Upcoming “reference tuning”

I spent the recent weeks mostly with trying to optimise my own target for headphones and IEMs trying to match the perceived spectrums I get from all of them which in the meantime seems close to impossible to me, but at least I was trying.

This definitely helped me a lot to optimise my recent own tunings since I started all this madness and make them simply more correct in order to really get at least a similar smooth kind of natural spectrum from all sources. The real problem are upper mids and treble, this is where all the realism is in a perceived sound. What I noticed in the meantime is that the smoothness of the spectrum is the most important factor aside from the correct timbre because your brain will quickly adopt to any sound despite some colorations although it will be much easier if the spectrum is smooth with no obvious bumps or peaks. Those bumps tend to hide and drone out  nearby details, a dominant frequency will mask softer frequencies, this also goes for frequencies at the opposite side of the spectrum, too much bass will influence the timbre of treble etc, it’s quite shocking how treble may change if you adjust only mids or bass….

Getting a real neutral reference, which means pink noise that is uncoloured, is also close to impossible or how often have you really heard pure natural pink noise which is the only signal which consists of all frequencies with the same constant ratio to each other aside the fact that the timbre of pink noise will change continuously depending on at which angle it enters your ear on top of this every ear will change this to another degree. The entire topic is quite complex, if it weren’t, all headphones and IEMs would sound great…

Therefore keep in mind all the optimisations I did lately were done with my own ear but also always verified with measurements after having found satisfying results, and then optimised further based on  the graphs, listening again to the new "improvements" after having done some cosmetic adjustments to a line. But after lots of trial and errors I finally reached a version where I have a hard time to improove it more. If I improve one part I somehow make the other part worse, you need to approach that optimum curve by only 10 filters and adjusting one filter will influence the others as well, but now finally having found that curve or an approximation of it which is more or less valid to my own ear and my ear resonance of around 6000hz. The individual ear canal resonance may vary continuously by shifting shift up or down. You can also influence this a bit with different insertion depth of an IEMs, notice how the sound will change if you either put it deeper or more shallow, or try closing your nose and doing a pressure equalisation for your ears and notice how upper mids will change significantly. Some days ago I was at my Otologist to let my ears clean again and my resonance shifted up to 6100-6150 now even slightly above that. Of course my previous tunings started sounding off, suddenly nothing sounded alike anymore, the speaker sounded weird, the headphone didn’t sound like the speaker anymore and the in-ear didn’t sound like the headphone. It’s a frustrating experience as all my work seemed to be just waste of time, but actually I managed to sort it out again and have something usable hopefully soon. I still need to give it some time before posting some half baked result, the next day the tuning I did before will also sound "weird"... it's a back and forth all the time and I really start to question if others will even hear the same thing, if already my own ears suddenly hear everything differently. The issue were not the tunings which still sounded quite alike with just a shift of 100hz up, but rather that they didn’t sound like my reference anymore, because my ears altered the sound of the speaker and headphone as well compared to before and I had readjust everything and now rather tried to only capture the general character of the spectrum and make it smooth instead of focusing on tiny details. It also helped me to create filtered pink noise with only a spectrum above 3khz as this is the area which is most important while frequencies below would be affected way too much by the room, placement of the speaker etc. and would distract too much from the frequency areas which need most attention. To get this really perfect one would need an anechoic chamber, perfectly flat speakers, tons of time and patients… thus everything I am doing is just a rough approximation and generalisation based on my own interpretation of “neutral”, keep that in mind.

I made huge progress since RC2 of “newtral” which I want to rename to “reference” now because there is simply no real correct “neutral” and there can be many interpretations for “neutral”, thus please treat this as my own interpretation of something called neutral and let me explain how I came to this. Harman’s method has way too much room influence and user preference together with too much generalisation. It can be just a rough guide for a curve but cannot be used for precise tunings. As I didn’t want any room influence to be captured, I derived my own IEM target or let’s call it tuning from various sources: Up to around 1khz a human ear hardly alters an incoming spectrum at same time the HD600 measures pretty flat and with additional EQ you can make it totally flat up to 1k, therefore I can be sure that what I hear from the HD600 up to 1k sounds “flat”, and can use this as reference for my tuning. Above that the result will start to become very individual, therefore here is where I started to equalize the HD600 in the upper midrange and treble with my filtered pink noise concentrating just on the part above 3khz and trying to match this to the filtered spectrum I would hear from a speaker. Imagine something between a Shhhh and the polish „Ś”, which is not that easy to reproduce properly. There is not a single headphone which will offer you exactly this out of the box. Having done this for myself I used this as base for creating my upcoming “reference” tuning for the Free Pro 2. Already having a solid base from my previous tunings I tried to optimise the settings even more and get the sound even closer to what I heard from my equalised HD600 and achieve that spectrum for the Free Pro 2 at least as close as possible, which took days of permanent listening and annoying my wife who begs that everything will hopefully come to an end soon…

I think I am really close to this optimum and have a final curve which looks quite reasonable to me and way more simple than my dip-theory some weeks ago. Obviously the human ear takes some averaging and doesn’t care so much about dips than peaks, which can result in double peaks if both start interfering with each other, more important seems to be the smoothness of the final curve, I repeatedly play sine sweeps to monitor if some parts will stand out in amplitude, or if there are some gaps in the spectrum. I was not able to match headphone and IEM let alone speaker exactly, but the general character is captured quite well I think and I managed to achieve that smoothness of this natural spectrum really well at least for me, but looking at the measurement it also looks quite sensible to me with no real anomalies or weird kinks (the range between 1-2khz are probably measuring artefacts, these will disappear with deeper insertion although will also result in a different response in that area, thus I will need your help to find that optimum with a response that I will either readjust towards my own measured curve or leave that particular range untouched in amplitude but rather concentrate on the shape of the overall curve. I will try to provide 2 such variants once I finished my final EQ. I am trying to make it decent over all the units I have and although they all differ all a little bit, they still resemble each other but may result in a more bass heavy or more muddy result for one or the other unit. Looking at the measurement you can see that the curve looks still pretty similar to my first "neutral" tuning in green and also "RC2" in blue:

Unfortunately with all my continuous filtering the resulting tuning became too soft, even softer than my previous attempts where people already complained about lack of loudness. Besides the lower and upper end of the spectrum got boosted too much, because the entire response shifted down in amplitude and I was afraid it would contain both too much lower and upper frequencies which I couldn't adjust further due to lack of filters. Therefore I wanted to try another attempt and start from scratch having achieved this curve now I tried to create an EQ based on my measurement simply adjusting all filters one by one until both curves align perfectly. You can see the measurement of my new filter settings on top of my previous one:

To my surprise they sounded still different. Although matched nearly perfectly aside from that lower bass boost and treble, it were those nuances that sounded different and made my new tuning sound worse, I am pretty sure the reason is because both these responses are achieved by completely different filters, although measuring exactly the same at this given insertion/ear canal length or whatever, thus slightly different insertion depths etc will make them drift apart and make them sound different, as this measurement is only a snapshot of a single state. 

I had to sit again for some days and optimise this new setting trying to make it sound even better. Those small resonances in the upper mids and treble are what makes the tuning sound good or bad. The rest of the spectrum can be exactly the same but if there is some slight nuance off it will ruin the entire sound. All frequencies must have the correct ratio to each other, otherwise it can sound OK but it will never sound really great. In the following measurement you can see my current setting inserted slightly deeper and more shallow, marked are the areas which need most attention as these will start to change most but also have the biggest influence on the final result. Although the entire rest of the spectrum will remain the same, it may be the reason for the overall sound to be simply off.

To cut it short: I have a setting now which is optimised for my current resonance of around 6100hz which I will post soon. I still hear a slight issue which I would like to correct before posting it to get some feedback. Douwe already tried an earlier version and loved it but I think my new one is even better and more correct although it may also be related to the fact that my ears simply hear differently again, what a mess!

There is one issue though: due to a completely new manual filter adjustment this setting became too loud now, I cannot even turn it past half volume because it would be too loud for me. I am not sure if it will be a problem for others, those who complained about lack of loudness before will complain now about lack or softness… Right now I listen at volume step 5-6 out of 16. I am not sure if I should start another attempt creating a new setting based on my current measured one, it will be hell of a work and I cannot promise that I will manage the same result, because as mentioned before even with 2 settings measuring the same the perceived sound can still be different as the response is achieved by different filtering. This was quite a shocking experience for me actually.

Although I really love my new personal setting there is still lots of work to be done, trying to get that curve over onto other ear canal resonances which cause the shift of the entire spectrum above 3khz, and still hope that this result will also sound similar to you although I am slowly getting in doubt that this is really possible if even 2 different settings measuring the same can sound different, and it’s exactly those nuances which define if the tuning is really perfect. It’s not the driver or some other technicalities how many will try to suggest you. It’s just tiny differences in the frequency response that can ruin one result while still sounding similar.

Those frequency shifts caused by different ear canal resonances could be probably even easily explained mathematically and an automatic calculated response created… but I am not a mathematician although that could be an interesting feature for a future product. I still wonder why nobody really did it yet although Apple’s AirPods Pro are claimed to have some automatic sound towards ear optimisation procedure, but not sure how well it really works, if you can disable it etc, because most optimisation routines I tried be it from Nura or the AKG N90Q were just unusable. Therefore my manual approach based on measurements which I can only verify for my own ear, and then trying to simulate that response for other ears with different ear canal resonances relying on a measurement system which I cannot verify for accuracy makes all this a bit questionable, but so far many seemed to have liked my results and I think you should love my latest version as well. Thus please be patient a bit until I sorted out everything. I was so busy with all this that I didn’t even manage to post my audiophile collection playlists yesterday, but I will catch up on that the next days!