Dysonics RAPPR Headphone Audio Virtualizer and RondoMotion Head Tracker
RAPPR Headphone Virtualization Software ($15) and RondoMotion Head Motion Tracker ($60)
Getting sound on headphones to sound like it's coming from outside your head is the holy grail of headphone performance. Virtual audio is going to be huge as we move forward. I've always thought this process would require modeling Head Related Transfer Functions (HRTFs) and how they change with head movement. Dysonics thinks otherwise.
The founders of Dysonics all have academic backgrounds in human audio perception and spatial sound capture/reproduction primarily at UC Davis. In 2012, Dysonics emerged from the UC Davis technology incubator (ETTC) with the goal of developing convincing 3D audio on headphones with motion tracking.
I had a nice long conversation with Robert Dalton Jr., co-founder and CTO of Dysonics, and it became immediately evident that this group of engineers/researchers are very familiar with HRTFs, in fact their group produce what was, at the time, the largest database of HRTFs on the planet. To clarify: an HRTF is a large set of frequency response plots (actually it's usually impulse response plots, which is the same thing but referenced to the time domain) of what each ear is hearing when sounds come from various angles.
In the course of their research and developing familiarity with HRTF data they came to believe that some things were very important for auditory localisation (inter-aural time delay, inter-aural level difference) and these data didn't change much from person to person, and some things were less important and/or did vary a lot from person to person (pinna reflections). For example, they found if they did use pinna reflections people who had pinnae that were similar to the modeled pinna would get good spacial perception, but people who's pinna were different would have worse spacial perception than if they were listening without the pinnae being modeled.
Slowly but surely they developed the idea that there may be a simpler way to synthesize virtual audio. Rather than the computationally intensive process of synthesizing the speakers, the room, and how the sound changes on an ongoing basis with head movement by selecting and interpolating between all the discrete angular HRTF measurements, they would simply measure the nature of the sound-field around the head with a microphone that had many of the characteristics of a real head.
The idea here is to put this head-sized microphone in front of a real sound system and record the nature of that transfer function at all 8 sensors around the mic. Special tones (something like MLS signals) are played through each speaker, and the received signal at all 8 mics are analyzed to develop the transfer function for that angle. Once done, a DSP filter exists for 8 positions around the head one of which can be output to your headphone earpiece depending on the position of the head. For head positions in between the 8 mic positions, special proprietary interpolation filters create solutions to modify the sound. While this is certainly not simple math, it is computationally less intensive than the HRTF/artificial room method, which should improve latency and performance on mobile devices.
Dysonics intends to produce a variety of products over time, including recorded music, spoken word, and other audio materials using their Rondo Microphone. At this early point in their lifetime, they have three simple products: RAPPR audio processing software for Mac, RondoMotion wireless head tracker, and RondoPlayer an iOS app that allows you to use the RondoMotion head tracker with Spotify on an iPhone/iPad. I only auditioned the RAPPR and RondoMotion products.
Downloading and installing the software was painless and as expected. Once installed, running the app causes a small Dysonic logo to appear in the top menu bar on your Mac. Any audio played on your system by any app will be routed through RAPPR. Clicking on the button brings up the user interface.
To the left of the control panel are buttons for selecting stereo music (two speakers), movie (three speakers, left/center/right), and 5.1 surround (five speakers). There is also a bass boost button.
To the right of the panel are two buttons for turning the head tracking function on and for reseting the angle to directly front. At the bottom is a button to enable/disable RAPPR.
At the center of the panel is a circular graphic. Inside the graphic are sets of semi-circular patterns representing the sound coming from the speakers. You will see different numbers of these graphics depending on which mode (music, movies, 5.1) you're in. These circles expand in and out with the program volume of that channel. You can also grab and drag the circumference of the large circle display to adjust how large you want the room to sound.
Two other panels are available in the app: one to pair and monitor the functions of the RondoMotion detector, and the other controls system output device, motion settings, and software settings.
Most are fairly obvious from the screens above, but one setting worth mentioning is the motion settings. Let's say you're working on your computer and you move to a different seat around the corner of the table. Because you've rotated your head, the sound will be coming from off to one side. The "Auto Reset Enabled" setting will cause the sound to recenter automatically to directly in front of you if your head has remained turned for more than 10 secondsor what ever time you've set on the "speed" slider.
The RondoMotion detector is a simple, roughly 1.5" square device that has a multi-axis rotational sensor (electronic gyroscope) within. It charges with a standard micro-B USB cable and will run for about 20 hours on a charge. A removable silicon rubber strap is mated with the RondoMotion and allows you to attach it to the headband (or any convenient place) on the headphones. It does not have to be mounted horizontally.
For testing sound quality I used the NAD VISO HP50, Audio Technica ATH-MSR7, and Sennheiser HD 800 headphone driven by the Simaudio Moon Neo 430HA headphone amp using the USB output from my MacBookPro. I'm very familiar with all three headphones and feel their transient performance excellent. Generally speaking, they deliver fairly good imaging and felt they'd be good headphones to experience the spacial performance of the Dysonics products. I listened to both music and movies.
In terms of spacial impression, I found that the image remained in my head and was located rather high in my head. It did, however, seem much more spacious than the straight music feed. As I moved my head slowly from side to side the sonic image did move to try to continue to appear to come from my computer screen. The problem is that because the center image was more up than forward, it felt like the music was going over the top/slightly-forward part of may head from side to side. This seemed a cognitive disconnect to me as I was rotating my head from side to side, but the music was traveling in an arch over the top of my head. Also, when I moved my head toward the left, and the sound moved toward the right ear, it appeared to be somewhat lower in elevation than when I looked straight forward and the sound was centered, but high in my head. This gave the strange feeling of the sound source moving up and down as I moved my head right and left. This problem seemed worst in music mode, and got less as I went to the movie mode, which seemed to image a bit lower due to the center speaker.
After listening with the RAPPR on for quite a while and getting acclimated, then disabling the RAPPR, the sound field in my head collapsed hard and I was left with sense of a very small, shallow image relative to the sense of space with the RAPPR on. The thing is, after about 5 minutes my head would re-adjust to my normal headphone listening mode and the audio image would fill my head and be essentially as "big" as the RAPPR image. I put big in quotes because neither really seem to occupy real space, but rather some imaginary spacial dimension in my head.
I found the tonal quality of the RAPPR software to be quite poor. In all settings I heard a significant emphasis in the low mid-range creating a boxy coloration, and I heard a significant loss of treble energy. Speech intelligibility was degraded due the elevated mid-range hump pushing back the treble cues. It's rather hard to characterize it in any more detail as the colorations did seem to be somewhat program dependent. I also found these colorations to get worse as you dragged the circle outward to put more distance to the speakers and create a larger room. Most of my critical listening was done with the circle at its smallest diameter. Probably the best way to explain what I heard was that Dysonics did manage to give me a relatively convincing sense of listening to speakers in a room, but it was a lousy sounding room.
I did also have a few other problems. I experiences some random clicks in the audio. It did entirely loose audio once. It did stop producing head movement changes a couple of times. In all cases, shutting RAPPR down and restarting it fixed the problem. I'd say it's still a little buggy.
As I talked with Robert Dalton about my experience I sensed he was very interested in my impressions. I found his probing questions of my experience and his detailed comments pointing out things I might listen for to be right on the money. It was obvious to me these folks really understood the issues at hand. Audio virtualization on headphones is extremely difficult. I felt Dysonics attempt with he RAPPR and RondoMotion an interesting and successful (for this early stage) attempt to virtualize headphone audio, but I also think there's a long way to go.
I felt the ability to produce the kind of effects I was hearing through a process completely different the HRTF DSP methods a stunning revelation. The effect I heard was as good as any synthetic headphone 3D I've heard to date in terms of spacial atributes. The tonality needs to be better, and the image needs to be down and forward more, but it does produce a strong sense of space.
One aspect of this effort not yet readily apparent is the upcoming RondoMicrophone. While it was used to develop the filters needed to synthesize an artificial space in the RAPPR app, it can also be used to directly record a performance in a room. This is an eight channel recording that when played back through the RAPPR app using the RondoMotion sensor allows you to experience music all around you. You can literally be immersed among the musicians and "look around" at musicians placed around the room. I was supplied with a few of these files for audition and it worked much better than when synthesizing the space artificially. The image was a little more spacious, but the tonality was hugely improved.
I'm going to keep abreast of Dysonics efforts. They're obviously very well versed in the problem of virtual audio on headphones; their current early products are clearly delivering a strong sense of space on headphones. My guess is that we'll see some very cool things happening over time as the folks at Dysonics keep working on the technology. The folks working on virtual reality goggles need virtual audio on headphones, it will be difficult, but it will happen. Dysonics is certainly someone who might pull it off well.
At this time, due to the significant tonal colorations, I do not recommend these products for improving the listening experience of movies and music. I do think, however, they provide a fun demonstration of the technology and should you be curious to have the experience the Dysonic products are a fun and relatively inexpensive purchase.