Oratory1990 has REMEASURED a load of headphones and updated his pdf's

[bobbooo]

Well hopefully Oratory will reply to my follow-up question and perhaps point to the point you make & maybe some others: https://www.reddit.com/r/oratory1990/comments/k7ptcc/_/getjrj0 I have a feeling he won't want to answer the question beyond his earlier general response, combined with my experience in the past that he replies quite quickly & I'm yet to hear back, but we'll see.

In my experience he usually always replies, but he likely gets so many messages he might miss one or two sometimes and not realize for a while - I just got a reply from him to a comment I made over a month ago

 

[kkeretic]

Have you compared it to the settings from Oratory's own measurements? Here:
https://www.dropbox.com/sh/654wxtmh...l-JW7mOnu_9a?lst=&preview=Hifiman+Sundara.pdf

Of course, I tried them first before I found the results for Sundara with new pads. I don't like the sound with any of these params. The problem actually is that all of these params from Oratory and AutoEq draw different curves depending on the equalizer software used. MathAudio Headphone EQ plugin for Foobar2000 seems to draw closest curves to the corresponding equalization curves that come with Oratory and/or AutoEq params. FabFilter Pro-Q 3 draws completely different amplitudes in a second half of the spectrum. Equalizer APO does the same for AutoEq params. What is interesting, Equalizer APO with MathAudio Headphone EQ VST plugin draws a slightly different curve than in Foobar2000 with Foobar2000 plugin, exactly the same params! So I made my params for all of these equalizers and for all of my headphones based on the Oratory frequency response curves using AutoEq and MathAudio Headphone EQ plugin for Foobar2000.

 

[Robbo99999]

Of course, I tried them first before I found the results for Sundara with new pads. I don't like the sound with any of these params. The problem actually is that all of these params from Oratory and AutoEq draw different curves depending on the equalizer software used. MathAudio Headphone EQ plugin for Foobar2000 seems to draw closest curves to the corresponding equalization curves that come with Oratory and/or AutoEq params. FabFilter Pro-Q 3 draws completely different amplitudes in a second half of the spectrum. Equalizer APO does the same. What is interesting, Equalizer APO with MathAudio Headphone EQ VST plugin draws a slightly different curve than in Foobar2000 with Foobar2000 plugin, exactly the same params! So I made my params for all of these equalizers and for all of my headphones based on the Oratory frequency response curves using AutoEq and MathAudio Headphone EQ plugin for Foobar2000.

(EqualiserAPO draws the same EQ Curve as that described in Oratory's pdf file whenever I've compared....so I'm confident that EqualiserAPO is applying the parametric filters properly/accurately.)

In my experience he usually always replies, but he likely gets so many messages he might miss one or two sometimes and not realize for a while - I just got a reply from him to a comment I made over a month ago

No reply from him yet, I'll leave it a few days and then send him another message.

 

[Suuup]

Are you the administrator of that database? There are a lot of people working to produce headphone measurements. What criteria do you look for in the measurement setup or data?

Oh did you create the website and graph comparison tool? If so, thanks, fantastic job! It seems you can't currently use it to directly compare measurements between different rigs though, which could be useful in seeing differences in e.g. responses of various couplers etc. Could this be implemented? You might want to ask 'Resolve' (aka Andrew Park) of headphones.com if you could use his measurements, I believe he uses the same GRAS 43AG as Oratory but with the 'hi-res' RA0402 coupler instead of the RA0045. Plus @crinacle has a growing database of measurements using the same rig as headphones.com, but charges for showing more than one headphone and a target at a time on his own graph comparison tool, so maybe if you had the same limitation he'd allow you to use his data.

Yes, I created the site. I've already talked to Resolve, and he says he's willing to share their data, so hopefully that's coming at some point.

I've also talked to Crinacle, but we haven't found a way to do it yet where both parties are happy.

Tbh I don't have a specific set of criteria, Oratory knows more about setups than I ever will, so I would let him decide.

 

[kkeretic]

(EqualiserAPO draws the same EQ Curve as that described in Oratory's pdf file whenever I've compared....so I'm confident that EqualiserAPO is applying the parametric filters properly/accurately.)

Well, it seems so. Found an interesting article about equalizer differences: Understanding EQ Curves – Why Identical EQ Settings Can Sound Different.

 

[Dreyfus]

https://www.reddit.com/r/oratory1990/comments/k7ptcc/_/gexoxhl

I use a Gras 43AC for in-ear headphones and a 43AG or 45BC (mostly the latter) for everything else. Aside from sealing which is easier on the 43AG, those will deliver identical results and the setup hasn't change in some time.

Looks like the deviation seen in post #1 is just within the standard error induced by seal and placement.

 

[bobbooo]

Of course, I tried them first before I found the results for Sundara with new pads. I don't like the sound with any of these params. The problem actually is that all of these params from Oratory and AutoEq draw different curves depending on the equalizer software used. MathAudio Headphone EQ plugin for Foobar2000 seems to draw closest curves to the corresponding equalization curves that come with Oratory and/or AutoEq params. FabFilter Pro-Q 3 draws completely different amplitudes in a second half of the spectrum. Equalizer APO does the same for AutoEq params. What is interesting, Equalizer APO with MathAudio Headphone EQ VST plugin draws a slightly different curve than in Foobar2000 with Foobar2000 plugin, exactly the same params! So I made my params for all of these equalizers and for all of my headphones based on the Oratory frequency response curves using AutoEq and MathAudio Headphone EQ plugin for Foobar2000.

AutoEQ's settings do not properly follow the Harman target, nor do they take into account non-minimum phase parts of a headphone's frequency response which cannot be effectively EQed out. Oratory's settings do however, so just use the filters in his pdfs using Equalizer APO with Peace GUI. Click the little green graph button below 'Effects' in Peace GUI once you've finished entering in the filter parameters and compare it with the 'EQ Curve total' on Oratory's pdf - if they're different you've entered something in wrong, and post a screenshot of your Peace GUI settings so we can see what's wrong exactly.

 

[Jose Hidalgo]

AutoEQ's settings

Speaking of AutoEQ, I'm having some trouble installing it properly. If anybody can help...

 

[3125b]

Unfortunately the Superlux measurements are still completely wrong due to lack of seal.

 

[kkeretic]

AutoEQ's settings do not properly follow the Harman target, nor do they take into account non-minimum phase parts of a headphone's frequency response which cannot be effectively EQed out. Oratory's settings do however

As I already wrote, I don't like the sound from Oratory's settings on any of the headphones I use, especially on ATH-M50x. So I use a custom settings based on FR curves from Oratory's pdfs + AutoEq corrections. It was a little more challenging to adjust all those parameters (curves) through three different equalizers, but the result seems like my end game, at least until I get some new headphones.
.

 

[kkeretic]

Speaking of AutoEQ, I'm having some trouble installing it properly. If anybody can help...

Which CPU? Tensorflow library could be tricky to make it work if you don't have some newer CPU. You can find Tensorflow libraries (even those for older processors that do not support avx2 instructions) here: https://github.com/fo40225/tensorflow-windows-wheel. This one works for my ten years old Core i7 860, should also work for you, try to install it manually with:

pip install https://github.com/fo40225/tensorflow-windows-wheel/raw/master/2.0.0/py37/CPU/sse2/tensorflow-2.0.0-cp37-cp37m-win_amd64.whl

For the specified version of tensorflow library to work, you also need this: Microsoft Visual C++ Redistributable for Visual Studio 2015, 2017 and 2019.

 

[solderdude]

Unfortunately the Superlux measurements are still completely wrong due to lack of seal.

I have seen this with Tyll's measurements as well. The plots show the HD681 is bass shy. Those that heard it know better. I haven't measured a single HD681 that didn't have elevated bass. In Tyll's case it certainly looked like a seal issue. Can't imagine Oratory making such a mistake yet boosting the elevated bass about 10dB opposite 1kHz makes no sense. I even reduce the bass to get proper sound.

 

[pozz]

Yes, I created the site. I've already talked to Resolve, and he says he's willing to share their data, so hopefully that's coming at some point.

I've also talked to Crinacle, but we haven't found a way to do it yet where both parties are happy.

Tbh I don't have a specific set of criteria, Oratory knows more about setups than I ever will, so I would let him decide.

I have a list I keep of databases and review sites with measurements. Many will be long shots when it comes to your project, but still worth contacting in my mind. As long you have some sort of process for adjusting different types of compensations it makes sense.

• Clarity Fidelity/Speakerphone
• DIY-Audio-Heaven (@solderdude)
• Head-Fi
• Headflux (@bartzky)
• Headphone Test Lab
• Klaus Eulenbach Blog
• Reference Audio Analyzer
• SBAF
• Sonarworks
• SoundStage! (@Doug Schneider)
• SoundStage! Solo (@Doug Schneider)

There are also some good Korean websites:

• Ear-Fi
• UnderKG
• 0dB

Not sure if you've gotten your answer about InnerFidelity yet, but the reviews are being slowly reuploaded to Stereophile, which is owned by AVTech Media Ltd. So private ownership and likely copyright issues.

 

[3125b]

Those that heard it know better.

That's what I thought when I got my 668-B (granted, it has maybe a little less bass than the 681, but not much).
I was listening to it, looked at the graphs and was like: have I gone deaf?
Your measurements reflect my subjective impression. The bass is certainly ample.

 

[bobbooo]

As I already wrote, I don't like the sound from Oratory's settings on any of the headphones I use, especially on ATH-M50x. So I use a custom settings based on FR curves from Oratory's pdfs + AutoEq corrections. It was a little more challenging to adjust all those parameters (curves) through three different equalizers, but the result seems like my end game, at least until I get some new headphones.
.

You said his settings produce different results with different equalizers which obviously shouldn't be the case, so I was just making sure you have definitely checked that you got the correct total EQ curve matched exactly with the ones on Oratory's pdfs when using Equalizer APO / Peace GUI before making a judgement on their sound.

 

[Robbo99999]

https://www.reddit.com/r/oratory1990/comments/k7ptcc/_/gexoxhl


Looks like the deviation seen in post #1 is just within the standard error induced by seal and placement.

Ah, yes, Oratory has replied to my post over on reddit re explaining how the new HD600 results are quite different and also answering the question re the different measuring equipment reference numbers. I'm satisfied with his answer now, he's really done a ton of measurements, see the following pic! And his explanation on the different measuring equipment reference numbers eased my prior doubts as well. Good job Oratory!

I do like the EQ I did based on this his newest measurement (more so than the older one), so I think his extra units he's measured along with multiple seatings has increased the accuracy of his measurement (although I'm aware he did multiple seatings on his old measurement too).

So we've got our answers to the questions I was posing in the OP.

 

[Feelas]

Speaking of AutoEQ, I'm having some trouble installing it properly. If anybody can help...

Seeing you have x86-64 OS (judging by Program Files (x86)), you're on proper OS version so the CPU shouldn't matter much - some stuff won't run and TF could run slow w/o AVX2. Yet, this is not the problem. It seems you're using x86 python (win32 wheels are downloaded). Try grabbing Windows x86-64 executable installer of Python and installing the x64 version instead, then modifying %PATH% to mention the new installed version. Please ping back if it works.

 

[Jose Hidalgo]

Seeing you have x86-64 OS (judging by Program Files (x86)), you're on proper OS version so the CPU shouldn't matter much - some stuff won't run and TF could run slow w/o AVX2. Yet, this is not the problem. It seems you're using x86 python (win32 wheels are downloaded). Try grabbing Windows x86-64 executable installer of Python and installing the x64 version instead, then modifying %PATH% to mention the new installed version. Please ping back if it works.

Hi Feelas. kkeretic has helped me a bit towards resolution. Apparently the problem lies in my old CPU (Core i3 530) + Windows 7, which don't allow me to download tensorflow 2.0 which is required by AutoEQ. I guess I'll have to move to Win 10 sooner or later. Let me know if I missed something. You guys are the experts !

 

[Feelas]

Hi Feelas. kkeretic has helped me a bit towards resolution. Apparently the problem lies in my old CPU (Core i3 530) + Windows 7, which don't allow me to download tensorflow 2.0 which is required by AutoEQ. I guess I'll have to move to Win 10 sooner or later. Let me know if I missed something. You guys are the experts !

I think moving to x86-64 Python should be good enough, so I recommend to try it anyways

 

[Jose Hidalgo]

I think moving to x86-64 Python should be good enough, so I recommend to try it anyways

With that said, you have just given me a download link for Python 3.9, and AutoEQ states clearly that even Python 3.8 doesn't work yet.