Meta Getting Closer to Showing Reality
Much of our time at the 8K Association is taken up with looking at display systems – the displays themselves, the driving electronics and the systems that deliver and decode the pixels. Like most of the industry, we tend to spend less time on the other side of the interface. Displays are, after all, just an output interface that is designed to work with an existing biological input interface, the human visual perception system (HVS). The better the output system matches to the input system, the more effective and efficient the interface.
So, you might expect that the industry would spend almost as much time and energy looking at the input side as the output side. However, there’s not much engineers can do about the input side (although I’m very grateful to surgeons and doctors that have recently worked on my visual input system!). The emphasis tends to be on the output devices, rather than on the HVS, although as we regularly point out, it was testing on viewers by Dr Masaoka of NHK that led to the development of 8K as an international standard.
Is 20/20 Vision The Ultimate?
It has often been a frustration to the author that even some highly technically qualified engineers and scientists get hung up on ’20/20 vision’ as the level that sets ‘visual perfection’. The reality, of course, is that 20/20 vision is merely ‘average vision’ and indicates that the viewer can read detail on a Snellen chart at 20 feet that is the same as an average viewer can see at that distance (or 6/6 Vision in metres). If the viewer can read smaller detail than the average, the ratio will show that. Really good vision might be down to 20/10 – that is to say that the viewer can read at 20 feet what a normal viewer can read at 10. It is estimated that 10%-15% of the population has better than 20/20 vision.Â
The limit of human vision (without correction) and measured with a Snellen chart is estimated to be around 20/10 or 20/8. According to Wikipedia, some birds of prey have acuity of around 20/2!Â
Now, this idea of acuity being defined by the ability to identify black and white shapes on a chart is a quick and useful tool for optometrists and others. However, it is a dramatic simplification of the ability of eyes to resolve images in the real world. For example, humans can be extremely good at detecting when two lines are slightly misaligned, a phenomenon known as vernier acuity. Â The natural world is rarely so geometric and clearly black and white.
Changing 20/20 to CPDs
Now, the 20/20 measure of acuity is generally for distance vision. To have a metric that is less sensitive to viewing distance, a common alternative is to look at the ability to detect alternating patterns of light and dark lines, known as grating acuity. It’s the kind of pattern used to measure the resolving powers (MTF) of lenses and is used in the characterization of resolution in the IDMS display testing standard.
The units to measure acuity in this way are ‘cycles per degree’ (cpd), in other words, the number of changes from black to white (or the reverse) in one degree of the human visual field. Typical values would put 20/20 vision at around 30 cpd and 20/10 vision at 60 cpd. This metric is especially useful when trying to understand the resolution requirements at long viewing distances, for example when projection is being used or large video walls specified, or when the eye is very close, in XR applications.
The News from Meta at Siggraph
All of that is a long preamble to get to the item of news in this article. At the recent Siggraph event, Meta showed a new prototype headset called ‘Tiramisu’ that demonstrated an amazing angular resolution of 90 pixels per degree or 45 cpd. In recent years, Meta has deliberately tried to extend what can be done in headsets by developing prototypes that are intended to be the state of the art in one particular visual element (field of view, brightness, HDR etc). This is part of a long term quest to meet what it calls the ‘Visual Turing Test’ – the same concept that Dr Masaoka had, of trying to understand when humans could no longer tell the difference between a displayed image and reality.Â
So, although Tiramisu is described by those that have seen it as having ‘stunning’ levels of clarity (it also had high brightness and great optics), the need to put a lot of resolution in a small area meant a limited (33 x 33 degree) field of view. It’s also very big and heavy.
The Tiramisu headset does not try to get to any sense of ‘balance’ in specification, aiming at great resolution, contrast and brightness. There is a capture of images from the headset that gives some sense of the quality shown on the Meta blog. Those that have seen the headset in action point out that the capture ‘doesn’t do justice to the full impact’ of the brightness and HDR of the unit.
At Siggraph, Meta did show a lens stack similar to the one in the prototype that might be used in a ‘Tiramisu 2’ which would be half the brightness (700 cd/m2) and just 60ppd, but might be more practical. (Note that Meta tends to quote ‘pixels per degree’ rather than ‘cycles per degree’, so as you need two pixels – one black and one white – per cycle, their number are typically double. 60 PPD is 30 CPD = 20/20 vision)
Meta Creates a New Flat Panel Laser Display
The Meta blog article about the visual Turing test also highlighted that ‘Lasers could ultimately prove impractical for VR’ so it was also intriguing this week to see an open access article in Nature that described a new laser display architecture that works with an LCOS imager to potentially create a flat panel laser display. The researchers mentioned as authors work at Meta’s Reality Labs. Of course, it can be a long way from a paper in Nature to a practical product, and the Nature article really highlights AR rather than VR, but it’s intriguing to see the work that Meta is doing.
