Are HRTF’s Necessary in Dolby Atmos Elevation Speakers?

There's been a lot of talk lately about Dolby Atmos Elevation (aka. Dolby Enabled) speakers and how through proprietary DSP in an Atmos A/V receiver along with a "specialized speaker design," they can create the illusion of height without having to place discrete speakers on the ceiling.  Dolby Elevation speakers feature an up-firing driver(s) placed around ear height on an angled baffle (usually about 20 degrees) pointed towards the ceiling to bounce the sound back down to the listening area.  Atmos Elevation speakers are typically placed right on top of their companion front or surround channels though some designs, such as the new Atlantic Technology 44-DA module, offer more placement flexibility as a stand alone unit.

Dolby makes some claims about Atmos Elevation speakers being equivalent in performance to discrete ceiling mounted speakers as can be seen in the following pull quotes from their website.

"Dolby Atmos enabled speakers produce slightly diffuse overhead audio that is quite lifelike and, in some cases, preferable to the sound that comes from overhead speakers."

"You’ll hear the same overhead sounds whether you’re using overhead speakers, Dolby Atmos enabled speakers, or a combination."

While this certainly hasn't been our experience via direct comparative listening tests between discrete overhead vs up-firing reflective speakers thus far, it's prudent to do a little objective analysis on the technology behind Atmos Elevation speakers to better understand if equivalency is really possible.

Recommended Reading Materials to further familiarize yourself about Dolby Atmos and Dolby Atmos Speakers:

Dolby Atmos for the Home Overview

Dolby Atmos Elevation Speakers Explained

Dolby Atmos Elevation Speakers Explained Part 2

What is a HRTF?

Before getting into the specifics of the Head Related Transfer Function (HRTF), a short historical perspective is in order to understand its origins (courtesy of Dr. Floyd Toole).

Dr. Edgar A.G. Shaw became best known for his research in understanding the acoustics of the external ear.  It was he who scanned the sound field in human ears using a tiny probe microphone with sounds arriving from different angles, and plotted the details of what we now call HRTFs.  This is all documented in Journal of the Acoustical Society of America papers, and Dr. Shaw received the Raleigh Medal Award for his work.  The cues for sounds arriving from above are associated with directionally sensitive resonances in the external ear in the frequency range of about 7 - 12 kHz. They can be very different for different people because our ears are all physically different.  Taking an average of many ears though, one finds a general trend indicating that as a sound source is elevated on the median plane (directly forward in this case) there is an increase in sound level reaching the eardrum at frequencies around 7-8 kHz.  There is another directionally sensitive resonance in the external ear around 12 kHz, but it is much less predictable because as frequencies get higher smaller physical differences between pinna result in greater response variations between individuals.

Dr. Shaw estimated that the dominant height cue could be modeled by a resonance centered at 7.5 kHz with a Q of about 3. A level increase of about 10dB would correspond to an elevation of about 45 degrees.  To put this into perspective, this work was done in 1972, 42 years ago, so none of this is breaking news.  A very perceptive final comment was “For this to work without special attention to the idiosyncrasies of the subject [the listener] it would probably be necessary to use a broadband source (e.g. white noise, clicks, etc.).”  Clearly this phenomenon was well understood many years ago.

The Dolby analog crossover integrated into speakers with the Dolby Atmos certification on them is designed to mimic the HRTF to theoretically help trick your brain into thinking the sound is coming from an elevated position, despite the fact that the speakers are at about ear level.  The graph below represents the shape of the filter response but it’s also important to note that there may also be some element of EQ processing that Dolby Atmos is doing in A/V receivers and processors to help bolster its effectiveness. 

HRTF Frequency Response of Dolby Elevation Speaker Crossover

Are HRTF’s Needed in Loudspeakers?

The curve from Fig. 14B in combination with the acoustical properties of your head and the spacing between your ears is what’s supposed to make the magic work. But is it really necessary to apply the HRTF to a loudspeaker crossover?  How effective will a smallish +2dB bump centered at 7kHz be at producing a heightened sense of elevation since Dr. Shaw demonstrated at least a +10dB rise in amplitude response was necessary to give a 45 degree angle of elevation?

As with all HRTF-based processing, the effectiveness of the simulation will greatly depend on how closely matched your own HRTFs are with the generic model used.  As we’ve found from our preliminary testing of Dolby Elevation speakers, there seems to be a fairly narrow sweet spot in which the elevation illusion would be most effective. This is dependent on room acoustics, ceiling height and flatness, proximity to the Atmos elevation speaker module, and receptiveness of the listener.

I wanted to dig a little deeper into the topic of the HRTF before analyzing the crossover so I called upon Dr. Floyd Toole to get his take.  Here is what he had to say.

Head Related Transfer Functions, HRTFs, are measurements of how sounds are modified by the head, torso and external ears as they make their way to our individual eardrums.  Sounds from different directions are modified in distinctive ways, giving our hearing system some clues about where the sound is coming from.  However, most of the directional information comes from differences in the sounds arriving at both ears, the inter-aural amplitude and time differences as functions of frequency - the binaural information. All of this gets combined with additional information from head movements to allow us to perceive the direction from which a sound is coming.  In headphone listening all of these clues are missing, and consequentially we hear things close to or inside our heads. With head movement tracking, things can be better.  Binaural music recordings have those clues encoded by using a dummy head microphone.  Sophisticated games use sounds encoded with synthesized binaural cues and generic HRTFs to try to deliver the information the brain needs to perceive the desired directions.  The dummy heads and simulated binaural/HRTF clues will not be a perfect match for any individual listener, but for many people they work well.      

However, in conventional loudspeaker reproduction, none of this is necessary.  The loudspeakers are real sound sources and all of our binaural hearing, HRTF and head movement clues are right there to enable us to identify directions and distances.  There is no need to add anything because our considerable capabilities are already fully functional.  In fact, because all of our ears are different, and therefore the HRTFs are all different, adding a generic HRTF-like equalization to a playback system is not only redundant, but it has a good chance of simply degrading the sound quality because of the mismatch between the individual’s specific HRTF and the generic one used in the product.

So in reality, adding a HRTF response to the loudspeaker crossover can actually degrade the sound quality per Dr. Floyd Toole.  We will re-examine this point in our forthcoming Dolby Atmos Crossover article.

HRTF Analysis

After reading through the Dolby Atmos Speaker patent and claims we decided to do some analysis of our own.

See: http://recherche.ircam.fr/equipes/salles/listen/download.html

This research has HRIR’s (which can be converted to head related transfer functions) for several different people at various azimuth/elevation positions.  You can convert these WAV impulse responses (using adobe audition or cool edit) into frequency response to try it yourself. It is evident that sounds originating from overhead generally have a little extra energy at 7kHz and a dip around 12kHz.  As Dr. Floyd Toole previously pointed out, positional audio engines for headphones etc use the HRTF’s ability to localize sound in the human brain.  This is how headphones give listeners the sense of a helicopter flying over head.  

Dolby claims they are incorporating this frequency response change (per Fig 14B) into their up-firing elevation speakers to try to trick the brain into thinking the direct sound from the speaker is coming from above.  As Dr. Toole pointed out, the reflection, which is actually coming from the ceiling, will already have the response change due to the shape of our ears.  In agreement with Dr. Toole’s comments, employing a HRTF into the loudspeaker crossover appears to be an inaccurate and redundant way of accomplishing this, because:

A.  It assumes the severely off-axis direct sound is linear at the listening position to start.  Considering the upward firing driver is 70 degrees off axis with respect to the listener, the higher frequencies are already rolled off by a considerable margin.  Perhaps a better solution would probably be to try to increase the directivity by using a larger than necessary driver or waveguide and roll off the lower frequencies that have spherical radiation.  But this remains a topic of further discussion in a different article altogether. 

B.   The reflected and direct sound will arrive at different times which can add spaciousness but can also penalize clarity of sound.

C.  The 7kHz and 12kHz response changes shifts to lower frequencies for overhead sounds less than 90 degrees.

D.  The 7kHz and 12kHz do not represent how all listeners hear overhead sounds. It is just an average best guess sort of thing.  Depending on the shape of your ear, shoulders, etc., the response may look quite different.  Shown below are two screen shots of the same location for two different listeners from the website above, you can see their HRTF’s vary quite a bit.

First Listener

Second Listener

As you can see, the first listener might have a better Atmos elevation experience than the second due to the missing peak at 7kHz for the second listener.  These are just some of MANY reasons why up-firing reflective speakers generally cannot come close to discrete ceiling mounted speakers in terms of fidelity or accuracy of delivering height information, despite Dolby claiming otherwise on their website. 

Designers in the headphone arena have been using HRTF to design positional audio for quite some time (ie. Creative Labs sound cards for games etc).  It hasn’t been applied to loudspeakers because reflections kill the sense of height.  We can determine azimuth easily because we use binaural cues: the inter-aural time difference (ITD) for low frequencies and the inter-aural level difference (ILD) for high frequencies. We actually use ITD at high frequencies too, but on the envelope of the sound, not on the “carrier” frequencies where the wavelengths are very short.  This is how we localize transients.  Using the HRTF to position audio in a room with loudspeaker sources is ineffective or at least unpredictable.  

What About the Magic Solution?

Consumers often want the "magic" solution. Even if it is not correct, these HRTF filters may have the ability to provide some directional cues to some people, and that may be enough to sell a product. If manufacturers can use HRTF to provide "suspension of disbelief", then it is possible to have a successful product. This is not high fidelity, it is creating an illusion.

Atmos Elevation speakers are NOT the first to employ HRTF’s.  There are several speakers with HRTF pre-emphasis (and de-emphasis) such as passive sound bars from the likes of Definitive Technology (ie. SSA-50, SSA-42, XTR-SSA-5, etc...).  They have the HRTF built into the crossover and also Inter Aural Crosstalk Cancellation (IACC).  The circuits were designed based on the patents of Mathew Polk. While it was agreed by most loudspeaker designers employing HRTF’s that the surround channels seldom sounded like they were behind the listener, the side image was convincing enough to be plausible. Are these high fidelity compared to separate speakers?  No. Are they high fidelity compared to cheaper sound bars. Yes. Definitive Technology sells a lot of them for good reason. 

Definitive Technology also has an in-ceiling speaker that is designed to de-emphasize the sense of height. This technology is described in US Patent 8,014,545 B2.   One useful thing found in the patent is that it cites a lot of prior art.  The reality is that, for many people, the illusion works and the speaker continues to sell well. 

To make a good living, design engineers often have to do things that are somewhat "impure", or at least things they would rather not do. There are other factors besides pure acoustics (heresy?). Consumers want the convenience of sound bars. The industrial design has to be attractive and offer high WAF.  Many people want the speakers to be invisible. These concerns can affect the acoustics, but they are essential attributes of a successful product. A loudspeaker engineer’s job is to make their speaker sound as good as it can in spite of any other factors. If manipulating HRTF results in a product that generates millions in sales and brings joy to thousands of consumers, is it a bad thing? Or is it better to design a technically perfect speaker that is never sold because it is too expensive/ugly/huge or impractical?

How is the Reflected Sound Perceived as Elevated Height?

It’s important to better understand how reflective speaker technology works independent of whether it has a HRTF implementation in the crossover or not.

Atmos Elevation Speaker Reflection Diagram

As you can see in the illustration above for a speaker firing up towards the ceiling with the axis 20 degrees off of vertical and an 8 foot ceiling, the axis of the reflected sound will be 35" in front of the speaker at seated ear height (35").  The path length for the reflected sound will be 116". The reflected sound will arrive 6ms after the direct sound, and due to distance will be attenuated 10dB (ref 1m). This is within the domain of the Precedence effect (Toole, Sound Reproduction: Loudspeakers and Rooms, Figure 6.16) which in normal situations would mean that we localize to the direct sound from the loudspeaker and the reflection is not perceived as a separate spatial event.

Dolby Atmos Diagram - Courtesy of Pioneer

However with the spectrum of directed and reflected sound being so vastly different, the Precedence effect breaks down and both the originating loudspeaker and the  point of reflection may be heard as coexisting sound sources having different spectral balances.  In the case of an Atmos reflective speaker, this helps the reflected sound at higher frequencies to be localized as a separate event.  The fact that transients play an important role in localization in complex sound fields further explains why the bounce effect can work as well as it does. 

However, as the listener moves further away from the speaker, he/she will move progressively off-axis to the reflected sound and on-axis to the direct sound, thus reducing its ability to be localized as a discrete elevated sound source. At a listening distance of 10 feet (3m) and an ear height of 35" the listener will be 27 degrees off axis for the reflected sound. The reflected sound will arrive 3.3ms later than the direct sound and will be attenuated 3dB by the distance ratio.  As the listener moves closer to the reflective speaker, the Precedence effect dictates the direct sound being dominant making it more localized at the source of origin rather than an elevated event.   

Bottom Line

Since this is a rather complex topic of discussion, I thought it would be useful to recap the main points herein as follows:

• Head Related Transfer Function (HRTF) was first documented in 1972. Its incorporation into a Dolby elevation speaker is not a unique application.
• HRTF varies among individuals and are not predicable but can be generalized.
•  If you want a sound to come from a discrete location, its most effective to put a speaker in that location rather than bouncing soundwaves off a ceiling.
• HRTF encoded into speakers is redundant and may degrade the sound quality.  Timbre matching speakers is important especially when directional and distance cues are involved.  Choosing speakers of similar quality and dynamic capabilities for all of your home theater speakers is recommended.
• The listening window for a reflected speaker varies significantly within the room and careful aiming is required to generate the proper effect.  Sounds with insufficient high frequencies will be localized at the origin speaker locations producing those sounds.  
•  A controlled dispersion speaker can create the illusion of elevated sound for a narrow listening area.

Conclusion

It seems very apparent that the deployment of a HRTF in a loudspeaker crossover is not only problematic but likely not necessary as well.  While they are useful in applications like headphones, and sound bars, adding them to an actual discrete external sound source like a loudspeaker playing in a room, and specifically in this case, a Dolby Atmos Elevation speaker module, can in fact impact performance in a negative way. Moreover, this also unnecessarily drives up the complexity of the speaker’s crossover and thus its associated cost.  The human ear already has the benefit of its own HRTF customized to each listener.  Simply selecting a loudspeaker with narrow and controlled dispersion, along with careful placement relative to the seated area can create the illusion of elevated sound for a narrow listening area, though not as precisely or as consistently as having a discrete sound source located in the position of origination it was intended to mimic. 

Acknowledgements 

I would like to personally thank the following people for their contributions and/or peer review of this article, all of whom are true experts in their respective fields. Their contributions enabled us to make the most comprehensive and accurate article possible about the new Dolby Atmos Elevation Speaker technology.

• Dr. Floyd Toole, retired VP Acoustical Engineering, Harman International ; Published Author of: Sound Reproduction Loudspeakers & Rooms

• Tim Gladwin, Loudspeaker Engineer of Warkwyn Associates