Resolution analysis for HMD helmets
Resolution analysis for HMD helmets
by Marc Bernatchez
Last revised March 24, 2007
Copyright © 1995 - 2007
All Rights Reserved
Buying a HMD can be a real problem. Due to the variety of configurations that are available on the market, it is practically impossible to evaluate and compare different models. The technical sheets are often confusing due to the large amount of numerical data that, some times, seems contradictory. The following document is meant to give a different approach concerning the representation of these numerical specifications for the HMDs. The best way to evaluate the perceived image quality from the user point of view is, in my opinion, to use the angular resolution approach. This measure offers the big advantage to make it possible to compare HMDs that are strictly different from a hardware point of view (i.e.: LCD Vs CRT based HMDs, different resolutions of displays, ...).
The author of this article desire to bring reader's attention to the fact that the content of this page doesn't constitute, in any case, a complete evaluation of the HMD's performances reviewed. In no case can the author be held responsible of any damages caused by the use of the present information in this document. Read our disclaimer please.
The fundamental principle
The evaluation of resolutions that follows takes into account two important factors:
• The raw horizontal and vertical resolution of the internal display system (LCD or CRT) in pixel units
• The horizontally covered field of view (FOV) in degrees
The resolution is measured in terms of minutes of arc per pixel, which includes these two factors. We can now compare these HMDs with a normalized measure, which will give us a better idea of the overall visual quality perceived. Practically speaking, this means that with this measure unit, we can compare two HMD that use different FOV or raw displays resolution just by comparing their normalized resolution results.
Let us first define what an arc minute per pixel is. One arc minute represent 1/60 of a degree. So the angle defined by the left and right extent of a picture element (pixel) define a value in arc minute per pixel. The figure 1a illustrate this definition.
For the human, the maximum angular resolution of the eye is around 1 arc minute for the smallest point that can be seen (it can be as little as 1/3 minutes of arc per pixel in some specific circumstances). This high visual acuity region is localized in the fovea on the retina, where there is the highest density of cones. These cones are mostly responsible for the central color vision.
In the next section, we will analyse HMD specifications in regard with what we just discussed. We will try to compare their performances in term of perceived visual resolutions.
HMD list and their specifications
|Manufacturer||ModelName||Hresolution||Vresolution||HFOV||VFOV||Overlap||Price||Date Last updated|
|CyberMind||Visette45 SXGA||1280||1024||36,00||27,00||36,00||9649||06-07-05 00:00|
|eMagin||X800 3D Visor||800||600||32,00||24,00||32,00||799||01-16-06 00:00|
|eMagin||Z800 3D Visor||800||600||32,00||24,00||32,00||549||05-14-06 00:00|
|Fifth Dimension Technologies||5DT HMD 800-40 3D||800||600||32,00||24,00||32,00||9895||01-27-06 00:00|
|iO Display Systems||i-glasses PC/SVGA-3D Pro||800||600||21,00||16,00||21,00||949||01-27-06 00:00|
|Kaiser Electro-Optics Inc.||ProView SR80||1280||1024||63,00||53,00||63,00||31500||05-14-05 00:00|
|nVision||nVisor ST||1280||1024||40,00||30,00||40,00||34800||06-06-05 00:00|
|nVision||NVISOR SX||1280||1024||48,00||36,00||48,00||23900||06-10-04 00:00|
|Rockwell Collins||ProView VO35||800||600||28,00||21,00||28,00||5500||05-14-06 00:00|
|Rockwell Collins, Inc.||Proview XL50||1024||768||40,00||30,00||40,00||16995||05-14-06 00:00|
|Rockwell Collins, Inc.||ProView XL50 STm||1024||768||34,40||25,80||25,46||49000||01-27-06 00:00|
|Rockwell Collins, Inc.||Sim Eye SR100||1280||1024||100,00||50,00||30,00||ND||01-07-06 00:00|
|SaabTech||Saab AddVisor 150||1280||1024||36,80||27,60||36,80||95000||12-13-04 00:00|
|SEOS||SEOS HMD 120/40||1280||1024||120,00||67,00||40,00||62500||01-07-06 00:00|
|Virtual Research||VR1280||1280||1024||48,00||36,00||48,00||15500||01-27-06 00:00|
|Virtual Research Systems||V8||640||480||50,00||38,00||50,00||ND||06-10-04 00:00|
|Virtual Research Systems||VR4||480||240||48,00||36,00||48,00||ND||02-04-05 00:00|
|Manufacturer||Model Name||Adjusted HFOV (One eye)||Angular Res||Fhar||Ffov||Human Match Factor||Fprice||Global Factor|
|eMagin||X800 3D Visor||32||2,40||76,00||15,24||45,62||87,16||59,47|
|eMagin||Z800 3D Visor||32||2,40||76,00||15,24||45,62||89,96||60,40|
|Fifth Dimension Technologies||5DT HMD 800-40 3D||32||2,40||76,00||15,24||45,62||59,89||50,37|
|iO Display Systems||i-glasses PC/SVGA-3D Pro||21||1,58||84,25||10,00||47,13||85,73||59,99|
|Kaiser Electro-Optics Inc.||ProView SR80||63||2,95||70,47||30,00||50,23||45,14||48,54|
|Rockwell Collins||ProView VO35||28||2,10||79,00||13,33||46,17||67,09||53,14|
|Rockwell Collins, Inc.||Proview XL50||40||2,34||76,56||19,05||47,81||53,06||49,56|
|Rockwell Collins, Inc.||ProView XL50 STm||29,928||1,75||82,46||16,38||49,42||39,42||46,09|
|Rockwell Collins, Inc.||Sim Eye SR100||65||3,05||69,53||47,62||58,58||0,00||0,00|
|SaabTech||Saab AddVisor 150||36,8||1,73||82,75||17,52||50,14||30,80||43,69|
|SEOS||SEOS HMD 120/40||80||3,75||62,50||57,14||59,82||36,26||51,97|
|Virtual Research Systems||V8||50||4,69||53,13||23,81||38,47||0,00||0,00|
|Virtual Research Systems||VR4||48||6,00||40,00||22,86||31,43||0,00||0,00|
The first parameter we are going to consider is the angular resolution, which gives us a good idea of the perceived image resolution. The angular resolution of an HMD is a very critical factor. It enables us to evaluate the perceived visual acuity of any HMD, regardless of the field of view it uses. The angular resolution could almost be called a "normalized resolution" for these devices. So, no matter how the optics rearrange the images you see inside the HMD, the angular resolution will tell you exactly what is the image resolution you will see. This is true because this measure is taken after the image has traveled trough all the optics of the helmet. As a result, it represents what the user's eye really sees. We can thus compare HMD models, side by side, using this unit.
Always keepin mind that smaller values represent better visual quality for this measure unit. Ideally, the angular resolution for a given HMD helmet should be lower than 4.43 minutes of arc per pixel. This value is equivalent to the image perceived at 2 feet of a 17" computer monitor displaying an 800 x 600 image. This is considered as a minimum these days for most applications requiring object and text viewing.
In the 90's, many low cost HMD were well above that value, making text reading an almost unattainable goal. As we can see from the above figure, things have improved since then. Today, HMD manufacturer tend to try to have an average angular resolution of 2,30. Considering that many of these HMD, which have angular resolution under 2,30 also have FOV below 25 degrees (horizontal), it seems fair to say that the initial pure VR immersion approach lost over more down to the ground requirements such as being able to read text for example. For the low-cost market, this makes obvious sense. Low-cost HMD imply low-resolution LCD displays. The best marketing trade-off for these HMD is to value classical 2D tasks, like reading text and watching video sources, instead of stereoscopic viewing and immersive high field of views. Meanwhile, these low-cost HMD devices are bringing the general public to know this technology. The public education aspect is very important in a so specialized market. It will hopefully stimulate the use of this technology and in turn, favour its growth in terms of technical improvements. We are still far from the ideal HMD that many have been dreaming of for the last decade.
Field of view (FOV)
The second parameter we must consider is the field of view (FOV). You can find the values for each HMD model in the previous table. It's quite difficult to build upsuch a table of FOV values for different HMDs. This is due to the fact that there is no actual standard in the way manufacturer gives these numbers. In this analysis, FOV values should always represent a 100% stereoscopic overlapconfiguration unless otherwise specified. Note that you need at least 20 degrees of overlapto suit the human visual system.
The above figure gives the total horizontal field of view as given by the manufacturers. As you may have noticed, there is also an "adjusted HFOV" column in the previous tables. This column's purpose is to take into account the visual overlapof HMDs when evaluating angular resolutions. The adjusted FOV column gives the FOV for a single display element (single eye) of the HMD device. A more detailed explanation about overlapping follows in the next section.
What is stereo overlapping?
Stereo overlapping is a process by which the individual field of views of each eye are voluntarily shifted horizontally so that the global HMD FOV is larger. What does this mean? The following figure illustrates the general principle:
In classical HMD designs, particularly low-cost implementations, the stereo overlapis set at 100%. This means that both eyes see the entire field of view of both LCD displays. This configuration has the advantage of providing a stereoscopic scene over the entire FOV of the HMD device. The inconvenient of using such scheme is that the maximum usable FOV is the FOV of each single LCD display making the HMD. In other terms, if the HMD optics is such that each eye sees a 50 degrees wide scene, the total FOV seen by both eyes will also be 50 degrees wide.
High-end HMD makes use of overlapping techniques exactly to overcome this. That is, what if we could get more FOV and still use the same displays? The answer to that question is to shift each display FOV horizontally so that only a partial region of each falls in the central region of the total HMD's FOV like seen on figure 5. This technique will effectively enable a wider total FOV. The inconvenient here is that we lose the ability to display stereoscopic information on some areas of that extended FOV. For the two side regions, as shown on figure 5, only the image of one of both displays reaches that area. For stereoscopy to be achieved, two images are required for each pixels element displayed in the total FOV region of the HMD. As previously mentioned, the human visual system only requires stereoscopic information in a limited central region of its total field of view. This is to say that the humans visual system is an overlapping stereoscopic visual system.
It has been reported that a central stereopsis region of about 20 degrees is sufficient to provide a good sense of depth to the user, while providing other types of cues in the peripheral regions such as speed and movements. The high-end HMD models make uses of that principle to maximize the immersion effect in term of maximizing horizontal FOV while maintaining the perception of depths in the scene. This is not an ideal solution unfortunately. In most implementations using overlapping, the overlapregion is fixed at the center of the global HMD FOV. The problem arises when the user looks on the sides by turning his eyes inside the HMD. Since the stereoscopic central region does not move with eye movements, the users will perceive disturbing effects while watching in these situations. Fusion of images may be momentarily lost, thus causing a break in the immersion sensation. To limit this problem, a possible partial solution is to voluntarily set the overlapregion to a value greater than what the human vision system requires. This is basically providing a safety margin in case the user's gaze changes. Some very costly HMD devices uses active eye tracking systems to move the stereoscopic region along with the user's eye movements. The moving overlapping region is often called a "high-resolution inset".
For an excellent information source about stereo overlapand other advanced HMD topics, see this document:
Helmet Mounted Displays: Design Issues for Rotary-Wing Aircraft
Edited by Clarence E. Rash
Visual Sciences Branch
Aircrew Health and Performance Division
U.S. Army Aeromedical Research Laboratory
Fort Rucker , Alabama
U.S. Army Medical Research and Materiel Command
Fort Detrick, Maryland
With a foreword by COL Cherry L. Gaffney
Human match factor
Now that we have categorized the two principal factors enabling us to compare HMD's, that is, the angular resolution and the field of view, we still need a convenient way to evaluate and rank each HMD with a single quantitative value. That is where the human match factor comes in. As its name implies, its purpose is to quantify the degree to which a given HMD technical performances match the human visual parameters.
To do so, we will use the Fhar and Ffov results from the previous tables. The calculation are as follows:
Fhar(horizontal angular resolution factor): 0 HAR = 100% to 10 HAR = 0% (linearly)
Ffov (field of view factor): 210 degrees = 100% to 0 degrees = 0% (linearly)
Since these values are based on human factors instead of technical specifications, the following results remain valid based on the assumption that the worldwide average human visual system performances does not change over time.
Let us take an example. If a certain HMD ranks at 75%, it means that this HMD match 75% of the human maximal specifications for the angle of vision (FOV) and the image quality (angular resolution). A HMD ranking at 100% would present us a scene equivalent to what we can observe in our everyday life surrounding, with maximal visual acuity.
Global quality factor
Even if we consider the angular resolution and the field of view (FOV) of a HMD, it is still difficult to really visualize the general value of a HMD with an other. For most people, there is always a remaining question: "which HMD gives the best value for the money?". Bellow is a graph that will summarize all these points taking into account the three following considerations:
• The human match factor (angular resolution and horizontal field of view)
• The price
For each of these points, a certain percentage as been computed for each HMD. The mean of the results obtained for each given HMD represent the global quality factor. Take note that this graph only represents the author's opinion. You should interpret it accordingly.
To get a 100% rank, an HMD would need to match the human acuity performances and the price would have to be near zero. This is nearly an impossible case of course.
So, the pure global quality factor value doesn't really matter. What do matter is that one HMD has a higher global quality factor rank than an other one. The global quality factor is given only as a tool to visualize differences in price / visual performance ratio among various HMD's. It's upto the reader to choose what is the most important factor for him (i.e. price, FOV or har) then consider the global quality factor of those HMDs which are in the desired range.
The Fprice factor is computed using the following function:
Why a logarithmic function? By evaluating the performance increase relative to the price of HMD's, we can see that it follows a logarithmic function. This is related to the amount of research and development costs required. High-end HMD will require much higher R&D costs. The market being more restrained in volume for high-end models, the cost also has to be higher to cover the extra R&D efforts.
Here are the results, taking into account the technical as well as the financial aspects of each reviewed HMD:
* Ranks of 0,00 in the above figure represent HMD for which insufficient data was available to compute their global factor value.
Some more in-depth considerations about these rankings
There are a few factors that can induce error in this evaluation:
• The pixels arrangement on the display device used inside the HMD
• The spacing between each of these pixels (let's call it the dark zones or inter-pixel gaps)
• Taking or not into account the depixelation filters effect
• Acknowledging that not a single human has the same visual angular acuity
The pixel arrangement used by a given display element (applies mainly to LCD's and some CRT's displays) are, to my knowledge, of two possible kinds: the linear and the delta arrangement. Thus:
While the linear arrangement permits straightforward calculations, the delta method is somewhat more difficult to work with. Some people use the KELL factor that compensate for the triangular pattern used in the delta configuration. I had several chat with people on that point way back when I started doing this page. The problem is that it's very difficult to have reliable information from the HMD's manufacturer in most cases. It is difficult to know the exact LCD model that is used in a particular HMD. I'm fully aware that there is an error included in this page due to that fact but it's out of my control. This page contains only specification data as given by manufacturers.
There is also the spacing between the LCD pixels on the display element. This represents a small percentage of the overall display surface and it probably account for some error in the previous results.
The use of depixelation filters may also dramatically change the overall visual quality of an HMD. It can enhance it or it can degrade it. This is an even more difficult parameter to incorporate in such analysis. Trying to get technical data about a filter from the manufacturer would be very difficult. How do you put these filters parameters in numbers? What kind of parameters should you include? That's yet an other point where we must bear with the small error it may induce in the final evaluation.
We do recognize that each individual has its own visual acuity limits but we can't possibly take that into account. It's known that the average human has a visual acuity of about 1 minute of arc per pixel (pixel here being the smallest naked eye visible element). True, there is a possibility to resolve down to 1/3 of an arc minute per pixel on some particular conditions... but note that those apply to all individuals, in general. Taking into account that most people will have at best a 1 minute of arc per pixel visual acuity is fair enough for this analysis to give a good idea of what we can expect from these HMDs.
Important things to keepin mind:
• This analysis is meant to give the reader an idea of what to expect from a given HMD. It can in no way give you a definitive verdict as whether an HMD is good or bad .
• This analysis is not meant to be a substitute to a full buying decision-making process. If you plan to buy an HMD, by all means, try before you buy .
• This analysis doesn't constitute a complete one. Many important factors such as the ergonomics and the weight of these units were not considered here .
• Even if a given HMD ranks lower in the global factor above, it may still be the best choice for a given application .
In the event where you would like to see other models of HMD on this page, please add a comment with the link at the bottom of this page. Also, please inform me of any inaccuracy showing in this page. Thanks.