It's like the Force. Can't be measured. No one believed in the midi-chlorian tricorder thing anyways.

Is this supposed to be a headphone thing, or mainly a DAC or amp thing? I guess all...

I've been blindly looking about it a bit. For headphones, could it have anything to do with being able to have a frequency response that is free of spikes and that is together (preferably flat) as much as possible?

If you look at the "raw data" for the HD800 here: http://www.innerfidelity.com/images/SennheiserHD800B.pdf you can see that the FR can generally be followed in a curve for most of it, though most just have a treble spike which will cover a bit but perhaps not all of the treble spectrum: http://www.innerfidelity.com/images/BeyerdynamicDT250250.pdf

AKGs are pretty cool like this with the treble as well: http://www.innerfidelity.com/images/AKGK712.pdf

(I realise reported treble response can vary a lot with different measurement methods though...)

I guess quick decay is great as well. No idea how you could check attack on a graph.

Honestly, over all no idea but I would like to find out.

I think it has nothing to do with frequency response.  Especially since some certain amps do better than others with this and they all measure flat.  No AKGs or Beyers come close to the HD800 retrieval either.  Decay is helpful but not all of the equation.  SS and Imaging is important as well.

Is it a function of low distortion, decay characteristics, attack characteristics, or is there something more to it that cannot be seen in measurements?

IMO (but am told am mostly very wrong about it) it has to do with low distortion, fast decay characteristics, fast attack characteristics, frequency range, 'flatness' in the treble area.
I suspect it has a lot to do with 'break-up' in the driver membrane/cone as well as membrane weigth + motor force.
Perhaps cone break-up could be made 'visible' with a very high speed camera + strobelight.
Another aspect is pads and acoustics.
Most of this is difficult to prove in plots but I feel spiky treble may be a sign for it.
I am also convinced it can be measured BUT not with simple THD and FR measurements and one has to have an entire suite of measurements. All of these must be judged in relation, not on 1 or 2 aspects.
'micro-detail' is basically a very small signal (excursion) superimposed on larger swings of the rest of the signal and the driver must have the ability to accurately 'follow' those.
Unfortunately small signals are ALWAYS masked by larger ones on ANY type of absolute value measurements and thus cannot be made visible that easily.

Also needle impulses tell a lot but perhaps the plot should not be linear but have a log type amplitude scale as our hearing is log and not linear.
For the same reason scope pictures of a sinewave aren't really enlightening when looking for small THD values.
If you can see it, it is VERY measurable and audible as well.

When looking at innerfidelity plots you must realise that everything above say 1kHz must not be considered as 'flat'.
Meaning treble that is at '0dB' in these plots is a peak far above '0dB' in reality.
So above 1kHz the treble 'energy' in the plots isn't accurate in an absolute sense.
He re-did a few plots with 'Olive Welti' correction and these were more accurate IMO but he would have to re-do loads of pdf's which IMO isn't worth the effort.
This is due to chosen corrections.
Of course his plots (pdf's) are excellent in any other aspect and are very valuable for comparing relative to the other plots.

The plots on this site (as well as GE IMHO) are more accurate in treble 'level' relative to low/mids.
But you have to realise that ALL acoustical measurements can be/are 'off' in an absolute and even relative sense due to lots of acoustic errors/compromises.
Take ALL acoustic measurements with a grain of salt, an indicative measurement at best that may be of help if you know how to interpret all relevant measurements TOGETHER.

Just my opinion of course.

'micro-detail' is basically a very small signal (excursion) superimposed on larger swings of the rest of the signal and the driver must have the ability to accurately 'follow' those.
Unfortunately small signals are ALWAYS masked by larger ones on ANY type of absolute value measurements and thus cannot be made visible that easily.

We could create a signal that has say, a 1khz component with amplitude X, and a 1.01khz component with amplitude Y, and keep decreasing the amplitude of the next component until we hit the point where the lowest amplitude component cannot be reproduced correctly.

Quote from: Solderdude on October 27, 2014, 06:00:07 AM

'micro-detail' is basically a very small signal (excursion) superimposed on larger swings of the rest of the signal and the driver must have the ability to accurately 'follow' those.
Unfortunately small signals are ALWAYS masked by larger ones on ANY type of absolute value measurements and thus cannot be made visible that easily.

We could create a signal that has say, a 1khz component with amplitude X, and a 1.01khz component with amplitude Y, and keep decreasing the amplitude of the next component until we hit the point where the lowest amplitude component cannot be reproduced correctly.

I wouldn't suspect this test to be terribly indicative of perceived microdetail. The FFT binning and noise floor of the measurement system would dominate.