Head-Fi.org › Forums › Equipment Forums › Computer Audio › Calibrating graphic EQs, the easy way
New Posts  All Forums:Forum Nav:

Calibrating graphic EQs, the easy way

post #1 of 6
Thread Starter 

There was a really good thread on tuning parametric equalizers

http://www.head-fi.org/t/413900/how-to-equalize-your-headphones-a-tutorial

 

but for those who rely on the more common graphic equalizer there isn't a similar thread for setting it up.

 

I took ideas from that thread and started applying them to graphic equalizers.

 

Let's take the classic 10 band EQ with bands at 31, 62, 125, 250, 500, 1000, 2000, 4000, 8000 and 16,000Hz.

 

Apart from tuning by ear on music, one method some people tried was to play sine tones at those ten frequencies and adjust the EQ until all the tones sound equally loud.  There are two problems with this approach:

1. The human ear does not hear tones of the same SPL at different frequencies as having the same volume.  This is explained here:

http://en.wikipedia.org/wiki/Equal-loudness_contour

 

 

2. A band on the graphic EQ affects more than the labelled single frequency.  The 10-band equalizer is also known as a 1-octave equalizer because each band affects a whole octave.  Within that octave, especially in high frequency bands for headphones, there can be wild ups and downs in the frequency response of the headphone-ear system.  So a tone at 8000Hz may be very unrepresentative of the loudness of the EQ band which spans e.g. from 6000 to 12000Hz. (I don't know the exact start and end frequencies, but the latter is double the former.)

 

To fix both of these problems, I designed the following test tones.  Pink noise is filtered through an EQ matching the equal loudness contour at a musical volume, then it is bandlimited at different ranges of frequencies corresponding to the 10 EQ bands to produce 10 test sounds.  You just need to adjust the EQ until the ten test sounds all sound as loud as each other at the volume you usually you play your music at, and the result should be a well compensated headphone frequency response.

 

http://www.mediafire.com/?esk01upu37deu

 

The sounds are labelled 10-xxx.wav, where xxx stands for the center frequency of the corresponding EQ.  There's also a "10 bands" file, which is just all the 10 test sounds put together, going from 31Hz to 16kHz.

 

There's a zip file that contains all the test tones:

http://www.mediafire.com/?93d3ux1xolowa61

 

I recommend starting with 10-1000 and adjusting the volume until it sounds about as loud as you usually play music.  Then go through the other test sounds adjusting the corresponding EQ sliders until they all sound equally as loud as the 1000 test sound.  For safety's sake I recommend turning down the 8k and 16k EQ sliders before you start, as with treble-happy earphones the test sounds for those bands can be really piercing.  Also if your EQ has a preamp slider, push it a ways down before beginning the calibration; this will give you headroom to boost frequencies without clipping.

 

I have also made a 5 band version of the test sounds for the 5-band Android Equalizer app:

http://www.mediafire.com/?bg3qs8xg52sbsd3

although for this first version I made the top band seems too hot so you may end up EQing the 14kHz band down too much by referencing these test tones.  Use these test tones to calibrate Equalizer by loading the tones into a media player, disabling the media player's EQ, enabling Android Equalizer, then play the test tone in a loop.


Edited by Joe Bloggs - 3/19/13 at 10:00am
post #2 of 6

Definitely will try this...I have gotten very good results from the parametric equalizer approach to headphone equalization mentioned in Joe Bloggs posting above, and this would allow a similar benefit for situations where parametric equalizers are not available.

 

For anyone who has not tried the original parametric equalizer approach - the htread makes fascinating reading and (for me anyway) really increased my enjoyment of a wide range of music through headphones. I have also used the equalizer settings that result from the process to produce alternative 'equalized' versions of my favorite albums in a variety of formats for those situations where NO equalizer is available.    Ian

post #3 of 6
Thread Starter 
Played around with the test tones I generated and found thatthe EQ on Poweramp for android has some really funky characteristics. The top treble bands are shifted down one octave in the frequencies they control (eg 16k actually works on around 8k and 8k on 4k) and cuts to those bands dont really work (if you want a treble cut you really have to boost all the other freqs instead). Explains all the problems I had with it before!
post #4 of 6
Thread Starter 

I did a version of the graphic EQ test sounds for 23 bands

http://www.mediafire.com/?h7ulxyi0sdoxwwu

 

But there's no graphic EQ out there with 23 bands.  So what gives?

 

I think these tones can be used to calibrate a multiband parametric EQ like Electri-Q and using these 23 test sounds will yield results that are indistinguishable from the method I've been using all along with pure tones (and that I tried to explain in my never-finished beast of an EQ tutorial thread...) or PiccoloNamek's method.

 

With testing using pure tones (especially with full size headphones) one can hear countless peaks and dips in the 6-12kHz range, such that one can never hope to EQ them all out.  Never mind that the peaks and dips would shift each time you put on the phones differently.  And to top it off, our pitch perception in this range isn't that good, and in music there is rarely pure tones in this range such that even if the EQ could be made that precise, it probably can't be heard.  I'd experienced this myself, replacing the precisely sculpted peaks and dips of the SHE3580 EQ with a gentle curve approximating the loudness level of the general range of frequencies, and the difference in the perceived sound is minimal.

 

Using these files to calibrate a parametric EQ like Electri-Q is simple:

1. Download Electri-Q and George Yohng's VST wrapper and install them for foobar2000

2. Put Electri-Q into foobar's DSP chain

3. Load the group of test sounds into a playlist in foobar

4. Make 23 control points in Electri-Q with frequencies and BW settings corresponding to the filenames of the test sounds: e.g. 60 1.8.wav stands for making a point at 60Hz with BW=1.8

5. Play the 1000 1.5.wav tone and adjust the volume until it matches the volume at which you listen to music

6. Go through the list of test sounds and adjust the corresponding control points on the EQ (hold down shift while clicking on a point to change the gain / cut of each band by dragging up and down without changing the frequency by accident) until all the sounds sound equally as loud as the 1000 1.5.wav sound.  Make sure that the whole EQ curve stays below the 0dB line.  If boosting above 0dB is needed, create a "gain only" control point (right click on a new point->Basic->Gain only) to drag the whole curve down and increase the playback volume to compensate (so that 1000 1.5 plays as loud as before)

 

That's it!  No need for VST hosts, audio loopback or anything and less than a quarter of the steps of the previous method!

post #5 of 6
Thread Starter 
bump?
post #6 of 6

Hmmm... Can I use this technique with RockBox and Fiio's EQs in X3 and X5?

New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Computer Audio
Head-Fi.org › Forums › Equipment Forums › Computer Audio › Calibrating graphic EQs, the easy way