Raw Frequency Response

We use a Brüel & Kjær HATS Type 5128-B (B&K) measurement rig in a room with a low noise level and minimal sound treatment. Once the headphones are placed on our artificial head, that's when the magic begins. The testing levels are calibrated with a 500 Hz test tone normalized for the selected headphones to generate 94 dB SPL. This is compliant with the IEC 60268-7 standardized headphone and earphone testing procedures, allowing us to validate our results with manufacturer specifications and other published frequency response graphs that comply with this widely accepted standard. We then play a test signal (also called a stimulus) from our PC, which is generated by an Audio Precision APx517B Acoustic Analyzer. This signal is a two-second 20-20kHz log sweep at a constant level previously determined in the calibration sequence. The headphones reproduce this signal, and the audio is captured by the B&K's ear simulator (type 4620). This data is then fed back to the analyzer.

Our B&K test rig, ready to go.

We smooth our plots and calculations at 1/12th octave since this amount of smoothing removes the very fine fluctuations in the frequency response that aren't perceptible to human hearing while preserving the details that are audible to humans. For Bluetooth headphones, we use the Bluetooth Duo module from Audio Precision to feed the wireless signal.

We don't take measurements just once, though! We also re-seat over/on-ear headphones five times and three times for in-ears. This movement helps us capture the small deviations in sound due to the natural variation of headphones' fit, positioning, and seal. Each time we move the headphones, we do another round of measurements. In our previous methodology, we used in-ear microphones worn by actual humans for on-ear and over-ear headphone measurements in the bass region. We now rely on the B&K HATS for the whole spectrum as it's a more standardized approach, and the ear simulator is fully specified to cover the full range of human hearing. The human measurements aren't completely gone but are now only used in the Frequency Response Consistency test.

If you're a seasoned audiophile, you may be able to read the raw frequency data alone and come to your own conclusions. That said, if you need a little more guidance, you'll want to use a target curve. Target curves are representations of what most people consider 'good sound' based on research and data. There are many out there to pick from, but personal preference also plays a large role in whether you'll like a particular frequency response. Using a target curve can help orient you toward your own preferences, though.

We use a target curve that's inspired and influenced by work done at Harman^1,2,3 as well as other studies in the audio industry⁴. Their research is, in turn, derived from how sound reaches your ears when using good speakers in a semi-reflective room. Their targets have been instrumental in informing our own. That said, our test rig isn't the same one used by Harman—they use a GRAS rig, which doesn't behave quite the same as our Brüel & Kjær HATS Type 5128 (B&K). Both rigs produce similar, valid measurements but differently, and ultimately, there isn't a way to directly reproduce their test conditions and in-room response to our scenario.

We've updated our curve based on the diffuse field response of our B&K rig. From there, we added a 6 dB downward tilt based on our speaker target and added a 4.3 dB bass shelf at 105 Hz (Q. 0.707) to achieve a target response that aligns with Harman's studies without being a direct conversion. Their headphones research shows a significant variance in listener preference, including the bass amount. A majority of listeners prefer Harman's target level in this range. In our own listening tests, we also noticed a preference for bass boost, so we want to reflect this in our own target curve.

The target response is then smoothed to 1/6th octave to reflect the fact that a target curve is about general tonality and not small frequency deviations. Our target curve philosophy is rooted in preference variability; small peaks and dips in any given response ultimately fall within preference bounds, so reducing their presence won't significantly impact your own assessment of sound quality.

So, there you have it: our target curve. This target is the same for all headphones; in the past, we had separate IEM and over/on-ear targets, but we've simplified them into one target. It isn't fully understood yet why IEMs tend to show more differences in bass response between different setups, i.e., between IEC 711 couplers (like you'll find in rigs like GRAS and HMS II.3) and type 4620 ear simulators (including our B&K rig).

An IEC 711 coupler⁵ used in an ear simulator.

A B&K Type 4620 ear simulator⁶.

A standardized target curve has its limitations, though. While our own target reflects a sound most people will like, at the same time, it's not an absolute measure of a 'correct' sound signature. Your personal preferences play a role in understanding what's right for you; a target curve is just one tool to help you make that decision more confidently.

You'll want to check out our R&D article to get more details on our target curve. In it, you'll discover how we came to our target curve, our research (including a listening test!), and observations related to the general state of measurements versus preference.