Best daily deals

All products featured are independently chosen by us. However, SoundGuys may receive a commission on orders placed through its retail links. See our ethics statement.

Harman research targets the Bruel & Kjaer 5128

A first look at the latest headphone response targets from CanJam NYC 2024.
By

Published onApril 2, 2024

TL;DR
  • Dr. Sean Olive, senior researcher at Harman, performed a pilot study using the Bruel & Kjaer 5128 measurement fixture for in-ear headphones. The data was presented at CanJam, New York.
  • The investigation included an exploration of listener preferences for in-ear headphones.
  • Six headphone frequency response target curves were compared in the study.

At this year’s CanJam in New York City, Harman researcher Dr. Sean Olive, a senior fellow at Harman International (a Samsung subsidiary), presented his findings from an investigation into what the Harman Targets for the Bruel & Kjaer 5128 test head should look like. Considering the 5128 is now regarded as the de facto standard for headphone measurements, findings from this investigation have significant implications for review outlets using this equipment. While the audio industry has adopted this new test apparatus, sufficient work hasn’t been done yet to establish verified headphone and earbud target curves for the new fixture.

This has led to a “wild west” era of headphone reviews, where anyone using the new test fixture must either post measurements without the context a validated standard provides, use a benchmark measurement, or develop their own target — which can result in readers coming away confused about how a product sounds. After all, many of these reviewer-created target curves don’t agree, nor are they easy to understand for a naive reader.

Following the Senselab study from late 2023, which validated the SoundGuys Headphone Preference Curve for over-ear headphones, Dr. Olive’s pilot investigation is our first look at independent test results contextualizing in-ear performance on the new test fixture. Though the listener sample set was small, the results identified areas worthy of further study — especially of interest to readers of headphone reviews.

A note on statistics and jumping to conclusions

While the temptation to hang a “Mission Accomplished” banner for our preference curve scoring so well is strong, we can’t. Because this was a pilot experiment with a small sample size, the conclusions we can draw from it will be limited — including how well each curve performed in relation to one another.

Additionally, some of the caveats discussed in the presentation bear repeating here:

  • Headphones can have similar predicted preference scores and sound very different.
  • Target curves don’t take distortion or spatial reproduction into account.
  • These models don’t account for listeners’ personal tastes, hearing loss, or acoustic effects of your physiology.
  • Target curves serve as a guideline for comparing differences between headphones.

What happened?

A phtoo of the Bruel & Kjaer 5128 head and torso simulator, facing the viewer with its cold, unseeing stare.
Though it’s been used for years in headphone reviews, the Bruel & Kjaer 5128 was not used in developing the Harman Target.

Before headphone manufacturers switch to new measurement equipment, plenty of research needs to go into establishing standards so that the data generated is understood and verified to be accurate. While it’s easier for review sites like ours to simply buy equipment and start posting results, it’s essential to the industry writ large that potential shortcomings of any new instruments are identified and addressed. It’s also important to confirm the advertised benefits of using such a test head are true.

To see what the Harman IE 2019 Target would look like on the Bruel & Kjaer 5128, Dr. Olive created new versions of the in-ear target, along with a beta version of the over-ear target for the new fixture. This latest beta curve used the Bruel & Kjaer 5128’s Diffuse Field correction curve but used a 6.4dB shelf filter below 105Hz and a -1.4dB shelf filter above 2.5kHz. After selecting four other target curves, the group was put to the test.

Using thirteen listeners, Dr. Olive used a similar process to his validation of the original Harman Target — albeit on a much smaller scale. Using modified Sennheiser Momentum M2 earphones with mics to confirm fit, each listener was presented with different level-matched, randomized tunings for the same audio samples. Then, each respondent was asked to rate each on a scale of 0-100 in a double-blind test. Results were compiled and analyzed to see what patterns emerged.

A plot with six of the compared targets studied by Dr. Olive for Harman.
Adapted from Dr. Sean Olive / Harman International
Though most response curves tested have strong bass and treble, the SoundGuys Headphone Preference Curve has notably less output in these ranges than the other curves.

The candidate curves were:

  1. A fixture-conforming version of the Harman IE 2019 Target.
  2. A proposed curve (Harman Beta 2024) based on the calculated diffuse field response of the B&K 5128 test head, with the bass and treble adjusted to the Harman AE Target 2018.
  3. The SoundGuys Headphone Preference Curve.
  4. Headphones.com’s projected target (HP.com IE) based on their preference-bounds approach derived from the diffuse field response and the 2015 Harman filters.
  5. The diffuse field response (DFR).
  6. The proposed modification to the 2018 Harman Target put forth by Senselab (APHarman2018Ver2).

Unfortunately, the “HP.com IE” score resulted from a misunderstanding, as that particular blend of DFR and filters is not what that outlet uses, so it should be disregarded.

What we learned

Though Dr. Olive was uncomfortable making many definitive statements from the pilot test, he condensed some findings for the CanJam audience. Though this experiment has gotten some audiophiles abuzz with familiar arguments about target curves, it turned up some valuable tidbits of information that might get lost in the shuffle.

Headdie-with-the-AirPods-3rd-Gen
The Bruel & Kjaer 5128 is here to stay.

Dr. Olive summarized his main conclusions in a slide, which included the following:

  1. The B&K 5128 (Type 4.3) test fixture represents the impedance of the human ear more accurately and will produce different results than a Type 3.3 at low and higher frequencies.
  2. The IEC 60318-4 pinna produces 2-3dB more high-frequency emphasis than human ears.
  3. Over-ear headphones measured at the blocked canal on mannequin test fixtures measure less low-frequency levels due to leakage and higher high-frequency levels than those measured on humans.
  4. A comparison of different 5128 target curves indicates that the diffuse field response is not liked despite its status as the current IEC target curve.
  5. Some target curves are more preferred than others; there’s some evidence of segmentation in listener tastes.
  6. The Harman 2019 IE Target scored the highest in blind tests (Ed. note: see disclosure above).
  7. The Harman Beta 2024 needs some more work.
  8. Preference ratings are strongly correlated to deviations from the target curve.

Generally, this is good for headphone reviewers using the B&K 5128 to test both over-ears and in-ears. Though there are some deviations, they’re at least predictable — which is reassuring.

The most interesting discussion for the audiophile crowd was the comparison of target curves. Final scores for each of these curves fell largely in four groups of statistical ties, so don’t place absolute faith in rankings; just take the groups as a ballpark. Below are the scores and the highest and lowest grouping achieved by that target. Everything in Group A is statistically tied with each other, and so on.

TargetPreference scoreHighest groupLowest group
Target
Harman IE (2019)
Preference score
70.795
Highest group
A
Lowest group
-
Target
SoundGuys
Preference score
68.436
Highest group
A
Lowest group
B
Target
Harman Beta (2024)
Preference score
64.179
Highest group
A
Lowest group
B
Target
HP.com IE
Preference score
59.090
Highest group
B
Lowest group
C
Target
APHarman2018ver2
Preference score
50.051
Highest group
C
Lowest group
-
Target
BK5128DF
Preference score
21.167
Highest group
D
Lowest group
-

It’s clear that for this gut check, the Harman IE 2019 Target did very well. Though nobody should draw definitive conclusions (i.e., the X curve is objectively more popular than the Y curve) at this stage, we were interested in how well the SoundGuys Headphone Preference Curve fared, as the Senselab study did not investigate in-ear performance. We use this universal target for over-ear headphones and in-ears, while other targets are typically explicitly intended for each product type. Interestingly, our notion of good sound has performed well in independent tests for both audio product types without modification, even if more extensive studies are needed to verify what we’ve seen.

A chart showing the difference in listener preference scores for each target curve, broken down by class of listener.
Adapted from Dr. Sean Olive / Harman International
Class 1 listeners were mostly trained, whereas Class 2 listeners were not.

Dr. Olive further divided the listener population into two classes: mostly trained and untrained. The data showed that Class 1 (mostly-trained) listeners preferred the SoundGuys, Harman IE (2019), and Harman Beta (2024) targets over others — with the SoundGuys curve posting the highest score from this group. However, the Class 2 listeners (untrained) had less consistent results, and they often preferred more bass and treble than Class 1 listeners. You can see that the error bars in their results are pretty wide compared to those of the Class 1 listeners. It’s important to reiterate that it’s possible that a larger sample size might give us different results — or reveal more classes of listeners with different preferences altogether.

We were fortunate enough to have the presentation slides provided to us by Dr. Olive. Still, the entire presentation is worth watching if you’re interested in the history and science behind the Harman Target. Plenty of topics of interest were contained within, and further areas of investigation were introduced, for example, using 3D-printed ears to test ear impedance.

Given the nature of the headphone industry and the resources pouring into listener preference studies, it will be interesting to see what else we learn in the coming years.

 

You might like