The study is entitled: A Preferred Headphone Target Curve Defined for the B&K 5128 and GRAS45CA-10 Test Fixtures. And can be downloaded from the AES website here…
1 Like
There was a Head-Fi thread on this topic with Dan Clark in there as well.
To be honest, I feel like target curve chasing like this is a bit like chasing ghosts. Maybe I don’t understand the science/didn’t read it too carefully but the part that sketches me out is the use of a few headphones to create the transfer function and then extrapolating that all headphones can fit in this paradigm with these rigs that have more than a few differences. And you know what they say about extrapolating…
3 Likes
We discussed this study at length in one of the episodes of the Noise Floor. The discussion starts here: https://www.youtube.com/watch?v=Cmz3pSQUgHA&t=40m52s
The biggest takeay from all of this IMO is the slide Dr. Olive posted on social media, which didn’t seem to make it into the study from what I could tell.
His further commentary on the implications of all of this on target curves is something I thoroughly agree with, so I’ll post that here as well.
All of this really just depends on what your goals are. If it’s to preserve the ‘known good’ target curve evaluative methodology, the conclusions drawn from this study are generally reasonable. But if the goal is to understand how these devices actually perform across people, and to understand what people actually prefer, the conclusion here should be the stuff Dr. Olive is posting about above. And really, that the single line target curve methodology makes no sense given the response variation across heads.
Important to all of this is also the recognition that even the industry standard GRAS systems that people currently use, like the 45CA or 43AG don’t use the same pinna that was used in the Harman research. Those who focused on those system with the hardcore “it must match harman” ethos likely assumed that headphone behavior would trend similarly across the different ears being used, but that’s not the case. Keep in mind, for the largest part of the research, they used a K712 Pro, which isn’t as low acoustic z as would be ideal anyway.
So… IMO this is yet further debunking of the more militant target adherence evaluate practices we see in various places. It probably felt good to give simple answers to complex problems, but in reality they were the wrong answers, they just didn’t know it.
2 Likes
This. So much this.
I don’t know how many times Sean has to tell people to stop extrapolating the conclusions from this testing to circumstances outside those that were tested, but uh… this study confirms rather bluntly that using a smoothed, EQed headphone response (which is what every Harman target functionally is) to judge other headphones is not something we should be doing.
2 Likes
Appreciate the replies and additional info on this from the three of you. There are already threads on this in a few different headphone forums, and it seemed like there should be one here as well.
There is alot to digest in this study. I’ll give the Noise Floor and other info above a look though. And then maybe will have a few more things to add on this.
I agree with many of the points made so far. However, I still think that generic targets can be useful, if they’re done right and applied right. And I don’t think that differences in acoustic impedance or individual measurements necessarily make them obsolete. The difficulties translating the original Welti target to different rigs in the study is interesting though.
I’m still sort of trying to find the new generic targets for the GRAS 45CA-10 and HBK 5128 mentioned in the title of the study. And suspect they are the KB5000_7, KB5000_20, 5128_7 and 5128_20 curves shown on both graphs in Figure 2.
As I’ve said many times though on this forum (and elsewhere), the primary reference curve I’m interested in is one based on in-ear measurements of neutral speakers in semi-reflective rooms (like in your home). And would hope that some of those could also be included in any future studies.
Rather, it’s that it makes the way people typically use these conventional target curves erroneous. When people are correcting their closed-back headphones to perfectly match the Harman target, that’s simply not the sound they’re actually getting, even though they think it is. Moreover, with the HpTF behavior differences we’ve been talking about for ages now, the target being used isn’t even yielding people the actual Harman target - they don’t know what that actually sounds like.
So if you use it as a way to disqualify the truly awful products out there, it’s going to do a good job of that. But that’s also not the typical way people in places like ASR tend to use it.
2 Likes
This explains why Harmon the eq settings avaialble on various sites never sounded right to me.
I couldn’t resist my curiousity and tried some once and it was kind of a trainwreck. I couldn’t say what seemed wrong other than “this doesn’t sound right”.
2 Likes
Yeah, even if people actually had the Harman target equivalent for the GRAS systems, that’s only going to be consistent for open back headphones.
1 Like
Unfortunately, not even “open back” is a sufficient enough condition for reliable replication. The AKG K712 Pro used as a replicator for the study seems considerably higher impedance than other open back headphones based on the deviation between the two rigs.
The Harman target is only really replicable in the condition it was tested in, which means you need an AKG K712 Pro with a modified headband and an EQ profile correcting its deviation from the target on the Welti pinna-equipped 45CA with RA0045 ears.
This is yet another reason why basing target/baseline methodology on Diffuse Field with a range of preference adjustment represents a much less asterisk-laden approach to both judging headphone responses as well as EQing them. The target being a load-sensitive condition (Harman being an EQed headphone response) instead of a load-insensitive one (speakers at a distance) introduces too much opportunity for earfield scrungles.
2 Likes
I agree with many of the points made so far. However, I still think that generic targets can be useful, if they’re done right and applied right. And I don’t think that differences in acoustic impedance or individual measurements necessarily make them obsolete. The difficulties translating the original Welti target to different rigs in the study is interesting though.
This is totally fair and I agree - I don’t think our newer understanding of measurements invalidates the original Harman target. If anything, it’s more of a reminder that research outputs are to be taken within the context in which it is presented. In other words, the community as a whole probably should’ve been a little more careful in taking the original Welti target and applying it to different rigs. TBH I myself just learned that this was the case. I’ll admit I’ve historically taken frequency response visualizations for granted and with many grains of salt in the first place, so I never looked too closely into it.
Also, I should probably clarify that my joke about extrapolations is a little facetious. In a way, the reason we do research is so that we might be able to extrapolate its results - we just have to be conscious that our samples (i.e. the chosen headphones) may or may not be properly representative of the population (i.e. every headphone that we measure later).
The nice (or not-so-nice thing) is that I think in general, what we’ve learned is that compared to pure DF 1) Getting the shape of the 2.5 - 3 kHz pinna structure right is pretty important; 2) People like a bit of bass; 3) People want to tone down the treble. I think it’s nice because it opens up a range of possibilities beyond strict target adherence which allows me (selfishly as a reviewer) to look at it from the perspective of the headphone’s overall presentation and how it colors the music in a way I like or dislike. I think this is a part that gets overlooked in the Harman research - while the Harman target was overall preferred, there were a few other targets that statistically tied it.
The not-so-nice part is that we don’t get any strict and easy answers. But I feel like that’s a major part of the hobby. To struggling and think about your music and enjoyment and come to terms with what you really like. Ultimately, I think this is what separates this hobby from other tech-related ones where you can read off a spec sheet. Not to get too poetic, but it provides another avenue to finding a part of your own humanity.
1 Like
The Harman target has always sounded slightly off to me as well. Some of that could be due to translation issues. But I think there’s probably a bit more to it than just that. And I’m not quite ready to hop on board with the new versions of the curve yet for that reason.
I’ve always sort of viewed the Harman target as a signpost pointing in the right direction. But not as the final destination.
I think the broad conclusions of the research are probably correct though. That most people are likely to prefer a response similar to the in-ear response of neutral speakers in a typical semi-reflective room.
I appreciate your perspective on this, Fc-Construct.
Yes, and no, I think. I was curious to see how closely some of the new 5128 curves match a sound power slope though, after compensation with HBK’s original 1/3-octave DF. And can post a couple graphs showing this if anyone’s interested. And if it does not violate any forum rules. (I made my own 1/3-octave plot of one of the new curves for this purpose.)
The 5128_7 curve is in the general ballpark. But there is still a fair amount of unevenness in the curve imo. It looks overly bright in the 5-7k and 1.5k ranges. And there’s not enough warmth around 200 Hz in the bass imo. These features make the curve a non-starter for me. And I think the dip in the bass is mostly an artifact of the 3-knob shelving approach of Harman’s preferred in-room curve.
Some folks who like a brighter V-shaped response might like this target. I think it’s probably a bit too close to the default sound sig of my Beyer DT-770 though, in some ways. Haven’t done a line regression, but I’d estimate the average slope is in the -1.0 to -1.2 dB/octave range with the stock 1/3-octave DF compensation.
Compensating the 5128_7 curve with Oratory1990’s revised 5128 DF produces a flat DF, with a >6 dB sub-bass boost, some rolloff in the high treble, and a rather sizable notch at 3k btw.
I mostly agree. Though I think many people simply use the Harman target as a starting point.
Perhaps. I suspect that there are a few headphones out there which are fairly close matches to the updated targets in this study. My question though is whether these new targets sound neutral.
The more I look into this, the more I’m convinced the folks at Sonarworks approach regarding EQ is as good as any other approach. I can report that the EQ curves for the Sennheiser HD 600 series do provide improved playback.
1 Like
Some headphones with responses similar to the 5128_7 curve in the study…
AKG K371
Has a slight darker overall tilt. And less energy around 4-6k.
Astro A50X
Has less energy around 4-5k, but otherwise very similar.
AudioTechnica M50X
Overall levels are somewhat similar. M50X has less energy around 6k and 320 Hz. And more energy in the bass around 125 Hz.
DCA E3
Has less air at 16k, and less energy around 5-6k. There are also some minor level differences in the mids between 700 Hz and 4k. But overall its somewhat similar.
DCA Noire X
A little less energy around 5k, but otherwise pretty similar.
HFM Sundara 2020
Has less energy around 6k, 2k, and the sub-bass because it is open. But the overall levels are somewhat similar between 100 Hz and 20 kHz.
HyperX Cloud IIIS
Not as much energy around 4-7k, but otherwise pretty similar.
Sennheiser 490 Pro - Mixing Pads
Very similar in mids and treble. But more rolled off in bass due to open design.
Sennheiser HD550
Has less sub-bass due to open design. And a slightly darker tilt in the mids and treble. But otherwise similar.
Sennheiser HD560S
Very similar in mids and treble up to about 11-12 kHz. 560S has less air above that, and less sub-bass due to open design.