• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

New AES paper from Facebook Reality Labs: Listener-Preferred Headphone Frequency Response for Stereo and Spatial Audio Content

thewas

Master Contributor
Forum Donor
Joined
Jan 15, 2020
Messages
6,742
Likes
16,175
When spatial audio content is presented over headphones, the audio signal is typically filtered with binaural room impulse responses (BRIRs). An accurate virtual auditory space presentation can be achieved by flattening the headphones’ frequency response. However, when presenting stereo music over headphones, previous studies have shown that listeners prefer headphones with a frequency response that simulates loudspeakers in a listening room. It is as yet unclear if headphones that are calibrated in such a way will be preferred by listeners in the context of spatial audio content as well. This study investigates how listeners’ preferences for headphone frequency response may differ between stereo audio content and spatial audio content, which was rendered by convolving the same stereo content with in-situ-measured BRIRs of loudspeakers in a room.

Authors: Engel, Isaac; Alon, David; Scheumann, Kevin; Mehra, Ravish
Affiliation: Facebook Reality Labs
AES Conference: 2020 AES International Conference on Audio for Virtual and Augmented Reality (August 2020)
Paper Number: 1-5
Publication Date: August 13, 2020 Import into BibTeX
Permalink: http://www.aes.org/e-lib/browse.cfm?elib=20868

Source and full free PDF download: https://www.aes.org/e-lib/browse.cfm?elib=20868
 

tomtrp

Member
Joined
Jan 14, 2021
Messages
73
Likes
51
Great to see more research developing in this area, the model and statistical method is great but the sample size is only 21 listeners(12 Facebook employees, 9 normal people) with no trained listeners which makes the generalisability of the results very poor no matter how high the confidence level is.
I understand the difficulty of conducting repeated larger sample experiments, but I don’t think AES is as strict as top level science journals.
I don’t know how many other AES research results are concluded from such limited experiments, but at least, we have Olive’s research with 268 people(including 12 trained listeners, but the sample has poor randomness and no repeated experiment on another group to confirm the result) which gives us the Harmen Targets, probably the most credible frequency response target.
 

pkane

Master Contributor
Forum Donor
Joined
Aug 18, 2017
Messages
5,631
Likes
10,205
Location
North-East
When spatial audio content is presented over headphones, the audio signal is typically filtered with binaural room impulse responses (BRIRs). An accurate virtual auditory space presentation can be achieved by flattening the headphones’ frequency response. However, when presenting stereo music over headphones, previous studies have shown that listeners prefer headphones with a frequency response that simulates loudspeakers in a listening room. It is as yet unclear if headphones that are calibrated in such a way will be preferred by listeners in the context of spatial audio content as well. This study investigates how listeners’ preferences for headphone frequency response may differ between stereo audio content and spatial audio content, which was rendered by convolving the same stereo content with in-situ-measured BRIRs of loudspeakers in a room.

Authors: Engel, Isaac; Alon, David; Scheumann, Kevin; Mehra, Ravish
Affiliation: Facebook Reality Labs
AES Conference: 2020 AES International Conference on Audio for Virtual and Augmented Reality (August 2020)
Paper Number: 1-5
Publication Date: August 13, 2020 Import into BibTeX
Permalink: http://www.aes.org/e-lib/browse.cfm?elib=20868

Source and full free PDF download: https://www.aes.org/e-lib/browse.cfm?elib=20868

It makes sense that binaural recorded content would work best with a flat transfer curve, while stereo content produced and optimized for speakers would work better with the Harman target.

This actually made it click for me. I've said it a few times before that I prefer a flat-ish EQ target that I produce myself with in-ear mics, and that I prefer this to Harman's. What I failed to realize was that a lot of the content I listen to on headphones is actually recorded binaurally. I prefer binaural recordings as they sound natural to me, with much better sense of space and natural tonality. I didn't think about the target needing to be different depending on how the recording was made, but this paper opened my eyes (ears?)
 

ernestcarl

Major Contributor
Joined
Sep 4, 2019
Messages
3,106
Likes
2,313
Location
Canada
It makes sense that binaural recorded content would work best with a flat transfer curve, while stereo content produced and optimized for speakers would work better with the Harman target.

I don't currently have an individualized flat target curve for my DT880 pro -- conveniently right beside me.

Anyhow, I did a short little A/B test using Morphit using two pre-set targets with just a little adjustment in the bass:

1613698806371.png


1613698818065.png


Dr. Chesky's Sensational, Fantastic, and Simply Amazing Binaural Sound Show

In most tracks such as Back Home in Indiana... I did not find that I had a clear preference either way. Though, in a few of the tracks like Pamafunk, it was quite clear that "Generic Studio" (which I presume the same in description to "generic flat" in the paper) preserved more of the spatial qualities from the binaural music recording.

Maybe if I made my own proper "individual flat" EQ for this headphone my preference for more of the tracks would sway towards that individual flat target curve? Perhaps... but I don't listen to headphones (and binaural tracks) that much to care enough to do it.
 
Top Bottom