Contravert perfidiousness: Eschew obfuscation!
3 Likes
Iām finding it difficult to rank them well enough to narrow it down to the top 2 so please allow me instead to attempt to hit the main points as I see them instead. Apologies to the original authors but by definition this means it will not 100% represent all of the originally intended information - not even close.
- How can we say that headphone X scales when moved from amplifier A to amplifier B when amplifier B has a lower SINAD?
- Past a certain point SINAD gets to āgood enoughā and does not entirely capture the quality of sound an amplifier produces anyway. No one number does. Many more numbers help (graph) but either thereās something we arenāt measuring yet (or canāt measure yet) or we donāt know how to interpret the measurements sufficiently to accurately predict sound quality. Now that Iām thinking about it, this could be a limitation of measuring the amp, the headphone, or both.
- An amp can measure with higher distortion but sound subjectively better, especially of the source sounds bad in a way that the distortion mitigates. A tube amp is the classic example although I donāt think it was mentioned.
- Measure all you like. Itās not a waste of effort but no amount of measurements is a substitute for listening to the equipment in question. Human perception can be fooled and think that equipment A is better than equipment B when objectively it is not. However itās better than drawing conclusions from just the measurements because of the issues with measurements above.
- What do we mean when we say headphone X āscalesā anyway? How can we discuss scaling when we canāt precisely define it and how can we compare scaling when we canāt measure it?
- It just means it sounds better with better amplifiers. Donāt overthink it. Itās a subjective quality so it is not required to be measurable.
5 Likes
Understandable. Thanks for compiling this summary, which I will noodle over.
2 Likes
This is why I ignore THD data since going by them the ER4XRās midrange should sound off from it being 0.8 ~ 1.6% on most units. The ER4XR sounded just as clean/detailed to a ER4SR I had that rated <0.3%.
I wonder if this incudes post EQ as well since I had no issue doing EQ on the ER4XR/ER3XR?.
2 Likes
numbers, numbers
golden slumbers audio crumbles resolution tumbles
suck your thumb personal grumbles
Itās all the same
Itās all different
numbers, numbers mumble mumble toil&trouble
flutter wow and table rumble THD a BFD
maybe not itās just a stumble
numbers, numbers
9 Likes
Very impressive!
2020
1 Like
Eminem, is that you? 
3 Likes
LOL
Will the real slim shadystac please stand up?
3 Likes
@Tchoupitoulas, get rid o dis, no shadystac you smokin crack? And slim slim, trim and dim, was fat not fat but round like that. @deafenears it appears M&M has years and years of proven talent, peanut plain and caramel rhyme, better beat, reference sublime.
MY DOGGERAL IS CHASING THE CATERAL.
4 Likes
More importantly, this stuff also isnāt completely subjective - even though thereās absolutely room for preference. The limitation on subjectivity is the fact that we all generally have ears. Even though there may be some variation on how people want their music to sound (or potentially even how things are heard), there are certain facts about human hearing that need to be taken into consideration. Moreover, we also know the kind of balance that most people prefer, along with what the other groups prefer. So while that doesnāt say anything about how individual people are going to enjoy things, we can at least reliably predict how itās going to be received by most people, and this gives a strong foundation for being able to evaluate headphones as well.
6 Likes
Fletcher Munsen curves/Equal loudness contours would be a good topic for a video when it comes to setting EQ. Iām guessing that Harman had some method of standardisation but many listeners embarking on this kind of thing will dial in what they think is goodā¦ Then, one night, they feel a bit saucy and turn it up and all of a sudden it is āoh no this sounds badā. Basically people need to have different eq for different volume settings due to the way our frequency curve preferences change as we crank that nob. otherwise they might end up erroneously blaming other stuff. If anyone out there could design an output EQ that compensates for such phenomenon without being destructive, thatād be great
4 Likes
To me the problem isnāt objective or subjective. I think you made a valid point that while Harman research came up with a curve that represents roughly what most people in their survey found likeable, you canāt predict what any individual will like. But there are further complications that Sean Olive freely admits. Headphones need to form a tight seal or the userās perception of the musicās frequencies will vary but there is no objective way of determining how much leakage exists on your or my head. When earphone placement on oneās head varies even a small amount oneās perception of the frequency curve can vary significantly. Iāve been experimenting with this recently and it amazes me that movement of a few mm up, down, left or right will significantly alter sound stage, detail, timbre, etcā¦ And then KMann from Audeze has pointed out a study which showed that there is almost no correlation between what a sophisticated test dummy records at its test dummy eardrum and recording apparatus placed at the eardrum (how do they do this???) in live humans. And to further complicate things, KMann pointed out that even the most sophisticated measuring dummy is inaccurate below 100hz and above 10k. To even further complicate things there was a fairly wide range in preference of Ā± 4db in the bass region from the āHarman Curveā in their study population. And lastly the Harman research points out that musical preference has an impact on what curve people in their study group preferred.
What annoys me is the reference to the Harman Curve as scientific. Humpty Dumpty famously told Alice āwords mean what I say they meanā so if Sean Olive wants to call his work scientific and I thought in your interview with him he scoffed at āunscientific approachesā thereās nothing I can do about it except to ask anyone who calls it thus to define what they think science is. A better word for Harman research is consumer research, like coke v pepsi. All this to say there isnāt IMO much that is objective in this research.
You only have to read through this forum, Head-fi, SBAF and others to see that there is huge disagreement among headphone listeners. Some extol Abyss 1266 while others scoff. Many love Susvara and others sell them. Some find Empyrians great while others think theyāre over rated. While I like the idea that some reference curve can be a starting point to help zero in on what headphones someone may like I think the evidence is that things just donāt work that way. Again as you pointed out, the Harman Curve may be a statistical indicator of what a select population may find pleasing, headphones arenāt purchased by populations, theyāre purchased by individuals. And as such consumer research curves are a hit or miss proposition in helping any individual in selecting their next pair of HP.
Woa. 
3 Likes
I donāt want to assume you havenāt read the research, but Iām curious as to which study in particular youāre referring to with the suggestion that it isnāt scientific. Because, one of the key things with the research is how well the various different studies validate and corroborate the results. I agree that there are some limitations there, but much of that is in some of the more peripheral studies where their goals were slightlyā¦ adjacent to the key pieces.
Moreover, I personally donāt find the research to be exhaustive, but the hill I will die on is that this stuff is absolutely within the realm of scientific inquiry. I think the difficulty some people have with it is that it involves preference, and that sounds more like an opinion poll, but thereās so much more to it than that.
With regards to the other study mentioned, Iāll have to go back and check out what his conclusions were on it, but in my view we shouldnāt expect a DRP measurement to be the same between a HATS and on an actual human, the HRTF wouldnāt be exactly the same as a given individual either. Butā¦ this doesnāt throw any doubt into the spokes of the Harman research as far as I can tell.
In any case, yeah the issue of individual vs the masses is a tricky one that anyone reading graphs is going to have to recognize. Like, itās not there to indicate what you will like, but rather what most people do like - itās descriptive not prescriptive (although the statistical model paper does try to predict listener preferenceā¦ I have some issues with the model, but at no point does it say you should like this). This is also why itās extremely useful to help evaluate headphones in relation to. I think the wrong step to make is the one that says any and all deviations from this target are bad, thatās missing the forest for the trees in my opinion. Instead, all that means is that itās not going to appeal to the Group 1 as much as something else that doesnāt deviate is likely to.
But as far as the connection between the research and a ācoke vs pepsiā type of study, itās a question of whatās prudent for companies to aim for when it comes to tonal balance and sound quality in headphones, and even Harman companies like JBL will go for tunings that deviate in the bass because theyāre going after a different audience for example. And, as we know, there are all kinds of other ways companies can sell headphones that have nothing to do with achieving the sound quality that most people prefer (Beats/Apple seem to have good marketing strategies, as do many audiophile companies). Soā¦ in my view, sure it can be used like this - but itās also the best we have for an objective reference point with the conditions of 1/3rd octave smoothing and the KB5000. Maybe weāll have 5128 research soon with higher res results, that would be cool.
7 Likes
This isnāt quite true - you can (and @Resolve does, as do some others) measure at your own ear entrance with a microphone. RTingsā tests of headphones use in-ear microphones on a number of peopleās heads and the bass they plot is an average, for example.
This is really overreaching the conclusion of Christensen et al 2013. There are narrow (high Q) variations in high-frequency HRTF and HpTF between individuals, sure - thatās meaningful when we start looking too closely at very thin spikes or, particularly, dips in headphone response relative to our target response. However there are quite meaningful trends in headphone response in the most significant areas of FR. The mannequin used in Christensen et al 2013 is also a non-standard, aftermarket modified head, so itās not entirely clear that its HRTF matched the population average to begin with. HammershĆøi & MĆøller 2008 also compared headphone measurements on mannequins and human ears, and found relatively strong agreement relative to the diffuse field HRTF.
This is true - in an acoustic impedance sense - only with current mannequins. The BrĆ¼el & KjƦr 5128 has accurate ear impedance from 20hz to 20khz, although impedance is less significant than positional variation at high frequencies.
The Harman research underpins a pair of statistical models which quite consistently predict listener headphone preferences. If your definition of scientific research doesnāt include empirical work which produces functional predictive modelsā¦well, thatās gonna exclude a lot of science.
10 Likes
Thanks for your and Resolves lengthy reaction to pick apart my rant. Iām not driven enough to respond to the various points because its not the technicalities that really interest me though I do feel there are plenty of points of argument to be made but Iāll leave you with the last word.
I believe that the Harman research is fundamentally aimed at manufacturers saying āthis is close to what many people will like, this how you should try to tune your HP products. Itās clear that Samsung owned companies apply this to some degree at least. Itās also clear that despite knowing about the Harman curve many HP manufacturers choose to deviate from it or just seemingly ignore it in tuning their HP for reasons of their own which donāt include ignorance despite what the model says they should do.
But for me the point where we really disagree is how much this research and āthe curveā are useful to a buyer setting out to upgrade his/her current headphones. You guys feel like its a useful starting point to help limit the myriad of choice down to a manageable size. Look at the HP FR curves, compare them to Harman and know how your own taste is similar or differs from Harman. Use that to narrow down your options. I think Sean Olive would say from his research donāt trust reviewers but I suspect at least Resolve, who is an HP reviewer wouldnāt go that far.
I think many audiophiles including myself just feel that the statistical model and oneās known preferences in relation to it is cumbersome and unnecessary and prefer to just read/watch reviews and try to focus in on as many reviewers at review sites and forums who share your musical taste and have similar equipment to see what they think about a specific HP. I believe this technique gives a much more nuanced view of any individual HP than a statistical model and its what everyone does in the end anyway. For myself I canāt look at a FR graph and think, boy am I going to like those headphones on the final crescendo minutes of the finale of Mahlerās 1st symphony or let me forget for a moment that Iām listening to headphones and find myself sitting in the Pawnshop in Stockholm listening to Arne Demnarus.
Is this what people are doing? I have never looked at a headphone frequency response graph before purchasing a headphone. I usually just listen to what Zeos Pantera has to say in his videos and then go out and buy the headphone. The only other time that I have looked at a headphone frequency response graph was when they were used in videos produced by āThe Beyerdynamic DT990 Guyā aka Nouvraught.
Actually that brings up an important point. Far too frequently, people will evaluate headphone measurements without fully understanding how the measurement systems being used function. And, to be perfectly honest I canāt blame them.
Certain things look compelling, and without the context and full understanding of the equipment, itās bound to lead to some wild judgments. Even with the more basic stuff like the KB5000 concha feature at 9khz - those of us conducting the measurements take this for granted, but itās bound to look weird to anyone who isnāt aware of that feature. Or another example would be the fact that the target is coarse grained and the FR results of headphones is usually much more fine-grained (like 1/12th or 1/24th octave smoothing vs 1/2 or 1/3rd for the target).
You can also take this to a broader context of DIY measurement rigs or the various differences that exist even among the 711 couplers. I think Judeās recent video on this underscores the point. This is also why I think people will sometimes prefer to just watch the more concise measurement-free takes, and I certainly canāt blame them for that.
2 Likes
I donāt want to put words in their mouths but, I imagine theyād say that itās a good idea if the manufacturer wants to make stuff that sounds good to most people. But from the research itself, I donāt get any of what we could call the ānaturalistic fallacyā going on.
For me the key aspect about the target itself is that it achieves a good balance between fundamental and harmonic - and I have to imagine that part of this comes from the in-room target. Maybe some people have ears such that they donāt hear it this way (Iāll leave room for that, but itās likely to be close), but in my mind when it comes to the ear gain region, some sort of balance thatās similar to what the target achieves is desirable and even āneutralā, without any particular harmonic dominating the range. People can prefer less ear gain, more ear gain, more bass, more treble whatever, as long as that balance is kept reasonably intact, at least if the goal is for instrument tones to sound somewhat normal to how they sound in real life. If thatās NOT the goal, by all means go hog wild.
2 Likes
I was curious as to peopleās insistence on HRTF variation so I did some research, and while it isnāt a super well studied subject, the only significant large-scale variation I found was that many listeners in Southeast Asia, especially Japan have a different enough hearing curve in the bass that there is a specific Equal Loudness curve variation on fletcher-munson in use in Japan. It has appreciably less bass than the 40 phon F-M curve. I could theorize about why this is (less bass instruments in traditional forms of music?) Many Japanese recordings are brighter and thinner and many Japanese horn speakers are quite treble emphasized.
That said, I didnāt turn up anything to support the idea that people hear in some radically different way outside of anomalies here and there. Largely, human HRTFS and ear curves fall within a broad average and tend to equalize to within a broad average. Preference can be a good individual guide, but no one personās personal preference is necessarily of importance in large studies.
4 Likes