Class D audio amplifiers with Hi End quality

[Picuino]

I recently read an article about a new Class D audio amplifier module with excellent harmonic distortion and noise performance. I’ve always associated Class D with low quality.
What is the current state of the art? Can I consider building small amplifiers using inexpensive, high-quality Class D chips? Or is this limited to very expensive equipment?

[jonpaul]

Easily available for mnay years, from NL.

https://www.hypex.nl/

[tom66]

It's been the case for at least 15-20 years that class D can produce THD+N below that of human perception.  Will that be lower than an analog amplifier?  Probably not, but I challenge anyone to distinguish the difference between 0.01% and 0.001% THD+N.  Most audio scientists don't think people can distinguish beyond 0.01% THD+N, especially given even a premium loudspeaker will introduce more distortion anyway.

[BigBoss]

I have never understood this phenomenon in audiophiles.

THD is not a figure of merit stand-alone, OK It should be as low as possible
Noise is one of the key parameters; that's OK too.
They say, "Damping factor shouldn't be neglected/omitted," alright...

Why do they pay 10K or even more than 25K for tube amplifiers?? They are heavy, bulky, power-eager, and have very low efficiency.
They also exhibit very high distortion and have a leakage of microphonic noise. They can supply 60-100 W max. that is so low level to drive high-quality loudspeakers.
And tubes exhibit emission failures in time, and they have to be replaced with closer mates and this is not easy. I heard once a tube amplifier manufacturer owner has been selecting appropriate
candidates in a lot of tubes.
So, there are many bottlenecks in tube amplifiers, but they still prefer them.
Strange and unusual.

[tom66]

People choose tube amps not for low distortion, but specifically to get higher distortion due to the nonlinearity of the tubes.  It is possible to replicate this distortion in DSP to a point where the difference between DSP and real tube is inaudible, but... why do people like riding an old motorbike or driving a classic Jaguar over a modern car?  Because they like to know that there is something old there, even if the modern car is probably better at the A to B and even the performance in most respects are better. 

I put it another way, if fools want to spend their money on high end audiofoolery, then let them do so; if nothing else, they keep engineers in business developing these solutions.

[Randy222]

It's been the case for at least 15-20 years that class D can produce THD+N below that of human perception.  Will that be lower than an analog amplifier?  Probably not, but I challenge anyone to distinguish the difference between 0.01% and 0.001% THD+N.  Most audio scientists don't think people can distinguish beyond 0.01% THD+N, especially given even a premium loudspeaker will introduce more distortion anyway.

+1
Monster makes speaker cables that produce better sound, although std zip cord is 99.99999% of the Monster cables. 

The argument is about the same as 4k vs 8k vs 12k TV panels. Std HD is very good for most eyes, 4k better to some/many, 8k 12k likely meaningless. And for TV's, if the content doesn't match the TV native canvas, then you are not seeing things as they should be displayed. 4k content is very little, 8k 12k content is just about non-existent.

[oPossum]

Easily available for mnay years, from NL.

https://www.hypex.nl/

Also  https://purifi-audio.com/

Both designs by Bruno Putzeys

https://en.wikipedia.org/wiki/Bruno_Putzeys

[paul cotter]

I repaired a Peavey class d years ago(can't remember the model number, it was a slimline 1 unit high job) and it sounded pretty good to me, though I would describe myself as tone-deaf . I always thought the Carver amps had a great sound.

[themadhippy]

I’ve always associated Class D with low quality.

Try an  ice module ,  https://icepoweraudio.com/  spent several years using a pa with them inside and they didnt sound low quality.

They can supply 60-100 W max. that is so low level to drive high-quality loudspeakers.

someone forgot to inform macintosh that there  350W amp  aint possible,and whilst there  also let ampeg know that 400W custom bass head  dont exist

[TimFox]

For amplifiers with "well-behaved" distortion, i.e. the non-linearity can be modeled accurately with a polynomial that deviates from a straight line, THD measurement is a useful way to estimate the parameters.
For actual listening, the little bit of extra harmonic content from a violin note (for example) is not important, compared with the inherent harmonic content of the sound from the vibrating string.
Real music is more than single notes, and the intermodulation distortion from unwanted combinations of tones/notes is more important, for the same reason that 3rd-order IM is important in RF communication systems.
The practical advantage of THD measurement is that is well-defined, while IMD measurement protocols vary substantially.
The traditional THD measurement, where a filter rejects the fundamental frequency at the device output and measures everything else, works for analog amplifiers, but is problematic for Class-D, since "everything else" contains a large amount of switching-frequency harmonic content, which should not be important acoustically, but really fouls up the measurement without careful extra filtratioon.  Measuring the harmonics on a good spectrum analyzer gives better results.

[tom66]

The traditional THD measurement, where a filter rejects the fundamental frequency at the device output and measures everything else, works for analog amplifiers, but is problematic for Class-D, since "everything else" contains a large amount of switching-frequency harmonic content, which should not be important acoustically, but really fouls up the measurement without careful extra filtratioon.  Measuring the harmonics on a good spectrum analyzer gives better results.

The vast majority of the switching harmonics are above the audible range.  They get notched out pretty heavily by the output LC filter, and in any case don't produce any audible components since they're well above both the operating range of the driver and the range of human hearing.

You will have some sub harmonics if there's jitter on the PWM but I'd argue any half decent class D amplifier has the jitter below the audible noise floor.  That wouldn't even be particularly onerous.  If your PWM is at 500 kHz (pretty typical), then 0.01% noise from jitter would be ca. 200ps peak to peak jitter.  Or, well within the range of a normal crystal or RC oscillator (on the die).

[TimFox]

The traditional THD measurement, where a filter rejects the fundamental frequency at the device output and measures everything else, works for analog amplifiers, but is problematic for Class-D, since "everything else" contains a large amount of switching-frequency harmonic content, which should not be important acoustically, but really fouls up the measurement without careful extra filtratioon.  Measuring the harmonics on a good spectrum analyzer gives better results.

The vast majority of the switching harmonics are above the audible range.  They get notched out pretty heavily by the output LC filter, and in any case don't produce any audible components since they're well above both the operating range of the driver and the range of human hearing.

You will have some sub harmonics if there's jitter on the PWM but I'd argue any half decent class D amplifier has the jitter below the audible noise floor.  That wouldn't even be particularly onerous.  If your PWM is at 500 kHz (pretty typical), then 0.01% noise from jitter would be ca. 200ps peak to peak jitter.  Or, well within the range of a normal crystal or RC oscillator (on the die).

Yes, the switching harmonics should be inaudible.  However, they interfere with the traditional THD measurement.  I had to build an external L-C filter (with air-core chokes and NP0 capacitors to avoid non-linearity) to reduce them for a measurement with an -hp- 339A analyzer, even with its built-in (switchable) low-pass filters, since the built-in output filter did not suffice for measurement purposes.

[Alex Eisenhut]

Yes another case of mature technology. I have a few classic solid-state class AB amps and also a small Aliexpress Class D thing that looks like it will fall apart. Sounds fine.

But the look of the equipment is also important (to me) to enjoy music. 

[TimFox]

Yes another case of mature technology. I have a few classic solid-state class AB amps and also a small Aliexpress Class D thing that looks like it will fall apart. Sounds fine.

But the look of the equipment is also important (to me) to enjoy music. 

I remember a "PDQ Bach" concert by the late Peter Schickele, where he included a concerto for lute and bagpipes.
He commented that the lute was such a quiet instrument that if there were a bagpipes in the same room, being played or not, you could not hear the lute.
However, he pointed out that the lute was a beautifully crafted instrument, lovely to look at.
Something we miss in this age of recorded music.
"Remember that, while listening to the bagpipes."

[Simon]

I once made a class A amp with a single BJT, sounded great to me, ok I ran it for under a minute before getting bored and it was probably terrible.

Class D is PWM, I suspect that replicating timing is as easy as replicating voltage if not easier, the rest is physics, filters followed by the greatest filter of all, our ears. I listened to some distorted audio once and then viewed it in audacity. I was truly astonished at how bad it looked compared to how it sounded. It was like it had no more than 4 bits of range and was heavily clipped, yet it was perfectly understandable if a bit crap.

human phycology always deems the old better than the new, how many people do you hear say the past was better compared to people saying the past was worse than how we have it now? that old tech was better? people love V8 engines over electric propulsion simply because of the sound and lets face it the memories of when they had a V8 engine in their car. The circumstances at the time become mixed and intertwined with the technology to the point that it was down to the particular technology they reminisce about as though they would not have been enjoying themselves like that if it had been with the new tech we have now.

[paul cotter]

Personally I try not to listen to bagpipes!

[Andy Chee]

I recently read an article about a new Class D audio amplifier module with excellent harmonic distortion and noise performance. I’ve always associated Class D with low quality.
What is the current state of the art? Can I consider building small amplifiers using inexpensive, high-quality Class D chips? Or is this limited to very expensive equipment?

Decades ago when MOSFET/transistor switching speeds were slow, you'd be right.

But now we have high speed devices, and that's without considering exotic devices like GaN and SiC.

Similar to power supplies, switchmode has made technological progress into domains once dominated by linear topologies  (arguably motivated by energy efficiency).

[Picuino]

This is the module that I have read about:
https://shop.icepoweraudio.com/product/icepower-sc400a2-new/

[SiliconWizard]

I recently read an article about a new Class D audio amplifier module with excellent harmonic distortion and noise performance. I’ve always associated Class D with low quality.
What is the current state of the art? Can I consider building small amplifiers using inexpensive, high-quality Class D chips? Or is this limited to very expensive equipment?

Apart from what is now almost "niche" applications in audio (some HiFi or instrument amplifiers), most amplifiers you'll see these days are actually class D, from your mobile devices to multi-kW stage amplifiers.

Low THD is not a problem with "modern" power MOSFETs and proper sigma-delta converters. The tricky part is mostly about EMI, and output filters require large inductors with proper specs that can also be pretty expensive depending on the max output power. There's a small TI app note as a starting point: https://www.ti.com/lit/pdf/ssztbo2

The typical achievable THD+N for a good implementation is around 0.01% over most of the audio spectrum, which is certainly not as good as what we can get with a class A amplifier, but apart from low-power applications, class A amps are not really practical. Class A power amps are bulky, draw a lot of current and heat up like crazy. The more common class AB can be better than a good class D amplifier in terms of THD+N, while not as good as class A, but as others have said, except if you need a power amp for measurement purposes, your own ears are unlikely to tell the difference. Below about 0.1% THD+N, it becomes hard to tell just by listening. Also, however low the THD figure is, once it's connected to actual speakers, you'll rarely get anything lower than this - 0.01% - and that is with high-quality speakers.

As to the module you mentioned, they claim "0.0002% THD+N at 100W", which I highly doubt can be true, in general for a class D amp, but all the more given its apparently very compact form factor.
I'd be curious to know what kind of IC they used and see real measurements.

[paul cotter]

I was hoping for a schematic in the link Picuino kindly provided.

[themadhippy]

Personally I try not to listen to bagpipes!

there fine in small doses,especially when a rock band is trying to drown em out

The more common class AB can be better than a good class D amplifier in terms of THD+N, while not as good as class A, but as others have said, except if you need a power amp for measurement purposes,

or yer having to lift a flight case full of the buggers in an out of the truck

[Gyro]

The first Class D - well Class T, amplifier IC that I designed into a piece of consumer A/V gear was a Tripath one, back in the late '90s when they were a start-up. I don't remember the part number but we got away with minimal output filtering. People used to get curious about the meaning of the 'tripath' bit, but the true reason was that the company was founded by a very nice gentleman named  Mr Dr Tripathi. 

[David Hess]

Yes another case of mature technology. I have a few classic solid-state class AB amps and also a small Aliexpress Class D thing that looks like it will fall apart. Sounds fine.

I have a couple of Fosi amplifiers which use the Texas Instruments TPA3255, which is about 0.01% up to 100 watts.  My speakers and especially my subwoofer would cost at least ten times the cost of the amplifier.

I wish there was a matching parametric equalizer.

[thm_w]

Audiosciencereview is where to go: https://www.audiosciencereview.com/forum/index.php?threads/topping-mini-300-amplifier-review.68581/

You can see from his measurements plenty of cheap $100 Class D chinese amplifiers will beat $6000+ commercial amps.
Anything over 70-80dB SINAD is plenty, ~0.03% THD.