40.2/Dewald
SID 00 DIGEST •
1
Sequential Color Recapture and Dynamic Filtering: A Method of Scrolling
Color
D. Scott Dewald, Steven M. Penn, and Michael Davis
Texas Instruments Incorporated, DLP
TM
Products Division, Plano TX, USA
Permission for publication, courtesy of Society for Information Display
Abstract
Scrolling color has long been a goal of the projector industry, as
it enables the most efficient use of light in a single panel display.
Current methods of implementing scrolling color use the
techniques of splitting the light into primary colors, and
manipulating that light on the modulator. The authors present
the techniques of dynamic filtering and sequential color recapture
(SCR) to achieve the same result with no moving components
other than a color wheel, showing that the efficiency of 3-
modulator systems can be approached with one modulator.
Analysis of the technique applied to DLP
TM
projection displays,
and results of prototype projection systems using the techniques,
will be presented.
1. Introduction
Creating a full color display with a single modulator has been a
goal of the projector design community for some time. Currently,
the accepted method used is sequential color, or “field sequential
color” [1,2]. In this method a sufficiently fast modulator (DMD
or LCD) creates three or more modulated images in primary
colors per video frame, which are switched sufficiently fast to
create a full color image in the human visual system. This is the
method used in many 1-chip DLP
TM
projectors in the market
today.
Another method of full color display using a single modulator is a
scrolling color method, several types of which are described in the
literature [3, 4]. This method has several advantages, the first
being that all colors are present on the modulator at the same time,
so the waste of light caused by field sequential color is avoided.
The second advantage is the reduction of “color separation
artifacts”, which are caused by quick eye movements or a fast
changing scene when viewed on a field sequential color display.
This is caused by successive colored fields of the image being
sufficiently displaced on the retina to yield a “fringe” of colored
light at a white-black boundary for instance.
Scrolling color optics in the literature use numerous optical
surfaces and moving optical components to manipulate the color
bands on the modulator. This is quite complex mechanically, and
has been shown to be much less efficient than theory predicts.
The authors show sequential color recapture (SCR) as a solution
to the scrolling color problem with a minimum of optical
components.
2. SCR Components
SCR was made possible by advances in dichroic and metal thin-
film technology, the most important of these being the
photolithographic patterning of dichroic coatings. Similar to
today’s 1-chip DLP
TM
projectors, a rod-type integrator and color
wheel are used with modifications as described below.
2.1 SCR Integrator
Integrators for SCR are similar to standard rod-type integrators
with the addition of an “entrance aperture”, formed by mirroring
the input end of the integrator, with the exception of a circular
transparent area of approximately 1/3
rd
the cross-sectional area of
the integrator. The size of the input aperture can be adjusted to
maximize the efficiency or the projector and is a function of the
etendue of the lamp and the device, as will be described below.
SCR integrators have been fabricated out of solid glass as well as
high-reflectance “light tunnel” material. See Figure 1.
Figure 1. – SCR Integrator (solid)
2.2 SCR Wheel
The SCR wheel is created from RGB dichroic coatings arranged
in a “spiral of Archimedes” pattern. The pattern should be sized
such that one RGB pattern covers the cross-section of the output
end of the integrator (Figure 2). The Archimedes spiral, defined
by the equation R=aθ, has the property that the boundary between
colors moves at a constant speed in the radial direction. This
causes the RGB pattern to move at nearly a constant speed over
the output face of the SCR integrator described above. For best
performance, transmission of the dichroic coatings should be
maximized in-band, and reflection maximized out-of-band. It is
also possible to include a “white segment”, or a clear area which
can be used to increase luminous efficiency in non-saturated
images.
40.2/ Dewald
• SID 00 DIGEST
2
Figure 2 – SCR wheel and integrator
The spiral pattern is manufactured using a proprietary lift-off
photolithographic technique that results in a very accurate
patterning of the 3 dichroic coatings. The number of RGB stripes
determine the speed of the wheel rotation, e.g., a wheel with many
RGB sets can rotate at a slower speed for a given “scan rate” (the
frequency at which a given color makes a full sweep of the
modulator). A photograph of an SCR prototype wheel is seen in
Figure 3.
Figure 3 – SCR Wheel
The SCR process also allows smaller wheels to be used in
compact projectors. Depending on the ability of the modulator to
handle curved color boundaries, the wheel size may be limited by
the actual motor diameter and the integrator size. The authors
have investigated SCR color wheels of less than 35mm diameter.
2.3 Relay Optics
The SCR wheel is placed such that the reflective dichroic spiral is
very close to the output end of the integrator. A relay optic
system then creates an image of the output end of the integrator
onto the light modulator at the proper f/# and magnification. As
the SCR wheel turns, it can be seen that slightly curved bands of
each primary color move across the modulator at a near-constant
speed. With the proper formatting of data, a scrolling color image
can be realized.
The importance of the integrator-wheel interface should be noted.
The gap between the two components should be as small as
possible, as light “leakage” can occur around the perimeter of the
interface. Also, the runout of the wheel/hub/motor assembly will
cause a variation of the spacing with time which can cause a
visible pulsing of the image if the motor speed is below the
threshold of human vision, approximately 3000RPM. Modeling
of the SCR process qualitatively indicates that the total runout
should be less than .1mm.
3. The SCR process
3.1 Recycling of input light
Light from a small-arc lamp is focused onto the input aperture of
the integrator. The light that passes through the input aperture is
homogenized by multiple reflections off the wall of the integrator.
If the integrator is sufficiently long the light distribution at the
output end will be fairly uniform. When the white light reaches
the SCR wheel, light of a given color will transmit through the
section of the wheel with the corresponding transmissive coating,
while reflecting off the remaining 2/3
rd
of the illuminated area. In
other words, red light will pass through the area of the wheel
covered by the red dichroic segment, but will be reflected back
towards the opposite end of the integrator by the blue and green
segments. This effect occurs continuously with light of all three
colors.
The light that is reflected by the wheel continues to reflect off the
walls of the integrator, further homogenizing the rays, until the
input aperture is reached. At this point 2/3
rds
of the light is
reflected by the mirrored surface, and the remaining 1/3
rd
passes
through the aperture to the lamp, and is assumed to be lost
(though tests have indicated that a fraction may return to the
integrator).
The remaining “2/3rds” mentioned above is homogenized again
on the way to the output end of the integrator, where red light is
allowed to pass through the red segment of the wheel while
reflected by the other segments, as described above. This process
is repeated several times until all the light that entered the input
aperture from the lamp is either transmitted to the modulator,
absorbed, or scattered (Figure 4).
Figure 4. – The SCR recycling process
3.2 Quantification of SCR Gain
SCR gain is described as the increase in intensity of the RGB
segments of the wheel, when imaged onto the modulator, over the
case where no light recycling was present. When no recycling is
present the scrolling system and a standard field-sequential system
should have identical efficiency (provided the F/S wheel has equal
size segments and no clear segment) since the duty cycle of a
particular color on a given pixel of the modulator is exactly 1/3.
Lamp arc is focused onto
opening in resonator input
end (requires small arc)
Light is homogenized
Color wheel transmits
RGB, reflects CYM back
toward input end
RGB stripes are imaged
onto DMD and scroll.
Color Wheel slowly rotates
Relay
Optics
Input aperture reflects 2/3
of CYM light back towards
color wheel.
Light is homogenized
40.2/Dewald
SID 00 DIGEST •
3
Also, when no recycling is present, the light that is not transmitted
to the modulator through the relay optic is returned to the lamp
and turned into heat. Simply put, when red is present on the
modulator blue and green light must be wasted.
Consider an SCR system with an input aperture of 1/3 the area of
the integrator. The light from the lamp which matches the red
wavelengths is homogenized and reaches the wheel, and 1/3 of the
area of the integrator is covered by red-transmitting coatings, so
1/3
rd
of the red light passes through the wheel to the modulator.
The remaining 2/3 of the red light returns to the input
aperture/mirror, and 2/3 of that light is returned to the output end
where 1/3
rd
of the red light is matched with the red coating and is
transmitted. This process is repeated several times. The series
1/3[1+(2/3)
2
+(2/3)
4
+….] represents the fraction of light of a
given color to eventually pass through the wheel. The series
converges to 1/3[1.8], meaning that an increase in efficiency in
the order of 80% can be realized.
Predictions of the expected efficiency gain can be more accurately
predicted using the following formula
()()
∞
=
ù
ê
ë
é
÷
÷
ø
ö
ç
ç
è
æ
−
÷
÷
ø
ö
ç
ç
è
æ
−=
0
21
21
11
n
n
INTINT
RR
A
A
A
A
GAIN
(1)
where A
INT
is the area of the integrator, A
1
and A
2
are the areas of
the input aperture and one of the color segments of the wheel,
respectively, and R
1
and R
2
are the total reflectance (including
absorption and scattering losses) of the input aperture mirror and
color wheel segment, respectively. As can be seen, the recycling
efficiency will increase as the ratio of the input aperture to the
integrator cross-sectional area decreases. For instance, if the input
aperture is 25% of the integrator area, the best-case theoretical
efficiency boost is 2.0. Even though the efficiency boost can be
increased by reducing the input aperture size, the reduced
coupling of the arc into the smaller area may reduce overall
efficiency of the projector.
3.3 Thermodynamic Considerations
Even though the predicted efficiency gain of 1.8 is much lower
than the theoretical efficiency gain of 3.0 for a conventional
scrolling system, practical application of the SCR optics decreases
the advantage of conventional color scrolling. The SCR optical
system has the same number of components as a conventional
DLP
TM
1-chip optical system, namely, the wheel, rod integrator,
and 3-5 element relay optics. On the other hand, a conventional
scrolling-color system will require 3 color splitting dichroics,
mirrors for positioning the 3 beams, optics for focusing each color
onto the modulator, and optics for manipulation of the beams to
cause the scrolling action. Each focusing step has the effect of
increasing the etendue of the lamp due to irreversibility of optical
aberrations, as well as providing opportunities for light to be lost
due to absorption, scattering and misalignment. Also, properties
of the scanning process can cause the color segments to change
size as they travel over the modulator resulting in an overscan
condition where areas outside the modulator array are illuminated.
The combination of these losses brings a practical value for
scrolling color efficiency gain down to a level on par with SCR
predictions. It should also be noted that the mechanical
complexity and volume required for conventional scrolling optics
is much higher than for SCR illumination.
3.4 Etendue Considerations
The SCR light recapture process has the effect of increasing the
etendue of the lamp arc by factor determined by the ratio of the
rod exit area to the input aperture size, nominally 3.0. This agrees
with earlier work [4] in that the lamp arc etendue must be
decreased by a factor of three to achieve similar coupling
efficiency as a non-scrolling system, since the arc must be focused
onto an area of 1/3
rd
the area of the modulator.
Figure 5 shows a graph of light collection for a UHP-type lamp
(1.3mm arc gap) with an f/1 elliptical reflector as a function of
aperture diameter (A1). This would correspond to the light that
would be admitted into the SCR integrator for a given input
aperture size. Also in the figure is a graph of equation (1) with
the following assumptions:
Modulator: .7”XGA DLP
F/# of illumination/projection: 2.4
Modulator etendue = 19.5 Str-mm
2
R
1
=R
2
=.95; A
2
/A
INT
=1/3
Figure 6 shows the product of the curves of Figure 5. Note that a
value for A
1
can be selected corresponding to maximum overall
projector efficiency. Note that the maximum efficiency occurs at
an aperture diameter of 4.2mm, which is much larger than the
3.6mm value, which would represent 1/3
rd
of the integrator area.
This is because the lamp arc etendue is much larger than would be
desired for high projector efficiency.
Figure 5 – Lamp Light Collection and Recycling Gain
Figure 6 – Projector efficiency vs. Input aperture diameter
0.8
0.85
0.9
0.95
1
1.05
1.1
22.533.544.555.5
Input aperture diameter (mm)
Relative Projector Efficiency
1000
1500
2000
2500
3000
3500
4000
4500
22.533.544.555.5
Input Aperture Diameter (mm)
Lumens collected into aperture
1
1.2
1.4
1.6
1.8
2
2.2
Recycling gain
Lumens
Gain
40.2/ Dewald
• SID 00 DIGEST
4
4.0 Prototype System
During the summer of 2000 the Advanced Optics Technology
Team of Texas Instruments’ DLP
TM
Products began work on a
Prototype SCR projection system.
4.0.1 Prototype System Parameters:
Modulator: 0.9” SXGA 13.8 micron pitch DMD
Illumination: f/3 telecentric
Integrator: 4.6x6.2mm solid BK7 with 3.6mm aperture
deposited on the input end.
Lamp: Osram 120W VHP with f/1.0 elliptical reflector
SCR Wheel: 126 segment spiral rotating at 660 RPM
Figure 7 shows the prototype optical system. A special formatter
was designed to allow scrolling data to be displayed on the DMD
device.
Figure 7 – SCR Prototype Optical System
4.1 Predicted Performance
Based on the color wheel filters used, the predicted “color wheel
efficiency” of the spiral wheel was 0.342. This is essentially the
average photometric transmission of the wheel with an adjustment
for “white boost”, which is the use of the regions between the
color bands, which are marginally cyan, yellow, and magenta
light, to enhance the brightness of white content.
To calculate the predicted lumen output of the prototype SCR
projector, one must multiply the following components of the
light budget:
Lumens captured by the input aperture = 3300
DMD Optical and Electronic efficiency = 0.68
Optical efficiency including projection lens = 0.70
SCR Gain Estimation = 1.67 (from gain formula with
5% loss at each end)
Color Wheel efficiency = 0.342
The product of the above 5 numbers is 897 lumens, the predicted
output of the SCR prototype projector.
4.2 Measurements
The SCR prototype was first measured in June 2000 in Plano, TX.
ANSI brightness measurements of the image averaged 902
lumens, which was very close to the predicted value. Due to the
non-flatness of the SCR wheel used, there was a variation of the
spacing between the integrator output and the SCR wheel, which
caused a noticeable pulsation in the picture. Clearly, this is an
indication of the SCR effect. The lumenous efficiency of the
prototype was 7.5 lumens/watt(lamp), which approaches 3-chip
pSi projectors with a similar color gamut. Surprisingly, the ratio
of screen lumens to input lumens (light entering the input
aperture) was equal to that of 3-chip DLP
TM
projectors, implying
that with a sufficiently small lamp arc or sufficiently high
modulator etendue, the efficiency of 3-modulator projectors can
be attained with this technique.
5. Future Experiments and Discussion
The Advanced Optics Technology team plans to continue
experimentation with SCR and related optical components. The
authors believe that 10 lumens/Watt(lamp) with SMPTE C colors
can be optained, and that >13 lumens/Watt can be demonstrated
with a single-DMD system with a less-saturated color gamut and a
clear segment in the SCR wheel.
6. Acknowledgements
The authors would like to thank Dan Morgan, Don Doherty, Paul
McFarland, Greg Hewlett, Claude Tew, and the employees of
Texas Instruments who made the first SCR demonstration
possible.
DLP and Digital Light Processing are trademarks of Texas
Instruments, Incorporated.
7. References
[1] U.S. Patent 5,448,314 (Issued 09/05/1995) Texas
Instruments, Inc.; Heimbuch, et. al.
[2] Florence, James M. and Yoder, Lars A, Display system
architectures for digital micromirror device (DMD)-based
projectors. Proc. SPIE Vol. 2650, p. 193-208, Projection
Displays II(1996)
[3] U.S. Patent 5,410,370 (Issued 04/25/1995) Philips
Corporation, Jannsen, Peter J.
[4] Stupp, Edward H., and Brennesholtz, Mattew S., Projection
Displays (Wiley, 1999) pp. 229-230.
ISSN0000-0966X/00/3001-0000-$1.00+.00 © 2001 SID