Can you tell which headphone is more compliant to Harman Research?

[IAtaman]

Below you can see two raw FR graphs of two separate headphones, measured with the GRAS measurement rig that was used in the preference research, plotted together with Harman AE/OE target (green) alongside the error curve (red) showing how much they deviate from the target for each frequency.

Question is simple : in your opinion, which of these headphones are more preferable according to Harman research?

There are no points for getting it right or wrong, so would appreciate if you could vote, and vote as per your best guess please

Headphone 1

Headphone 2

If you have access to the A Statistical Model to Predict Listeners’ AE/OE Headphone Preferences paper, I'd appreciate if you didn't vote in the poll please.

Thank you.

 

[IAtaman]

You can only see the results after you vote, so that your vote is not influenced by what most people think.

And this is not a test of your ability to read FR graphs Have a look please , and let me know what is your take by voting in the poll.

 

[preload]

I've read that paper (and series of papers) ad nauseum so I will not vote, as you had requested.
I'm curious if you're trying to make the point that folks can't reliably predict headphone preferences simply by "eyeballing" the FR curves.

 

[Bugal1998]

Below you can see two raw FR graphs of two separate headphones, measured with the GRAS measurement rig that was used in the preference research, plotted together with Harman AE/OE target (green) alongside the error curve (red) showing how much they deviate from the target for each frequency.

Question is simple : in your opinion, which of these headphones are more preferable according to Harman research?

There are no points for getting it right or wrong, so would appreciate if you could vote, and vote as per your best guess please

Headphone 1

View attachment 334972


Headphone 2

View attachment 334973

If you have access to the A Statistical Model to Predict Listeners’ AE/OE Headphone Preferences paper, I'd appreciate if you didn't vote in the poll please.

Thank you.

I was contemplating possible answers, when I submitted a response on accident. That said, I could think of possible reasons that would lead me to conclude any of the three answers could be correct.

Translation? I don't know the answer.

 

[IAtaman]

I've read that paper (and series of papers) ad nauseum so I will not vote, as you had requested.
I'm curious if you're trying to make the point that folks can't reliably predict headphone preferences simply by "eyeballing" the FR curves.

I am not trying to gauge anyone's ability to eyeball. Allow me to elaborate later, once the poll is closed please. It is not anything smart really.

 

[IAtaman]

33 people voted so far - thanks to all who did.

If you haven't done so yet, just have a look at the graphs and decide which one looks more complaint to the Target please. No math(s) or calculation is required

 

[OnLyTNT]

Intriguing, just voted. By the way, is the paper worth reading and publicly available (guess not)?

 

[IAtaman]

Intriguing, just voted. By the way, is the paper worth reading and publicly available (guess not)?

Thanks for voting.

Yes, it is worth reading and yes, it is publicly available but not for free. You need to be an AES member or pay them I believe $30 or sth to get it.

 

[phoenixdogfan]

I don't like the way the curve on headphone 1 is normalized. I thing you could and should move all points on the curve down 2 db so the 800-2k region hugs the standard. If you do that what's revealed is a phone that slopes off it bass starting around 200 hz, has a 2db scoop out of its mids from 3k to 5k, and then sees a precipitous drop in it's highs after 6k. It's definitely the less Harman compliant of the two phones.

 

[IAtaman]

I don't like the way the curve on headphone 1 is normalized. I thing you could and should move all points on the curve down 2 db so the 800-2k region hugs the standard. If you do that what's revealed is a phone that slopes off it bass starting around 200 hz, has a 2db scoop out of its mids from 3k to 5k, and then sees a precipitous drop in it's highs after 6k. It's definitely the less Harman compliant of the two phones.

Appreciate you sharing your thought process. I'd be more than glad to address your comments in a few days after the poll is closed.

 

[IAtaman]

54 responses so far which is great, and one more day to go.

 

[preload]

I am not trying to gauge anyone's ability to eyeball. Allow me to elaborate later, once the poll is closed please. It is not anything smart really.

The results of your poll seem to suggest that people aren't terribly reliable when it comes to "eyeballing" curves. Curious what your original intent was, however.

 

[olieb]

That said, I could think of possible reasons that would lead me to conclude any of the three answers could be correct.

Same here, so I "voted" equally preferable (as compliant with Harman) but the vote is already closed of course and I will see what is the verdict.
My guess is, I would prefer HP1 as a listener.

 

[Bleib]

HP1 needs less EQ, in the bas region. This is easier to get correct. Human hearing is the most sensitive higher up, mostly in frequencies where human voice dominates

 

[IAtaman]

Thank you to everyone who voted.

Allow me to first correct the percentages, albeit a minor correction, by removing myself from the votes. FYI, I voted so that I can see the voting results, and voted equally preferable not to affect the votes in any way, not sure whether it matters.

If I remove my vote, this is the percentages we end up with.

Which of these headphones are more preferable according to Harman research?
Headphone 1	24	39.3%
Headphone 2	33	54.1%
Equally preferable	4	6.6%
Total	61	100.0%

I have taken these graphs directly out of the research paper, for criticism, comment and reporting purposes, which I am hoping is inline with fair use policy.

They show two of the "average frequency response and error curves for headphones in four different categories based on their preference rating".

Headphone 1 is the category Good, and Headphone 2 is the category Fair. I have removed the text from the graphs for testing purposes. Here they are, untouched.

So, people who thought Headphone 1 is more preferable were correct. FR of HP1 is significantly more preferable according to the research.

And yes, there is indeed a typo in the paper that was missed. Level is written as Lwcwl.

I thought of doing this because while I was reading the paper, I was surprised to see Good was rated higher than Fair as Fair is tracking the target much better in the 500 -1000Hz range, and it is not deviating from the target too much on highs and treble either. (Please keep in mind the range that is correlated with preference is 50Hz - 10KHz range so deviations below 50 and above 10K is not found to correlate with preference)

Error curve for fair is also much closer to zero than that of Good, so how come Fair seems to be closer to the target yet it is less preferable?

I do not have a clear answer for this yet. Maybe it is a plotting error? Maybe it is our inability to eyeball curves? Maybe something else is in play?

I do not know and I was hoping maybe we can think about this together and figure it out maybe.

 

[IAtaman]

As I tried to explain, when I saw these two graphs, I also thought the second one would be more preferable as it is tracking the target much better in the 500 -1000Hz range, and it is not deviating from the target too much on highs and treble either. Plus, the error curve for the second one is also much closer to zero than that of the first one.

For those who thought the second one was better, what was your thinking?

 

[OnLyTNT]

I chose FR2, because more compliance with the target which led me to think it would be more preferred.

Those graphs reminded me 7Hz Zero vs Truthear Zero "a bit", particularly around bass region, I am not a bass head but I prefer Truthear for daily casual listening and these both products are revised with updated FR response favouring a bit darker taste . Both 7Hz Zero 2 and Truthear Zero Red (with adapter) FR looks like a combination of FR1 (treble region) and FR2 (bass region) which make sense in my opinion.

 

[IAtaman]

In a way 33 people / 54% made the expected choice and thought, incorrectly, HP2 is more compliant, like I did.

What about the people who chose HP1, what made you choose HP1 over HP2?

 

[TurtlePaul]

I chose headphone 1 but this was tough.

HP2’s error looks distinctly V shaped, which I don’t like. Particularly that midbass boost should leak into lower midrange and impact tone (up to 300 hz!).

Visually HP1 has a lot of its error in low bass and above 10khz which matters very little. But it hugs the target in the midrange. If you set the dashed trend line 2 dB higher it would be obvious how good HP1 is but it would look even more bass deficient.

If you look at the bottom graphs there is ‘parallel lines illusion’ which makes the 3khz+ shortfall of HP1 look much worse than the normalized graph.

 

[isostasy]

I chose headphone 1 but this was tough.

HP2’s error looks distinctly V shaped, which I don’t like. Particularly that midbass boost should leak into lower midrange and impact tone (up to 300 hz!).

Visually HP1 has a lot of its error in low bass and above 10khz which matters very little. But it hugs the target in the midrange. If you set the dashed trend line 2 dB higher it would be obvious how good HP1 is but it would look even more bass deficient.

If you look at the bottom graphs there is ‘parallel lines illusion’ which makes the 3khz+ shortfall of HP1 look much worse than the normalized graph.

@IAtaman this was my thought process for picking HP1 too. I imagined the dashed trend line a few dB higher as described and then guessed the large excess of bass in HP2 might bring the score down lower than the shortfall in sub-bass of HP1.