No scientific engineer would say that. There is zero scientific evidence that the sound of an audio device can be completely described by industry standard measurements.
What some engineers do believe is that if distortion and noise are low enough then any physical differences in sound should not be audible to humans.
Where different engineers disagree is how much distortion and noise are 'low enough.' Moreover, there is scientific evidence that THD+N is a pretty useless metric. There have been attempts by researchers to adopt a new industry distortion standard, such as the Gedlee metric proposed by Earl Geddes. Unfortunately, the industry is seemingly not interested in improved metrics.
In addition, there is the mostly unspoken of problem of noise modulation, which can be an audible defect despite low conventionally measured noise.
EDIT: Also of possible interest, Benchmark Media, the makers of the SOA power amp AHB2, have an ABX listening room at their design facility. They definitely do listen. ESS is another company that has described doing blind listening tests of their dacs to detect audibility of noise floor modulation.
Last edited:
"There is zero scientific evidence that the sound of an audio device can be completely described by industry standard measurements."
There are so many qualifiers in that statement that it becomes meaningless. But, an amplifier, for example, shouldn't have "a sound", and it becomes quite a simple matter of doing a null test to make sure it has as little as possible.
There are so many qualifiers in that statement that it becomes meaningless. But, an amplifier, for example, shouldn't have "a sound", and it becomes quite a simple matter of doing a null test to make sure it has as little as possible.
I have been withholding talking about this, but I have a Benchmark AHB2 which is used along with Sound Lab electrostatic speakers. Recently we did a listening test to compare a modified Aragon 8008 MkII amplifier with AHB2.
It was obvious to me that AHB2 had less audible IMD than the Aragon. What surprised me was that AHB2 did not audibly reproduce some small details of sound that the Aragon did. When people described such things to me in the past I rationalize that they must be hearing some distortion artifacts that are being confused with small musical details. Not so, as it turned out. I now believe AHB2 has noise floor modulation. That's what it sounds like to me. While I still like it for its ultra low distortion, I don't like the hidden-in-very-smooth-sounding-dynamically-changing-noise loss of musical details.
It was obvious to me that AHB2 had less audible IMD than the Aragon. What surprised me was that AHB2 did not audibly reproduce some small details of sound that the Aragon did. When people described such things to me in the past I rationalize that they must be hearing some distortion artifacts that are being confused with small musical details. Not so, as it turned out. I now believe AHB2 has noise floor modulation. That's what it sounds like to me. While I still like it for its ultra low distortion, I don't like the hidden-in-very-smooth-sounding-dynamically-changing-noise loss of musical details.
Last edited:
Another subjective listening impression. Don't get me wrong, you are entitled to it.
I have never seen noise modulation published other than by ESS. However, dac chip-level designers do measure it since the mechanism is more well understood in sigma-delta dacs. Seems likely ESS talked about it because they feel they have a competitive advantage in that area. Perhaps like Purifi appears to talk about hysteresis distortion because they feel they have an advantage with it. Probably shouldn't expect anyone to bring up a non-standard measurement when they don't see a direct benefit from doing it.
Last edited:
What is the mechanism that you presume to be at work here? How would a amp be able to modulate the noise floor?
EDIT: The mechanism in DACs should more precisely be called "signal dependent noise floor generation", a wholly different story IMHO
Last edited:
I think the subjective effect sounds like noise floor modulation, but what actually occurs is distortion floor modulation.
A dac designer found that an audio preamp sound changed depending on the AC power cord used with it. He found that noise floor varied with the audio signal. He found that the AC cord was acting as a filter for noise on the AC line coupling into the preamp. The preamp was fitted with a noise filter on the AC input and the problem was fixed.
In other cases, some work may need to be done to show what mechanism(s) are involved. Probably not much to be gained by speculating about causation here. If you want to swap ideas by PM maybe that would be okay.
Regarding noise modulation verses distortion modulation, on an FFT either one may look to some degree like noise, could look totally like noise. Things sometimes tend to be described in terms of how they appear in measurements.
Last edited:
Good engineer at work
Scope FFT revealed MHz spikes on my soundcard output, modulated by current draw of whatever it was powering, and it was related to music signal. Thus, HF noise modulation. Turned out a switching DC-DC converter was 1cm away from output jack. It was powering the audio chips, current draw correlated with signal, thus noise modulated by signal. Replaced with LM317, sound improved. This is not rocket science.
Note the amount of trash from mains conducted into your amp varies with the conduction time of the rectifier diodes, which means it depends on current drawn. When diodes are off, no noise is conducted. More noise modulation.
Yeah, someone said, "you should measure what is relevant, not what is easily measurable"...
Gedlee says the higher the second derivative of the transfer function at zero, the worse it sounds. It's absolutely logical.
I generalized it by replacing "at zero" with "near zero" which means, measure THD on low amplitude sine wave, with a DC offset, step the DC offset so it results in a DC current that sweeps the whole crossover in small steps, and pick the measurement that gives the max THD.
Measurements incoming in a few weeks if all goes well.