ADUHF
Headphoneus SupremusMusic And Measurements
Is your closest achievable neural more neutral than my definition of neutral?
My plot of the Harman curve is impossible to achieve either, but can a brother gets his point across?
I'm not there yet, but I'll give you my best guess at the moment on the above question.
I use diffuse field btw, because it's easier to use those as a basis for comparison with the sound of actual loudspeakers in a room. So here is what the 2018 Harman Over-Ear target looks like with diffuse field compensation, as a starting point for this discussion...
I suppose it depends just how far off of neutral your headphones are to begin with. But fwiw, I seem to be finding curves that are a little closer to the actual behavior and characteristics of loudspeakers in a room (especially their directivity) a somewhat better model for a neutral sound than the above shape.
In broads strokes, the overall shape of the curve should be a downward slope toward the treble of about -10 to -15 dB in total. Which works out to approximately -1.0 to -1.5 dB per octave. So somewhat similar to the above curve. But perhaps with some of the following tweaks or adjustments...
There should be a somewhat flatter area or "plateau" between about 3.5 and 8.5 kHz. Some bumpiness is ok there though. Most of the better headphones will have a few small peaks or hills there, usually at the 3.5, 5.5-6.0, and 8k frequencies. Which is totally ok.
Less roll-off in the treble. Perhaps as little as only a few dB, or 5 dB by 20 Hz, instead of the rather pronounced roll-off in the higher frequencies shown above. I go by the levels of the peaks in the treble btw, rather than using some sort of average of the peaks and troughs there. Because there are typically large notches at around 10k, 15k and 20k. Those notches don't count toward what I'm talking about here. And need to remain intact.
A depression of at least a few dB or possibly more at around 2k. Imo, the depression in that area on the Harman curve is positioned a bit too high in frequency. So it's not really having quite the desired effect. (And is probably making the upper mids between 1.5k and 2k sound too bloated by comparison.) I'm still experimenting with just how deep and how broad this depression should be though. Because the optimal shape seems to vary a bit with different recordings. The directivity indices and sound power curves of loudspeakers may be a good guide here, since this correction is partly to compensate for the diminished dispersion at the crossover of the midrange and tweeter drivers in this range. This diminished dispersion will manifest as a dip in the sound-power curves, and as a small peak or bump in the directivity index of a speaker's spin-o-rama plot. More about this below.
The sub-bass should be on the order of 10 dB higher (give or take) than the somewhat flatter area in the upper-mids/low treble between 3.5 and 8.5k. (Audiophiles who prefer more mids and treble may prefer a bit less bass though. And "bassheads" may prefer a bit more.)
And instead of a fairly confined depression in the upper bass at around 200-250 Hz, there should probably be a very modest sag or smile in the overall shape of the curve from approximately the mid-bass to the somewhat flatter area in the upper mids/low treble. Which give the overall curve a subtle U-shape. (Plus a little roll-off or flattening out at the ends. And the dip at 2k.)
Hard to picture all of this in one's head. But it should become a bit clearer once I'm a little more confident on all this. And able to post a few plots of headphones with a similar response. The basic message though is room curve, with a dip at 2k. Maybe a little roll-off in the treble. Notches in treble left intact.
This applies mainly to Oratory's diffuse field plots btw. Since I haven't looked at this type of data from other sources. To see just the diffuse field data for the headphones, you have to select it from the drop-down list of "Target response" options. And also turn off the "Include raw" option, by selecting No there, to hide the headphones raw response.
You can also find the spinorama measurements for a number of loudspeakers here...
https://www.audiosciencereview.com/forum/index.php?pages/SpeakerTestData/
Use the scrollbar on the right side to scroll down to more speakers. And double-click on "Review & Measurements" to see the spinorama graphs for each speaker.
Although he doesn't really deal with the topic of headphones per se, this video by Floyd Toole (which is about an hour long) explains how the spinorama speaker data works. And the difference between the direct sound curves (which should always be fairly flat), the sound power curves which help to predict the speaker's in-room sound. And the directivity index curve.
The directivity index will be kind of a mirror image of the sound power, or room curve (which is actually calculated from the direct, early reflections and sound power curves). And it gives you an idea how even the dispersion of the speaker is across different frequencies. That's important because it effects how loud the speaker is at different frequencies when placed in room. Greater dispersion = more reflection from the walls & surfaces = greater loudness.
The directivity index curve turns this characteristic upside down though, and really shows you the areas where the response of the speaker is more highly directional. And less broadly dispersed. (Which is a bit confusing.) To put this into context though, the dispersion of a speaker is always the widest at the lower frequencies in the bass. And that's why it's always the lowest area on a directivity index curve. Because the speaker is least directional in that range. (That should make perfect sense to alot of speaker buffs. But it frankly took me awhile to grasp this concept.)
Another model I use for my headphone's response are the reference levels for home theaters, which include a 10 dB increase in the lower frequencies. There is a link in my sig which explains that in more detail.
Most of the better open-back headphones also follow a model roughly similar to what I have outlined above in the treble and midrange. And they only begin to divert from this below around 200 Hz in the bass, due to the physical limitations in their design. So they also follow a "room sound" analogy as well. You will find diffuse field plots for many of the popular open-backs in Oratory's graphing tool as well, so you can confirm this.
I am not sure which of the headphones on Oratory's graphs are the most accurate yet. But some of these mostly closed-back headphones might be in the general ballpark...
https://headphonedatabase.com/oratory?ids=72,210,68,36,11
Sennheiser HE-1
PSB M4U (ANC on) - maybe a bit too rolled off in the treble, and bit too bumpy in the bass.
Onkyo A800 - this is pretty well-extended in the bass for an open-back, and maybe also a bit too rolled off in the treble.
Hidition Viento - the response is a bit uneven in the upper mids and treble, but pretty good in the bass and lower mids.
Apple APM (Vocal - Moderate) - too depressed in the low treble, but otherwise fairly good.
https://headphonedatabase.com/oratory?ids=205,275,56,73,141
Beyerdynamic Custom One with (Velour 3/4 Bass) & (2/4 Bass) - probably not depressed enough around 2k, and a little bumpy in the bass, but otherwise pretty good.
Focal Radiance - probably not depressed enough around 2k, and too depressed around 4k, and too much of a bump in mid-bass, but in the ballpark generally.
AudioTechnica M40x - Too rolled off in the sub-bass, and possibly a tad too withdrawn and U-shaped in the mids versus the levels at 200 Hz and 8 kHz, but not too bad otherwise.
Audeze Mobius & Audeze Sine (Cipher Cable) - both are too withdrawn in the low treble, and maybe not quite depressed enough around 2k, but otherwise in the ballpark.
AKG K371 - Too depressed in the high treble and at 4k, and maybe not depressed enough at 2k, but otherwise in the ballpark
You can use the links above to also look at raw responses of these headphones versus the Harman target (and other compensation curves). Just select "Harman 2018" from the Target response pulldown menu.
Last edited: