Hypex nCore vs Class A amps

[BrokenEnglishGuy]

and what about the class A+D ? NuPrime are making class A+D amplifier, class A for highs and d for lows seems to be a nice deal

 

[Julf]

and what about the class A+D ? NuPrime are making class A+D amplifier, class A for highs and d for lows seems to be a nice deal

Class D is fine for highs too.

 

[VintageFlanker]

and what about the class A+D ? NuPrime are making class A+D amplifier, class A for highs and d for lows seems to be a nice deal

NuPrime’s sonic character stands well apart from – and indeed above – the soft, cushy sound we often associate with most tubed amplification and the dry, clinically edgy sound we often associate with solid-state and switching technology. Rarely does an amp achieve an ideal balance of richly textured timbres and harmonics, bottom-end authority, startling dynamics, exquisite low-level detail, and a fully revealed, fully dimensional sound stage against a backdrop of mile-deep silence.

I'm afraid this is bullshit. It's only Class A input buffer + D amplification. The few measurments available for their amps (here the ST10) are terrible so far...

 

[BrokenEnglishGuy]

Class D is fine for highs too.

yeah but what about the taste, A is supossed to be more smooth than neutral

 

[VintageFlanker]

yeah but what about the taste, A is supossed to be more smooth than neutral

Did you even read the thread, or the entire forum whatsoever?

Seems like Class A failed to "smooth" Nuprime highs, at the end...

 

[BrokenEnglishGuy]

Did you even read the thread, or the entire forum whatsoever?

Yes to both, do you even know what is the difference between neutral and smooth? we are talking about coloration.
Or you gonna start to say all the amplifier sound the same and there is only different technical differences?

 

[BrokenEnglishGuy]

Did you even read the thread, or the entire forum whatsoever?

Seems like Class A failed to "smooth" Nuprime highs, at the end...

also, nu prime launch the ST-10M
'' bla bla... he ST-10M sounds more full bodied and rich, conveys greater immediacy and power as compared to the ST-10. The improved power delivery and reduced interference gives the instruments richer overtones and more natural decay. The soundstage is deeper and creates a greater sense of space. Imaging gains more solidity over the ST-10 which already has an impressive placement. ... bla bla ''

The new generation of nu prime seems to be more in the smooth side..

 

[VintageFlanker]

do you even know what is the difference between neutral and smooth? we are talking about coloration.

Haem haem. Did I miss something? Seeing your posts, you're clairly a newbie here as you seem to be regarding audio in general. No issue with that, but better watch your tongue, then...

Or you gonna start to say all the amplifier sound the same and there is only different technical differences?

Would you tell me what's not "technical" about the sound in general? I'm all ears.

The new generation of nu prime seems to be more in the smooth side..

Then believe what Nuprime says. They sure should be right.

 

[Sergei]

The sampling theorem, simply removing or adding samples to a digital data stream is not BW limited in the analog domain. More to the point what does simply trimming digital bits off of a waveform have to do with how any actual signal chain or device operates (unless broken)?

From https://en.wikipedia.org/wiki/Nyquist–Shannon_sampling_theorem:

"Strictly speaking, the theorem only applies to a class of mathematical functions having a Fourier transform that is zero outside of a finite region of frequencies."

"Shannon's version of the theorem states:

If a function x(t) contains no frequencies higher than B hertz, it is completely determined by giving its ordinates at a series of points spaced 1/(2B) seconds apart."​

As Hayabusa already correctly noted in a response to this thread, a pulse "contains all frequencies". That is, its Fourier transform is not zero outside of a finite region of frequencies.

Moreover, see, for instance, https://en.wikipedia.org/wiki/Bandlimiting:

"A bandlimited signal cannot be also timelimited. More precisely, a function and its Fourier transform cannot both have finite support unless it is identically zero. This fact can be proved using complex analysis and properties of the Fourier transform."

So, mathematically, the only "music piece" which is unconditionally a subject to the sampling theorem is a recording of complete silence. In all other practical cases, the "Fourier transform" is just an approximation.

Qualitatively (not precisely and not always), fourier-transforming a sequence of a thousand samples results in approximation with a precision (defined differently in different contexts) of about one thousandth, which is 0.1%, or -60 dB in power terms.

We encounter this value over and over again as an approximation of an accuracy threshold below which audio distortions become inaudible. A thousand cycles of a sinusoid with constant or slowly changing amplitude is typically something with which the practical analysis and synthesis methods using Fourier transform deal well.

Transients - by definition - are not such sinusoids. In a very literal and precise mathematical sense, the Sampling Theorem doesn't apply to them. They can be still mathematically analyzed in the infinitely wide spectral domain, and theory of Class D amplification uses such mathematical analysis.

Since real-life audio signals contain transients, the amplifiers have to somehow amplify them as well. In high-end Class A and Class AB amplifiers, the problem is solved by further extending the spectral domain they operate in: it is typical to have an open-loop amplification bandwidth of up to 5 MHz in power amplifiers, and up to 50 MHz in audio-grade operational amplifiers.

Then of course a significant part of this wide bandwidth is used up in trading off for reducing harmonic distortions, yet not all of it. The high-end Class A and Class AB amplifiers still retain some of it to ensure their seemingly unreasonably huge slew rate, which may be an order of magnitude higher than it is necessary for just accurately reproducing the highest-audible 20 KHz thousand samples long sinusoid.

Qualitatively, this is akin to digital oversampling, so that instead of a thousand original samples being the minimum "safe length", it is improved to about a hundred samples, and thus the audible distortions in transients are pushed away. The seemingly unreasonably high professional studio equipment sampling rates, and extended frequency response tweeters, aid the same objective - reducing transient distortions.

Modern Class D amplifiers do it too. The early ones - the first commercially sold Class D amplifier was released in 1964 - used lower switching frequencies. Theoretically, just 88.2 KHz would be sufficient for cleanly reproducing the long sinusoids. Yet only when the power transistors capable of up to 800 KHz clean switching moved to affordable territory, the Class D amplifiers really took off.

Once again, a part of this "bandwidth overkill" is used for reducing distortions in regular sinusoids. Yet another part of it can be also used for reducing the transient distortions. Qualitatively though, to achieve the level of the best Class A and Class AB amplifiers in this regard, the straightforward Class D amplifiers would also need to go to switching frequencies of up to 5 MHz. Gallium Nitride transistors help moving in that direction (see e.g. https://epc-co.com/epc/cn/GaN技术杂谈/P...Sound-Quality-and-Efficiency-to-Class-D-Audio).

Yet there is also another way. Instead of considering the objective being to amplify the sum of slowly changing sinusoids, with audibly tolerable phase shifts of individual sinusoids allowed, the new objective is to amplify the original waveform, with as much precision as practically possible, given the size and cost constraints.

In addition to Fourier analysis in the infinitely wide frequency domain, achieving such objective can be also aided by analysis employing Control Theory, and the work of the Hypex and then Purifi teams, as well as of others who are advancing the science and art of the Class D amplifiers design, is a testament to that.

If we use analogy with Class A and AB amplifiers, expressed in classical terms, a "killer" Class D amp that professional studios could switch to en masse would amplify a signal bandlimited to 96 KHz with THD+Noise below 0.1% in its full power operational range. A variation: bandlimited to 60 KHz with THD+Noise below 0.03%. I believe such goals are in the ballpark of necessary to finally make the A and AB class amps obsolete.

 

[BrokenEnglishGuy]

Ok anti bs crusader, your posts are a contribution for the humanity, sorry i will watch my mouth i didn't notice i was talking with the most important human in the internet, again sorry, i have only few post asking for a problem with my amplifier and in the end was the amp plug and something irrelevant too about tektons and... yes, i have only 25 post. Nothing more.

Every post of you here it's a masterpiece of pro audio here and you can judge every people that you don't even know xD, you are another fanatic just like someone from head fi or something ? no problem with that, internet is full of fanatics of his opinions

I didnn't even see that ST-10 amplifier, i didn't ask for that amplifier but you try to generalice the company from one amplifier, i saw NuPrime launch new amplifiers and for that i maked my question, but the holy paladin of audio apparently is angry. I'm only want to know what the people think about the new NuPrime, u came with the st-10 which is the old one, so you don't have a clue about the new NuPrime's.

Sadly there isn't full profesional of every new NuPrime amplifier, or you say watch your tongue because you already have all of the test of the new NuPrimes!? :O

 

[VintageFlanker]

Ok anti bs crusader, your posts are a contribution for the humanity, sorry i will watch my mouth i didn't notice i was talking with the most important human in the internet, again sorry, i have only few post asking for a problem with my amplifier and in the end was the amp plug and something irrelevant too about tektons and... yes, i have only 25 post. Nothing more.

Every post of you here it's a masterpiece of pro audio here and you can judge every people that you don't even know xD, you are another fanatic just like someone from head fi or something ? no problem with that, internet is full of fanatics of his opinions

I didnn't even see that ST-10 amplifier, i didn't ask for that amplifier but you try to generalice the company from one amplifier, i saw NuPrime launch new amplifiers and for that i maked my question, but the holy paladin of audio apparently is angry. I'm only want to know what the people think about the new NuPrime, u came with the st-10 which is the old one, so you don't have a clue about the new NuPrime's.

Sadly there isn't full profesional of every new NuPrime amplifier, or you say watch your tongue because you already have all of the test of the new NuPrimes!? :O

My bad, accidentally pressed ignore button. Cheers.

 

[BrokenEnglishGuy]

The most important human block me, :O Cheers saint god

 

[SIY]

From https://en.wikipedia.org/wiki/Nyquist–Shannon_sampling_theorem:

"Strictly speaking, the theorem only applies to a class of mathematical functions having a Fourier transform that is zero outside of a finite region of frequencies."

"Shannon's version of the theorem states:

If a function x(t) contains no frequencies higher than B hertz, it is completely determined by giving its ordinates at a series of points spaced 1/(2B) seconds apart."​

As Hayabusa already correctly noted in a response to this thread, a pulse "contains all frequencies". That is, its Fourier transform is not zero outside of a finite region of frequencies.

Moreover, see, for instance, https://en.wikipedia.org/wiki/Bandlimiting:

"A bandlimited signal cannot be also timelimited. More precisely, a function and its Fourier transform cannot both have finite support unless it is identically zero. This fact can be proved using complex analysis and properties of the Fourier transform."

So, mathematically, the only "music piece" which is unconditionally a subject to the sampling theorem is a recording of complete silence. In all other practical cases, the "Fourier transform" is just an approximation.

Qualitatively (not precisely and not always), fourier-transforming a sequence of a thousand samples results in approximation with a precision (defined differently in different contexts) of about one thousandth, which is 0.1%, or -60 dB in power terms.

We encounter this value over and over again as an approximation of an accuracy threshold below which audio distortions become inaudible. A thousand cycles of a sinusoid with constant or slowly changing amplitude is typically something with which the practical analysis and synthesis methods using Fourier transform deal well.

Transients - by definition - are not such sinusoids. In a very literal and precise mathematical sense, the Sampling Theorem doesn't apply to them. They can be still mathematically analyzed in the infinitely wide spectral domain, and theory of Class D amplification uses such mathematical analysis.

Since real-life audio signals contain transients, the amplifiers have to somehow amplify them as well. In high-end Class A and Class AB amplifiers, the problem is solved by further extending the spectral domain they operate in: it is typical to have an open-loop amplification bandwidth of up to 5 MHz in power amplifiers, and up to 50 MHz in audio-grade operational amplifiers.

Then of course a significant part of this wide bandwidth is used up in trading off for reducing harmonic distortions, yet not all of it. The high-end Class A and Class AB amplifiers still retain some of it to ensure their seemingly unreasonably huge slew rate, which may be an order of magnitude higher than it is necessary for just accurately reproducing the highest-audible 20 KHz thousand samples long sinusoid.

Qualitatively, this is akin to digital oversampling, so that instead of a thousand original samples being the minimum "safe length", it is improved to about a hundred samples, and thus the audible distortions in transients are pushed away. The seemingly unreasonably high professional studio equipment sampling rates, and extended frequency response tweeters, aid the same objective - reducing transient distortions.

Modern Class D amplifiers do it too. The early ones - the first commercially sold Class D amplifier was released in 1964 - used lower switching frequencies. Theoretically, just 88.2 KHz would be sufficient for cleanly reproducing the long sinusoids. Yet only when the power transistors capable of up to 800 KHz clean switching moved to affordable territory, the Class D amplifiers really took off.

Once again, a part of this "bandwidth overkill" is used for reducing distortions in regular sinusoids. Yet another part of it can be also used for reducing the transient distortions. Qualitatively though, to achieve the level of the best Class A and Class AB amplifiers in this regard, the straightforward Class D amplifiers would also need to go to switching frequencies of up to 5 MHz. Gallium Nitride transistors help moving in that direction (see e.g. https://epc-co.com/epc/cn/GaN技术杂谈/Post/13752/Gallium-Nitride-Brings-Sound-Quality-and-Efficiency-to-Class-D-Audio).

Yet there is also another way. Instead of considering the objective being to amplify the sum of slowly changing sinusoids, with audibly tolerable phase shifts of individual sinusoids allowed, the new objective is to amplify the original waveform, with as much precision as practically possible, given the size and cost constraints.

In addition to Fourier analysis in the infinitely wide frequency domain, achieving such objective can be also aided by analysis employing Control Theory, and the work of the Hypex and then Purifi teams, as well as of others who are advancing the science and art of the Class D amplifiers design, is a testament to that.

If we use analogy with Class A and AB amplifiers, expressed in classical terms, a "killer" Class D amp that professional studios could switch to en masse would amplify a signal bandlimited to 96 KHz with THD+Noise below 0.1% in its full power operational range. A variation: bandlimited to 60 KHz with THD+Noise below 0.03%. I believe such goals are in the ballpark of necessary to finally make the A and AB class amps obsolete.

Any experimental data yet?

 

[SIY]

The most important human block me, :O Cheers saint god

What might be useful is a review of actual listening test data. That will make interactions with people who understand the science much smoother for you.

 

[BrokenEnglishGuy]

What might be useful is a review of actual listening test data. That will make interactions with people who understand the science much smoother for you.

Well, good to know there is people how claim his self a profesional of science in the forum, some people just think in another forum that here are just people with expensive toys who don't know how to measure, of course i have my opinion.
You are talking about something useful, do you tested or have measurements of the new NuPrime's ?

 

[SIY]

Well, good to know there is people how claim his self a profesional of science in the forum, some people just think in another forum that here are just people with expensive toys who don't know how to measure, of course i have my opinion.
You are talking about something useful, do you tested or have measurements of the new NuPrime's ?

You’ll want to look at things more generally.

 

[Julf]

Well, good to know there is people how claim his self a profesional of science in the forum, some people just think in another forum that here are just people with expensive toys who don't know how to measure, of course i have my opinion.

You know what they say about opinions...?

 

[Julf]

yeah but what about the taste, A is supossed to be more smooth than neutral

If it is coloration/distortion you want, DSP is a better tool for that.

 

[Sergei]

... This "analog" waveform is no longer single tone ... Therefore, it does not require your hypothesis to explain what you said you hear.

Thank you for the effort to generate and post the graphs. Let's discuss now what they really represent. First, let's start with yet another experiment. Let's render exactly same sequence of samples, but at 48 KHz sampling rate. Thus, the audio signal would be at 12 KHz.

If we follow the interpretation you described, then, with the shortening of the sequence, more and more frequency lobes will be appearing; at some point they'd become audible; and as the sequence will be approaching minimum length, what we'd hear would be some kind of wide-band noise with evolving spectral envelope, mixed with original, yet diminished, peak at 12 Khz.

Yet this is absolutely not what is happening in real life. In real life, you'd be hearing the 12 KHz tone diminishing in loudness as the sequence shortens, and at some point it will become inaudible. Or it will be inaudible right away, if the system sound level is not high enough for you to hear it. There will be no wide-band noise. When the sequence shortens to the minimum length, you may hear a tone-neutral click, once again depending on the system sound level.

Why the contradiction? Because in reality no multi-tone independent components magically appear. What we see on the graphs are artifacts of digitized Fourier representation of a signal which is time-limited within the analysis window. Such artifacts are more noticeable than the usual ones we are all accustomed to seeing - such as the "spectral leakage" lobes appearing on a Fourier transform graph of a single tone, even if it is not time-limited within the analysis window.

These artifacts are not errors: if you apply reverse Fourier transform to them, using the phase values as well, you will indeed get the original sequence of samples. The artifacts just aren't real-life independent sinusoids that you can take a magnitude of, and based solely on this, estimate how loudly they'd sound.

Why do we use the Fourier representation for analysing the performance of electronic and mechanical components then? Because under the right conditions it is helpful. What are these conditions? The input signal must be a mix of constant-amplitude sinusoids represented by a long enough sequence of samples. SINAD measurements on this site are good examples of that.

It is not a coincidence either that the SINAD sinusoids are separated by 1/3 of an octave: this is the distance along the frequency axis that makes the sinusoids largely independent from each other, from the human hearing system perspective. In this case, we can take their magnitudes and estimate how loudly they'd sound, disregarding their relative phases.

 

[Mulder]

I have three different class D amplifiers, and they all sound a little different. I also have a tube amplifier that goes in class A and I have two traditional transistor amplifiers that go in class AB. In one case, it has a built-in preamplifier with tubes. All of these sound different. There are differences between different class A, class AB and class D. It is not possible to generalize. There are amplifiers of all classes that sound great. There is also a separation line between transistor amplifier and tube amplifier. The latter are often class A. Class D at its best, is clean, dynamic, handles bass exceptionally well, the treble is smooth and natural, and it has a very low noise floor. When class D is at its worst, the treble sounds unpleasantly sharp and lifeless and the sound becomes flat and dull. Ideally, one could not distinguish the sound from different typologies, but in practice this is not always the case. And not everyone thinks that a neutral representation is desirable. In fact, many people like tube amplifiers, but these never sound as clean as Class D or a good traditional transistor amplifier.