[jaakkopasanen]

Been using your work since the early day now, nice stuff to have.
I am wondering if I can push my own raw measurement into this project. It won't be as simple as just adding the file and execute the command isn't it?

I'm afraid that not supported. I have the selected sources in AutoEq and can't realistically support other random measurements. You can of course use your own measurements to create eq settings for yourself but it's not possible to get them shared to the community via AutoEq.

 

[Banbeucmas]

You can of course use your own measurements to create eq settings

This is actually what I meant when I said pushing measurement into the project. Though I assume you mean creating EQ settings to other earphone within the same source than to a scientific target right? (Providing if I my raw target are compliment to IEC-711 with in ears for example)

 

[jaakkopasanen]

This is actually what I meant when I said pushing measurement into the project. Though I assume you mean creating EQ settings to other earphone within the same source than to a scientific target right? (Providing if I my raw target are compliment to IEC-711 with in ears for example)

You can save your measurement to my_data, for example my_data/AYA Siren/AYA Siren.csv. Then just point the input dir to it --input_dir=my_data/AYA Siren. AutoEq can read almost any CSV file as long as the first column is the frequency data and the second column is the amplitude (gain) data.

You can use Harman targets if you're measurement system is compatible with them, otherwise you need to figure out the target yourself.

 

[Banbeucmas]

You can save your measurement to my_data, for example my_data/AYA Siren/AYA Siren.csv. Then just point the input dir to it --input_dir=my_data/AYA Siren. AutoEq can read almost any CSV file as long as the first column is the frequency data and the second column is the amplitude (gain) data.

You can use Harman targets if you're measurement system is compatible with them, otherwise you need to figure out the target yourself.

I see, thanks for the heads up

 

[Heldaeus]

Is there a way to calculate Harman preference scores for different target curves and export those results to a separate .md file? Let's say, for example, the "harman_over-ear_2018_wo_bass" target.

 

[jaakkopasanen]

Is there a way to calculate Harman preference scores for different target curves and export those results to a separate .md file? Let's say, for example, the "harman_over-ear_2018_wo_bass" target.

You can change the target curve file path in results/update_indexes.py#L251. After than just run the script python results/update_indexes.py. It will take some time because it updates everything, including the HeSuVi ZIP.

 

[Heldaeus]

You can change the target curve file path in results/update_indexes.py#L251. After than just run the script python results/update_indexes.py. It will take some time because it updates everything, including the HeSuVi ZIP.

(venv) C:\AutoEq-master>python results\update_indexes.py
Traceback (most recent call last):
File "results\update_indexes.py", line 15, in <module>
from measurements.manufacturer_index import ManufacturerIndex
File "C:\AutoEq-master\measurements\manufacturer_index.py", line 6, in <module>
from rapidfuzz import fuzz
ModuleNotFoundError: No module named 'rapidfuzz'

What's going on here?

 

[jaakkopasanen]

(venv) C:\AutoEq-master>python results\update_indexes.py
Traceback (most recent call last):
File "results\update_indexes.py", line 15, in <module>
from measurements.manufacturer_index import ManufacturerIndex
File "C:\AutoEq-master\measurements\manufacturer_index.py", line 6, in <module>
from rapidfuzz import fuzz
ModuleNotFoundError: No module named 'rapidfuzz'

What's going on here?

Aha. You need to install some extra dependencies. There's another requirements file in measurements folder.

 

[Heldaeus]

Aha. You need to install some extra dependencies. There's another requirements file in measurements folder.

As I figured, thank you.

 

[HiFiHawaii808]

It seems I have totally forgotten to update the project status here. Since the last replies of this thread a lot has happened. The number of covered headphone models is now at over 2400. I've added Rtings, oratory1990, Reference Audio Analyzer and now Crinacle's measurements and pre-computed results for them.

Different type of equalizers have gained support too. AutoEQ produces settings for basically all types of equalizers. There are parametric equalizer filters optimized, fixed band (graphic) equalizer filters optimized and impulse responses for minimum phase and linear phase filters. This means that any device or platform which has some type of equalizer is covered. However out of all possible fixed band equalizers only 10-band EQs are supported in the pre-computed results but others can be generated with little effort.

This is a fantastic project. Well done.

I am trying to figure out the best way to test this AutoEQ presets for my HD800S headphones. I see that there is a 10-band graphic equalizer profile. A 5-band parametric EQ profile and a 10-band parametric EQ profile.

I listen mostly to Tidal Hifi and all of my devices stream over wifi instead of using USB. I will first test AutoEQ with my laptop using Peace and EqualizerAPO as originally recommended. But, I need a long term solution that does not use USB.

I am thinking a DAP connected to my Hugo 2 would be the best solution. I was wondering if you could provide any feedback on which implementations work the best. I could also purchase a hardware graphic equalizer.

I've seen the Wavelet App, but that requires Android 9 and most DAPs only support 8.1. The new Hiby R6 2020 supports Android 9, but does not do parametric EQ. It does do graphic equalizer. And, there is a thing called Convolution that someone uses these AutoEQ settings.

Anyway, can you provide any input on which implementations either work the best or sound the best? Does Wavelet or Convolution do your project justice?

 

[batfier]

@jaakkopasanen

Thanks so much for your pretty nice work! Love it!

Having these measurements consolidated in one structure is a gift.

Getting EQ settings out of it is even better.

Before I discovered AutoEQ, I was using the oratory1990 para EQ setting for the Stax 009 and already liked it. The AutoEQ Convolution *minimalphase.wav files are even better! Feels more clear, precise, 3D to me.

Currently I use EQ (parametric, Convolution) in Roon and it works perfectly. Can also recommend the paraEQ implementation in RME ADI-2 pro fs and in Auralic Aries.

 

[Heldaeus]

Aha. You need to install some extra dependencies. There's another requirements file in measurements folder.

I got everything working properly and was able to compute the rankings for several different targets. So, thanks for that!

One question, I created a hybrid target that combines the Harman 2013 bass with the mid-range and treble of the Harman 2018 target and I saved the .csv file as "harman_over-ear_combined.csv". I tried to compute the preferance rankings based on this custom target but I ran into some problems. Here is the error report and what do I need to change to solve this?

(venv) C:\AutoEq-master>python results\update_indexes.py
Traceback (most recent call last):
File "results\update_indexes.py", line 318, in <module>
main()
File "results\update_indexes.py", line 314, in main
write_ranking_table()
File "results\update_indexes.py", line 262, in write_ranking_table
row = ranking_row(fp, harman_overear, 'onear')
File "results\update_indexes.py", line 235, in ranking_row
fr.compensate(target, bass_boost_gain=0.0) # Pre-computed results are with Harman target without bass
File "C:\AutoEq-master\frequency_response.py", line 1066, in compensate
tilt=tilt
ValueError: operands could not be broadcast together with shapes (689,) (695,)

 

[jaakkopasanen]

This is a fantastic project. Well done.

I am trying to figure out the best way to test this AutoEQ presets for my HD800S headphones. I see that there is a 10-band graphic equalizer profile. A 5-band parametric EQ profile and a 10-band parametric EQ profile.

I listen mostly to Tidal Hifi and all of my devices stream over wifi instead of using USB. I will first test AutoEQ with my laptop using Peace and EqualizerAPO as originally recommended. But, I need a long term solution that does not use USB.

I am thinking a DAP connected to my Hugo 2 would be the best solution. I was wondering if you could provide any feedback on which implementations work the best. I could also purchase a hardware graphic equalizer.

I've seen the Wavelet App, but that requires Android 9 and most DAPs only support 8.1. The new Hiby R6 2020 supports Android 9, but does not do parametric EQ. It does do graphic equalizer. And, there is a thing called Convolution that someone uses these AutoEQ settings.

Anyway, can you provide any input on which implementations either work the best or sound the best? Does Wavelet or Convolution do your project justice?

Parametric vs convolution should not make that big ofa a difference. Wavelet is very good app. It doesn't use convolution but some different implementation, maybe large number of high shelf filters, I'm not sure. Qudelix 5K DAC could be an option, it has 10 band parametric eq. MiniDSP also has some hardware solutions. I guess Hiby R6 would work if it's compatible with Wavelet.

@jaakkopasanen

Thanks so much for your pretty nice work! Love it!

Having these measurements consolidated in one structure is a gift.

Getting EQ settings out of it is even better.

Before I discovered AutoEQ, I was using the oratory1990 para EQ setting for the Stax 009 and already liked it. The AutoEQ Convolution *minimalphase.wav files are even better! Feels more clear, precise, 3D to me.

Currently I use EQ (parametric, Convolution) in Roon and it works perfectly. Can also recommend the paraEQ implementation in RME ADI-2 pro fs and in Auralic Aries.

Glad to hear!

I got everything working properly and was able to compute the rankings for several different targets. So, thanks for that!

One question, I created a hybrid target that combines the Harman 2013 bass with the mid-range and treble of the Harman 2018 target and I saved the .csv file as "harman_over-ear_combined.csv". I tried to compute the preferance rankings based on this custom target but I ran into some problems. Here is the error report and what do I need to change to solve this?

(venv) C:\AutoEq-master>python results\update_indexes.py
Traceback (most recent call last):
File "results\update_indexes.py", line 318, in <module>
main()
File "results\update_indexes.py", line 314, in main
write_ranking_table()
File "results\update_indexes.py", line 262, in write_ranking_table
row = ranking_row(fp, harman_overear, 'onear')
File "results\update_indexes.py", line 235, in ranking_row
fr.compensate(target, bass_boost_gain=0.0) # Pre-computed results are with Harman target without bass
File "C:\AutoEq-master\frequency_response.py", line 1066, in compensate
tilt=tilt
ValueError: operands could not be broadcast together with shapes (689,) (695,)

Could be that AutoEq is not interpolating targets. You might be able to get your target interpolated to the standard frequencies by doing one mock equalization using it as the input and setting --standardize_input. That interpolates the input data and replaces the original file.

 

[Heldaeus]

Could be that AutoEq is not interpolating targets. You might be able to get your target interpolated to the standard frequencies by doing one mock equalization using it as the input and setting --standardize_input. That interpolates the input data and replaces the original file.

This didn't affect the results. I'm still recieving the same error message. I compared the new interpolated data (BTW I created the hybrid target by stitching the 2013 and 2018 .csv data together, not by using WebPlotDigitizer) against pre-computed .csv data (harman_over-ear_2018) by subtracting the frequency entries from both data sets from one another; all results were "0".

Secondly, I replaced the "update_indexes.py" file with an unedited back-up and again, same errors.

 

[Okiba]

Amazing job! Thank you! I own some old Ultrasone HFI-550 headphones. I couldn't find it on the list. I assume because those are a bit too old to have response graph available? that's the basic thing I will need to produce EQ results?