One thing to be careful about is reliance on simple measurements such as frequency response, or THD.
Ideally, a measurement such as EVM that is used in the wireless industry would be a more appropriate measure of the quality of the amplifier.
One example of the problem with measurements like THD is that its easy to crank up the amount of feedback to create good THD measurements. However that will smudge out the transients in a signal and cause a dull lifeless sound. That is why on Class A/AB designs the better sounding amps tend to be the ones with zero global feedback.
Going back to Class D designs, my experience with several ICE amplifiers has been they tend to distort the transients - particularly the quieter signal levels. Perhaps this is due to the deadband issue that is described on the Mark Levinson website - I don't know. Its a type of digital nonlinear effect and so can't be described by simple 2nd order/3rd order products.
I recently got a Spectron digital amp, and haven't noticed this issue at all so far, although I have been using it in balanced mode, which perhaps serves to reduce the effect to the point of not being noticeable.
I am also intrigued by the idea of the Devialet amp - which combines class A with a class D to capture accurate transients - although I have yet to hear one.
Ideally, a measurement such as EVM that is used in the wireless industry would be a more appropriate measure of the quality of the amplifier.
One example of the problem with measurements like THD is that its easy to crank up the amount of feedback to create good THD measurements. However that will smudge out the transients in a signal and cause a dull lifeless sound. That is why on Class A/AB designs the better sounding amps tend to be the ones with zero global feedback.
Going back to Class D designs, my experience with several ICE amplifiers has been they tend to distort the transients - particularly the quieter signal levels. Perhaps this is due to the deadband issue that is described on the Mark Levinson website - I don't know. Its a type of digital nonlinear effect and so can't be described by simple 2nd order/3rd order products.
I recently got a Spectron digital amp, and haven't noticed this issue at all so far, although I have been using it in balanced mode, which perhaps serves to reduce the effect to the point of not being noticeable.
I am also intrigued by the idea of the Devialet amp - which combines class A with a class D to capture accurate transients - although I have yet to hear one.