Improving my headphone target and recording quality. (Patreon)

Having finally received several pairs of completely new microphones a while back which should allow for more consistent recordings due to better build quality, I finally took my time to improve my entire recording setup including creating a better and more neutral headphone target which comes along with a matched compensation for my recordings. Therefore I had to re-measure everything and concentrated on a single pair of microphones as for now. In case these break I will have to repeat this for the next pair. I did first tests with 3 different pairs and they all gave slightly different results, so I settled on a single pair now which somehow sounded best to me with the best channel matching etc. The last 2 weeks I spent several days doing new measurements of my KH80, my headphones etc. as well as performing listening sessions to get some satisfying results and would like to demonstrate my outcome here. Everything is a kind of work in progress, but I am pretty happy with what I achieved so far and wanted to share some background about all this.

The main goal is to offer you the most natural and realistic binaural recording given you listen over a headphone tuned towards a neutral target. I am still not 100% sure how well my target will work for others and even when I was thinking that I already reached my goal a while back I discovered that my older headphone EQs sound simply off to me now. This may be of course a result of having gained more listening experience over time or my new microphones giving better and more consistent results but also the headphone (in my case the HD58X) changing its response over time due to pad wearout etc, but in the end it can be a result of all factors taken together and I simply had to bring everything up-to-date.

So for the start we need to define what "neutral" sounds like. I don't want to discuss about diffuse field, freefield etc here, my goal was rather to make the headphone sound as close as possible to a neutral speaker pair playing in a room. This is also what Harman's target is all about, although their results are more then questionable to me, and no headphone I ever tried so far sounded "right" to me. The ear may get used to a given sound after a while, but I am trying to get it as close as possible to the sound I am hearing from the speakers changing back and forth between both and am mostly adjusting by ear using additional measurements done with my ears to have some further reference. It turned out that trying to simply match 2 different responses to each other won't make them sound equal, therefore it took me a lot of time to approach this "optimum". I can spend several hours a day doing listening tests, which becomes pretty exhausting after some while and you cannot be sure you are really able to evaluate sound differences properly as the ear immediately searches for similarities between 2 different spectrums instead of detecting the differences, therefore I need some rest and can continue another day to finally notice some new issues all of a sudden. This usually can take up to 2-3 weeks. I had also noticed that my hearing is completely different in the morning after waking up than in the evening. This is probably the result of the ear canal shape being different in the morning, maybe it's more narrow or more relaxed, but all my settings I thought were sounding OK in the evening, started sounding off in the morning. To verify this I did several sine sweep tests and while getting smooth transitions between 7-9khz in the evening I suddenly heard some slight peak or dip in that area in the morning. But if I can hardly get it correct for myself, how can I think it will work for someone else? At least I want to get it as correct for myself as I manage, so maybe it will also give an improvement for someone else.

Let's get back to "neutral" again. Using the KH80 as reference, placed nearfield 1m away, with the speakers 1m apart in an optimum stereo triangle I repeated the placement several times on different days and always performed measurements to achieve some kind of "average". I also noticed that the result will be quite different if the speakers are not placed exactly at ear height but slightly below, any change in distance from the speakers will also change the response significantly like shown below where you can see measurements done with my ears. White should be more or less within the sweet spot, red would be too much to the front, green too much to the back of the sweet spot and blue the average of all 3.

You can see that there is lots of change in sound if you just move back and forth like captured in the sound sample attached below, where in the beginning you can hear my binaural pink noise sample I usually use for playing back pink noise over headphones instead of uncorrelated pink noise (thanks again to the guy who provided this file to me some years back). Then it changes to pink noise played back through the KH80 with me moving back and forth around the sweetspot. The resulting response compared to the calculated average from above is shown below, with light blue being the dynamic average from the sample:

Doing an average over both these again in dark grey, we can put this against a measurement done with a moving microphone around the listening position and averaged over several takes in light grey:

So the response I measure with my ears (also averaged over left and right ear to make things more simple, otherwise the effort would be way too big) differs quite a bit from the response I measure with a traditional measuring microphone.

Now let's have a look at some older headphone curves where I tried to match the perceived spectrum I heard from the speakers to that of the headphone. In green my original HD58X neutral target from nearly 2 years back and in blue an updated one I was using until recently and already thought it sounded close enough.

Both still differ quite a bit from the speaker response captured with my ears, so below again different variants I created recently trying to stay as close to the measured response of the speakers in blueish tones, while in the end the currently final version in white gives me the best and smoothest result so far:

It is still not perfectly inline with the speaker measurement, but sounds closest to me at least as for now. When I was trying to approximate the areas even closer to the measurement of the speakers it started to sound more off, therefore I had to find some compromise and always needed to listen if the new adjustment is actually an improvement or not.

The toughest part is the area around 8khz which with my ears is measured as a dip in case of the HD58X even more so with the HD800S which even gives me near to null response in this range. Trying to even out this area just results in some perceived peak, but the area should be heard as smooth transition without any obvious changes in amplitude. As mentioned before achieving this perfectly is close to impossible because my ear will hear differently in the morning or evening like mentioned before.

Here my new "targets", neutral in blue and red with in-room bass:

This way the headphone sounds closest to how I perceive neutral speakers playing in a room and I will use this as reference for any future headphone projects. I am pretty satisfied with the micro detail I achieved. This may be improved even more, but it's very hard as tiny changes within the response in some area may compromise another area. If all frequency ratios are not correct to each other, it won't sound natural. It's hard to explain but this takes longest to get perfect results and I will add another listening session soon for a final adjustment and will settle for this as my final headphone target to be perceived as being neutral at least for my own ears.

Based on this response I did further iterations to improve the compensation of my recordings. Playing back the sound I had captured from the speakers through the headphones and captured this again I had to verify that the resulting response will still be the same as the original response of the speakers or as close to this as possible. You can see below with blue being the original response and yellow the same response played back over the headphones again.

You can hear the result of my new compensation with the KH80 playing "Analog" from Soulpersona&Princess Freesia at the beginning, then switching to my previous compensation and back to the new one before changing to the original recording at the end.

As for now I ditched the idea of offering EQs for various headphones, simply because the effort is way too big to achieve the results that I really think are worth it. But with all being done I can finally continue my work on the Oluv'Sone headphone again and hope to come up with something usable soon. I have several units which I need to measure and check for consistency and can then decide if a public release would make any sense.

All my upcoming videos will have my new compensation applied, which should offer smoother and more neutral results than my old recordings regardless of used headphone.

Thanks for all your support so far!