Frequency response is mainly a result of the output impedance vs. frequency, I was used to look at the frequency response but with time I started to prefer the output impedance. They are two ways of looking at the same thing.Would you care to elaborate please? I'd considered the exact same measurements, but came to the conclusion that, assuming purifi's "1m! on the scale means 1 miliohm, the purifi has much less output impedance, and much less rise in output impedance, relatively speaking - however even though the pascal has a greater amount and a bigger rise, its quite far up in the frequency spectrum, and if anything, rising output resistance would reduce, not increase treble?
But on the other hand if the Purifi's rise in output impedance is actually great enough to be audible, the way it graphs, it starts so early that under a low resistance load, if it rolled off frequencies in that same inverse manner, that would I think correspond pretty well with what I heard...
Because assuming I'm correct in saying rising output resistance, would reduce voltage/linearity in the given area, relative to the rest, i.e. would roll off upper frequencies, and more so the lower resistance the connected load is, and assuming the output resistances of both amps are high enough to affect linearity to an audible degree, due to the non-linear nature of the resistance (which I'm aware would be dubious, as both with conventional "good enough" reasoning, should be low enough as a whole and would be best to actually test.), and presuming the non-linearity of either amps impedance is equally audible (fx. if the purifi scale is ohm, not miliohm), simply looking at where they would be rolling off:
Then the purifi would theoretically start seeing some degree of roll-off or reduction in relative level already from 3-400Hz, increasing at the frequencies rise, which would produce a warmer/more rolled off sound, more so than the Pascal arguably, which only begins to really have rising resistance at 2-3kHz,, an only steeply rising much later.
I don't know if that is what you hinted at, but assuming all of the above is true, and is audible, then I think it would line up with, and likely explain what I was hearing.
It would also possibly explain why the difference was more audible on my JBL Ti10k vs. the Dynaudios, as the JBL is very sensitive, and from 70hz up is around 4Ohm, where as the Dynaudios are not efficient, but fairly stable at 6Ohm, going to 5 at their lowest - in other words, the JBL's lower impedance, would exacerbate the above issue, relative to a higher impedance speaker like the Dynaudio.
Though this may also simply be a result of the Dynaudios low efficiency, which I know has been correlated to speakers being much less audibly affected by characteristics of the chain they're connected to...
1. The frequency response:
The difference between the Purifi and the Pascal is that the Pascal is not load invariant. We see that from 5 kHz, the response of the amplifier is changing with the load attached. Since the speaker impedance is far to be flat, some frequencies above 5 kHz will be emphasized (high speaker impedance), others attenuated (low speaker impedance). For an amplifier that is not load invariant, the Pascal is well behaved, in the sense that it doesn't amplify anything. Since in general speaker impedance is still low around 5 kHz but is rising further away, one can assume that frequencies in the 5-7ishes kHz may be attenuated. Higher frequencies will be less attenuated due to higher speaker impedance, lower frequencies the same but due to the load invariance. So a first source of coloration comes here. In comparison, since the Purifi is completely load insensitive, it will not add any coloration.
2. The output impedance:
Depending the frequency, the Purifi output impedance is between 0.05 and about 0.65 mOhm. For a speaker impedance of 4 Ohm, this corresponds to an attenuation between 0.0001 and 0.0014 dB.
In comparison, the Pascal output impedance is between 6 and about 150 mOhm. Still for the 4 Ohm speaker impedance, this is an attenuation comprised between 0.013 and 0.3 dB.
Since the speaker impedance is not constant, some frequencies will be attenuated, some other not. We consider that level matching is mandatory because human ear is able to discern differences of 0.1 dB. You can see that this threshold is reached in the case of the Pascal. A coloration is introduced.
While we have a tendency to consider that level matching erase all differences, it is only the case for comparison between gear with flat frequency response. The problem is that amplifiers loaded with speakers do not have this property. If levels are matched at a frequency where amplifier A attenuates and amplifier B amplifies from a flat response, instead of being erased, differences will be highlighted.