What modern MOSFETs for class D?

[GK]

I may have a need to construct a class D amplifier whose duty it will be to deliver a 200Vp-p, continuous-level 20 kHz sine wave at 1.9kW input max.* from a single DC power supply rail of 160V. Given supply constraints it will therefore need to be a bridged design.

A switching frequency of 100 kHz will be sufficient. What model MOSFETs should I be considering? I don't think the voltage is high enough to make IGBTs worthwhile.



* With no worse than 1.5kW delivered to the load.

[peter.mitchell]

At 100khz you can get away with nearly anything if you have a good gate driver.

[GK]

This is what the worse case load current (1500W) waveform will potentially look like.

[lewis]

What's it for Glen out of curiosity? This seems a departure from your usual uber-quality audio designs...

[fcb]

That doesn't look much like a sinewave..

I'd certainly consider IGBT's at that power level, there are some nice generation-4 parts that work well at 20KHz (I'm using them on a project).

[GK]

The voltage is sinusoidal; the load is non linear.

[GK]

What's it for Glen out of curiosity? This seems a departure from your usual uber-quality audio designs...

High frequency power distribution (for small transformers); nothing at all to do with audio.

[dannyf]

Not sure if it is modern enough for you, but check out IRF's class D audio amp reference designs. Lots of "new" mosfets used there.

[diyaudio]

I may have a need to construct a class D amplifier whose duty it will be to deliver a 200Vp-p, continuous-level 20 kHz sine wave at 1.9kW input max.* from a single DC power supply rail of 160V. Given supply constraints it will therefore need to be a bridged design.

A switching frequency of 100 kHz will be sufficient. What model MOSFETs should I be considering? I don't think the voltage is high enough to make IGBTs worthwhile.

* With no worse than 1.5kW delivered to the load.

I have successfully developed a class d amplifier capable of  1600 watts RMS into a 4 ohm load with a regulated SMPS power supply (all around one comparator,  and a IR2011) . here is a screen shot of what it behaves like in simulation, im busy improving the compensation loop for linearity and stability.

You can evaluate the following:

TO-220 parts:
-----------------------
IRFZ48V 100V 36A .021ohm 42nC/41nC
IRFB4019 150V 17A .080ohm 13nC/160nC
IRFB4321 150V 83A .012ohm 71nC/300nC
IRFB4020 200V 18A .080ohm 18nC/280nC
IRFB4227 200V 65A .020ohm 70nC/430nC
IRFB4233 230V 56A .031ohm 120nC/760nC
IRFB4229 250V 46A .038ohm 72nC/840nC
IRFB4332 250V 60A .029ohm 99nC/820nC


With the power levels you describing its no simple task achieving those power levels.  PCB design is mandatory, forget everything you understand concerning class-AB theory and star ground layout.
   



 

[GK]

, forget everything you understand concerning class-AB theory and star ground layout.

Really? So, for example, my ground return wire to the power supply can double as my signal ground with no ill effects??

I've seen numerous Class D audio amplifiers cocked up because the builder cobbling the thing together from an IC datasheet or application note didn't understand the basics of building a conventional "analogue" amplifier.

[GK]

I'd certainly consider IGBT's at that power level,

IGBT's have their place (HV switching) but I don't think they would perform better than MOSFETs in this application (LV, high current - 100A peak). An IGBT conducts in the reverse direction through the internal "hyper fast" anti-parallel diode. A MOSFET, on the other hand, conducts with low rds on in both directions and body diode conduction is actually something that is deliberately avoided (the body diode is generally a "slow" switcher).

The lower d-s voltage drop when conducting in the reverse direction makes MOSFETs more efficient.

[megajocke]

You can evaluate the following:

TO-220 parts:
-----------------------
IRFZ48V 100V 36A .021ohm 42nC/41nC
IRFB4019 150V 17A .080ohm 13nC/160nC
IRFB4321 150V 83A .012ohm 71nC/300nC
IRFB4020 200V 18A .080ohm 18nC/280nC
IRFB4227 200V 65A .020ohm 70nC/430nC
IRFB4233 230V 56A .031ohm 120nC/760nC
IRFB4229 250V 46A .038ohm 72nC/840nC
IRFB4332 250V 60A .029ohm 99nC/820nC

That table looks familiar. There is some more explanation, for example of the xx nC/ yy nC values, over where I saw it the first time:
http://www.diyaudio.com/forums/class-d/112065-nice-mosfets-class-d-ir-how-they-try-impose-us-directfet-thrash.html

The body diodes of those are quite good compared to old-school types like IRFP260.

Many of those IRF parts also have TO247 versions, for example the IRFP4227PbF if you can tolerate higher parasitic inductance to get a larger contact surface for thermal transfer. There are also some parts with even lower on resistance. For IR FETs there is a table in the following document on page 38 (PDF page 42), though I'm not sure this is the current version:

http://www.irf.com/product-info/selection-guide/IRProdGuide.pdf

I know some makers (for example Fairchild and IXYS) make IGBTs with 300 V ratings, but in this case the voltage is probably low enough that MOSFETs are a better alternative if you want high efficiency.

How are you planning to design the output filter and how good does the sine wave need to be? It seems nontrivial to generate a 20 kHz sine wave voltage using a 100 kHz switching frequency if the load current looks like that... but who doesn't like a challenge? 

If the load current is guaranteed to look like that, then maybe even a buck preregulator + class B amplifier could give quite good efficiency.

[GK]

How are you planning to design the output filter and how good does the sine wave need to be? It seems nontrivial to generate a 20 kHz sine wave voltage using a 100 kHz switching frequency if the load current looks like that... but who doesn't like a challenge? 

Yeah. I was contemplating some kind of resonant arrangement but overall it’s not looking particularly feasible/practical at the moment. Even with an unlimited switching frequency, that load is a real pooper.

[johansen]

, forget everything you understand concerning class-AB theory and star ground layout.

Really? So, for example, my ground return wire to the power supply can double as my signal ground with no ill effects??

eliminating noise with high inductance star grounding using cotton insulated pure silver wires isn't going to help when 100nH is enough to blow up your gate driver...

in any case there isn't much difference between a class D audio amp and an induction heater. its just a matter of one is a buck converter, the other drives a hf transformer.

[GK]

, forget everything you understand concerning class-AB theory and star ground layout.

Really? So, for example, my ground return wire to the power supply can double as my signal ground with no ill effects??

eliminating noise with high inductance star grounding using cotton insulated pure silver wires isn't going to help when 100nH is enough to blow up your gate driver...

 

Inductive loops in the power switching section need to be kept minimal for obvious reasons, but the routing of the actual analogue signals still needs to follow basic (hierarchical rather than the idealised “star”) grounding rules applicable to the "optimal" implementation of any other amplifier.

What is it with audio design that so reliably brings out the evangelists with their oversimplified and dogmatic statements of "fact"?

[nickm]

Infineon is usually the best for modern fets, especially high voltage (600V) ones.  But with only a 160V rail their optimos series would be your best bet. 

[GK]

OK, I'll keep that in mind for perhaps some other project in the future, but this particular one is canned now at the concept evaluation stage due to impracticality.


--------------------------------------------------------------------------------------------------------------------


Just out of interest, has anyone out there ever tried running some piece of test/industrial equipment with a conventional linear (transformer+bridge+capacitor filter) power supply from a "pure sine wave" 50/60Hz inverter, towards the inverters maximum power output?

Looking at the filter designs in some commercial examples of "pure sine wave" inverters I can't see how they would be too happy with such a (non linear) load.




 

[diyaudio]

, forget everything you understand concerning class-AB theory and star ground layout.

Really? So, for example, my ground return wire to the power supply can double as my signal ground with no ill effects??

eliminating noise with high inductance star grounding using cotton insulated pure silver wires isn't going to help when 100nH is enough to blow up your gate driver...

 

Inductive loops in the power switching section need to be kept minimal for obvious reasons, but the routing of the actual analogue signals still needs to follow basic (hierarchical rather than the idealised “star”) grounding rules applicable to the "optimal" implementation of any other amplifier.

What is it with audio design that so reliably brings out the evangelists with their oversimplified and dogmatic statements of "fact"?

evangelists ? oversimplified and dogmatic statements of "fact"??
dogmatic  -> arrogant assertion of unproven or unprovable principles

Really now? You ask for help then you start to insult really, mature.

Let me correct your foolish response for the last time , now that I understand what your level of understanding is regarding this subject and frustration cause you cannot seem to get anything working.

A class-d amplifier is no different than a switch mode power supply or high power switching converter, one of the main differences is the modulator is made to modulate to an incoming audio signal and then based on the complexity of the modulator perform quantization and drive a pair of FETs in their non linear region, they either on or off then demodulated with a suitable low permeability magnetics, if you understood this from the very beginning you would not even start to doubt the advice that was given to you. (This is not an audio evangelists or audio-fool point of view)

As for the star grounding a class-d OR SMPS (as we in the experienced area know) carries high frequency loops, high current loops and are subjected to many many design issues (some of the important ones we spoke about, "excluding you"), good class-d PCB designs are always done with a continuous ground plane 2 layers or more, signal loops as well as high current loops need to be kept in a confined area. This is not class-ab or your breadboard understanding of star ground loop.

I recommend you start with a simple 50-watt calss-d first before undergoing such ambitious projects.

[GK]

, forget everything you understand concerning class-AB theory and star ground layout.

Really? So, for example, my ground return wire to the power supply can double as my signal ground with no ill effects??

eliminating noise with high inductance star grounding using cotton insulated pure silver wires isn't going to help when 100nH is enough to blow up your gate driver...

 

Inductive loops in the power switching section need to be kept minimal for obvious reasons, but the routing of the actual analogue signals still needs to follow basic (hierarchical rather than the idealised “star”) grounding rules applicable to the "optimal" implementation of any other amplifier.

What is it with audio design that so reliably brings out the evangelists with their oversimplified and dogmatic statements of "fact"?

evangelists ? oversimplified and dogmatic statements of "fact"??
dogmatic  -> arrogant assertion of unproven or unprovable principles

Really now? You ask for help then you start to insult really, mature.

Let me correct your foolish response for the last time , now that I understand what your level of understanding is regarding this subject and frustration cause you cannot seem to get anything working.

A class-d amplifier is no different than a switch mode power supply or high power switching converter, one of the main differences is the modulator is made to modulate to an incoming audio signal and then based on the complexity of the modulator perform quantization and drive a pair of FETs in their non linear region, they either on or off then demodulated with a suitable low permeability magnetics, if you understood this from the very beginning you would not even start to doubt the advice that was given to you. (This is not an audio evangelists or audio-fool point of view)

As for the star grounding a class-d OR SMPS (as we in the experienced area know) carries high frequency loops, high current loops and are subjected to many many design issues (some of the important ones we spoke about, "excluding you"), good class-d PCB designs are always done with a continuous ground plane 2 layers or more, signal loops as well as high current loops need to be kept in a confined area. This is not class-ab or your breadboard understanding of star ground loop.

I recommend you start with a simple 50-watt calss-d first before undergoing such ambitious projects.

Here we go....... I currently have approx a dozen SMPS designs under my belt, from up to 2kV supplies of a few watts to a 500W 13.8V bench supply, and a few RF amplifiers including VHF radio using Lateral MOSFETs.

And I said absolutely nothing to contradict the fact that high frequency loops need to be kept minimal in any kind of high frequency switching design. I do take issue with the statement that absolutely nothing applicable to grounding/signal wiring in class AB design is applicable to class D design, and I have already party explained why. By making such a statement it is YOU who betrays a significant lack of understanding. And you didn't answer the question I asked of you either.

But anyway, you're the expert now that you've simulated a class D amp in LTspice.
   

[codeboy2k]

* codeboy2k pulls up a chair

[GK]

I should have rather entitled the thread "What modern MOSFETs for a power inverter".

[BravoV]

I should have rather entitled the thread "What modern MOSFETs for a power inverter".

You still can change the thread's title, just edit it at the first post.

[dannyf]

What is it with audio design that so reliably brings out the evangelists with their oversimplified and dogmatic statements of "fact"?

Religion. To them, physics doesn't exist. Only faith does.

At that point, it makes no sense to reason with them.

My strategy is not respond to any of them.

[diyaudio]

What is it with audio design that so reliably brings out the evangelists with their oversimplified and dogmatic statements of "fact"?

Religion. To them, physics doesn't exist. Only faith does.

At that point, it makes no sense to reason with them.

My strategy is not respond to any of them.

Funny how you didn't have a constructive answer after our last discussion, now you rallying support.  I anticipated this from you  

[dannyf]

But anyway, you're the expert now that you've simulated a class D amp in LTspice.

There was this famed "class D" audio amp designer in the 1990s (?) whose Class D amp would blow up its mosfets, only at very low load. He had no clue what was happening,

The amp received wide praises from users and online sites (6moon?) and was considered revolutionary.

You just cannot have a rationale discussion with those guys.