• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

List of Amir's Headphone PEQ filters

Here is a thing I have been racking my brains over.

I inserted a GraphicEQ.txt for my DT-770 into Equalizer APO, changed "variable" to "31-band" and tried to apply these 31 filters to the 31-band plugin in foobar. The last two filters should be "16000: -13.14" and "20000: -16.5" but in the plugin you cannot go lower than "-12", so
I added +4.5dB to every 31 filters (thus changing the least "-16.5" to "-12"). Then the "Auto Volume" (to prevent a distortion, I suppose) added "-3.6 dB."

And I am now thinking: did my mathematical exploit make any sense at all after that "Auto" thing got everything back? Shouldn't I have simply put the original 29 filters as it is, and the last two made "-12"?

Thanks in advance.
 

Attachments

  • auto.png
    auto.png
    16 KB · Views: 177
Here is a thing I have been racking my brains over.

I inserted a GraphicEQ.txt for my DT-770 into Equalizer APO, changed "variable" to "31-band" and tried to apply these 31 filters to the 31-band plugin in foobar. The last two filters should be "16000: -13.14" and "20000: -16.5" but in the plugin you cannot go lower than "-12", so
I added +4.5dB to every 31 filters (thus changing the least "-16.5" to "-12"). Then the "Auto Volume" (to prevent a distortion, I suppose) added "-3.6 dB."

And I am now thinking: did my mathematical exploit make any sense at all after that "Auto" thing got everything back? Shouldn't I have simply put the original 29 filters as it is, and the last two made "-12"?

Thanks in advance.

Let's break this down...

Solution #1:
In Equalizer APO, I added a GEQ with variable filters, then imported your linked 127-band profile, and it showed me a peak gain of -0.2dB (which means that a pre-amp of +0.2dB is needed for these filters). Which means it won't digitally clip regardless of how loud you turn it up.
1613441950185.png
Then, I converted this to a 31-band GEQ (by clicking on the "31-band")
1613442108396.png
Here's the generated 31-band GEQ profile:
20 -9.7
25 -10.3
31.5 -10.88
40 -11.3
50 -11.4
63 -11
80 -8.99
100 -4.2
125 -8.89
160 -7.2
200 -1.53
250 -1.77
315 -3.47
400 -4.07
500 -4.38
630 -4
800 -4.58
1000 -5.73
1250 -4.2
1600 -4.3
2000 -4.45
2500 -5.07
3150 -0.93
4000 -3.68
5000 -5.43
6300 -9.96
8000 -8.4
10000 -8.59
12500 -10.27
16000 -13.14
20000 -16.5

Which changed all the filters accordingly, and it also changed the peak gain of -0.9dB (which means a pre-amp of +0.9dB is needed for these filters) since I saw:
1613442155191.png
This means that regardless of how loud you listen to this profile, it will never digitally clip.

Now, like you said, Foobar2000 has a limitation of only allowing a -12dB of gain for each filter.
Which means the last 2 frequencies are over the limit now...
So we have to change:
"16000 -13.14"
"20000 -16.5"
to
"16000 -12"
"20000 -12"
To make it compatible with Foobar2000.

Now, let's graph this profile in Foobar2000:
1613442466308.png
This shows a pre-amp of +0.9dB is needed for these filters, which matches the Equalizer APO's pre-amp of +0.9dB as well.

The problem, however, is that the last 2 filters weren't meant for -12dB.

Basically, solution #1 won't reduce the last 2 high frequencies as much as we'd like.
So let's take a look at solution #2.


Solution #2:
Like you had the idea, we offset all filters by the difference of the last frequency (which is -16.5dB going to -12dB, which is +4.5dB).

So, I just offset all filters by +4.5dB and got:
20 -5.2
25 -5.8
31.5 -6.38
40 -6.8
50 -6.9
63 -6.5
80 -4.49
100 0.3
125 -4.39
160 -2.7
200 2.97
250 2.73
315 1.03
400 0.43
500 0.12
630 0.5
800 -0.08
1000 -1.23
1250 0.3
1600 0.2
2000 0.05
2500 -0.57
3150 3.57
4000 0.82
5000 -0.93
6300 -5.46
8000 -3.9
10000 -4.09
12500 -5.77
16000 -8.64
20000 -12
Then I imported into Equalizer APO and saw that peak gain of 3.6dB (which means a pre-amp of -3.6dB is needed for these filters).
1613443671956.png
Which of course will match Foobar2000's pre-amp of -3.5db (close enough) as well.
1613443919312.png

Solution #2 gives you a more accurate profile, since the last 2 frequencies aren't truncated like in solution #1.

I hope that helps. :)
 
The program comes with a "pair of filters, pair of shelves and two mid-frequency bell filters."
So...make the filters manually.
I understand that. What I don't understand is what all these "shelves" and "bells" mean, and how one correlates those things with PEQ and GEQ filters, but thanks anyway.
 
I understand that. What I don't understand is what all these "shelves" and "bells" mean, and how one correlates those things with PEQ and GEQ filters, but thanks anyway.
If I understood correctly...
1613457776493.png
When the program shows => It means
Low cut => low pass filter (or maybe a high pass filter just for lower frequencies)
High cut => high pass filter (or maybe a high pass filter just for higher frequencies)
Low shelf => low shelf filter
High shelf => high shelf filter
Low-Mid Bell => peaking filter #1 (for certain range of frequencies)
High-Mid Bell => peaking filter #2 (for certain range of frequencies)

But I can't verify, since I didn't install it.

Obviously, this is not as flexible as the PEQ (or even GEQ) profile, so the functionality is limited.
There's no algorithm that I know that will create a profile with these requirements, so just do the best you can to match this program with original PEQ profile.
 
Last edited:
OBSERVATION...

Remember how I showed you how to use 'Equalizer APO' to calculate the pre-amp needed for a given set of PEQ filters?
Here's the recap:
1. Take the PEQ profile for 'Sony MDR-ZX110' that Amir created
Filter 1: ON LS Fc 40 Hz Gain 8.0 dB Q 1.0
Filter 2: ON PK Fc 1000 Hz Gain 5.0 dB Q 3.0
Filter 3: ON PK Fc 2500 Hz Gain 5.0 dB Q 2.0
Filter 4: ON PK Fc 5900 Hz Gain 7.0 dB Q 2.0
2. Import into 'Equalizer APO'
3. Here's the result
1614026481157.png
Which shows you a pre-amp of -8.6 dB is needed.
4. The first question I had was...is this really -8.6 dB...let's zoom in to tell...
1614026610280.png
Yes...you can see by the mouse pointer that the highest value is 8.6 dB, which corresponds to a frequency of 23.2hz

Now, this is new...
Yesterday, 'PEACE' interface just got updated with a new version 1.5.7.0 (changelog).
1. Open Peace
2. I imported the same profile into 'Peace' I saw this:
1614026757359.png
And if you graph it:
1614026804243.png
3. Notice the 7.9 dB value...doesn't that look like the pre-amp that we need?
Except it doesnt match 'Equalizer APO' peak gain of 8.6 dB.
4. I wanted to verify that this is indeed the pre-amp, so I adjusted the pre-amp to -7.9 dB and graphed it again, then offset the graph to see that it worked.
1614027068540.png
As you can see, it properly offset the filters, and now changed the value on top to be '0.0 dB'.

Interesting finding!

So what's the conclusion?
1. 'Peace' changelog didn't say anything, it now includes an automatic pre-amp calculator.
2. This explains what's going on.
3. I'm going to stick with 'Equalizer APO's automatic calculation, just like before, for consistency, of course.
 
Here are my subjective comparisons of Sennheiser HD650 and Beyerdynamic DT990 Pro (I am using Amir's PEQ filters).
HD650 provides better timbre of voice reproduction and clearer sound. DT990 Pro provides better vibrato reproduction and therefore gives a more pouring sound.
1955 Rigoletto (Serafin, Di Stefano, Gobbi, Callas) I prefer with DT990 Pro.
1955 Lucia di Lammermoor (Karajan, Panerai, Callas, Di Stefano) I prefer with HD650.
 
Improvements...
How to use 2 DACs on 1 Windows 10 PC, with separate PEQ profiles?
Let's say you want to use separate DACs (and have headphone amplifiers & headphones OR studio monitors hooked up) on 1 Windows 10 pc and want the ability to EQ both DACs separately, using 'Equalizer APO' for Windows.
I've created a how-to-guide here.
 
Improvements...
What happens if I install 'Equalizer APO' and setup everything up correctly, but it still doesn't work?
Backstory:

1. While setting up 'Equalizer APO' for a friend, I noticed that 'Equalizer APO' wasn't working as intended.
2. He had a pair of studio monitors (with XLR female inputs), and connected them using "dual XLR male to 3.5mm TRS male cable" to his desktop computer's "Line Out" port on the motherboard. His motherboard had a Realtek audio chipset.
3. After installing, properly setting up 'Equalizer APO'. I imported a PEQ profile as I normally would, and tested it. However, I noticed that togging the PEQ profile on/off didn't change the original sound.
4. Of course, I checked to make sure that he had the latest Realtek audio drivers installed. That didn't fix the problem.
Solution:
1. Open "Configurator" (located in C:\Program Files\EqualizerAPO)
2. Enable Check "Troubleshooting Options" box at the bottom
3. Select your primary device (in my case, it was just the Realtek Audio device)
4. Change the drop-down menu from "Install as LFX/GFX (recommended)" to "Install as SFX/EFX (experimental)".
5. Press "OK", and restart computer.
6. After the restart, 'Equalizer APO' immediately worked.
I don't know whether this is just related to Realtek chipset or what. I hope this helps some of you.
 
Last edited:
This is a Parametric Equalization (PEQ) guide based on Amir's excellent headphone reviews! ...

I have a tangential question to the PEQs published in this thread. Specific to how they are calculated. Do you happen to know if there is a "Harman Target CSV file" that Amir uses in his graphs, in the form that AutoEQ uses for an input compensation file, that I can download from somewhere and use?

The reason I ask is the target curve he publishes in the frequency response images in his headphone reviews posts has nowhere near as steep high frequency rolloff, as the compensation files provided in the AutoEQ distribution. Wondering if the target Amir is comparing his measured frequency responses to, and EQ matching to, is indeed different than the .csv target files I find in the AutoEQ github distribution for Harman or Oratory or Crinacle 'targets'. And if it is different, I'd like to run some AutoEQ passes and try his target out, and see if I like that better than the targets I've been using.

Hope this doesn't read as a word salad above and you understand what I'm asking about. Much Thanx.
 
I have a tangential question to the PEQs published in this thread. Specific to how they are calculated. Do you happen to know if there is a "Harman Target CSV file" that Amir uses in his graphs, in the form that AutoEQ uses for an input compensation file, that I can download from somewhere and use?

The reason I ask is the target curve he publishes in the frequency response images in his headphone reviews posts has nowhere near as steep high frequency rolloff, as the compensation files provided in the AutoEQ distribution. Wondering if the target Amir is comparing his measured frequency responses to, and EQ matching to, is indeed different than the .csv target files I find in the AutoEQ github distribution for Harman or Oratory or Crinacle 'targets'. And if it is different, I'd like to run some AutoEQ passes and try his target out, and see if I like that better than the targets I've been using.

Hope this doesn't read as a word salad above and you understand what I'm asking about. Much Thanx.
I don't know what target curve Amir is using (@amirm would know best)... but why not just digitize Amir's "target curve" using webplotdigitizer, then export frequency/db as file.csv, then use this csv file as the new "compensation" curve, and run the AutoEQ algorithm against that?

Yeah, it's not as perfect as having the raw curve that Amir uses, but it's better than nothing, and it will take you about 5min to do it.
You could compare it to the other "compensation" curves and see how close it is.
 
I don't know what target curve Amir is using (@amirm would know best)... but why not just digitize Amir's "target curve" using webplotdigitizer, then export frequency/db as file.csv, then use this csv file as the new "compensation" curve, and run the AutoEQ algorithm against that?

Yeah, it's not as perfect as having the raw curve that Amir uses, but it's better than nothing, and it will take you about 5min to do it.
You could compare it to the other "compensation" curves and see how close it is.
Ha! Just so happens I'm doing that, for another reason. Yep. Sounds good. ;)
 
I found out something new about 'Peace' and how it interacts with 'Equalizer APO'....:D

1. So let's say you just install 'Equalizer APO' by itself (no 'Peace'), then if you open it up, it shows you that it's using a file called config.txt from "c:\program files\equalizer apo\config". You can modify this file using either the GUI or the config.txt directly. Nothing new here, as I've showed this previously in my video. For example, in the video, I showed that I was using the akg 371 amir.txt, where the PEQ profile was stored.
So my config.txt of course would be:
Include: akg 371 amir.txt
2. Now, let's say you install 'Peace', and launch it, and as soon as you do, this 'Equalizer APO' configuration file called config.txt changes from it's previous configuration to new code:
Include: peace.txt
What is this peace.txt? This is the PRIMARY EQ profile that you have in 'Peace'. If you have dozens of them, like I mentioned in the video, they are stored in the same directory as 'files.peace'.
My peace.txt looked like:
Device: all
Channel: all
Filter 1: ON HP Fc 120 Hz
Which is my primary EQ profile in Peace.
This tells me that 'Peace' is taking control over 'Equalizer APO' and is now the primary source of filters, as expected. So even if you open 'Equalizer APO' when 'Peace' was running, the system would still work normally, since 'Equalizer APO' simply referenced the new peace.txt.
3. The new functionality I discovered today is.... IF you quit Peace (say by accident), your previously referenced PRIMARY EQ profile is still running in the background. Meaning, even if you quit Peace, it's last EQ profile doesn't actually stop working.
What's actually happening in the background when you quit 'Peace'?
'Equalizer APO' of course takes over and it looks at the file 'Equalizer APO' configuration file called config.txt, which still has the code:
Include: peace.txt
Which you can see...is still referencing the last known 'Peace' EQ profile.

Cool! :D
 
Just a slightly off topic question on equalization...

Is there a difference between EQ-ing to the Harman target some headphones that start out close to the target, and some headphones that start out quite a way away from the target?

So, for example, the Sennheiser HE-1 is very close to Harman. The Stax SR-009 is not. I own the latter. If I equalize it to Harman, will it sound materially the same as some HE-1s equalized to Harman? (Leaving aside differences in the headphones - they are both very high quality, low distortion electrostats).
 
Back
Top Bottom