Often time in research, newer work is based on or builds on older work and sometimes some important limitations and assumptions are not repeated, glossed over or worse just forgotten. Now that so many consumers and reviewers use the research conducted by Harman Institute, I wish there is more talk that explains the limitations instead of just taking the work as gospel and basing headphone purchases on graphs, or worse reviewers just describing tonal response as they see on the graph with respect to a preference curve/target curve which can be misleading to the audience.
All of the research that lead to the Preferred Headphone Target curves was based on using virtual headphones. I.e, the real headphone such as a Audeze LCD-4 was not used, instead the response of the headphones was first measured using a modified 54CA, then the frequency and phase mimicked using a headphone such as HD518. Then the preferred response was arrived at by using a single headphone to mimic (i.e., eqd) all other headphones.
- Now ask yourself, why spend any money on headphone gear when you can just buy one single headphone and just EQ it to the target, because that is what the study did and the listener preference was based on the virtual headphone.
- The study assumes if a listener likes or dislikes a virtualized form of the headphone they will equally like or dislike the original headphone being virtualized. They did try to validate if this was true, but guess what, LCD2 was used in this study and LCD2 was virtualized worst! i.e listener preference of virtualized LCD2 did not correlate to the original LCD2. One would think this is a possible red flag. An AKG that was ranked second in this study behind LCD2 before virtualization, now became ranked 1 after virtualization.
- The study also assumes a preference curve developed using a specific virtual headphone such HD518 or Hd800, would be the same if a different headphone (say LCD-4) were used. I do not think there were studies conducted with listeners to try and determine the variability of a preference curve when different headphones are used. If a different headphone that does well in bass were used, or if the headphone were more transparent or resolved better, it would have affected how the listeners choose to change the response.
- The target curve was arrived by letting the listener choose between different target curves that were generally smooth, the listeners were not given fine control over what the can tune except for a bass or treble shelf. Those who EQ know that that is just not enough, yes you can make a headphone sound a bit better but given the variability of how headphones measure on a rig, using a nice smooth curve as a target will not produce optimal results. So at best, they should be treated as suggestions.
This is not meant as a criticism of the work done by the Harman Institute. I think the intent in trying to objectively evaluate headphones is a good one, and helps educate both the customers and helps manufacturers understand the customer preference better. However, it is equally important to understand the limitations of the research. A pragmatic approach over a dogmatic approach is better here.
While we have followed the research with great interest, our intent in tuning our headphones has not changed much, we want music to reproduced as is (flat, or equally loud across the frequency spectrum) and not intentionally color it to suit a target curve. Of course, there are limitations to how we can achieve this without compromising on other equally important qualities such as distortion, transparency, dynamics, soundstage etc but we keep working on pushing these boundaries.
thanks for doing it.
