Switched to Class D or get left behind?

Indeed, around here my ASR apologetics are "hollering down a well". I've been "called" on that and will be silent for now, however I reserve the right to defend ASR in future.

I would rather have great chats with you about gear and such and what you like / don't like BASED ON YOUR OWN OPINION and such than have you here to defend ASR. This isn't ASR. This is Audio Shark.

If we want to know what ASR thinks about stuff we can go over there ourselves and trudge through that ourselves.

Let's talk about hifi audio and not ASR.

Deal?
 
Last edited:
What ASR fails to point out and their readers are mostly ignorant to, Class D can compress, highs become harsh and bass weaken all the while still measure low distortion. This is why folks you need your ears. And, why specs isn't the only thing to rely on.

A good amp regardless of Class needs a good power supply, high reserves of capacitance and good current delivery. The amp has to be able to deal with the reaction of the speaker load. It's just fact, no debate.

I've often talked about how a receiver can spec lower distortion than lets say McIntosh. I think most listeners would prefer the Mac. Well now we know receiver manufacturers learned if they apply tons of negative feedback they could get super low THD measurements. We also learned that too much negative feedback or not applied correctly can be degrading to sound quality.

If you are going to be the one listening to a system and you don't use your ears that's about as foolish as I can imagine.

They are all too young to remember (or may be audiophiles in the USA skipped?) the Japanese THD wars.

When I graduated, I went home with my "dorm room system". Mission 70 loudspeakers, Creek 4040 integrated amp and Rega Planar 2 turntable. It was the budget system at the time. By the time I started work and making a little money to upgrade my system, 0.001% THD was all the rage - and Japanese amp designers were racing to see who could achieve more zeros. I was sold by my dealer into a Kenwood with SIX zeros - 0.00001% THD. And regretted it for the next few years before I could afford to trade it out.

Unfortunately, the Class D designers over the past 20 years have been doing down the same road. Higher and higher feedback in order to achieve lower and lower distortion (and higher efficiency). That's why the Genesis Reference Amp was noted as "Class D reaches maturity" when I launched it in 2006. I used the original UcD400 module and modified it with lower feedback and used a linear power supply.
 
Yes
I would rather have great chats with you about gear and such and what you like / don't like BASED ON YOUR OWN OPINION and such than have you here to defend ASR. This isn't ASR. This is Audio Shark.

If we want to know what ASR thinks about stuff we can go over there ourselves and trudge through that ourselves.

Let's talk about hifi audio and not ASR.

Deal?
,
Yes, it's a deal since I've already decide to desist from ASR apologetics.
 
I was having a conversation with AI about amplifiers and sound. It zeroed in on what I gravitate to. Then I thought it would be fun to ask if there was a Class D amp that fits my profile. Here's a snip:

But honestly, based on your descriptions, I suspect you still may prefer:


  • high-bias AB
  • sliding bias designs
  • or very carefully voiced hybrids

because you appear unusually sensitive to:


  • depth gradation
  • harmonic sustain
  • image body
  • transient texture after the leading edge

Those are precisely the places where many Class D amps still sound somewhat simplified compared with excellent linear amplifiers.


One thing that’s fascinating psychoacoustically:
many Class D amps reproduce the attack of notes extremely well, but not always the bloom and decay trajectory afterward. The brain uses those trailing spatial/harmonic cues to perceive realism and dimensionality.


That can make the sound feel:


  • fast but emotionally flatter
  • clean but less immersive
  • organized but less alive

The fact you value both:


  • dark backgrounds
  • and dimensional depth

puts you in a narrower target zone than many audiophiles realize. A lot of designs excel at one while compromising the other.

Although AI correctly suggested I may not be a good fit for Class D it did recommend, Avic, Mola Mola, Jeff Rowland AGD or Marantz Model 10 as possible amps to listen to.
 
I was having a conversation with AI about amplifiers and sound. It zeroed in on what I gravitate to. Then I thought it would be fun to ask if there was a Class D amp that fits my profile. Here's a snip:

But honestly, based on your descriptions, I suspect you still may prefer:
  • high-bias AB
  • sliding bias designs
  • or very carefully voiced hybrids
because you appear unusually sensitive to:
  • depth gradation
  • harmonic sustain
  • image body
  • transient texture after the leading edge
Those are precisely the places where many Class D amps still sound somewhat simplified compared with excellent linear amplifiers.

One thing that’s fascinating psychoacoustically:
many Class D amps reproduce the attack of notes extremely well, but not always the bloom and decay trajectory afterward. The brain uses those trailing spatial/harmonic cues to perceive realism and dimensionality.

That can make the sound feel:
  • fast but emotionally flatter
  • clean but less immersive
  • organized but less alive
The fact you value both:
  • dark backgrounds
  • and dimensional depth
puts you in a narrower target zone than many audiophiles realize. A lot of designs excel at one while compromising the other.

Although AI correctly suggested I may not be a good fit for Class D it did recommend, Avic, Mola Mola, Jeff Rowland AGD or Marantz Model 10 as possible amps to listen to.

Evidence, I think, that if you ask AI a leading question it will tell you what you want to hear.
 
I see the same at church. The people who proclaim the loudest to have God in their heart are usually the ones with the least.

The ones who never talk about it are usually the most true.
“‘Tis better to be silent and thought a fool than to speak and remove all doubt.”

“If I claim to be a wise man, it surely means that I don’t know.”

I would think it obvious these days. The loudest, most strident voices are generally those with the least credibility.
 
Evidence, I think, that if you ask AI a leading question it will tell you what you want to hear.

Except I didn't ask a leading question. The conversation was about certain brands of amplifiers and what I prefer. Once it sort of had my profile or preference then I asked if there was a Class D that fit my preferred sound. Before that no mention of Class. AI is not perfect or always correct. BTW it did give a list of Class D amps that would come the closest, hence, the Mola Mola, AVIC, etc.
 
Except I didn't ask a leading question. The conversation was about certain brands of amplifiers and what I prefer. Once it sort of had my profile or preference then I asked if there was a Class D that fit my preferred sound. Before that no mention of Class. AI is not perfect or always correct. BTW it did give a list of Class D amps that would come the closest, hence, the Mola Mola, AVIC, etc.
Ask yourself where AI has gained the knowledge to come up with those answers. It surely must rely on human opinions by way of (for example) reviews, comments on forums, etc.

AI has no ears, has no labs to test equipment so you must surely take little notice of what it says - better still. don't even bother asking this sort of question. Leave it to answer easy questions where it needs simple stats only.
 
Except I didn't ask a leading question. The conversation was about certain brands of amplifiers and what I prefer. Once it sort of had my profile or preference then I asked if there was a Class D that fit my preferred sound. Before that no mention of Class. AI is not perfect or always correct. BTW it did give a list of Class D amps that would come the closest, hence, the Mola Mola, AVIC, etc.

Well we can't be totally sure because you didn't report with told/asked the AI.

I'm not against posting an AI response, but I think if one wants to do so, one ought o post the question(s) asked and also name the AI engine consulted.
 
Well we can't be totally sure because you didn't report with told/asked the AI.

I'm not against posting an AI response, but I think if one wants to do so, one ought o post the question(s) asked and also name the AI engine consulted.
Yes. There was a posting here (or maybe another high-end forum) where someone asked whether one obscure amp sold in minute numbers would be an ideal match for an equally obscure speaker. Although AI could refer to specs of these obscure items, the only way it could possibly be able to answer the question is to refer to other users who just happen to have these 2 obscure items and have described their experiences. How could AI's answer be in the least helpful as the number of users of these 2 items in combination may be (let's be generous) 5, and of these 5, how many would describe their experience - perhaps 1 or 2, but likely nil - and how useful would these inevitably biassed opinions be?

Before asking any question of AI, one must ask yourself whether AI can possibly have useful information to provide a meaningful answer. In the above example (and probably the one being discussed here) the question should never have been asked as the answer would be likely (as someone observed) one to please the questioner!
 
Last edited:
Ask AI a question and get a long but perhaps comprehensive answer.

I asked ChatGPT the following question:

Given that the latest class D amplifiers measure near perfection in sonic performance, are there measurable performance aspects of class A or AB amplifiers that might account for the subjective preference of the latter classes by many audiophiles?


ChatGPT's response:

Yes — there are several measurable differences between even excellent modern Class D amplifiers and many Class A/AB amplifiers that could plausibly contribute to subjective preferences, even when conventional metrics like THD+N and frequency response appear essentially perfect.

The key point is that “excellent conventional measurements” may not fully characterize all aspects of amplifier/speaker interaction or human perception.

Some measurable candidates include:

1. Distortion spectrum and harmonic structure

Two amplifiers may both have vanishingly low THD, yet produce different distributions of distortion components.

For example:
  • Some Class AB or Class A amplifiers produce predominantly:
    • 2nd harmonic
    • low-order even harmonics
  • Some Class D amplifiers may exhibit:
    • higher-order residuals
    • more complex ultrasonic distortion products
    • noise-shaped artifacts
Even if these are extremely low in amplitude, some listeners may be sensitive to:
  • harmonic distribution
  • intermodulation character
  • time-varying distortion behavior
Many audiophiles subjectively prefer low-order even harmonics because they can slightly reinforce tonal richness or spatial impression.

2. Distortion vs frequency and load

A major historical weakness of Class D has been load dependency.

Modern Purifi and Hypex designs have dramatically reduced this, but measurable differences can still exist in:
  • distortion under reactive loads
  • phase shift into complex impedances
  • HF linearity with real loudspeakers
A resistor dummy load does not fully simulate:
  • crossover back-EMF
  • reactive phase angles
  • dynamic current demands
Some Class AB amplifiers maintain extremely linear behavior into awkward real-world speaker loads.

3. Ultrasonic noise and filtering behavior

Class D amplifiers inherently switch at high frequencies (typically hundreds of kHz).

Even when far beyond hearing range, residual ultrasonic energy can:
  • interact with tweeters
  • intermodulate in downstream electronics
  • alter behavior under complex loads
Measurable aspects include:
  • residual carrier leakage
  • output filter response
  • ultrasonic IMD products
Some researchers suspect that very small ultrasonic artifacts may subtly affect perceived treble texture or spatial cues.

This remains debated.

4. Feedback topology and transient behavior

Modern Class D often relies on:
  • extremely high feedback
  • sophisticated error correction
  • switching reconstruction filters
Some audiophiles argue that:
  • recovery from transients
  • overload behavior
  • dynamic settling
can sound different even when steady-state measurements are excellent.

Possible measurable areas:
  • transient intermodulation distortion (TIM)
  • multitone IMD
  • settling behavior
  • overload recovery
Certain Class A/AB designs may clip or recover more gradually.

5. Output impedance and damping behavior

Class D amplifiers often achieve:
  • extremely high damping factor
  • ultra-low output impedance
Counterintuitively, some listeners prefer the slightly higher output impedance of some Class A/AB designs because it can:
  • subtly alter bass behavior
  • slightly reshape speaker response
  • affect crossover interaction
This is measurable.

A speaker’s frequency response can vary slightly depending on amplifier output impedance.


6. Noise spectrum differences

Two amplifiers may measure equally quiet in broadband noise terms, yet have different spectral distributions.

For example:
  • Class D may exhibit ultrasonic noise shaping
  • linear amplifiers may have smoother wideband noise spectra
Human hearing is extraordinarily sensitive to:
  • low-level modulation
  • correlated noise
  • temporal masking effects
Conventional SINAD numbers may not fully capture perceptual significance.

7. Crossover behavior in Class AB

Ironically, one measurable imperfection of Class AB may contribute to listener preference.

A high-bias Class AB amplifier often has:
  • tiny residual crossover artifacts
  • low-order nonlinearities
These may create a subtly euphonic presentation.

This parallels why some listeners enjoy:
  • vinyl coloration
  • tube harmonics
  • transformer saturation
Absolute accuracy and subjective preference are not always identical.

8. Power supply behavior

Many classic high-end Class AB amplifiers use:
  • huge linear transformers
  • massive capacitor banks
  • enormous current reserves
This can produce measurable differences in:
  • burst power delivery
  • rail sag characteristics
  • dynamic headroom
  • clipping softness
Some listeners perceive this as:\
  • “weight”
  • “slam”
  • “ease”
Modern switching supplies can equal or exceed this technically, but implementation matters.

9. Measurement limitations themselves

This is perhaps the most important issue.

Common published measurements emphasize:
  • THD+N
  • SINAD
  • 1 kHz test tones
  • steady-state behavior
But human hearing may be more sensitive to:
  • multitone complexity
  • time-domain behavior
  • modulation effects
  • speaker interaction
  • spatial microcues
Research continues into whether current metrics fully correlate with perception.

Important perspective

The measurable differences between top modern Class D and excellent Class A/AB amplifiers are usually:
  • extremely small
  • often near or below established audibility thresholds
In controlled blind tests:
  • many claimed differences disappear
  • level matching becomes critical
  • expectation bias can dominate
But that does not necessarily mean all preferences are imaginary.

It may mean:
  1. Some subtle measurable factors are underappreciated.
  2. Some listeners are unusually sensitive to certain artifacts.
  3. Existing measurements do not perfectly map to perception.
  4. Some preferences are for pleasant coloration rather than strict accuracy.

Current consensus among many engineers

Designs from Purifi and advanced Hypex NCore modules are generally considered to have pushed Class D performance beyond the point where topology itself is likely the dominant audible factor.

At that level:
  • speaker quality
  • room acoustics
  • mastering
  • DSP/room correction
typically matter far more than amplifier class.
Frankly ChatGPT's response tends to favor the arguments of the "Golden Ears" more than hardcore objectivits.
 
I was considering giving Class-D another try so I looked at Mola Mola Ossetra. That was until I saw the MSRP of $25,100. View attachment 35713


Exactly, because quality power supplies and other factors to make a quality amp costs. What's happening some companies are putting out these flea weight Class D that measure low distortion beyond measurement with inflated power ratings but can't stand up under real world use because they don't have the power supply and current delivery. On the other hand I'd hope that could be achieved under $25k, LOL

The truth is Class D can sound like total crap and still measure well. That's the snake oil of Class D. What I mean by "crap" power rails sag, highs become harsh, sound compresses, due to not being able to produce the power and current needed, this is a fact. Some people like Here Here who has ultra high efficient speakers may never have taxed his Class D so may have a skewed opinion. Class D may work for some people who can keep them in a good working range or not able to hear when they fall apart. This doesn't have to mean human hearing it could also be inferior speakers.
 
Back
Top