Hypex nCore vs Class A amps

[RayDunzl]

Experiment related to Post #485 above:

A transient source - tapping a plastic pen on a plastic cutting board a couple of inches from the mic

Recorded at 24/48 with UMIK-1

Played back and recorded

Upsampled to 384k to better reveal the waves in the sample points

4ms of data

Conclusion:

(no DRC in use)

No voltage smear at amplifier - reproduction shows no visible error in this view

Some loss of very high frequency in the antique Martin Logan but the wave is followed well

Better HF in the JBL but some loss of the overall waveform - mic is below the axis of the speakers - 10 feet away but about 24" low

 

[boXem]

From https://www.pnas.org/content/97/22/11773 :

"The anatomical and biophysical specializations of octopus cells allow them to detect the coincident firing of groups of auditory nerve fibers and to convey the precise timing of that coincidence to their targets. Octopus cells occupy a sharply defined region of the most caudal and dorsal part of the mammalian ventral cochlear nucleus. The dendrites of octopus cells cross the bundle of auditory nerve fibers just proximal to where the fibers leave the ventral and enter the dorsal cochlear nucleus, each octopus cell spanning about one-third of the tonotopic array. Octopus cells are excited by auditory nerve fibers through the activation of rapid, calcium-permeable, α-amino-3-hydroxy-5-methyl-4-isoxazole-propionate receptors. Synaptic responses are shaped by the unusual biophysical characteristics of octopus cells. Octopus cells have very low input resistances (about 7 MΩ), and short time constants (about 200 μsec) as a consequence of the activation at rest of a hyperpolarization-activated mixed-cation conductance and a low-threshold, depolarization-activated potassium conductance. The low input resistance causes rapid synaptic currents to generate rapid and small synaptic potentials. Summation of small synaptic potentials from many fibers is required to bring an octopus cell to threshold. Not only does the low input resistance make individual excitatory postsynaptic potentials brief so that they must be generated within 1 msec to sum but also the voltage-sensitive conductances of octopus cells prevent firing if the activation of auditory nerve inputs is not sufficiently synchronous and depolarization is not sufficiently rapid. In vivo in cats, octopus cells can fire rapidly and respond with exceptionally well-timed action potentials to periodic, broadband sounds such as clicks. Thus both the anatomical specializations and the biophysical specializations make octopus cells detectors of the coincident firing of their auditory nerve fiber inputs."

What does it mean? If onset of a transient is blurred over 1 ms or longer time interval, the transient may not be perceived the same way. Instead of startling or energizing, it could be boring, or simply perceptually dissipate within the rest of the music.

Compared to a good Class A, a good Class D amp isn't likely going to be more than 20-40 microseconds behind in swinging from zero to the full amplitude. By itself, this doesn't seem critical. Just 2% to 4% of the critical transient detection window width. Consider this a small timing distortion.

Yet in combination with certain loudspeakers, containing "heavy" crossovers and transducers that further blur the transient, such difference may sometimes preclude the transient detection mechanism from reacting, or at least make the transient subjectively not as loud.

If the high-end speakers designers were tuning their creations using "nearly-perfect" classic monoblocks, they may be stopping at a point when they can no longer detect the transient anomalies. Likewise, mixing and mastering engineers, listening to their nearly perfect studio systems, would make the transients sound right too.

Yet a less-perfect amplifier could theoretically push the overall system to a condition where the transient anomalies are heard again. Corollaries:

(A) A system with much less than perfect loudspeakers would exhibit such anomalies no matter what, and the difference between amplifiers in this regard may not be noticeable at all. Excessive DSP may lead to such situation as well.

(B) A system with "super-perfect" loudspeakers, and/or competent DSP, would in effect "tolerate" timing idiosyncrasies of different amplifiers, and thus the timing differences between such amplifiers once again would not be noticeable.

(C) Only when the loudspeakers, amplifiers, DSP, and the music are "off" just enough, the substitutions of such components could lead to noticeable subjective differences in transients. These differences can be further exacerbated by certain rooms acoustics.

Could you please clarify how you come to the 20-40 us difference between class A and D?

 

[Sergei]

Could you please clarify how you come to the 20-40 us difference between class A and D?

For instance,
https://www.hypex.nl/img/upload/doc/an_wp/AN_UcD_and_IMD_DIM_slewrate.pdf
starts discussing it on page 4.

As I understood, the lower slew rate of Class D amp is caused by its output filter. Class A and A/B amps don't need one, and you can find examples of 10+ times overkill in their slew rate - not necessary for faithful reproduction of sinusoids within audible range at a given swing voltage, yet arguably helpful for reproducing short pulses more accurately.

Subjectively, such super-high-slew-rate amps tend to exhibit pronounced "rock and roll punch", also attributed to otherwise poorly measuring, yet ever-popular Yamaha NS10M studio monitor. Fortunately, serious people researched how NS10M's superior transient-handling characteristics helped its success (it is a fun read, including discussion of yet another creative use of toilet paper):

https://dt7v1i9vyp3mf.cloudfront.net/assetlibrary/n/ns10m.pdf.

 

[Julf]

For instance,
https://www.hypex.nl/img/upload/doc/an_wp/AN_UcD_and_IMD_DIM_slewrate.pdf
starts discussing it on page 4.

As I understood, the lower slew rate of Class D amp is caused by its output filter. Class A and A/B amps don't need one, and you can find examples of 10+ times overkill in their slew rate - not necessary for faithful reproduction of sinusoids within audible range at a given swing voltage, yet arguably helpful for reproducing short pulses more accurately.

I think we are mixing two very different issues here. The hypex paper discusses slew rate in the context of slew-rate-induced distortion - mostly an issue with class A and A/B. Simple linear bandwidth limiting of course limits simplistic dV/dt "slew rate", but only affects frequencies above the filter cutoff.

Subjectively, such super-high-slew-rate amps tend to exhibit pronounced "rock and roll punch"

Any objective evidence for that?

Fortunately, serious people researched how NS10M's superior transient-handling characteristics helped its success (it is a fun read, including discussion of yet another creative use of toilet paper):

https://dt7v1i9vyp3mf.cloudfront.net/assetlibrary/n/ns10m.pdf.

That paper only discusses transient response in terms of ringing/resonances - a completely different thing again. It also offers no objective evidence that the subjective perception of the speaker sound is in any way related to the transient response.

 

[boXem]

For instance,
https://www.hypex.nl/img/upload/doc/an_wp/AN_UcD_and_IMD_DIM_slewrate.pdf
starts discussing it on page 4.

As I understood, the lower slew rate of Class D amp is caused by its output filter. Class A and A/B amps don't need one, and you can find examples of 10+ times overkill in their slew rate - not necessary for faithful reproduction of sinusoids within audible range at a given swing voltage, yet arguably helpful for reproducing short pulses more accurately.

Subjectively, such super-high-slew-rate amps tend to exhibit pronounced "rock and roll punch", also attributed to otherwise poorly measuring, yet ever-popular Yamaha NS10M studio monitor. Fortunately, serious people researched how NS10M's superior transient-handling characteristics helped its success (it is a fun read, including discussion of yet another creative use of toilet paper):

https://dt7v1i9vyp3mf.cloudfront.net/assetlibrary/n/ns10m.pdf.

In addition to what @Julf wrote: you confirmed what I was suspecting: the difference you calculated comes from the differences of slew rates between types of amplifiers.
1. most speakers are unable to reproduce more than 50 kHz, so the slew rate at speaker output is anyhow limited. The potential difference between high and low bandwith amplifiers would be the order of the low pass filter formed by the association tweeter+amplifier.
2. there is something not clear in the extract of the "octopus cells" paper you cited: what is the detection threshold? If it's very low, then there would be no difference between various slew rates, and only pure delays would have an influence. In this case, all types of amplifiers have delays of a few ns. If it's high, then slew rate influence would be more significant, but then we are back to 1.

 

[tmtomh]

I think we are mixing two very different issues here. The hypex paper discusses slew rate in the context of slew-rate-induced distortion - mostly an issue with class A and A/B. Simple linear bandwidth limiting of course limits simplistic dV/dt "slew rate", but only affects frequencies above the filter cutoff.

Thank you for this explanation - I found it very helpful and clarifying.

It also seems like part of a pattern: whether it's this, or the more commonly cited issue of pre-ringing that @boXem | audio cites above (and also explains well), there seems to be a current of audiophile discussion where people are very concerned with nonlinearities/distortion that only manifest themselves at the ultrasonic level. Folks seem to be under the impression, for example, that pre-ringing caused by linear/fast filters is something that affects every transient in the audible spectrum, smearing all the transients, when in actuality (if I understand correctly) it affects only signals at the Nyquist frequency. Similarly, from what Julf says here, it appears this slew-rate issue is not relevant for the audible range, yes?

 

[Julf]

Thank you for this explanation - I found it very helpful and clarifying.

It also seems like part of a pattern: whether it's this, or the more commonly cited issue of pre-ringing that @boXem | audio cites above (and also explains well), there seems to be a current of audiophile discussion where people are very concerned with nonlinearities/distortion that only manifest themselves at the ultrasonic level. Folks seem to be under the impression, for example, that pre-ringing caused by linear/fast filters is something that affects every transient in the audible spectrum, smearing all the transients, when in actuality (if I understand correctly) it affects only signals at the Nyquist frequency. Similarly, from what Julf says here, it appears this slew-rate issue is not relevant for the audible range, yes?

Indeed. Slew rate is only an issue inside the amp feedback loop (and even then only if the slew rate is not sufficient to handle the input signal). Some of us are old enough to remember the whole TIM/SID thing that became a non-issue once people started understanding feedback theory better.

 

[Sergei]

I think we are mixing two very different issues here.

I think that in the context of this conversation some of us are mixing three facets of music audio signals into one. The three: sinusoidal (aka harmonic), impulse (aka transient), and noise. The paradigm of Linear Time-Invariant systems deals elegantly with the first one, while treating the second and third ones as theoretical abstractions (e.g. dirac deltas, and random noise with a known and stable spectral distribution).

The mammal hearing system, on the other hand, incorporates specific neuro-physiological apparatuses, dealing with those three aspects of audio signals commonly occurring in nature and in most genres of music. Interaction of these apparatuses results in dynamic non-linear behaviors, which further complicate the neural responses of the hearing system, which is neither linear nor time invariant even in the presence of a simple mix of pure sinusoids.

If we model the whole audio chain just like an LTI system, then sure, all we should care about is accurate reproduction of the sinusoids, in the happy assumption that noise happens to be below the hearing threshold in each FFT bin. In reality, there are also impulses - short non-periodic signals which in LTI paradigm can only be represented with infinite Fourier spectrum. And the noise may interact non-linearly with both sinusoids and impulse components, not to mention that it perceptually more accurately integrates over a critical band rather than over an FFT bin.

Theoretically, a sufficient condition of full-fidelity sound reproduction is the equivalence of time-domain waveforms created by the reproduction system and the live music waveforms, at the listener's two ears (I believe that's what RayDunzl was aiming at). We know that for the sinusoidal component, this is not a necessary condition: e.g. in general hearing system is tolerant to large phase shifts in two sinusoids separated by a critical band, if their amplitudes are constant or slowly changing.

For the impulse part, this may not be the case: in the LTI parlance, decorrelating of the dense harmonics of a sharp impulse signal may throw off the neuro-physiological detection of a time-coincidental front. Let me emphasize once again: per se, such decorrelation caused by the output filter of even a medium-quality Class D amplifier appears to be too low to be noticeable all by itself. Yet, since the time-coincidence detection is based on a sharp timing threshold, such small difference may "break the camel's back", so to speak, in the presence of other decorrelating hardware components, such as crossovers and parts of the transducers with non-trivial reactive impedances.

That's the most plausible explanation I could come up with so far for the unpleasant effects I observed with certain types of music played by a decently-measuring Class D amp on a less-than-perfect mainstream three-way loudspeaker equipped with somewhat weird passive crossover network, which I decided to still keep for specifically such tests. The RayDunzl experiment demonstrated something similar, even on a simpler and active crossover, and a presumably less distorting studio-grade transducers.

These effects were absent when I was AB-box-switching from Class D to one of two +-0.5 db @ 1 Khz level-matched modest A/B amplifiers: two of them were indistinguishable from each other with this loudspeaker, my ears, and the music I typically listen to - prog rock. All three amplifiers were rated +- 1 dB in the 20 to 20,000 Hz range, and below the perceptible THD level across the range of frequencies and sound levels. I don't believe the confirmation bias was considerably in play - I actually expected the more expensive and modern Class D amplifier to sound better.

I intend on keeping testing newer Class D amplifiers coming into my price range and time availability slots, on the weird crossover loudspeaker I happened to stumble upon, until I arrive to an amp that can deal with it just as well or better than one of the A/B class ones. For the reference: the weird loudspeaker is Pioneer Elite SP-EC73, and the decently measuring Class D amplifier that exhibited the symptoms most clearly was Yamaha WXA-50. Naturally, I'm open to alternative explanations of the phenomenon.

 

[SIY]

That's the most plausible explanation I could come up with so far for the unpleasant effects I observed with certain types of music played by a decently-measuring Class D amp on a less-than-perfect mainstream three-way loudspeaker equipped with somewhat weird passive crossover network, which I decided to still keep for specifically such tests. The RayDunzl experiment demonstrated something similar, even on a simpler and active crossover, and a presumably less distorting studio-grade transducers.
...Naturally, I'm open to alternative explanations of the phenomenon.

Most likely: the complete absence of even rudimentary controls in your experiment. Do a good experiment to try to validate or falsify your many, many hypotheses.

 

[RayDunzl]

If we model the whole audio chain just like an LTI system, then sure, all we should care about is accurate reproduction of the sinusoids, in the happy assumption that noise happens to be below the hearing threshold in each FFT bin. In reality, there are also impulses - short non-periodic signals which in LTI paradigm can only be represented with infinite Fourier spectrum.

In case you missed it, I compared Impulse computed from a sine sweep with an impulse recorded in the air (one bit full scale).

https://www.audiosciencereview.com/forum/index.php?threads/impulse-response.1765/#post-44352


Same for Step, calculated from sweep and recorded using a step signal.

https://www.audiosciencereview.com/forum/index.php?threads/impulse-response.1765/#post-44440


I thought the computational analysis that rather accurately creates "edges" from a signal (sinusoid) with no "edges" was a nice trick.

 

[pozz]

From https://www.pnas.org/content/97/22/11773 :

"The anatomical and biophysical specializations of octopus cells allow them to detect the coincident firing of groups of auditory nerve fibers and to convey the precise timing of that coincidence to their targets. Octopus cells occupy a sharply defined region of the most caudal and dorsal part of the mammalian ventral cochlear nucleus. The dendrites of octopus cells cross the bundle of auditory nerve fibers just proximal to where the fibers leave the ventral and enter the dorsal cochlear nucleus, each octopus cell spanning about one-third of the tonotopic array. Octopus cells are excited by auditory nerve fibers through the activation of rapid, calcium-permeable, α-amino-3-hydroxy-5-methyl-4-isoxazole-propionate receptors. Synaptic responses are shaped by the unusual biophysical characteristics of octopus cells. Octopus cells have very low input resistances (about 7 MΩ), and short time constants (about 200 μsec) as a consequence of the activation at rest of a hyperpolarization-activated mixed-cation conductance and a low-threshold, depolarization-activated potassium conductance. The low input resistance causes rapid synaptic currents to generate rapid and small synaptic potentials. Summation of small synaptic potentials from many fibers is required to bring an octopus cell to threshold. Not only does the low input resistance make individual excitatory postsynaptic potentials brief so that they must be generated within 1 msec to sum but also the voltage-sensitive conductances of octopus cells prevent firing if the activation of auditory nerve inputs is not sufficiently synchronous and depolarization is not sufficiently rapid. In vivo in cats, octopus cells can fire rapidly and respond with exceptionally well-timed action potentials to periodic, broadband sounds such as clicks. Thus both the anatomical specializations and the biophysical specializations make octopus cells detectors of the coincident firing of their auditory nerve fiber inputs."

What does it mean? If onset of a transient is blurred over 1 ms or longer time interval, the transient may not be perceived the same way. Instead of startling or energizing, it could be boring, or simply perceptually dissipate within the rest of the music.

Compared to a good Class A, a good Class D amp isn't likely going to be more than 20-40 microseconds behind in swinging from zero to the full amplitude. By itself, this doesn't seem critical. Just 2% to 4% of the critical transient detection window width. Consider this a small timing distortion.

Yet in combination with certain loudspeakers, containing "heavy" crossovers and transducers that further blur the transient, such difference may sometimes preclude the transient detection mechanism from reacting, or at least make the transient subjectively not as loud.

If the high-end speakers designers were tuning their creations using "nearly-perfect" classic monoblocks, they may be stopping at a point when they can no longer detect the transient anomalies. Likewise, mixing and mastering engineers, listening to their nearly perfect studio systems, would make the transients sound right too.

Yet a less-perfect amplifier could theoretically push the overall system to a condition where the transient anomalies are heard again. Corollaries:

(A) A system with much less than perfect loudspeakers would exhibit such anomalies no matter what, and the difference between amplifiers in this regard may not be noticeable at all. Excessive DSP may lead to such situation as well.

(B) A system with "super-perfect" loudspeakers, and/or competent DSP, would in effect "tolerate" timing idiosyncrasies of different amplifiers, and thus the timing differences between such amplifiers once again would not be noticeable.

(C) Only when the loudspeakers, amplifiers, DSP, and the music are "off" just enough, the substitutions of such components could lead to noticeable subjective differences in transients. These differences can be further exacerbated by certain rooms acoustics.

This is a case of taking research completely out of context. You are taking an experiment which tests neural responses in one type of cell and extrapolating consequences for the entire auditory system.

Why not relate that 1ms result to empirical data about click perception or the fusion interval? If you are going this deep, why not trace the complex transformation of acoustic pressure into its mechanical, hydraulic and finally electrochemical form? I guarantee you that a transient traveling that path will not follow that 1ms result. You also don't take into account that firing rates change when under continuous stimulus, and eventually saturate.

I can't pretend that I can draw you the entire picture. I'm no scientist and I haven't studied enough. But what you're talking about contradicts what's commonplace in psychoacoustics.

 

[boXem]

For the reference: the weird loudspeaker is Pioneer Elite SP-EC73, and the decently measuring Class D amplifier that exhibited the symptoms most clearly was Yamaha WXA-50. Naturally, I'm open to alternative explanations of the phenomenon.

The WXA-50 output filters are outside of the feedback loop. So the filters will have resonances (high Q) pending the loads (speakers). If the cutoff frequency is low enough and the Q high enough, this can impact the frequency response of the amplifier in the audible band.
It would be interesting if you could try the same experiment with an UcD/nCore/Eigentakt.

 

[Sergei]

In case you missed it, I compared Impulse computed from a sine sweep with an impulse recorded in the air (one bit full scale).

https://www.audiosciencereview.com/forum/index.php?threads/impulse-response.1765/#post-44352


Same for Step, calculated from sweep and recorded using a step signal.

https://www.audiosciencereview.com/forum/index.php?threads/impulse-response.1765/#post-44440


I thought the computational analysis that rather accurately creates "edges" from a signal (sinusoid) with no "edges" was a nice trick.

I like that. Nicely illustrates my point. If you stretch the first graph in time, you will see that the responses are qualitatively similar, yet not exactly the same. On the second graph, same deal - similar yet not exactly the same. In addition, it shows the ramp-up to about 10% of amplitude before the fast rise to 100%, instead of the signal going from 0% straight to 100%.

Even on such a nice system like yours, if I read the graph correctly, the initial ramp-up appears to be taking just under 1 millisecond, which is significantly more than the characteristic 10 microseconds attack times of some musical instruments (https://www.audiosciencereview.com/...es-dsd-sound-better-than-pcm.5700/post-171729), to which cochlear hair cells can react in less than 50 microseconds in the live music situation.

The just-under-1 millisecond ramp-up, complete with pre-ringing, most likely comes from the crossover and transducers. This nicely illustrates the idea that the speaker's designers were likely slaving over it until they reduced the overall transient response time to a value just below the transient detection duration threshold. In combination with a fast-slew-rate amp (up to 50v/us, like the one discussed here: https://www.diyaudio.com/forums/pass-labs/162348-xa100-5-slew-rate.html), this shall provide good subjective transients reproduction.

Now imagine that instead of your nice amp, you'd be using something with significantly lower slew rate. As is explained starting on page 5 of
https://d1.amobbs.com/bbs_upload782111/files_28/ourdev_548668.pdf
Class D amplifiers typically have an order of magnitude lower slew rate than high-end Class A or Class A/B amplifiers, for all the good reasons perfectly valid in the LTI paradigm.

Let's say you'd be using a Class D amp that would take 20 us to swing to 100 V, instead of your nice amp's 2 us. This could push your nice just-under-1 millisecond ramp-up to slightly over 1 millisecond. Moreover, the ramp-up shape could be transfigured too. These changes could potentially change what would be heretofore perceived as an honest transient into something resembling a short high-frequency sound - likely subjectively perceived as a distortion in the context of a particular music piece.

Now, while designing a new active speaker with Class D amps driving one transducer each, the transducer itself can be made fast enough, so that the slower slew rate of the amp doesn't push the overall system over the transient detection duration limit. It appears the Neumann KH 80 nicely illustrates what could be done in this respect: I've seen its shootout against Yamaha NS-10M, and listeners perceived the KH 80 as just as fast on transients.

As a counterexample, consider Adam Audio starting replacing their mass-market active studio monitors A/B amplifiers with Class D a decade ago, without materially changing their transducers. The supposedly new and improved Adam studio monitors sounded quite unpleasant to my ears, and I eventually sold all of them, switching to Neumann, who kept their traditional A/B amps until they finished developing/sourcing new transducers.

 

[RayDunzl]

I like that. Nicely illustrates my point. If you stretch the first graph in time, you will see that the responses are qualitatively similar, yet not exactly the same. On the second graph, same deal - similar yet not exactly the same. In addition, it shows the ramp-up to about 10% of amplitude before the fast rise to 100%, instead of the signal going from 0% straight to 100%.

I might attribute the "ramp" to the "Room Correction" - FIR and IIR filters - applied at that time.

Red, with DSP, black, without. The DSP settings may (or more likely may not be) exactly the same as in the example above, but should be very similar.

Calculation of Step Response from swept sine:

I'm sorry I didn't mention that condition.

 

[Sergei]

what you're talking about contradicts what's commonplace in psychoacoustics.

Depends on the definition of "commonplace" I guess. It is an active area of research, which, while anatomically known since 1960s, only came into prominence after 2012, when better neural simulation models became commonplace:

https://www.frontiersin.org/articles/10.3389/fncom.2018.00036/full

 

[SIY]

So many thousands more words, not even a hint of experimental testing of the remarkable hypothesis.

 

[Thomas savage]

So many thousands more words, not even a hint of experimental testing of the remarkable hypothesis.

So dull and uninspiring isn't it , they owe us a drink for wasting our time ha ha

Most likely: the complete absence of even rudimentary controls in your experiment. Do a good experiment to try to validate or falsify your many, many hypotheses.

Without these things it's hard to see what value talking online about it can bring ..

If we are really interested in the progression of knowledge those with such fervent hypotheses really do need to do some leg work , else it's all just a waste of time .

 

[Chromatischism]

The effect being described here would seem to apply a "smoothing" effect to all content. The fact that no one has yet identified that difference in blind testing should be a cause for skepticism.

 

[Thomas savage]

The effect being described here would seem to apply a "smoothing" effect to all content. The fact that no one has yet identified that difference in blind testing should be a cause for skepticism.

Healthy skepticism that can be challenged if one want to promote more than just their idle thoughts.

ASR is ment to aspire to more.

 

[Sergei]

The effect being described here would seem to apply a "smoothing" effect to all content. The fact that no one has yet identified that difference in blind testing should be a cause for skepticism.

It does apply the smoothing effect, yet it is too small to be noticed all by itself. You'd need a system which happens to be just below the edge of the transient front detection threshold, to notice the effect when substitution of one of the components pushes the overall system over the edge, and transients start sounding like short sinusoidal or noise bursts not present in the original music.

This is trivial to simulate with DSP, and those of us who dealt with it know how easy it is to reach the "over-smooth" condition, incidentally or on purpose. You may have heard something similar on over-compressed MP3/AAC files. Granted, this is a second-order-of-importance effect for the Class D amplifiers. I felt it could be interesting to discuss it now, given that Class D progressed beyond the first-order-of-importance shortcomings.

Some of the people actually designing Class D amplifiers are present on this discussion thread. My hope is that they'll take this effect into consideration: not in the sense that they'd need to search for ways to drastically increase the Class D slew rate, but in the sense that they'd even more carefully consider the rest of the system they are inserting their amps into.

A high-end, or just exotically weird, system with the transient detection budget already just below the acceptable threshold due to the heavy use of passive crossovers, massive transducers, and DSP, could be a worse measurement/demonstration vehicle for a Class D amplifier than a simpler system with excellent transient response well below the threshold.

Naturally, this consideration is relevant to consumers as well. I see some of the modern Class D modules being wrapped into heavy and pretty aluminum skins and sold at premium prices as amplifiers supposed to replace in place the traditional high-end Class A and Class A/B. It may not work that well in certain situations, and before ditching such Class D amp, some consumers could be well advised to first try it out with a simpler system.