Missteps in headphone science? (5-31-25 Noise Floor Disc.)

There was alot to chew on in the most recent Noise Floor livestream. Much of it was more on the esoteric side. And some I agreed with and some I didn’t quite.

I think Harman should have used diffuse field compensation for their measurements and target response curve. Because this would’ve made the results more broadly applicable. There were several problems with this though.

The biggest is that the rig that they used for most of their studies did not have a head or torso that could be measured in a diffuse field. I assume that they chose the GRAS system over some other options (with heads and torsos) because it was more affordable, and potentially more accessible to a broader swath of users and developers in the audio field. A DF curve would’ve been an invaluable tool though for translating their work to other measurement systems.

I think they also believed there was enough similarity between the different types of rigs that a raw target or response curve developed on one system would generally be applicable to others. This is understandable because most of the B&K, GRAS, and Head Acoustics rigs that were in use at the time were based on 711 couplers, and were also compliant with other standards for measurement. The shapes of the heads and pinnae were all somewhat different though. And they all had somewhat different resonant characteristics. So the raw target translated better to the systems closest to the one Harman used in their research. Namely, the GRAS. And less well to the other systems.

The other reason is that they had already ruled an unmodified DF curve (without a slope) out as a target preference curve. And probably did not want previous unsatisfactory targets like DF or FF to influence their final results.

The lack of DF compensation impaired the broader adoption of their target though.

1 Like

Much of the discussion in this livestream also focused on some of the minutiae of the Harman research, and how it has been misused or misinterpreted. If you focus more of your attention on the broad strokes though, then it’s less of a problem.

In broad strokes, all Sean was really trying to do was replicate Floyd Toole’s studies on speakers for headphones. And to determine what type of response people most preferred: DF, FF, or something else maybe a bit closer to the in-ear response of speakers in a room.

The latter idea had already been floating around in the ether when he began his research, most notably in the work of Paul Barton and his concept of “RoomFeel”. So based on Barton’s work and Harman’s previous work on speakers, it wasn’t a huge leap to assume that preferences would likely fall somewhere along those lines. If people preferred neutral speakers (with a flat on-axis response) in a semi-reflective room, then why not the same type of sound in a pair of headphones? … It just made sense. And for all intents and purposes that’s where the Harman headphone research has landed us.

There are plenty of flaws and shortcomings in the research though. And if you try to read too much into those, it will just lead you deeper into the woods and trees, so you can’t as easily see the forest.

I think more preference research would be useful. But I don’t think it’s the only way to go (as implied in the livestream). This video by JJ Johnston gives some insights as to why this is the case.

I’ll summarize my takeway this way. If you don’t know the answer to what you are looking for, then preference is the best way to go. If you already know the answer, then you can test instead for “accuracy”.

Imo, the Harman research gave us the answer to what most people prefer. Namely, the in-ear response of neutral speakers in a semi-reflective room. And now we can simply test headphones for how well they approximate that type of response.

In the livestream it was also suggested that the Harman target, or any target with a bass boost is inherently colored, because it does not match the response of existing HRTFs. And this is why we have to use preference.

It’s an interesting idea. But if the HRTF is based on the sound field of neutral speakers in a semi-reflective room, and that is also your target in-ear response for headphones, then there is no difference in coloration between them. They are both for all intents and purposes neutral.

1 Like

Well… I kind of see this more as a semantic platitude, which is sort of the point. Anything could be defined as ‘accurate’ or ‘neutral’ if it matches what it’s being referenced to.

But several things to consider there as it relates to preference, there’s quite a massive degree of variation in bass level preference, so depending on the segment you’re going to have different results. Moreover, when you look at the actual results of the responses people preferred, there’s a range of preferred bass to treble delta, and with varying ‘coloration’ throughout.

For the broad strokes concept, I tend to agree that what we got is reasonable, but that doesn’t mean this paradigm can’t be improved upon.

I mentioned this on the stream but I want to reiterate… many of the things we may have identified as ‘mistakes’ are really just a matter of what could’ve been done differently in hindsight. I don’t think any of us blame the researchers for doing things the way they did. It’s more of a wishlist for what we would’ve wanted to better future-proof it. The exception there might be to do with using the K712 Pro, but that also may have been mandated by someone else.

1 Like

Thank you for the reply, Resolve.

I take your points. And there are some semantics involved here. “Coloration” needs some kind of reference though. In HDTV, the standard was Rec. 709. HDR UHDTV uses Rec. 2100. And I think most computer monitors still use sRGB(?).

DF HRTF is not a good reference point in this case for defining neutrality or the absence of coloration in headphones, because most people find it unnaturally thin and bright (without some kind of tilt, slope, or bass-boost added).

Sean’s goal was to find a new response curve for headphones that sounded more neutral. He wanted a pair of headphones (like his Harman-tuned loudspeakers) that would not add any obvious coloration to his music when listening on headphones. And I think if you ask him, that is precisely what he believes he got from the Harman headphone research.

That doesn’t mean that everyone will agree. If you’re losing some of your higher frequency hearing, then you might prefer headphones with a little more treble. Or if you’re used to listening to the hopped up bass in your car, then you might also prefer headphones with more bass. Your listening habits and conditions might affect your preferences as well. Some people also like their TVs with more blue, or red, or green added.

What the Harman research gave us though was a common reference point or baseline to work from, that was better than the previous DF and FF. And most of the other preferences tend to fall somewhere around or close to that new reference point.

I think the paradigm can also be improved upon. But I think one of the better ways of going about that is to do some more comprehensive in-ear measurements of speakers in semi-reflective rooms. More preference studies would be great too. But in-ear measurement of speakers should also be part of the process imo.

Fixed my typo on “semi-reflective speakers” btw, and made a couple other minor edits for better clarity.