I spent several weeks to develop this EQ preset for Sennheiser HD800

[JIW]

As a quick side note, does anybody know which country Oratory1990 lives? I'm thinking of sending him my AKG K702 headphones, but I had a feeling he lived in the USA? (Hopefully he's in Europe).

Austria. For measurement and EQ see here.

 

[Robbo99999]

Austria. For measurement and EQ see here.

Thanks, not too expensive from UK to Austria then. Looks like I need to join Reddit then to communicate with him. I might send him my AKG K702 and perhaps the new version of the Sennheiser HD600 which I ordered today, as it might be interesting for him to test the new version of the HD600 vs the old version.

 

[Theriverlethe]

Harman Curve is based on average preference, but I believe it's based on average preference around an initial Diffuse Field curve (ie one that was created to simulate flat speakers in a room taking an average HRTF into account), so I think bass level was tuned on top of the Diffuse Field curve. So I would think Harman Curve is a good start then you tweak bass to your liking (using broad Low Shelf), and perhaps tune 10kHz+ with broad High Shelf filter too. It sounds like we agree on that broadly.

Yes, I agree about tuning to flat sound and the problems there, that is what I was getting at.

My point is average HRTF will not work at all for some people and only work extremely well for an equally small number.

 

[Patrick1958]

As a quick side note, does anybody know which country Oratory1990 lives? I'm thinking of sending him my AKG K702 headphones, but I had a feeling he lived in the USA? (Hopefully he's in Europe).

Austria

 

[Robbo99999]

My point is average HRTF will not work at all for some people and only work extremely well for an equally small number.

I agree. However there isn't an alternative beyond placing microphones in your ears and measuring response from perfectly flat speakers in a room, and even then the microphone would theoretically have to be right on your eardrum thereby creating your own person diffuse field target curve....so it's not possible. On top of that, even if you had determined your own diffuse field target curve (based on your own HRTF), then you'd still need to find a way of measuring your headphones that simulated your own ear & ear canal....again not really possible. So we're left back with the average HRTF associated with the Harman Curve along with measurements on the best rigs (eg. Oratory1990) which is again an average approximation of an ear. So it's basically 2 "averages" used together, so yes there's quite a bit of innacuracy at play when it comes to individuals, just I don't think it can be better than that?

 

[Theriverlethe]

I agree. However there isn't an alternative beyond placing microphones in your ears and measuring response from perfectly flat speakers in a room, and even then the microphone would theoretically have to be right on your eardrum thereby creating your own person diffuse field target curve....so it's not possible. On top of that, even if you had determined your own diffuse field target curve (based on your own HRTF), then you'd still need to find a way of measuring your headphones that simulated your own ear & ear canal....again not really possible. So we're left back with the average HRTF associated with the Harman Curve along with measurements on the best rigs (eg. Oratory1990) which is again an average approximation of an ear. So it's basically 2 "averages" used together, so yes there's quite a bit of innacuracy at play when it comes to individuals, just I don't think it can be better than that?

Don’t forget the sample variation in headphone drivers, so we have another unknown. I think you could apply Oratory1990’s headphone measurements to your individual HRTF target. Maybe this could be done with a script like AutoEQ.

Audeze has an interesting program where you upload a picture of your ear and it attempts to derive a personal target using AI/machine learning. It sounded pretty bad when I tried it, but I only had iSine 10’s when I started the free trial. Looking at Oratory1990’s iSine 10 measurements suggests that Audeze’s EQ curve for the iSine is pretty bad, so my experience may not haven been the AI’s fault.

 

[Robbo99999]

Don’t forget the sample variation in headphone drivers, so we have another unknown. I think you could apply Oratory1990’s headphone measurements to your individual HRTF target. Maybe this could be done with a script like AutoEQ.

Audeze has an interesting program where you upload a picture of your ear and it attempts to derive a personal target using AI/machine learning. It sounded pretty bad when I tried it, but I only had iSine 10’s when I started the free trial. Looking at Oratory1990’s iSine 10 measurements suggests that Audeze’s EQ curve for the iSine is pretty bad, so my experience may not haven been the AI’s fault.

Yes, you could definitely add sample variation from unit to unit as another variable. I suppose my point is that in my understanding it's not really practically possible to get better than Oratory1990 EQ's, and then you tweak bass using Low Shelf to personal preference (which is shown in the research as being a valid tool and comes down to individual preference). You can also tweak north of say 10kHz using High Shelf, but my intuition is that you'd start with the Low Shelf tweak first. But yes, there are a lot of variables, you can't really control them all, just use the best approximations.

(Individual HRTF's are pie in the sky as far as I know at this point).

 

[JIW]

You could listen to the songs I posted and try with Oratory’s settings yourself if you have a pair of HD800. Unless the original recording is awkward, I know which one sounds more natural.
Also my setup is indeed created with negative gain only. The reason that the curves crosse each other was simply that I ran the measurements at slightly different level aiming to indicate how sound signature really changes - when you are using pre-gain to match the overall level on top of the EQ. No one would listen to an all negative filter and think it sounds better than the original one as level difference actually lead one to think the sound is “better” or “fuller”, which all folks here know is not correct, right?

For an identical filter response, where is the difference between turning the volume up after using only cuts and turning the volume up after using a pre-gain cut to ensure boosts don't result in levels above 0 dB FS?

Anyways, my comparison was based entirely on relative values looking at the overall shape of the frequency response, so more positive gain is the same as less negative gain.

To the latter point, with highly non-neutral headphones, neutrality at lower volume may indeed be preferred. Based on the oratory1990 measurements, using cuts only, the HD800 would lose maybe 2 dB in loudness at 1kHz if cut to +-1 dB or so from Harman target (maybe 3-4 dB overall). Most of the cuts would be the excess treble including the resonance, the low-mid hump and the (relative) bass-mid hump between 80 Hz and 1.5 kHz. I'd wager, the EQ'd version would indeed sound better.

To your point about cuts not producing clipping, the linear phase filters also produce artifacts that alter the waveform in a way such that the amplitude can increase. Having downloaded the files and measured their RMS and peak sample values using a simple python program, for the first track the un-EQ'd version has -23.24 dB FS RMS and -7.16 dB FS peak while the EQ'd version has -22.41 dB FS RMS and -6.36 dB FS peak, for the second track the un-EQ'd version has -17.66 dB FS RMS and -6.43 dB FS peak while the EQ'd version has -16.34 dB FS RMS and -3.94 dB FS peak and for the last track the un-EQ'd version has -15.58 dB FS RMS and -2.17 dB FS peak while the EQ'd version has -14.11 dB FS RMS and 0.0003 dB FS peak. For the last track I have also plotted the waveforms. First, without EQ and second, with EQ.

You can use the pre-FX monitoring on the RME or maybe DIGICheck to see if the EQ introduces clipping.

 

[LaLaLard]

It is quite easy to compare the files if you have a pair of HD800. Being a member here we all know the importance of measurements, but how it sounds like is the key point after all. No matter how many times we discuss about the authoritativeness of someone’s work, it doesn’t change how a person feels about a sound in his ear.
Perhaps someone also have HD800 could just share his impression here about both EQ sets. Oratory’s presents should be quite simple to be applied.

 

[Theriverlethe]

Yes, you could definitely add sample variation from unit to unit as another variable. I suppose my point is that in my understanding it's not really practically possible to get better than Oratory1990 EQ's, and then you tweak bass using Low Shelf to personal preference (which is shown in the research as being a valid tool and comes down to individual preference). You can also tweak north of say 10kHz using High Shelf, but my intuition is that you'd start with the Low Shelf tweak first. But yes, there are a lot of variables, you can't really control them all, just use the best approximations.

(Individual HRTF's are pie in the sky as far as I know at this point).

My point is don’t take the Harman curve more seriously than was intended. An average preference curve doesn’t override personal preference. Oratory1990 himself said he listens to HD650 without EQ. IME, it made my LCD-2’s sound like a bathroom.

 

[xykreinov]

@LaLaLard, are you aware of AutoEq? It seems like a lot of time in those steps could have been saved by that program.
If I were you, I'd be tempted to start over and use orratory1990's measurements. As others have pointed out, the MiniDSP Ears is flawed equipment that can only get you so far (as anything else in its budget).
Nice work, though. I'm keeping my eyes peeled for a guide to loudness matching. While the Harman Target is definitely consistently better than the stock tonality of my headphones, it's still a bit bright to my ears. Toole and Olive said that low q band reduction in treble is an acceptable compromise, but, I want to generate my own curve using the Harman Target as a basis so I can have greater objective reference.

 

[Robbo99999]

My point is don’t take the Harman curve more seriously than was intended. An average preference curve doesn’t override personal preference. Oratory1990 himself said he listens to HD650 without EQ. IME, it made my LCD-2’s sound like a bathroom.

Well, yeah, if Oratory1990 EQ's sound like sh*t to someone, then they shouldn't use them. The point is I don't think you can really effectively EQ headphones yourself beyond broad Low Shelfs & broad High Shelfs if you don't have accurate/relevant measurements to base them around.

 

[Robbo99999]

@LaLaLard , I've just had a thought that might help you as an approach. If you don't like the Harman Curve then go ahead and import the measured raw frequency response of your headphones from Oratory1990's pdf file, import them into REW....then use REW to help you EQ those measurements to whatever variation on the Harman Curve you like. You could experiment with variations to any part of the Harman Target Curve which you'd apply as a target in REW, and then you'd use REW to either create manual filters or automatic filters to EQ to your new 'experimental' target curve. I suggest this approach because I think you can really trust Oratory1990's measurements on his fancy rig more than you can trust your measurements on your miniDSP EARS.....but this approach I described will still let you experiment with variations to the Harman Curve. I took a similar approach when I was EQ'ing my AKG K702 in terms of using REW to adapt filters, it's an extremely versatile & flexible piece of software. You can simply look at Oratory1990's pdf file and zoom in the raw frequency response and then write down using pen & paper the plotted points of frequency & dB that best describe that raw frequency response curve....you then put them into a notepad text file and import that frequency response into REW. You would also edit & import your modified 'experimental' target Harman Curve into REW in the same way. This way you're using reliable measurements as a basis for all your subsequent EQ work.

 

[zermak]

@LaLaLardYou can simply look at Oratory1990's pdf file and zoom in the raw frequency response and then write down using pen & paper the plotted points of frequency & dB that best describe that raw frequency response curve...

That's a lot of work... not that I didn't do it with rtings RAW measurements but there are easier way to do it... SPL trace inside VituixCAD software (another great piece of software indeed).

 

[Robbo99999]

That's a lot of work... not that I didn't do it with rtings RAW measurements but there are easier way to do it... SPL trace inside VituixCAD software (another great piece of software indeed).

I know REW from my RoomEQ experience (this year only), so I just used my own imagination & creativity to use REW in the way I described.....I didn't find it much work, like a couple of hours. There may indeed be easier/quicker ways of doing exactly the same thing with different software, but I really have no experience with any other software.

 

[zermak]

I know REW from my RoomEQ experience (this year only), so I just used my own imagination & creativity to use REW in the way I described.....I didn't find it much work, like an hour or two max. There may indeed be easier/quicker ways of doing exactly the same thing with different software, but I really have no experience with any other software.

I am into this a little longer, but not that long and I had the luck to find VituixCAD which is a very powerful tool for speakers designing and much more (it is nominated often in many threads indeed).

 

[gorman]

I find this discussion a very interesting read.
I've recently discovered AutoEQ, oratory1990's work and the effect they have on my old, trusty DT 880. Transformative it's the word I would use.

Now, I came to this thread since I was thinking of upgrading my headphones and HD800 are one of the models I'm thinking about (T1 and Amiron Home are the other two, mainly). It's just that with the enjoyment that equalizing to the harman curve has brought to my music listening... I'm not sure I would get much of an upgrade. Since you're here you're obviously not averse to EQ, so I'm asking: what would I gain by upgrading? Would I get an upgrade or a sidegrade if I simply plan to EQ to the harman curve whatever I'm gonna get? Thank you.

 

[LaLaLard]

I find this discussion a very interesting read.
I've recently discovered AutoEQ, oratory1990's work and the effect they have on my old, trusty DT 880. Transformative it's the word I would use.

Now, I came to this thread since I was thinking of upgrading my headphones and HD800 are one of the models I'm thinking about (T1 and Amiron Home are the other two, mainly). It's just that with the enjoyment that equalizing to the harman curve has brought to my music listening... I'm not sure I would get much of an upgrade. Since you're here you're obviously not averse to EQ, so I'm asking: what would I gain by upgrading? Would I get an upgrade or a sidegrade if I simply plan to EQ to the harman curve whatever I'm gonna get? Thank you.

In Sean Olive’s paper he indicated that untrained ears could tell the difference between a pair of (cheap-ish headphone) EQed to a expensive pair but the preference scores between the two are close. Not the same, but close. In other words, you should be able to find a difference. Whether that difference is great enough to be called “upgrade” you need to find out yourself. And even someone with both headphone would need to tune both headphones for a long time until the perfect state is arrived.

 

[preload]

So, are you going to post the filters you used?

 

[Erispedia]

I tried EQing Sundara. Rather than Oratory1990 settings, I prefer the sound from my own tweaking using Rtings FR measurement and FR target curve as a reference. But Oratory1990 setting for WH-1000XM3 is already good enough for my ears. Just ordered HD800, can’t wait to tame this bright headphone.