Class D amps with ext clock over 1MHz?

[Joerg]

Hello All,

Is there a fairly comprehensive list on the web about class D audio amps
that run above 1MHz, synchronized or externally clocked?

Ideally I am looking for something from the mainstream suppliers such as
National, TI, ST, Philips. Their web sites aren't all that helpful here.
Their product selectors are sometimes incomplete and don't list clock
speeds so you have to trundle through every single datasheet to pry out
that info. Done that for several dozen and the only one I found was the
MAX9712 but nothing for the big suppliers.

Regards, Joerg

 

[Pooh Bear]

Hello All,

Is there a fairly comprehensive list on the web about class D audio amps
that run above 1MHz, synchronized or externally clocked?

Ideally I am looking for something from the mainstream suppliers such as
National, TI, ST, Philips. Their web sites aren't all that helpful here.
Their product selectors are sometimes incomplete and don't list clock
speeds so you have to trundle through every single datasheet to pry out
that info. Done that for several dozen and the only one I found was the
MAX9712 but nothing for the big suppliers.

None that I know off offhand. Tripath's 'spread spectrum' design maxes out
at about 600kHz IIRC.

Haven't seen any device featuring sync either.

Any special reason to go that high ?


Graham

 

[Joerg]

Hi Graham,

Any special reason to go that high ?

I wanted to see if they can be used as precision pulse width modulators
up to 1-2MHz. With the THD levels they claim, these amps should be quite
precise. Most of all they would cost a whole lot less than rolling your
own from CMOS chips and discretes.

Regards, Joerg

 

[Winfield Hill]

Pooh Bear wrote...

None that I know off offhand. Tripath's 'spread spectrum' design
maxes out at about 600kHz IIRC.

One serious issue that's not often talked about is an asymmetric
power MOSFET turn-on and turn-off time. Moreover, FET turnoff
time has a slow recovery tail, and is memory dependent for short
time intervals. The distortions from this issue are exacerbated
at high PWM frequencies, and lead to a degraded performance.

 

[Pooh Bear]

Pooh Bear wrote...

One serious issue that's not often talked about is an asymmetric
power MOSFET turn-on and turn-off time. Moreover, FET turnoff
time has a slow recovery tail, and is memory dependent for short
time intervals. The distortions from this issue are exacerbated
at high PWM frequencies, and lead to a degraded performance.

Absolutely. Just noticed the same today looking at IGBTs.

Tripath get 'better than most' results by taking feedback from the output
to counter this. They have a programmable dead-time too. Needless to say -
the shortest dead-time gives the highest performance with the greatest
risk of cross-conduction.

Driving gates at high frequencies takes some current too !


Graham

 

[Ken Smith]

Pooh Bear wrote...

One serious issue that's not often talked about is an asymmetric
power MOSFET turn-on and turn-off time. Moreover, FET turnoff
time has a slow recovery tail, and is memory dependent for short
time intervals. The distortions from this issue are exacerbated
at high PWM frequencies, and lead to a degraded performance.

This sounds like a job for a feedback loop. You could measure the
difference between the turn on delay and the turn off delay and apply that
correction on the next cycle.

 

[Winfield Hill]

Ken Smith wrote...

This sounds like a job for a feedback loop. You could measure the
difference between the turn on delay and the turn off delay and
apply that correction on the next cycle.

Right, but then you're analog at a critical spot, and no longer a
cool completely-digital-to-the-speakers system.

 

[Joerg]

Hi Winfield,

One serious issue that's not often talked about is an asymmetric
power MOSFET turn-on and turn-off time. Moreover, FET turnoff
time has a slow recovery tail, and is memory dependent for short
time intervals. The distortions from this issue are exacerbated
at high PWM frequencies, and lead to a degraded performance.

With a monolithic solution the chip designer would probably compensate
at least for some of these transition differences. One of the Maxim
class D chips (MAX9712) is claimed to run at 0.01% THD when clocking at
1.1MHz. It doesn't go much above 0.02% for most of the output power
range. It might get worse when running it up to its 2MHz clock maximum
but probably not by a lot.

Regards, Joerg

 

[Tony]

This sounds like a job for a feedback loop. You could measure the
difference between the turn on delay and the turn off delay and apply that
correction on the next cycle.

Yes, dead time remains the critical missing block in good class D
design. Even a few ns can cause serious (to an audiophile) crossover
distortion, and a new ns of overlap can start to heat things up quite
a bit. Plus it all varies with load impedance, current, temperature
and devices, so there's really no practical option but to control it
dynamically (and quickly) to some criterion (peak punch-through
current?). LT did a neat bias IC for class B linear output stages
decades ago. Wouldn't it be great if IR could integrate dynamic dead
time control into one of their drivers? Then the differential
propagation delay specs wouldn't be so critical. If someone doesn't do
it I guess I'll need to before I finally get around to that big class
D monster I've been planning for years; either that or steal Crown's
method (which itself is not without complications).

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

 

[Ken Smith]

Ken Smith wrote...

Right, but then you're analog at a critical spot, and no longer a
cool completely-digital-to-the-speakers system.

I won't tell if you don't.

You could run the switching stuff into the timing circuit running at many
GHZ and then feed that number back into a DSP thus closing the loop
digitally.


Or, you could just enclose the whole thing in an analog servo loop and get
better linearity at the cost of some marketing boost.

 

[Ken Smith]

[... Class D and MOSFET switching ..]

Yes, dead time remains the critical missing block in good class D
design. Even a few ns can cause serious (to an audiophile) crossover
distortion, and a new ns of overlap can start to heat things up quite
a bit.

If you want real big fun, try it with bipolars.

Many years ago, I made a class D with bipolars as the switches. Even with
very complex Baker clamping circuits, the switching times were way
different. I ended up wrapping the switches with Schottkys and enclosing
the whole thing in a servo loop. The result sound quite good. AM radios
didn't like working near the circuit however.

 

[N. Thornton]

Ken Smith wrote...

Right, but then you're analog at a critical spot, and no longer a
cool completely-digital-to-the-speakers system.

I assumed that rather than implementing audio feedback, he meant just
measure the device turn on and turn off times, and feedforwardly tweak
the on and off times by that to cancel the problem. Thus all digital.

NT

 

[classd101]

Ken Smith wrote...

Right, but then you're analog at a critical spot, and no longer a
cool completely-digital-to-the-speakers system.

Hi,

Tripath's hysteretic modulators, or "spread spectrum frequency
modulator" can go up to over 1Mhz during idle/low signal conditions.
It's an analog modulator, with adaptive filtering to learn and
digitally control the timing of the output stage, I'm not sure it
qualifies as taking feedback from after the filter, don't think it
does though.

What is it that we don't like about fully analog, self oscillating
designs? There's a number of them out there, some of which are of the
highest quality out of any class of amp, considering the power levels
that is. Non switch near 1Mhz either.

Thanks,
Chris

 

[Pooh Bear]

I assumed that rather than implementing audio feedback, he meant just
measure the device turn on and turn off times, and feedforwardly tweak
the on and off times by that to cancel the problem. Thus all digital.

This is a 'yeah but' argument.

Switching times alter with load current and temperature. You can never fix the problem reliably that
way.

Graham

 

[Pooh Bear]

Hi,

Tripath's hysteretic modulators, or "spread spectrum frequency
modulator" can go up to over 1Mhz during idle/low signal conditions.

Maybe on newer versions. It topped out at about 600-800 kHz a few yrs back.

It's an analog modulator, with adaptive filtering to learn and
digitally control the timing of the output stage, I'm not sure it
qualifies as taking feedback from after the filter, don't think it
does though.

It takes feedback pre the output filter IIRC.

What is it that we don't like about fully analog, self oscillating
designs? There's a number of them out there, some of which are of the
highest quality out of any class of amp, considering the power levels
that is. Non switch near 1Mhz either.

Which were you thinking of ?

Agreed that none switch at 1 MHz or near that.


Graham

 

[Ken Smith]

I assumed that rather than implementing audio feedback, he meant just
measure the device turn on and turn off times, and feedforwardly tweak
the on and off times by that to cancel the problem. Thus all digital.

[/QUOTE]

Yes this is what I was arguing and I believe Win understood the point.
Win is quite right that the system would not be truely completely digital.
The measuement would contain at least a small analog section.

Assuming that, I will now disagree with Win (or maybe this is just a
misunderstanding):

Switching times alter with load current and temperature. You can never
fix the problem reliably that
way.

If we assume that the switching frequency is well above the highest signal
frequency and that the amplifier is never driven to clipping, I think we
can make the following simplifications:

(1)
The die temperature does not change much from on cycle to the next.

(2)
The current in the MOS-FET, the next time its on, will be predictable
based on the previous input data and the newest point.

(3)
The delay characteristics of a MOSFET at a certain temperature and current
changes only very slowly as it ages.


Based on this, I suggest that a clever enough circuit that contains the
following would work:

You will need some circuit to convert the large amplitude signal that
indicates whether the MOS-FET is conducting or not into a digital logic
level.

This logic signal would be connected to some sort of counter clocked at,
lets say, 5GHz. This counter would measure the time between the ideal
turn on time and the true turn on time and the ideal turn off to true turn
off.

either:

(A)
The numbers from this counter are fed back to the PWM circuit which is
also clocked at 5GHz. In the PWM circuit, the delay times are subtracted
from the ideal times and these corrected times are then used to drive the
outputs.

or:

(B)
The numbers from the counter are used to keep a table discribing the
MOS-FETS characteristics up to date. Values from this table are then used
to adjust the numbers before they are presented to the PWM output section.


The weakness in this idea is that it assumes that the MOSFET delays and
then switches perfectly after that delay. This circuit can improve
matters but without correcting for the actual turn on and turn off shapes,
the results will still be less than ideal.

 

[Genome]

| Pooh Bear wrote...
| >
| > Joerg wrote:
| >
| >> Is there a fairly comprehensive list on the web about class D audio
| >> amps that run above 1MHz, synchronized or externally clocked?
| >
| > None that I know off offhand. Tripath's 'spread spectrum' design
| > maxes out at about 600kHz IIRC.
|
| One serious issue that's not often talked about is an asymmetric
| power MOSFET turn-on and turn-off time. Moreover, FET turnoff
| time has a slow recovery tail, and is memory dependent for short
| time intervals. The distortions from this issue are exacerbated
| at high PWM frequencies, and lead to a degraded performance.
|
|
| --
| Thanks,
| - Win
|
| (email: use hill_at_rowland-dotties-org for now)

I'd like to know the basis for that one?

DNA

 

[Genome]

|
|
| Winfield Hill wrote:
|
| > Pooh Bear wrote...
| > >
| > > Joerg wrote:
| > >
| > >> Is there a fairly comprehensive list on the web about class D audio
| > >> amps that run above 1MHz, synchronized or externally clocked?
| > >
| > > None that I know off offhand. Tripath's 'spread spectrum' design
| > > maxes out at about 600kHz IIRC.
| >
| > One serious issue that's not often talked about is an asymmetric
| > power MOSFET turn-on and turn-off time. Moreover, FET turnoff
| > time has a slow recovery tail, and is memory dependent for short
| > time intervals. The distortions from this issue are exacerbated
| > at high PWM frequencies, and lead to a degraded performance.
|
| Absolutely. Just noticed the same today looking at IGBTs.
|
| Tripath get 'better than most' results by taking feedback from the output
| to counter this. They have a programmable dead-time too. Needless to say -
| the shortest dead-time gives the highest performance with the greatest
| risk of cross-conduction.
|
| Driving gates at high frequencies takes some current too !
|
|
| Graham
|

IGBTs are a different class of beasty.

DNA

 

[Ken Smith]

IGBTs are a different class of beasty.

They are not really that different. They look a lot like a N-MOSFET
driving a PNP. I think we should be developing class D audio amplifiers
using mercury ignitrons. Using the capacitively coupled quenching
circuits, you'd have to have some very tricky DSP code to get the PWM
right. You would have lots of power and that nice "warm tube sound" so
there would be a market.

 

[N. Thornton]

This is a 'yeah but' argument.

Switching times alter with load current and temperature. You can never fix the problem reliably that
way.

Graham

yeah but...
temp is slow changing compared to the 600kHz etc switching speed, so
thats not a problem
load current ditto, though its faster moving than theta.

NT