When quantitatively assessing how good a pair of headphones are, the main measurement looked at is the frequency response curve.
A frequency response curve plots how loud each frequency is, from the deepest bass the human ear can hear (approx 20Hz), to the highest treble (approx 20kHz or 20,000Hz). Here is an example for two different pairs of headphones:
For loudspeakers, the ideal frequency response curve is a flat line, when the speakers are measured in an an-echoic room. That is, the speakers should reproduce all the frequencies exactly as contained in the recorded material.
If we were then to place these completely ‘flat’ speakers in a room, the frequency response we would observe would no longer be flat, as lower frequencies would tend to reflect off surfaces in the room more than higher frequencies and appear amplified. Additionally, the shape of the human ear also makes frequencies around the vocal range appear louder. Needless to say different rooms and different people’s ears will sound different.
Harman International, a division of Samsung, who own brands including Harman Kardon, AKG, JBL have done some very extensive research to determine the ideal frequency response for headphones. In summary they have determined that a good pair of headphones should sound like a good pair of speakers in a good room. The resultant curve is commonly referred to as the ‘Harman curve’. A simplified version is as follows:
As the shape of the ear will influence the perceived frequency response of the headphone, one cannot get a good indication from simply using a microphone next to the headphone, or a microphone on a flat plate. Thus the industry standard test devices use a rubber human ear. To complicate matters further, different test rigs (even super expensive professional grade ones) will give different curves for the same set of headphones.
To overcome all these complications, my approach when developing the Homebrew Headphones has been to find an affordable headphone that gets very close to the Harman curve, purchase a pair, then using my MiniDSP EARS test rig, measure it and use the resultant curve as the target frequency response. Based on Tyll Hertsens’ article in InnerFidelity, I have chosen the NAD Viso HP50 as the benchmark.
This is a very brief, simplified summary of a very complex topic, so if you are interested in reading further, here are some good resources:
The following are some descriptions of the qualities of the sound of a pair of headphones – they are things to listen out for when evaluating a set.
Bright: the high frequencies are emphasised and clear
Coloured: not natural sounding, usually due to peaks and dips in the frequency response curve. These are generally caused by resonances in parts of the headphone construction.
Detail: the presence of fine details in the music which are lost with poor quality headphones
Soundstage: the ability of the sound to convey the location of the instruments, eg drummer in the centre, guitarist to the left etc. If you close your eyes and listen, you should be able to pick the locations of the instruments on the ‘soundstage’.
For further definitions, see this page.
Next: Deign Your Own Pair