Line for Line, Pixel for Pixel

ARTICLE

LINE FOR LINE, PIXEL FOR PIXEL
Linearity in Projectors and Monitors

by Peter H. Putman, CTS

When the topic of large screen displays comes up (as it often does these days), we hear plenty of acronyms tossed around (LCD, DLP, CRT, ILA, etc.), as well as terms like grayscale, uniformity, brightness, "native resolution", and so forth. We may even hear the words "minus 3 db bandwidth"" or "modulation transfer function" discussed, if the participants are really into the subject.

What we're not as likely to hear is a discussion about linearity, or the ability of an electronic display to faithfully reproduce all the parameters of the input signal driving that display. Yet linearity is the most important property of any electronic amplification circuit, whether it is a high-performance stereo amplifier or a three-gun CRT video projector.

We live in a linear world, and our senses respond in a linear fashion to aural and visual stimuli from all around us. How we respond to those stimuli depends on their linear nature: A horn blast that gets progressively louder tells us a train is approaching. A blinking light in the sky that becomes steadily dimmer indicates an airplane moving away.

We can also judge relative intensities of stimuli over a wide range, thanks to that marvelous computer inside our noggins. This dynamic range lets us process a great deal of information about the sounds we hear and the sights we see.

For any large-screen display to approximate visual realism, it must also exhibit a wide dynamic range. That's a tough job for many display technologies as they try to define images with low or no voltage levels (black), high voltage levels (white), and everything in between (shades of gray or grayscales). Let's take a second to define what makes up the dynamic range of a display.

PUMPING IRON

In any linear amplifier, there is a definable measure of performance - gain. Gain is simply the difference between the input signal level and the output signal level, measured in decibels (dB) or in volts (V). Gain is not particularly difficult to measure - instruments exist to take a wide range of gain readings at audio frequencies through very high radio frequencies. Simply connect the source signal, the measuring device, and an appropriate load or termination, and you're off to the races.

But gain alone is not a useful figure without also measuring the dynamic range of the signal amplifier. What good would an audio amplifier be if it only worked at certain frequencies, or was unable to deliver the rated power without creating distortion?

For that reason, dynamic range is usually defined as the range of input signals over which the rated gain figure is achieved, up to the point which the amplifier goes into compression. That occurs when the output drops by 1 dB, also known as the 1 dB compression point.

When compression occurs, the amplifier is no longer behaving in a linear fashion, and may in fact be modifying the input signal, resulting in distortion. These distorted signals are no longer faithful reproductions of input sine waves, but square waves. Square waves also create harmonics, which can mix with the input signal to create signals that never existed such as distortion products and picture artifacts.

EVERYTHING'S GETTING FUZZY

Dynamic range in audio and RF amplifiers is not difficult to calculate. Video displays are a different story, though! We must depend on test patterns as well as test equipment to tell us exactly what is happening in a given projector and monitor.

CRT-based displays (monitors and projectors) are analog displays, and it's not difficult to tell when they're operating in a non-linear fashion. A cathode-ray tube is made up of three or four elements - the cathode, which emits electrons, one or two grids, which behave like faucets to vary the flow of electrons, and the anode (plate), which is what the shaped electron beam strikes to form an image.

The dynamic range of an individual cathode-ray tube can be quite high, particularly when a second grid is used (known as G2; the control grid is G1). If the tube is biased for Class A or Class AB operation, the ratio of the output signal level to input signal level (gain) can easily exceed 100:1. Still, there will be a point at which the CRT's gain will flatten out and it will go into compression: Increases in grid current won't result in corresponding increases in anode current.

There exists a real chance that the anode current will run too high, drastically shortening the life of the tube. For this reason, many manufacturers incorporate a cathode current limiting circuit such as a ballast resistor into the projector or monitor. Current limiting ensures the tube won't be pushed outside its normal operating parameters, but in effect slaps an arbitrary upper limit on the tube's gain and its dynamic range.

EVERYONE'S BIASED

The control grid (G1) bias and screen (G2) bias determine a CRT's gain and linearity. Both of these controls can be adjusted to really "crank out the lumens", again at the expense of tube life. For this reason, G1 and G2 circuits are heavily regulated and may be current-limited.

Use too little G1 bias and the tube operates in Class B or even Class C mode, resulting in incorrect color levels, image distortion, and uneven grayscale response at low input voltages. Use too much bias and anode current soars, possibly crushing high input signals as the tube goes into current limiting. (Too much power dissipation on G2 may cause it to blow out!)

Grayscale tracking (the color of gray at different input signal levels) is also determined by the linearity of the individual picture tubes or projection CRTs, and will require additional fiddling with bias adjustments. For these reasons, CRT displays need to be calibrated to produce the widest possible grayscale, never the brightest image. The new 8-step ANSI test patterns and multi-step grayscales (available from Sonera Technologies and Extron Electronics) are very useful tools for performing this calibration.

It's not hard to see why a matched set of CRTs would be a worthwhile investment in a critical projection application. Hopefully the manufacturer of your particular projector has done his homework and matched the tubes beforehand, so that any small differences in performance can be tweaked with bias adjustments.

BITSTREAM BIAS

Digitally-modulated displays such as LCDs and DMDs are another story. These technologies use rapid changes between "on" and "off" states and the time intervals in-between those states to create grayscale images. In LCD panels, small transistors work as switches to turn on and off voltages to individual pixels. This creates a charge/discharge cycle that works fast enough to display video and many fast computer refresh rates.

DMDs are controlled by pulse-width modulation (PWM) circuits that accomplish the same thing, except there's no analog voltages involved. Mirrors are either "on" or "off", and the ratio of "on" cycles to "off" cycles in a given time interval determines the grayscale value created. This true on/off behavior is why digital light processing is referred to as the only "true" digital imaging technology.

Plasma displays are closer in behavior to LCDs. A voltage is applied to a cell, ionizing the gas within to a plasma-like state and creating ultraviolet light. This UV light excites the red, green, or blue phosphor inside the pixel, and the duration of the charge/discharge cycle determines the intensity or grayscale value of that pixel.

In all these cases, a sine wave is approximated by the rate and interval of on/off cycles. The higher the sampling rate, the more steps along that sine wave we can reproduce. The controlling electronics must digitally process the video/computer signal information by sampling it as three individual streams of 8-bit, 256-step monochrome information.

In a practical sense, 256 levels of gray should be sufficient to convey realism, although some manufacturers are tinkering with 10-bit sampling. But we're still dealing with square waves and must filter out or tolerate their associated artifacts, such as "aliased" curves and colorimetry errors. Video scalars are a good example of a digital product that interpolates a "linear" signal from an analog source - the best scalars use lots of filtering to correct for aliasing and other color/picture artifacts.

For LCD, DMD and Plasma technologies, linear display response continues to be a big obstacle. CRT-based displays still produce richer-looking blacks and wider grayscales. Part of the problem lies with the "off" states of LCDs, plasma and DMD displays - if they reflect or transmit any light at all when inactive, that semi-"on" state arbitrarily defines the lower limit of the display's dynamic range.

This effect is similar to turning the brightness up too high on a CRT display, resulting in a dark gray "no-signal" image, instead of a rich black. In plasma displays, what often happens is that "black" areas in video images become solarized with oddball colors. That's because the RGB signal processing in a plasma display doesn't understand that there's something below NTSC black (4 IRE), and its signal processing becomes non-linear, resulting in a color artifact.

A properly-adjusted CRT projector can reproduce a wider range of gray steps than we can even perceive, resulting in contrast ratios far beyond 255:1 (0 = black). The best LCD displays right now are hitting 150:1 average with peaks to 300:1. DMDs are still a little behind at about 100:1, while plasma displays vary from a low 75:1 to over 130:1 average.

Improvements in pixel transmissivity (aperture ratio) and mountings for pixel-based imaging systems will surely improve performance at low levels of gray and even black, although it's doubtful that LCD and DMD black levels will approach those seen on CRT-engined displays anytime soon.

The notable exception is liquid-crystal light valve technology (LCLV), which uses a CRT to "write" the scan lines on a single-pixel liquid crystal. Think of the 1.7" CRT writing the images to a LCLV as an extremely linear, low-power tube preamplifer (remember 12AX7s?).

The liquid crystal light shutter then becomes a high-power Class A or Class AB amplifier, providing a wide range of brightness and contrast without compressing the dynamic range of the display. Contrast ratios of 300:1 or more are routinely attained in this system, offered by both Hughes-JVC (Image Light Amplifier) and AmPro.

© Peter H. Putman & Intertec/Primedia
This article appears in the November 1998 issue of Sound & Video Contractor.


						ARTICLE
					LINE FOR LINE, PIXEL FOR PIXEL Linearity in Projectors and Monitors


					by Peter H. Putman, CTS When the topic of large screen displays comes up (as it often does these days), we hear plenty of acronyms tossed around (LCD, DLP, CRT, ILA, etc.), as well as terms like grayscale, uniformity, brightness, "native resolution", and so forth. We may even hear the words "minus 3 db bandwidth"" or "modulation transfer function" discussed, if the participants are really into the subject. What we're not as likely to hear is a discussion about linearity, or the ability of an electronic display to faithfully reproduce all the parameters of the input signal driving that display. Yet linearity is the most important property of any electronic amplification circuit, whether it is a high-performance stereo amplifier or a three-gun CRT video projector. We live in a linear world, and our senses respond in a linear fashion to aural and visual stimuli from all around us. How we respond to those stimuli depends on their linear nature: A horn blast that gets progressively louder tells us a train is approaching. A blinking light in the sky that becomes steadily dimmer indicates an airplane moving away. We can also judge relative intensities of stimuli over a wide range, thanks to that marvelous computer inside our noggins. This dynamic range lets us process a great deal of information about the sounds we hear and the sights we see. For any large-screen display to approximate visual realism, it must also exhibit a wide dynamic range. That's a tough job for many display technologies as they try to define images with low or no voltage levels (black), high voltage levels (white), and everything in between (shades of gray or grayscales). Let's take a second to define what makes up the dynamic range of a display. PUMPING IRON In any linear amplifier, there is a definable measure of performance - gain. Gain is simply the difference between the input signal level and the output signal level, measured in decibels (dB) or in volts (V). Gain is not particularly difficult to measure - instruments exist to take a wide range of gain readings at audio frequencies through very high radio frequencies. Simply connect the source signal, the measuring device, and an appropriate load or termination, and you're off to the races. But gain alone is not a useful figure without also measuring the dynamic range of the signal amplifier. What good would an audio amplifier be if it only worked at certain frequencies, or was unable to deliver the rated power without creating distortion? For that reason, dynamic range is usually defined as the range of input signals over which the rated gain figure is achieved, up to the point which the amplifier goes into compression. That occurs when the output drops by 1 dB, also known as the 1 dB compression point. When compression occurs, the amplifier is no longer behaving in a linear fashion, and may in fact be modifying the input signal, resulting in distortion. These distorted signals are no longer faithful reproductions of input sine waves, but square waves. Square waves also create harmonics, which can mix with the input signal to create signals that never existed such as distortion products and picture artifacts. EVERYTHING'S GETTING FUZZY Dynamic range in audio and RF amplifiers is not difficult to calculate. Video displays are a different story, though! We must depend on test patterns as well as test equipment to tell us exactly what is happening in a given projector and monitor. CRT-based displays (monitors and projectors) are analog displays, and it's not difficult to tell when they're operating in a non-linear fashion. A cathode-ray tube is made up of three or four elements - the cathode, which emits electrons, one or two grids, which behave like faucets to vary the flow of electrons, and the anode (plate), which is what the shaped electron beam strikes to form an image. The dynamic range of an individual cathode-ray tube can be quite high, particularly when a second grid is used (known as G2; the control grid is G1). If the tube is biased for Class A or Class AB operation, the ratio of the output signal level to input signal level (gain) can easily exceed 100:1. Still, there will be a point at which the CRT's gain will flatten out and it will go into compression: Increases in grid current won't result in corresponding increases in anode current. There exists a real chance that the anode current will run too high, drastically shortening the life of the tube. For this reason, many manufacturers incorporate a cathode current limiting circuit such as a ballast resistor into the projector or monitor. Current limiting ensures the tube won't be pushed outside its normal operating parameters, but in effect slaps an arbitrary upper limit on the tube's gain and its dynamic range. EVERYONE'S BIASED The control grid (G1) bias and screen (G2) bias determine a CRT's gain and linearity. Both of these controls can be adjusted to really "crank out the lumens", again at the expense of tube life. For this reason, G1 and G2 circuits are heavily regulated and may be current-limited. Use too little G1 bias and the tube operates in Class B or even Class C mode, resulting in incorrect color levels, image distortion, and uneven grayscale response at low input voltages. Use too much bias and anode current soars, possibly crushing high input signals as the tube goes into current limiting. (Too much power dissipation on G2 may cause it to blow out!) Grayscale tracking (the color of gray at different input signal levels) is also determined by the linearity of the individual picture tubes or projection CRTs, and will require additional fiddling with bias adjustments. For these reasons, CRT displays need to be calibrated to produce the widest possible grayscale, never the brightest image. The new 8-step ANSI test patterns and multi-step grayscales (available from Sonera Technologies and Extron Electronics) are very useful tools for performing this calibration. It's not hard to see why a matched set of CRTs would be a worthwhile investment in a critical projection application. Hopefully the manufacturer of your particular projector has done his homework and matched the tubes beforehand, so that any small differences in performance can be tweaked with bias adjustments. BITSTREAM BIAS Digitally-modulated displays such as LCDs and DMDs are another story. These technologies use rapid changes between "on" and "off" states and the time intervals in-between those states to create grayscale images. In LCD panels, small transistors work as switches to turn on and off voltages to individual pixels. This creates a charge/discharge cycle that works fast enough to display video and many fast computer refresh rates. DMDs are controlled by pulse-width modulation (PWM) circuits that accomplish the same thing, except there's no analog voltages involved. Mirrors are either "on" or "off", and the ratio of "on" cycles to "off" cycles in a given time interval determines the grayscale value created. This true on/off behavior is why digital light processing is referred to as the only "true" digital imaging technology. Plasma displays are closer in behavior to LCDs. A voltage is applied to a cell, ionizing the gas within to a plasma-like state and creating ultraviolet light. This UV light excites the red, green, or blue phosphor inside the pixel, and the duration of the charge/discharge cycle determines the intensity or grayscale value of that pixel. In all these cases, a sine wave is approximated by the rate and interval of on/off cycles. The higher the sampling rate, the more steps along that sine wave we can reproduce. The controlling electronics must digitally process the video/computer signal information by sampling it as three individual streams of 8-bit, 256-step monochrome information. In a practical sense, 256 levels of gray should be sufficient to convey realism, although some manufacturers are tinkering with 10-bit sampling. But we're still dealing with square waves and must filter out or tolerate their associated artifacts, such as "aliased" curves and colorimetry errors. Video scalars are a good example of a digital product that interpolates a "linear" signal from an analog source - the best scalars use lots of filtering to correct for aliasing and other color/picture artifacts. For LCD, DMD and Plasma technologies, linear display response continues to be a big obstacle. CRT-based displays still produce richer-looking blacks and wider grayscales. Part of the problem lies with the "off" states of LCDs, plasma and DMD displays - if they reflect or transmit any light at all when inactive, that semi-"on" state arbitrarily defines the lower limit of the display's dynamic range. This effect is similar to turning the brightness up too high on a CRT display, resulting in a dark gray "no-signal" image, instead of a rich black. In plasma displays, what often happens is that "black" areas in video images become solarized with oddball colors. That's because the RGB signal processing in a plasma display doesn't understand that there's something below NTSC black (4 IRE), and its signal processing becomes non-linear, resulting in a color artifact. A properly-adjusted CRT projector can reproduce a wider range of gray steps than we can even perceive, resulting in contrast ratios far beyond 255:1 (0 = black). The best LCD displays right now are hitting 150:1 average with peaks to 300:1. DMDs are still a little behind at about 100:1, while plasma displays vary from a low 75:1 to over 130:1 average. Improvements in pixel transmissivity (aperture ratio) and mountings for pixel-based imaging systems will surely improve performance at low levels of gray and even black, although it's doubtful that LCD and DMD black levels will approach those seen on CRT-engined displays anytime soon. The notable exception is liquid-crystal light valve technology (LCLV), which uses a CRT to "write" the scan lines on a single-pixel liquid crystal. Think of the 1.7" CRT writing the images to a LCLV as an extremely linear, low-power tube preamplifer (remember 12AX7s?). The liquid crystal light shutter then becomes a high-power Class A or Class AB amplifier, providing a wide range of brightness and contrast without compressing the dynamic range of the display. Contrast ratios of 300:1 or more are routinely attained in this system, offered by both Hughes-JVC (Image Light Amplifier) and AmPro. © Peter H. Putman & Intertec/Primedia This article appears in the November 1998 issue of Sound & Video Contractor.