Distortion Surface: old measurement, new approach

Next level of enlightenment.
Mentally visualize music harmonic content overlaid on a given distortion contour.

 

That’s some good shit

 

That's some baller shit right there. We're not worthy

 

Yeah. I can see "acoustically averaged" music spectra by genre to give an answer to the question, "Is this any good for classical/dubstep/metal/...?"

 

Mjolnir 3 is one incredible amp. Here are distortion surfaces and contours for it.







This one is a keeper.

 

It would be interesting to see this approach applied to IMD measurements as well, and perhaps try other types of equipment like DACs.

It looks like the contour plots for individual harmonics are more interesting than the THD+N, since the latter ends up dominated by noise in the lower regions.

 

Oh, this can be fun. Check this paper [1] for inspiration. Essentially, defining a penalty function for higher harmonics (finding the weights of course warrants a workshop with top minds and beer).

[1] http://cired.net/publications/cired2015/papers/CIRED2015_0392_final.pdf

 

Now that's interesting.

I suggest a workshop/meet where single harmonics are intentionally introduced.
The listeners could then score how objectionable each is. The SBAF-AD score could be derived.

Proceeding onto multiple harmonics and scores would be/could be impractical?

 

My take is that Distortion should be decomposed into the following component sweeps over level and frequency:
1) D2
2) D3
3) D4 through D10 as a single sweep, using the penalty weighting scheme suggested above
4) AC mains noise, fundamental + 5 or 6 harmonics
5) non-harmonic noise (everything else)

It may be a PITA to grok, but otherwise too much information is lost in the roll up.

 

short recap examining THD+N at 1 KHz 2Vrms vs surfaces and contours for

1) DSHA-3F
2) iDSD Diablo
3) Mjolnir 3

Evaluation of the three amplifiers as preached by the Cult of Sinadology:

Very limited view of amplifier performance.

The same three amplifiers evaluated by scientific advancements in measurement methodology:

Infinitely more information about the three amplifier THD+N distortion profiles than in the previous graphic.

Decomposition of THD+N into D2, D3, 4+HD, Noise AC Mains, Noise Everything Else is on my list of automation activities for data collection and visualization rendering.

 

Fantastic!

What's the 2D projection at the bottom intended to convey?

 

It's just a heatmap of surface. Same thing, different way of visualizing.

 

I'd be interested to know from an engineering standpoint what the difference in the smoothness between each surface might be attributed to, particularly between the iFi Diablo and MJ3. My gut says something to do with the power supply sections and the choke in the MJ3, but if so, why, specifically?

 

Coming more from a creative technical background as opposed to an engineering background, you could probably do something cool with the csv files using a combination of QGIS (use the csv and the grid interpolation tool to create a raster-based elevation map like a DEM file) and Blender.

QGIS can output the elevation map as grey-scale image at 16bits which can be used as displacement map on a surface in Blender. Blender would allow you to really examine the surface from a range of angles. Depending on the resolution of the CSV data, that surface shape could be pretty detailed. It would take some doing to get the scales correct, but this could be a cool way to visualize this data.

 

I've always meant to ask this question (out of complete ignorance), but keep forgetting. A question over the the MJ3 thread brought this back up in my mind.

How much does the 300R load play into these graphs vs a lower load (like from a set of focal cans)? Particularly curious how it would play into the data for the 3F amp since it was ostensibly designed to be synergistic with focal headphones...

 

I don't know yet. I am still wrapping my head around this greatly expanded view of amplifier distortion. Also keep in mind that two excellent sounding amplifiers may have surfaces and contour maps that differ greatly. I haven't used this method on amplifiers generally thought to fall in the category of sounding either meh, or not good. Even partially automated it remains a time consuming undertaking.

 

The answer is, it depends. Some general rules of thumb (that almost always have exceptions):
1) amps with higher impedance outputs are likely to measure differently for distortion between low and high impedance loads
2) high impedance loads want more voltage than low impedance loads
3) low impedance loads want more current than high impedance loads

Now add to the mix the complexity presented by many headphones:

Focal Clear represents a very interesting load rising to 500 Ω at driver resonance.
HD800 family rises to between 800 and 900 Ω.
HD600 rises to over 600 Ω.
HD650 rises to 500 Ω.

HFM HE-500 and ETA Mini-C both maintain their nominal low impedance over the frequency range.

So even though we often discuss headphones with respect to their nominal impedance, it is important to be mindful of impedance sweeps such as these.

 

@atomicbob any chance you have an HD595/598 impedance measurement in there? I seem to recall they started off 50's-ish and spiked up but don't remember if it got as high as the Focal.