An Exclusive Interview with Seth Steiling, NaturalPoint Product Manager for TrackIR
by Joe Keefe
What’s new in TrackIR 5, and how is it improved over TrackIR 4?
Seth: The most significant improvement is in resolution. Our new sensor is significantly more powerful than the TrackIR 4 — to the tune of 3x raw sensor resolution. So, on a foundational level, TrackIR 5 can see far more detail than ever before. On top of that, we’ve introduced a new type of image processing — Grayscale Precision. It gives TrackIR 5 the ability to see much more granularity in light, which means that when it’s measuring your movement, it can do so accurately down to minute portions of a pixel. So, on a direct 1:1 comparison, TrackIR 5 has three times as many pixels representing the physical space your head moves in during game play. But not only that — it also has the ability to see much more detail within each of those pixels. The combination of those abilities results in a 10-fold increase in effective resolution over TrackIR 4.
Other nice improvements include a hefty 12% increase in field of view — which expands trackable real estate pretty significantly, a magnetic base and USB cable anchor for reduced damage from wear and tear, Spurious Marker Rejection for improved ambient light filtering (available in software version 5.0), and adjustable status LED brightness.
Joe: On your TrackIR 4 vs. TrackIR 5 comparison page, you list some pretty enormous figures. 96,000 x 72,000 Reporting Resolution. 1/150th Pixel Sub-pixel Resolution. Do these numbers actually mean something, or is this an exercise in hype marketing?
Seth: Yes they mean something! The listed specs for the new TrackIR might seem over the top, especially when compared to past generations, but they are legitimate, and they reflect intentional and meaningful improvements by our developers. NaturalPoint is driven by superior engineering, and we only endorse what we ourselves believe.
Joe: So, why have you chosen to introduce this terminology into TrackIR specs? It’s not been there before.
Seth: Raw sensor resolution alone provides an incomplete picture of the effective resolution of a device. If we are measuring movements that are smaller than a pixel — and previous generations of TrackIR also did so, but to a lesser degree — then sub-pixel and reporting resolutions provide a way to compare the improvements between them. Our specs seek to clarify this and reflect the actual performance of the device, instead of just the raw components used to build it.
The basic principal of our technology is tracking the location of markers using infrared light. So, for instance, it’s inaccurate to compare it with a consumer digital camera where the goal is to take a picture composed of pixels. The 96,000 x 72,000 resolution isn’t referring to how many pixels our sensor has. It’s also not the result of software interpolation. It’s a unit of measurement for describing how finely the TrackIR can detect movement.
Our resolution specs relate to an ability to take a tracked object which spans 10’s to 100’s of pixels and extract its position. Because our objects span multiple pixels, their weighted center position is likely to fall between pixels. From that starting point the fidelity of the tracking is determined by the granularity of the image processing techniques used.
Think of how a ruler with standard inch marks is used to measure something. Sometimes the object doesn’t line up with an inch mark exactly and you want to measure more accurately. Then you need sub-inch units, like sixteenths of an inch.
Similarly, our sensor is 640 pixels wide. But, we can actually measure movement that is much smaller than a pixel — in short, sub-pixel. In fact, we can repeatedly and stably measure movements that are 1/150th of a pixel. If you spread that over the whole of the sensor’s width, that gives us a range of 96,000 locations that an object could be detected at (the “reporting resolution”). And yes, we really have tested, verified and repeated these numbers. Skeptics are welcome to conduct their own testing using our free OptiTrack™ SDK and a micrometer.
By the way — the opposite of what we do is also possible. A manufacturer, due to cost or inexperience, could choose to not perform any sub-pixel analysis and discard resolution on a device. Throwing away resolution results in increased noise and coarser tracking, which then requires excessive software smoothing (with the penalty of high latency) to make it stable. That is a poor solution to an unnecessary problem. Enabling minute, sub-pixel tracking within the hardware is a far more effective solution.