Distortion Surface: old measurement, new approach

BBC weighting would be good for a single number or graph that rules them all. However, my personal preference is seeing distinct D2 and D3, and if needed D4+ (for most transducers, D4 and higher calms down, but for those edge case instances where it does not calm down, show it). In other words, THD of 95% of transducers tends to be highly dominated by D2 and D3.

 

The number crunching actually isn't that hard at all. Using the BBC method (but taking into account only D2 and D3 with D3 being penalized with x 9/4 multiplier)

Pinnacle P1


Ara


I think the results above are obvious, but after experimenting with this (with older graphs), I'm not convinced a single graph that measures THD(BBC) tells a good enough story. Second and third harmonic distortion sound very different (and even thenm, has a different effect in the lows to highs). I feel that D2 and D3 are best presented separately rather than "combobulated". I'm also not convinced penalizing the third harmonic more than the second is justified. It's different for sure, but not convinced that it's always worse.

Here are some distortion patterns typically seen, relating to personal preference.

 

With the patterns, talking about all gear - although dynamic including isodynamic drivers do tend to fall within the first triode-like pattern, unless the design is really screwy / cheap.

 

Playing around a bit more. This is basically the 3D version of what I was doing before with the distortion charts. This is distortion in absolute db terms instead of % of the fundamental.

JAR600
Second Harmonic (db) absolute vs Frequency vs Output Level (db @ 500-1kHz AVG)


JAR600
Third Harmonic (db) absolute vs Frequency vs Output Level (db @ 500-1kHz AVG)


In many ways, I feel this is more intuitive for ordinary people to understand compared to THD%. For example there's no "rise" in % as we move to lower output levels where we run into dynamic range limits, which can kind of throw people off. As we increase output levels, distortion rises along with it. It also gives us a better D2 to D3 perspective. There is still some noise at the lower output levels where distortion is low, especially in the lows.

 

Same visualization technique as above but with RS2X

RS2X F-Pads
Second Harmonic (db) absolute vs Frequency vs Output Level (db @ 500-1kHz AVG)


RS2X F-Pads
Third Harmonic (db) absolute vs Frequency vs Output Level (db @ 500-1kHz AVG)

 

CFA Ara IEM. This is a BA type IEM which I consider one of the cleanest and clearest sounding around. It uses 7 drivers. BA or balanced armature types will have third harmonic that tends to group together tightly regardless of output for this visualization. The second harmonic barely moves, probably because so many drivers are sharing the load. The third harmonic starts out highish from the lower output levels but tends to stay stable, that is rise only slowly.

@Biodegraded

D2 absolute


D3 absolute

 

Indeed, however I don't think this software will do that. I'd have to write that custom.

 

I'm going to see if I can swing a 5128B head.

I like what @Resolve has proposed and I plan on following his lead:
https://forum.headphones.com/t/developing-a-new-headphone-reference-target/20094/34?u=purr1n

The gist is that dummy heads or head/ear rigs will come with DF comps. Thus headphone FR plots should have DF comp applied. What this allows is comparable results between GRAS (whatever 45 variant), B&K 5128, or whatever GRAS will come out next to compete with B&K, China clones, etc.

The resultant DF comp can then be measured against various targets. For example, the 1db/oct slope (typical FR at listening position at mix stages or speakers flat at 3M, placed in a living room with furniture). Variants of this could be B&K AES1970 which is fairly flat to 200Hz with then 1db/oct slope. Other targets could be Audeze house sound. And of course Harmon 2013 target (who Resolve translated to DF comp). Heck, I could even provide a translation of the flat plate coupler to DF too.

Point is cross comparability. Sure there will be variations, but it won't be like Crin taking raw from whatever he's currently using to Crin target, Amir using raw from his specific GRAS targeting to whatever Harmon year, Oratory comping to Harmon target, whatever RTings is doing, probably Harmon 2013, etc. It's basically all over the place and no one can compare anything.

All plots compesated to DF. Then then targets are easy. I could draw a dark line on the B&K in my FR plots because that's the studio neutral reference which I prefer. But I can also draw lighter lines for Audeze house sound, Crin target, Harmon 2013, Harmon 2017, Harmon 2018, Harmon 2023, Harmon 3022. Then basically everyone can decide for themselves what target they would like to use for themselves,

P.S. People are dumb and a few are already up in arms over "Resolve's Target". OMG, the gaul of going against Dr. Sean Olive P.h.D. and his peer reviewed (it's not) research, when it's not even anything like that at all!

 

Yay, figured out some stuff on the AP.

SW51
300-ohm load
90kHz bandwidth
low-Z output
Distortion in dBu (absolute)

H2


H3


H4-10

 

0dBu distortion at the 0dBu output level means 100% distortion.

Likely function of transformers that just prefer certain areas to operate in, "sweet spots". The increased distortion at the ends at higher output levels is typical for transformers. @Zampotech jammed these in using special techniques.

Sure. I may try one more thing first: try THD% instead of dBu for distortion. It's a bit easier for amps to do this compared to headphones.