All the more reason to test and publish all amplifiers' residual noise levels in uV. The spectrum of the hiss should be commented upon. It's good see people are finally seeing through the hype, listening for themselves and hearing anomalies in these Class Ds.
There is no anomaly. The SNR results tell us that the top Class D amplifiers have ridiculously low levels of noise and only a few AB amplifiers can top them (Benchmark AHB2, McIntosh MC462). You are right here though:
As
@pma states, inputs are shorted and then you measure the residual. Sure, it can be brickwalled at 20kHz for the Class Ds or A-WTD, but if you
A-WTD and/or brickwall the measurement, do the same for Class ABs as well or the playing field isn't level.
One should use the same weighting. Regarding the hiss, a properly built NC500 or Purifi amplifier has no hiss. And hiss coming from the sensitive input line stages and them amplified by the power stages can happen in any topology. If an amplifier amplifies a signal, it amplifies also a hiss.
We also know Audiophonics is probably lying about the noise figure (11.5uV) on their amplifier. They have quoted the Purifi spec sheet noise and yet they have a switchable gain stage up front. Note, that is an A-WTD figure, not unweighted. We know the noise will be higher than 11.5uV except on perhaps the "bypass" position. You cannot have 7.2dB or 12.5dB of gain up front without impacting the published module-only noise figure.
It is true that they quote directly form the Purifi spec sheet, in a complete amplifier it is rather difficult, if not impossible to have only 11.5uV of noise. However, if the input buffer's noise is low enough, you can easily have roughly the same total output noise whether unity, 7.2dB or 12.5dB of gain are used. Also because the noise from the buffer does not
necessarily depend from the gain.
This said,m even using a state of the art buffer like the Neurochrome Purifi/NCore buffer, we have
The 12.8Db gain of the 1ET400A represent a voltage ratio gain of 4.365. now let us have a look at the noise from the buffer
- 4.0 µV RMS at 13 dB gain -> the contribution from the buffer becomes ~ 17.5 µV RMS
- 1.6 µV RMS at unity gain -> the contribution from the buffer becomes ~ 7 µV RMS
in the first case (assuming Purifi also gives RMS values) the total RMS noise becomes ~ 21 and ~ 13.5 µV RMS.
So giving the value as 11.5uV for the whole amp would be wrong. But also
not orders of magnitude wrong.
Of course, it is all to be demonstrated that the Audiophonics amps have such good values – anecdotal evidence seems indeed to point to the contrary, and looking at the poor cable routing would also suggest that.
Roberto