This task was to create a large field-of-view (FOV), high-performance head-up display (HUD) to be used in the study of HMI related issues for the ACAS project. Main features of the HUD were to be high brightness and uniformity of the virtual image, high FOV, and reconfigurability of the image source. The HUD would use largely existing individual technologies, but in combination would yield a system with performance levels not yet demonstrated in previous concept HUD development projects. Specific goals and objectives are:
The intended use of this HUD was to be part of driver interface studies, therefore it was important to create a very high-quality HUD virtual image. This would then reduce the influence of HUD image viewability on driver performance data to be gathered during close-course testing. Thus, an essential part of this project was the development of display source and backlight. The future trend of HUD image content strongly suggests reconfigurability, and this project would demonstrate an image source having color, high resolution, high contrast, high brightness and good brightness uniformity.
The first design iteration of the HUD used the Cadillac Seville as the target installation vehicle. As in all automotive HUD development exercises, some compromises had to be made with regard to the original specification. For example, the display source had to be changed. The originally specified Sony display was not available in sufficient quantity and with appropriate drive electronics, so instead a Seiko-Epson display was used for the initial design. The main difference in using this source was a loss of vertical display resolution, which went from 5.3 pixels per milliradian to 2.2 pix/mrad. (Horizontal resolution remained adequately high.)
The horizontal field-of-view (HFOV) of the virtual image had to be reduced to 4.5º from 5.0º. The original spec actually called for only 4.0º HFOV, but an internal goal was set at 5.0º based on a very preliminary assessment of package size and other estimated optical parameters. Similarly, the projection distance (PD) of the virtual image was reduced from 4000 to 2500mm. Both the PD and HFOV were reduced to enable better overall optical system performance, particularly vertical disparity. This is the angular difference strictly in the vertical direction between two eyes trying to resolve a single point on the virtual image, and it is considered a key measure of overall image quality.
Vertical disparity (VD) tends to be increasingly worse as one looks away from the center of the virtual image, meaning that it is worst at the edges of the image. Reducing the FOV helps to alleviate the maximum VD values seen in the earliest 5.0º-wide designs. Similarly, reducing the required projection distance in turn reduces the system magnification and VD, resulting in better overall optical performance.
Lastly, the lookdown angle was reduced from 6º to 3º. This was purely a packaging issue as there is no way to physically position a HUD in this particular vehicle such that the nominal driver has a 6º lookdown angle. In fact the upper edge of the dash top pad is designed to stay below a 5º lookdown angle, implying that 6º down will be looking where there is no appreciable windshield area to be used as a combiner, and some or all of the HUD image will be below this lower limit. 3º down was chosen as a maximum value that could allow drivers of various heights (99th percentile) to see all of the HUD image from the specified eye motion box. Either the 3º lookdown was to be accepted against the 6º specified requirement or a new vehicle would need to be chosen.
The benchtop demonstration originally called for using the installation-intent unit (the Cadillac Seville design). It was determined, however, that this would require a premature design freeze on the Cadillac design in order that an in-vehicle version and a benchtop version could be fabricated in parallel. Instead, a "borrowed" HUD from a different, unrelated program was utilized for the benchtop demo. This allowed full design optimization and completion for the preliminary Cadillac design while the bench demo was built as a now dissimilar parallel project.
The main objective of this benchtop demonstration was to show the installation-intent LCD and representative backlight, not necessarily to show the whole HUD system and its other optical performance parameters. The HUD used was not originally designed for a 1.3--diagonal LCD, so there was some degradation in distortion and vertical disparity. However, the unit showed installation-intent performance from the display and the single halogen bulb backlight in terms of color, brightness, contrast, image size and resolution. This demo hardware was later used to evaluate various potential backlighting solutions for the final system installation.
Because the lookdown angle was so "shallow" in the Cadillac Seville design (3º rather than 6º), an alternative vehicle was sought. It was thought that a sport utility vehicle would provide a more ideal HUD seating and viewing arrangement from a human factors perspective. It appeared at the time that a Chevy S-10 Blazer would be available for the project, thus a new design for this vehicle was developed (the vehicle would become unavailable at a later time).
The resulting design achieved a 5º lookdown angle, much closer to the specified 6º. The image projection distance was raised to 4100mm, much more in line with the 4000mm requirement. Also, overall optical performance, including VD and distortion, were improved thanks to a spherical fold mirror introduced into the design.
Mechanically, a larger HUD package did result from the increased projection distance, but the Blazer had more dash volume available so it was of little impact. However, alignment of internal optics was more critical with the powered fold mirror. A flat fold mirror is much more forgiving in regard to alignment as it serves the same function at any range of positions and angles, while a curved mirror is intended to function at a very specific location and angle in the optical path.
Although the Blazer was the better of the two vehicle-specific designs, the vehicle to be used became unavailable and the Cadillac design was eventually built. However, the Blazer hardware did get used in the intermediate operational demonstration.
Between the benchtop demonstration and the full vehicle installation, an operational demonstration was put forth to have a first turn at this type of HUD in a vehicle. This would allow for any intermediate lessons learned to be incorporated into the final design. The demo was done in the Blazer while the vehicle was still available. The HUD system showed installation-intent performance in image qualities such as FOV, image location, resolution and brightness. Showing this image and the HUD opto-mechanical unit itself was the primary focus of the demonstration.
A number of other things were not as intended for the final vehicle presentation, such as:
The main requirement relative to backlighting was that uniformity must be such that no greater than a 2:1 variation in brightness across the area of the LCD is present. 2500 ft-L was established as a goal for brightness given initial known quantities such as bulb/HID brightness, LCD transmissivity and HUD optical efficiency. The first system trial was thought to be best of all possible approaches, that being a combination of 2 HIDs (high-intensity discharge lighting) and one halogen bulb delivering light via a glass fiber bundle. The intent was to use the HIDs during daytime and the halogen bulb at night. It was understood that there are challenges to using HIDs in automotive applications over more conventional lighting techniques, but it was thought that the highest brightness would be achieved this way.
The challenges that became apparent were cost, packaging size, and dimming difficulties. Dimming cannot take place as it does with other lighting techniques by simply lowering voltage or varying signal pulse width; the source light must be only fully on or off. This led to the use of a polarizer that could be rotated to modulate the amount of light reaching the LCD. This approach worked adequately except for its sensitivity to heat (discolored and/or melted polarizers) and its non-linearity of dimming relative to the linear rotation of the polarizer.
Following the operational demonstration in which this approach was used, a single halogen bulb-only approach was revisited. Originally, the single bulb was not considered because of its resultant nonuniformity at the LCD (nearly 10:1). However, a “light tunnel” was developed that not only captured and recycled stray light, but also redirected the distribution of light such that significant uniformity gains were realized. When combined with a diffuser, the final system uniformity was 1.5:1, while achieving brightness levels very close to that of the HID system. With significantly less risk involved using the single bulb approach, this was the final backlight system used.
This step was essentially a refinement of the preliminary design created for the Cadillac Seville. With the Blazer vehicle unavailable, the Cadillac was indeed the final target vehicle. Recall that the virtual image projection distance (PD) was limited to 2500mm in the preliminary design since optical performance was degraded at higher values of PD. This preliminary design used an eye motion box size of 175mm in the horizontal direction, while the spec only called for 140mm. 175mm is used in most typical automotive HUD applications, including high-volume production programs. Reducing it to 140mm enabled an increase in the PD to 3600mm while maintaining good overall optical performance. This result was much closer to the original goal of 4000mm PD.
With the final design built and installed in the Cadillac Seville, the following performance parameters were measured (Table 3.16):
Table 3.16: HUD Performance Parameters in Cadillac Seville.
| Parameter |
|
|
VFOV |
|
|
| HFOV |
|
|
| Projection Dictance |
|
|
| Ghosting |
|
|
| Vertical Disparity |
|
|
| Brightness |
|
|
| Uniformity |
|
|
The increase in PD to 3600mm caused a increase in package size that was difficult to accommodate in the dash volume available, but proper integration for presentation was achieved. It was originally thought that the HID and fiber bundle approach to backlighting was superior to other methods. It was seen, however, that a single halogen bulb with an appropriate light tunnel interface to prevent light loss and increase uniformity could have very similar performance while involving significantly less risk, cost, and packaging constraints.
A significant challenge was the shallow lookdown angle present in the Cadillac design. With the only solution being an alternate vehicle, the Blazer was chosen and an even better overall design resulted, including a lookdown that increased from 3º to 5º, although not fully the recommended 6º. This solution was not fully utilized, though, as the Blazer couldn't be used for the final vehicle installation.
Nighttime brightness of the HUD was not measured in the demonstration vehicle, and it is possible that the backlight dynamic dimming range does not extend low enough for extremely dark conditions. This was not addressed during the program since the focus of the effort was on daytime and because of a lack of development time.
The lookdown angle was reduced significantly in this vehicle installation due to packaging constraints. This is largely a function of the particular vehicle with its particular windshield rake angle, available dash depth and volume, and location of other systems such as HVAC, steering column, etc. Two vehicles total (Cadillac Seville and S-10 Blazer) were adequately evaluated for suitability to this HUD application. The Blazer was more favorable in terms of lookdown angle, but was not available for the final build. Perhaps a number of other specific vehicles as well as generic vehicle types (passenger car, light truck, class-8 truck, etc.) should be assessed for HUD applicability. This list could then be referenced at the beginning of similar HUD demonstration projects at the vehicle selection stage, and the risk of surprise vehicle insuitability would be reduced.