I know these guys you mentioned all have credits. But I also have credit. I have a Ph.D in electrical engineering. So you need to understand why I said what I said.
DSP is good, but is not perfect. There was a company that claimed they can make a NHT speaker generate a perfect impulse response. I visited the company's booth in CES and asked the engineer just one simple question, does that mean the true quality of the speakers is no longer important? He refused to answer my question in front of other customers and wanted me to go away. The truth is like this. First, all speakers have characteristics changing between small signals and large signals. What is a bad speaker? A bad speaker is a speaker exhibit wild difference between small signal characteristic and large signal characteristic. Does the DSP fix the small signal characteristic or the large signal characteristic? DSP is a linear operation, it cannot fix both. But servo can fix both. Secondly, you know the car you drive does age and its feel is difference when it is cold vs when it is hot. Why do you think the subwoofer is any different? When it ages, the frequency response also changes. Does the DSP correct that? No. When the voice coil in the subwoofer is cold vs when it is hot, the frequency response is also different. Some called it thermal compression. I prefer to call it thermal memory effect. What is memory effect? If you press a memory foam, you will notice the impress takes some time to recover. The thermal effect on voice coil works the same way. You will hear the term thermal stress very often to explain how a voice coil can be fried. Now think of how the thermal comes and goes in voice coil. Does DSP correct that? No. Some subwoofers have such as poor memory effect, people think the subwoofer sounds slow with respect to the other channels. That slowness is a profound characteristic. Otherwise, all we have to do is move the subwoofer closer to us to correct it. Lastly, we all heard the term "unit-to-unit" variation. I can even extend that to "batch-to-batch" variation. How many times people find the actual frequency response is different from published spec? In fact, these variations do exist and is the explanation of what had happened (assuming the manufacturer does not make up the spec in the first place) . How does DSP address that? It cannot. Do you really think they will give each batch a different DSP correction? That will only create more service problem. As a result, you will see some frequency response measured by third party that does not look like smooth curve at all. It is a clear indication that imperfect EQ has been applied and as a result wavy frequency response is measured, especially when the reviewer pick a random unit from the market place without manufacturer's acknowledgement. It is imperfect because the correction is only valid to the sample unit in engineer's hand. If they find the production units are different from the sample unit, what do you think they will do? You would have guessed it right. Ship it. The customers will buy that DSP rubber stamp. Servo is different. It is a closed loop feedback system. It takes the unit-to-unit/batch-to-batch variation into account and adjust the needed correction. So now you can guess how can DSP generate a perfect impulse response? It can, but only at one particular level that the engineer samples. That is the reason I was sent away because the R&D already know the limitation. If TV is selling that DSP to me, I will chanllenge the same thing. DSP only works with the assumption of a model. Who double check if the model is correct? If no one does, it is just a rubber stamp. Not only that, you know that two bad characteristics in the system, instead of just one.
Now in terms of amplifier class, we also use class-D amplifiers. So I don't think there is a problem with class D amps. But I do think the transformer based amplifiers will sound better in bass and the argument does not come from weight. It is more from the technical difference if you care to understand. In your transformer-based design, the wall power goes to the transformer first and is rectified and then pass to large power capacitors. These capacitors gets charged at a rate of 120hz. Now even though I say it is at 120hz, the actual charge time is only 1/10 of the 120hz (or about 1ms duration). For the other 9/10 of the 120hz cycle, the energy comes from the power caps themselvs only (as if the wall power is gone). Moreover, you notice that the bass frequency we talk about, say 20hz, takes a span of 6 charging cycle. This is how you notice the bass signal can have a different impact than a 1khz signal on the power supply becasue the latter can be completely covered in one charging cycle. Now the most important thing for us to take note is there is no resistance between the power capacitors and the power supply of the amplifier because it is just a wire connection. So the power amplifier can take whatever is in power capacitors without restriction. So much as that I can short the power amplifier's power supply to get the max amount of current. To be more precise, I can short the power amplifier supply and the power supply unit can still survive. Not only that, you can get an electrical arc if you do with a contact, confirming the current supply is almost infinite. Certainly we don't want to do it too many times. But in short it will not be catastrophic. Now for the SMPS (switching mode power supply), the transformer sits after the power capacitors. So when transfer energy from power capacitors to the amplifier it goes through these components: MOSFET, transformer, and then rectifier. All of these components create IR drop (current times resistance) which restricts the current flow. Now try to short the power supplier of SMPS. Not only you can destroy the SMPS, you will never see the same electric arc because the current is limited. In short, it just does not supply the current in a snap.