HD600 - front neutral EQ - WIP (Patreon)

Ok here it is: my very first version of the new neutral EQ for the HD600 which was derived from my recent experiments trying to render wind noises as faithfully as possible. With this EQ you should basically hear exactly what I recorded, at the same time this EQ should offer some kind of “flat” response if “flat” is possible at all. Keep in mind that this EQ is optimised for my own HRTFs, to get an overall 100% faithful result you would need to equalise the headphone in the same way using your ears. I verified the final EQ by playing back my recording over the corrected HD600 recording this result again. Comparing both spectra again they should still be close.

I used this as reference spectrum for some kind of “smooth” natural response instead of pink noise which is hardly achievable in real life with that mostly diffuse and expanded character which contributes a lot how this spectrum will be captured with my ears while this short loop of the reference spectrum was captured with a normal dynamic microphone: https://drive.google.com/file/d/1Po0JeN0mQ2c25tEgHS8mhpWYZzreFwSC/view?usp=sharing

The result would be this binaurally captured final spectrum that was manually equalised to match the reference spectrum in frequency response for left and right ear separately: https://drive.google.com/file/d/1AqtkbRvhsIym8jwuXwpGbpKhyPgXHxlb/view?usp=sharing

I noticed that in case of the realism of my recordings I need to equalise my both ears separately due to their differences in the captured frequency response. Averaging the responses for left and right ear and equalising towards this resulted in a more blurry imaging and a more static artificial sounding spectrum.

I tried to match the measured responses of both reference spectrum and the binaural spectrum as close as possible above 1khz. I completely ignored the range below as usually HRTFs do not affect the incoming sound at longer wavelengths that much, but especially the range above 3khz is crucial to get a convincing final sound.

How was this EQ achieved?

My spectrally corrected binaural recordings were played back over the HD600 recorded again with the output equalised to match the initial spectrum again although this time not as precisely and averaged for left and right because the differences between left and right ear would become way too obvious when baked in into a generalised tuning. But I also plan to try a version optimised for left and right ear separately just to hear how much this may change again although equalising both ears separately is way harder and takes way more effort, and I would need to rely too much on graphs only, while in this case I adjusted a lot by ear trying to make it appear smoother and also closer to the real sound I was hearing outside. But maybe an own EQ for left and right could improve the final sound even more at least for myself as I could probably probably reproduce the effects of my HRTFs even better in the final response.

With this achieved EQ I listened to my recordings and compared the headphone output to the real thing manually adjusting some smaller aspects to make it appear even more correct sounding to me than just from a simple curve adjustment towards another curve. I didn’t want to deviate too much from the initial EQ but tried to improve for those areas which I thought didn’t quite sound like in real life. With the current version shown here I got as close as possible to this optimum although there may be still some areas for further tweaking and improvement I am already quite satisfie by the result. It has to be seen and heard if a more strict adjustment towards the initial spectrum would really get better results or if some averaging and blurriness would be better because of so many factors influencing the sound due to different positoning etc. From my experience so far it would seem to me that for the final EQ some generalisation of the final response leads to more convincing results than a strict graph matching. There is a single filter unused which I still left inside as I was not sure about it, but meanwhile I think it rather worsens the result.

Right now I am listening to Acoustic Alchemy which really gives me goosebumps.

I would welcome some comment about how others perceive this setting over their HD600. Apart from different ears there are of course other variables like different pads etc which may affect the final sound a lot. Therefore I am really curious if this setting may give any improvement to anyone with different ears.

The crucial point seems to be how the binaural spectrum was created and equalised primarily. In this case the spectrum used for this EQ was concentrated mostly on sound coming from the frontal direction with all other natural noises around me coming from all around. If I got more sound contribution from the sides and back as well, I would need to equalise the spectrum differently to match it to the the reference spectrum which in the end would also affect the response of the headphone and needs to be readjusted accordingly to still allow for the same correct and proper reproduction of the initial spectrum. I also want to experiment with this more and may provide EQ settings which would be rather averaged over a spectrum captured from all around. This would result in a darker sound because the spectrum would contain more information from all directions making it appear brighter with the need to readjust the headphone darker to compensate for this.

Update: I let Oratory1990 measure the HD600 on his standardised rig with my EQ applied (simulated measurement with just the HD600 measured raw and my EQ added as function on top) and you can see his graph overlaid with the original HD600 and the diffuse target:

This forced me to tweak my initial EQ which has my personal HRTF peaks too pronounced and reduced those close to 6k and 13k bringing the whole nearby frequency area into a better balance to have a similar timbre as before. V2 definitely sounds smoother also to me, although a bit more "wrong" compared to what I hear in real life, but I suppose V2 should also sound smoother to other ears. If I deviate too much from this curve, the sound would simply differ too much from my initial spectrum I captured. Please let me know what you think.

Update 2: Thanks to Mika who managed to reduce filter amount to 10. I retweaked the settings a bit to get a smoother response overall still trying to stay close to my inital settings. There are of course spectral differences between the reduced version and the original one, but I think the reduced sounds very convincing. I could try to optimise it even more. But both my binaural recordings sound pretty decent with it, just as pink noise and also music. This makes the EQ now usable with external devices like Qudelix-5K. I am listening to the HD600 through the Qudelix and the reduced EQ and I really like the result. If I manage some improvement I will post an updated version. 

Update 3: I renamed this posting to WIP, because all these settings are what they are: a work in progress. I was not aware how tough it is to get the it right and had to start from scratch already several times since I posted my last version, but I think I am finally on the right path and wanted to post the latest iteration of what I would call a linear EQ for the HD600. Unfortunately I noticed that exactly the same EQ values used in EqualizerAPO and the Qudelix 5k result in a significantly different sound. My settings were created with Equalizer APO, I will need to rework them to allow for the same sound in Qudelix 5K which of course is annoying. Therefore V3 is verified to sound correct only through EqualizerAPO, I couldn’t test other EQ apps.

Let me explain the new EQ a bit more: as a start the headphone output of the reference spectrum was manually equalized to match the spectrum as good as possible. It would seem as if not a precise match was important but rather staying overall closely to the target. I used psychoacoustic smoothing to adjust the EQ which seemed to work well and resulted in a more realistic sound than a precise EQ matching of all peaks. Then I compared a binaural recording of a speaker standing in front of me listened over with the HD600 and the EQ applied and did some manual corrections based on the differences I could still hear compared to the real speaker. Although I didn’t manage to reproduce the sound of the speaker by 100% I think I got quite a bit closer to what I achieved before. There are still some differences especially with mids which I couldn’t correct for. It was impossible for me to track down the differences precisely, as based on measurements I am following the original reference spectrum already quite well, but I can hear some harmonics in the mids which I just wasn’t able to recreate faithfully. Here you can see the original spectrum in black, the corrected binaural spectrum in red and the spectrum the equalised headphone is reproducing in blue. There is still a deviation below 1khz which I also wanted to account for creating 2 variants for this EQ version.

A is the more common variant with a more or less a flat measuring bass except in the lowest frequencies which are raised towards 30hz slightly. In reality this setting will result in too much upper bass if listened to a neutral flat spectrum and to compensate for this the headphone EQ needs to be adjusted in this region, which I did with B.

B is the more “correct” setting assuming that one tries to perfectly recreate a natural spectrum over the headphone. Like this the bass follows more the Harman target in this range of 200-300Hz although I tried to keep it rather neutral without too much boost below 100hz. B definitely appears slightly brighter than A due to some lack of energy in the 200hz range therefore it may need some further tweaking to achieve a better balance again.

Below you can see a simulated measurement on a GRAS system of both settings overlaid with the Harman target:

Update 4: I added optimised settings for the Qudelix 5K, which unfortunately with the same filter values will sound different than EqualizerAPO like shown in the measurement below (compensated for my target):

Black is the response of the HD600 with the original flat bass settings from EqualizerAPO, while blue is meaasured from the Qudelix with exactly the same values. You can see some drop to both bass and treble, which I tried to correct with an own tweaked setting for the Qudelix. I had to sacrifice the last filter which corrects for the peak around 15khz and added some high-shelf instead which raises the entire treble range to the correct level, unfortunately the peak at 15khz will remain which may bother some with younger ears.

I wonder about your opinion. Listening to my own recordings I can still notice some slight errors in the response, only noticeable with complex spectrums though, but the realism I managed to achieve is already quite stunning. It would probably need way more precise frequency matching for results fooling you, you are actually hearing the real thing and not just some recording.