Wow, I was busy for a few days and this thread just ran away. As the co-founder of NuForce and now CEO of NuPrime, let me say a few things to summarise what is going on.
In 2005, The Absolute Sound, for the first time (of any audiophile magazines), gave the NuForce Reference 9 Product of The Year award. In that article, Chris Martens made a controversial statement that Class-D has arrived. Well, it took the next ten years and countless iterations of designs to finally master it. There aren't many companies engaging in high power, high-end class D design, most of them purchase the module from Ice Power or Hypex and tweak it from there. And sure, those amp modules improved over time.
With every new models of UcD (now Ncore) or ICE, people raved about it. Likewise, NuForce plowed along with its own V1,2,3 series of amps.
A very layman explanation about Class-D concept: input analog signal is modulated into a series of pulses (think of your FM or AM signal). Unlike FM and AM, class-D uses the width of a pulse to represent the amplitude of the input analog signal that is sampled. So, you ended up with a series of pulses with the same amplitude but different width. The pulses are further manipulated to become pulses with bigger amplitude. And eventually the big pulses are then demodulate to become analog signal again to drive your speaker. Other than converting the initial analog signal into pulses, there is nothing digital about this type of implementation. It is very analog in nature (ah, you have many ways to shape the sound using analog techniques). There is another type of Class D implementation that is very digital where input signal is converted into 1s and 0s using DSP to manipulate (Devialet and many low power class-D chips). The advantage of going digital is that you can do a lot of processing to change the sound (increase the bass, change the EQ, combined with DAC, etc). So this is where Class D is often confused with all-digital amp. The drawback of this all digital approach is that it sounded "digital".
In the early days of Class-D "analog" implementation, engineers have not figured out how to improve switching frequency (of the pulses) and manipulate them, so people can detect various weaknesses (bass, HF extension, jitter, whatever). Solid state and tube implementations have weakness too. But engineers have worked on those designs for much longer period of time. Some people think that Class D is not high resolution and so it will always sound digital. Unless you are listening to live music, as long as music has to be recorded, along the way, resolution is "constrained", even with vinyl.
By now, a handful of engineers who specialised in Class-D designs can implement switching frequency at more than 500kHz (CD is 44.1kHz) and in our lab, we have designs running at 1MHz, 0.0x% THD at wide range of freq and power, low jitter, etc. Without going into details that get many people lost (and I can't explain anyway), our engineers can now produce class-D amp that is near perfect.
This is not to say that most class-D engineers know how to do that. There are many bad class D implementations on the market too.
Unlike Ice Power or Hypex, we are not in the business of making high volume amp modules at low cost. So we spend a lot of our effort in making different sounding amps. See http://www.nuprimeaudio.com/index.php/reviews/amp-comparison.html
Just like one of you who stated, high-end audio is like women (or men if you are one of those rare lady audiophile), not everyone share the same taste. I would say that MCH-K38 is like a very sexy blonde that most men like. Make sense?
Lets talk about amp characteristic. There is no perfect sounding amp. And some characteristics are contradictory to others. For example, to make an amp "Warmth" sounding, it must have more even order harmonics. And it won't be as neutral and clear as one without even harmonics. A recording engineer might use both types of amps. A super clean and detailed amp as a reference, and another amp to help shape the sound. Electric guitar amp is intentionally made to have a unique sound, but it is used to produce the sound, not for reproduction.
Ref 20 was the NuForce top end reference amp, and it was designed to be as smooth, clean, detailed, dynamic, etc as possible. At NuPrime, we made a huge leap forward in not only how to build a world class amp, but many different sounding amps. Going from making one "near perfect" design to making many "near perfect" designs is a huge step. Very often, you improve something and you lose something else. So we want a tube amp to sound like tube, but I don't want to lose the dynamic and speed. Is MCH-K38 better than Ref 20, no, otherwise we would not have kept Ref 20 with NuPrime logo. but if you watch a lot of movies and also listened to a wide variety of music, you might actually prefer MCH-K38 over Ref 20. If you are into big classical music, you will likely go for Ref 20 or ST-10. If vocal is your thing, you will likely pick MCH-K38 or STA-9.
I think MCH-K38 strikes the right balance that impressed John so much. Yes, MCH-K38 is really something.
We won't be bringing out a stereo/mono amp using the same K38 amp. To do that would mean our R&D has stopped working on amp.
Oh, I want to talk about pricing. We don't price something based on how good it sounds within our own family or versus the competition. We are an engineering company. We are not in the business of pricing our amps based on weight or extensive workmanship. We price our amps based on our cost. With more knowledge over time, the cost comes down while complexity and performance go up. I will use an example that most of you understand. You all know that switch resistors is an expensive high end preamp design. But we learned how to implement it using FPGA (it is a big logic chip) so this design that used to be in $5000 preamp, is now in every one of our integrated amp and DAC. We are talking about an order of magnitude in cost reduction (from a bank of expensive resistors to a chip). Similarly, our amplifiers use a naturally occurring self oscillating design principle to generate the PWM pulses for Class D sampling (instead of very expensive external PWM generator). As we learned how to increase the frequency to 1MHz, instead of selling you a $100,000 amp, we might be able to sell you a $649 amp that priced like the STA-9. To be very honest, some other parts of the amp now cost more than the amp module itself. Power supply is now the dominant cost.