Class D "digital" amps. Opinions on performance

[fkh006]

I love class D amps....Markbass Little Mark II and my
GK MB112 combo.

Little Mark II is an A/B amp with switch mode power supply as I believe.

 

[will33]

Following along here......learning by osmosis.

In a smps/class D design, is our signal not deconstructed and then reconstructed along the way? There's more going on than just remaining intact, although passed through many components, manipulated, gained up, etc....is it true that it's actually taken apart and put back together again? I could easily be confused there and certainly don't mind being corrected.

 

[dave_p]

The bottom line is that it's all in the signal the amp delivers to the loudspeaker. For an amp to change the sound, it has to change the signal. If two amps put out exactly identical signals into the same loudspeaker, they will sound identical. There isn't any sidechannel telling the loudspeaker to respond differently because the signal's coming from a certain brand or amp class or through BJTs or MOSFETs or glowFETs.

The more accurate the amp, the less it changes the signal, apart from multiplying it. The better designed the amp, the more accurate it probably has been made to be. So it's safe to predict that when you compare well-designed and well-made amps, it's highly possible that they're going to perform so closely to ideal that there are no audible differences. In fact, the biggest weakness in many casual comparisons (even unintendedly casual ones) is the usual assumption that they will unquestionably be audibly different.

Bob carver proved this. The carver challenge.

 

[fdeck]

Following along here......learning by osmosis.

In a smps/class D design, is our signal not deconstructed and then reconstructed along the way? There's more going on than just remaining intact, although passed through many components, manipulated, gained up, etc....is it true that it's actually taken apart and put back together again? I could easily be confused there and certainly don't mind being corrected.

No, but it's a perfectly valid question. The typical Class-D power amp is an analog circuit. The difference is that the control of the output signal is being done by high speed switching or modulation, but the switching frequency is way above the audio band (hundreds of kHz up into the MHz). Unfortunately it adds a conceptual layer of complexity that's hard even for experienced electronics guys to understand without studying it specifically.

 

[Rockman]

No, but it's a perfectly valid question. The typical Class-D power amp is an analog circuit. The difference is that the control of the output signal is being done by high speed switching or modulation, but the switching frequency is way above the audio band (hundreds of kHz up into the MHz). Unfortunately it adds a conceptual layer of complexity that's hard even for experienced electronics guys to understand without studying it specifically.

Yup. All analog. I designed a class D amp for my senior project in undergrad. It worked, the key is have the super high frequency "switching" in order to get really good granularity on your pulse width modulated signal. This is where the majority of the advancement has come in these amps, really really high frequencies. Also high frequencies are more efficient...but i digress.

 

[will33]

No, but it's a perfectly valid question. The typical Class-D power amp is an analog circuit. The difference is that the control of the output signal is being done by high speed switching or modulation, but the switching frequency is way above the audio band (hundreds of kHz up into the MHz). Unfortunately it adds a conceptual layer of complexity that's hard even for experienced electronics guys to understand without studying it specifically.

Thank you. After I posted that, I wasn't sure of I remembered the right things. Thought that the musical instrument signal may well stay together and was a power supply thing or whatever that was being "switched" way above the audio band/human hearing.

What is meant by "control" of the output signal?

Thanks,

-Will

 

[will33]

Yup. All analog. I designed a class D amp for my senior project in undergrad. It worked, the key is have the super high frequency "switching" in order to get really good granularity on your pulse width modulated signal. This is where the majority of the advancement has come in these amps, really really high frequencies. Also high frequencies are more efficient...but i digress.

Ya know....when you say stuff like "granularity", somebody's gonna come along and say these things sound "grainy" compared to good ol' big iron.....or output tubes.

 

[seamonkey]

Yup. All analog. I designed a class D amp for my senior project in undergrad. It worked, the key is have the super high frequency "switching" in order to get really good granularity on your pulse width modulated signal. This is where the majority of the advancement has come in these amps, really really high frequencies. Also high frequencies are more efficient...but i digress.

Just a nitpick but it's not too high. LF or MF at most. That's why simple analog test gear gets confused. It just needs to be set to human hearing ranges to correctly measure.

It's great to se a new generation of engineers comfortable with switching technology. It's here to stay!

 

[fdeck]

Thank you. After I posted that, I wasn't sure of I remembered the right things. Thought that the musical instrument signal may well stay together and was a power supply thing or whatever that was being "switched" way above the audio band/human hearing.

What is meant by "control" of the output signal?

Thanks,

-Will

Ah, that's a good question too. I just meant that the output transistors or tubes are pushing the output voltage up or down. But there is actually a bit more to that. The level of control achieved by most output circuits is poor, and so the output is governed by a circuit that compares the output signal to the input signal and "tries" to make them always proportional to one another. That's the concept of negative feedback -- a topic in and of itself.

You can see it in classic tube amp circuits such as the SVT, where there is a signal path through a resistor going all the way from the output terminal back to the cathode of an input tube. This is called "cathode feedback." Solid state amps typically employ a different feedback circuit made from a matched pair of transistors, but you'll still find that resistor sneaking from the output back to somewhere close to the input.

Feedback is like the finishing touch on the circuit -- which irons out irregularities such as nonlinearity and non-flat frequency response.

 

[Munjibunga]

I've talked to a lot of soundmen about them, and the majority seem to agree that they do work well and make roadwork a lot easier, but I hear the same thing out of a lot of them that I hear out of many bass players...if you're used to a certain amount of wattage in a lead sled, you will want to overbuy class D wattage to have that level of headroom. I've also heard from a handful of holdouts that they don't like the sound of class D power amps. With all due respect to Bob and the other folks who build power amps, these guys hear differences enough to hold onto their lead sleds and put up with the schlep. Granted, they're not doing arenas usually, and that would likely change their minds, but I think there's enough testimony out there from soundmen to believe there's something to it. Not a single one complains about neos, though

The sound at arena shows typically sucks so bad, it wouldn't matter what kind of power amps are used. I know I really loved the sound of my EAW mains and subs driven with QSC PLX amplifiers. I did make sure to have plenty of headroom - 1.5 to 2 times rated cab power, as recommended by EAW. Nowadays, I'm using JBL PRX600 powered speakers. Love those, too.

 

[David Jayne]

The sound at arena shows typically sucks so bad, it wouldn't matter what kind of power amps are used.

Agreed. No more for me thanks, club shows from now on.

 

[BbbyBld]

What's with that? Help a junior techie understand this. I'd think if the amp reaches 100% duty cycle, it's the same as temporarily passing DC through the MOSFETs, and then it's just up to the current rating of the devices. So, I'm trying to think of what else could go wrong: Perhaps the feedback circuit latches up, or the MOSFETs take an inductive discharge?

As the amp approaches 0 or 100% duty cycle the pulse width becomes so narrow that the devices are no longer switching on and off--they will be operating in the linear region. This is partially due to parasitic capacitance in the output devices, propagation delay in the control logic gates, and dead time delay circuitry. The devices will cross conduct and will often literally blow apart.

So, I think clipping isn't it. The technical root cause for a "signature sound" of Class-D, if there even is one, still eludes us.

Well...now that you mention it...One of the reasons for a characteristic clipping sound in class D amps is because most of them these days are the self-oscillating topology. The switching frequency decreases as the output current increases. This is advantageous because the efficiency increases when the switching frequency drops and the EMI decreases because of a spread-spectrum carrier, but the clipping does not sound pleasing at all--actually it sounds terrible. I didn't want to post that before because I know someone will use it as an argument against class D amps, but there are ways around it. It's also possible for the carrier to drop to the point where it's barely in the audio spectrum NEAR clipping, and I would say that it has a characteristic sound, but it can be masked.

 

[fdeck]

As the amp approaches 0 or 100% duty cycle the pulse width becomes so narrow that the devices are no longer switching on and off--they will be operating in the linear region. This is partially due to parasitic capacitance in the output devices, propagation delay in the control logic gates, and dead time delay circuitry. The devices will cross conduct and will often literally blow apart.



Well...now that you mention it...One of the reasons for a characteristic clipping sound in class D amps is because most of them these days are the self-oscillating topology. The switching frequency decreases as the output current increases. This is advantageous because the efficiency increases when the switching frequency drops and the EMI decreases because of a spread-spectrum carrier, but the clipping does not sound pleasing at all--actually it sounds terrible. I didn't want to post that before because I know someone will use it as an argument against class D amps, but there are ways around it. It's also possible for the carrier to drop to the point where it's barely in the audio spectrum NEAR clipping, and I would say that it has a characteristic sound, but it can be masked.

Thanks for filling me in! I knew there must have been a reason, but couldn't think of one.

 

[agedhorse]

Well...now that you mention it...One of the reasons for a characteristic clipping sound in class D amps is because most of them these days are the self-oscillating topology. The switching frequency decreases as the output current increases. This is advantageous because the efficiency increases when the switching frequency drops and the EMI decreases because of a spread-spectrum carrier, but the clipping does not sound pleasing at all--actually it sounds terrible. I didn't want to post that before because I know someone will use it as an argument against class D amps, but there are ways around it. It's also possible for the carrier to drop to the point where it's barely in the audio spectrum NEAR clipping, and I would say that it has a characteristic sound, but it can be masked.

Agreed, in well executed designs, these issues do not come into play because they are designed out of the equation. This is one aspect that defines a good versus poor design.

 

[seamonkey]

The focus switches again to the extreme use cases, and the dumb design that doesn't take these extremes into account. Most of the solutions are built in to the off the shelf driver chips available for class-d. It never hits some "signature sound". Every TV, Cell phone, laptop, ... 75% of the amps - signature sound - meh

 

[Bob Lee (QSC)]

True. That precision is in the spec of the components and built into the tolerance of the unit. It's also taken into account in the final product and sometimes used as a guide to know where to measure the unit for the best specs.

Difference in perception of identical amplifiers set to different gain is more a matter of how human hearing varies at different SPL than a matter of component specs.

 

[fdeck]

The focus switches again to the extreme use cases, and the dumb design that doesn't take these extremes into account. Most of the solutions are built in to the off the shelf driver chips available for class-d. It never hits some "signature sound". Every TV, Cell phone, laptop, ... 75% of the amps - signature sound - meh

Sorry about that. My curiosity got the best of me, and the thread was winding down anyway.