In this app note, we will walk through EQ-ing your headphone based on measurements taken with the miniDSP EARS measurement rig.
To implement the EQ, the app note will assume that you have a miniDSP HA-DSP headphone amplifier or another miniDSP device such as the 2x4 HD. However, the technique can be used with many other devices and also with software implementations, as long as they have a full parametric EQ capability.
For devices that don't have a full parametric EQ or that can't have an REW measurement signal run through them, see the related app note Headphone EQ for iPhone (or Android, or DAP).
Why EQ?
There are a lot of factors that affect a headphone's sound, but the frequency response is the primary one. It determines not just the overall balance of bass, midrange and treble, but also the distribution of harmonics that are part of every musical note.
Headphones (and IEMs) are complicated things when it comes to their acoustics. Unlike a speaker, which can be measured for accuracy in an anechoic chamber, headphones can only be measured when they are on a "head" (either real or artificial). Sound bounces around in the tiny confined space between the headphones and your ears, and travels down the ear canal until it eventually reaches the eardrum.
Professional artificial measurement "heads" attempt to simulate this environment, and as result are very expensive. Despite their cost, they still produce very different results, as shown in the graphs in this article by Audeze. Furthermore, we all have different anatomy: studies show the acoustic signal at the eardrum varying by 5 dB in the bass, for example. Factors such as the positioning of the headphone, the seal of the pad to the head, and so on, will change the measurements. Finally, different people have different subjective preferences.
So, when you run headphone measurements for EQ, it's important to treat the measurements as a guide. You always need to listen to the result of any EQ based on the measurements, and adjust it so that it sounds right to you.
Some additional points:
-
Because there are a lot of factors affecting the sound of a headphone, you probably won't be able to EQ a headphone you don't like into one that you do. It's best to start with something that basically sounds good to you.
-
The headphone designer(s) will have traded off all of those factors when designing the headphone. So correcting only the frequency response may not produce the expected result. Again, you need to listen for yourself and decide how to apply the EQ.
-
Remember Fletcher-Munson! Our ears' sensitivity to high and low frequencies depends on the SPL, so do your listening adjustments at the level you normally listen. (Don't be afraid to add some bass and treble if you prefer to listen at lower levels.)
-
In this app note, we suggest using just a few bands of EQ, say no more than 5. This will make it easier to make manual adjustments when auditioning.
-
Don't try and flatten every little bump and "wrinkle" in the measured response. Just work to make the response closer to flat.
-
Regardless of what the measurements say, work within the natural capabilities of the headphone. A bass-light headphone may not respond well to a lot of boost in the bass, for example.
-
Any modifications or "mods" to the headphone will change the response, so you may need to redo your EQ. Even different earpads change the response. (You can measure the change with EARS.)
The basics of parametric EQ
There are two main types of EQ filter. A peaking filter boosts or cuts frequencies around the center frequency. This example (from the HA-DSP) shows a 10 dB boost at 200 Hz with a wide bandwidth (Q=0.7) and a 10 dB cut at 2000 Hz with a fairly narrow bandwidth (Q=4):
A shelving filter boosts or cuts the signal above or below the set frequency. Shelving filters are typically used to "bring up the bass" or "take down the treble". This next example shows a shelving boost of 6 dB (Q=0.7) centered at 100 Hz, and a shelving cut at 2000 Hz (Q=0.5):
Measuring the headphones
The EARS page on the miniDSP website now has a compensation type named "HEQ", for "Headphone EQ". Go there now, enter your serial number and select HEQ to download the calibration files. These calibration files help simplify the process of EQing headphones.
Now run a measurement on your headphones, loading the file L_HEQ_xxxyyyy.txt into REW for the left and R_HEQ_xxxyyyy.txt for the right. (See the User Manual for details.)
Use the All SPL tab in REW to create an average of the left and right measurements. Optionally, you can take several measurements on each side with the headphone moved a bit each time, and average all of them.
The intent of the HEQ compensation is that, when it is used to measure a "neutral sounding" headphone, the resulting measurement will be approximately flat. It is similar in intent to the "preferred headphone target response" identified by Olive, Welti and McMullin (Ref. 1) but adapted for the EARS.
Let's look at a couple of examples. Headphone A is a good quality headphone that is generally considered to be "dark". Subjectively, we felt that it was withdrawn on vocals but also a little hot in the high treble. Bass is solid but it doesn't have a lot of sub-bass. Its response when measured with HEQ is shown in Figure 1 in red. You can see that it measures low (compared to the flat line) in the 2 to 7 kHz area, which explains the "darkness."
Headphone B is also a good quality headphone but a very different type of design. Reviews refer to this headphone as "crystal clear". Subjectively, we felt that it is indeed very clear (great for acoustic music) but that the whole treble region was lifted too high. Bass was good, but sounded like there is a broad bass hump with the sub-bass and midrange on either side of it missing. Its response when measured with HEQ is shown in Figure 1 in mauve, which shows the elevated treble and the bass hump.
Figure 1. Comparing two mid-priced headphones
Applying EQ
To calculate the EQ, view the averaged measurement (of left and right, optionally with additional measurements). Click on the EQ button to open the EQ screen.
-
Set Equaliser to MiniDSP 2x4 HD. This will work for the miniDSP HA-DSP. The parametric EQ settings will also work for many devices that have parametric EQ.
-
Set your target curve flat using the Target Settings values shown below. For the Target Level, start with 84 dB if you have followed the measurement procedure in the user manual. Or, click "Set target level" to have REW calculate it. You may need to experiment with this setting, as different target levels can result in simpler EQ curves.
-
Click the EQ Filters button:
-
Turn off EQ filters so there are five set to Auto:
-
Set the correction range that you want. To start with, try 20 Hz to 20 kHz. You may want to change this later.
-
Click on "Match response to target" to have REW calculate a set of correction filters.
That's your starting point. Now adjust the filters manually. The basic process is to change a filter from Auto to Manual, change its parameters (Type, Frequency, Gain, Q), then click on "Match response to target" to get REW to recalculate the other filters. Then repeat.
Here are some things to try:
-
Decrease the Q of a filter and decrease the amount of boost or cut. (REW sometimes generates filters with higher Q than they need to be, and manually forcing a lower Q can result in a simpler EQ.)
-
Change a filter to a shelving filter (type LS or HS). This adjusts the level of all frequencies above or below the set frequency. (REW is not able to use shelving filters in its auto-EQ.)
Don't worry if there are some "peaks and valleys" left (with many headphones, it's unavoidable anyway). Also, don't try too hard to fill in the valleys.
Once you have a set of filters, copy the EQ settings over to your HA-DSP or other device. If copying a shelving filter (type LS or HS), set Q to 0.7.
Important: you will also need to reduce the input gain by the maximum amount of boost you have used in your EQ, to avoid clipping. For example, if you have a maximum boost of 6dB in your EQ, set the input gain to -6 dB.
Now have a listen. You may find that you want to "back off" on some of the filters. Or, add additional filters to alter the sound more to your personal taste. (See the page Frequency Response on diyaudioheaven.com for a great explanation of the audible effect of different frequency ranges.)
An example EQ
After following the above procedure, here is our EQ screen for headphone A. It shows the EQ curve in cyan and the predicted response in red (compare this to Figure 1):
When listening, we "backed off" the broad boost centered at 4 kHz from an initial setting of 8 dB down to 6 dB. Here are the EQ filters we ended up with:
Here is the same set of filters loaded into the HA-DSP:
Since the maximum gain in the EQ is 6 dB, we set the input gain to -6 dB, to avoid any possibility of clipping.
You will notice that we have a low-shelving filter here that boosts the low end response by 3 dB. This mild bass boost is built into the HEQ curve, so a headphone that normally measures with a flat bass response will tend to "droop" with the HEQ compensation. If you prefer to not have this type of boost, simply leave out this filter or adjust your EQ to droop a little in the bass.
If you wish, remeasure the headphones with EQ applied. (This is not really necessary except to check that you have not made any mistakes, as REW's predicted response is accurate.)
Finally, don't forget that you can use the input bass and treble controls of the HA-DSP for quick tonal adjustments:
Wrapping up[Top]
That's it for this app note! Have fun, and please let us know of your headphone EQ results in our forum.
References
- "Listener Preferences for In-Room Loudspeaker and Headphone Target Responses," Sean E. Olive, Todd Welti, and Elisabeth McMullin. AES Convention Paper 8994, October 2013. Figure 17.
