typically 1500V.

Tam

 

[Walter Harley]

WRT your 120V spec, do you need to drive your speakers to +/-120V?
For an 8-ohm load that's 1800 peak watts without clipping, more
with some distortion. Seems a bit much.

Hardly. Been to a pop concert lately? Even small clubs are pumping that
much these days. (I say this as a musician and a professional sound man,
not as a curmudgeon.) I personally own two class-D power amps with output
at least that much. Partly it's because everyone (but me) evidently wants
things louder. I have a theory that partly it's because over the last
couple decades amplified pop music has pushed about an octave lower in the
frequency spectrum. People are building and playing instruments intended to
emphasize fundamental frequencies in the range of 28 to 32Hz, almost
subaudible. Loudspeakers are less efficient and ears less sensitive down
there.

I suppose this is not actually a new thing, but rather a reversion to an
older thing: the last time we had high acoustic energy at very low
frequencies in pop music was with big pipe organs in churches.

 

[John Woodgate]

I read in sci.electronics.design that Walter Harley
about 'Half-bridge MOSFET drivers for class D audio?', on Thu, 15 Apr
2004:

I suppose this is not actually a new thing, but rather a reversion to an
older thing: the last time we had high acoustic energy at very low
frequencies in pop music was with big pipe organs in churches.

There aren't many pipe organs that will go below 32 Hz fundamental. You
can simulate C in the lowest (64 ft) octave by sounding C and G in the
next octave together (which make the second and third harmonics of the
low C), but it takes an awful lot of wind.

 

[Tony]

Hi Win,

Yes, it was the MOSFET turn-off times (delay and transition) that I was
concerned about in the original post; I'm still trying the weigh up the options.
The later comment was in answer to Rob's concerns.

Tony wrote...

Another issue to watch out for is MOSFET turn-off-time delay.

Thanks,
- Win

whill_at_picovolt-dot-com

Tony (remove the "_" to reply by email)