| Video Noise August 2009
What Makes a Picture Good (Accurate) or Bad? Part
Two: Gamma
 When I
began this series, I thought two months would be enough to cover this topic. But I decided
to expand the coverage of gamma and color accuracy, making this a three-part series.
Gamma is a number calculated based on the luminance
(brightness) of each grayscale step. Figure 1 contains three sample gamma curves. The
vertical axis is in foot-lamberts (fL, a measurement of screen luminance, or the amount of
light coming from the screen). The 30fL peak white level in this example is about where
you want 100% white to be when you’re viewing a panel display in a dark room. This
setting avoids eyestrain from an overly bright display. 30fL is the peak white level at
the end of the curves in the upper right corner of the graph. The black point (0% white
and 0fL) is in the lower left corner. This graph is referred to as a luminance graph
or luminance curve because luminance (fL) is on the vertical axis and the grayscale
steps are on the horizontal axis. The shape of the luminance curve is an indicator
of what gamma is doing. The perfect shapes in this graph are, of course, idealized. But
the closer you get to these ideals, the better the images will look.
Figure 1
The red line represents a gamma of 1, and 50% white (which
is a shade of gray) is half as bright as 100% white -- that’s 15fL in this example.
The 10% step is 3fL, or 10% of the peak level (30fL). Limitations in early video
technology resulted in video cameras and TVs having complementary gamma curves that look
like the green line (camera gamma) and the blue line (TV gamma). The camera records
everything brighter than you would see in real life. To make these images appear correctly
in your home, your TV has to compensate for the camera gamma by having the opposite curve.
This may not look right to you on the graph, but the blue 2.2 gamma line is the
mathematical opposite of the green camera line. Subtract the blue TV curve from the green
camera curve and you get the red line. Today’s technology would permit a video
standard that was recorded, broadcast, and viewed with a gamma of 1 (red line), but
because video technology initially needed gamma (and because gamma allows efficiencies on
the record and broadcast end of things), gamma remains a part of our video system today.
Gamma does not affect black or white -- you set the black
level with a PLUGE (picture line-up generation equipment) pattern from a test/setup disc
using the brightness control, and you set white using the contrast control either by eye
or with a purchased meter. Everything between black and white is defined by gamma. You can
imagine the damage to an image if something that’s supposed to be 40% white (a middle
gray shade) is displayed as 46% white, especially if there are many errors at different
grayscale steps. In general, the image quality will be best if each step from 10% to 90%
has the same gamma: 2.2 to 2.3 for most displays. There’s a lot of controversy over
the actual value of display gamma, but one thing is certain: a range of 2.2 to 2.5 is
considered valid or acceptable gamma. I’ve run a number of different displays at 2.4
and 2.5 gamma, and everything appeared too dark -- even after I tried acclimating to the
high gamma for two days or more. A 2.5 gamma curve would be darker (lower on the graph)
than the blue 2.2 gamma curve.
Figure 2
Each of the two curves in Figure 2 has an average gamma of
2.2. But it’s clear that the curves are not the same shape. The blue curve in Figure
2 is the same as the 2.2 gamma curve in Figure 1. It shows the uniform shape of the curve
when you have the same measured gamma (2.2 in this case) for every grayscale step from 10%
to 90%. The orange curve shows what happens when the display has poor gamma performance.
You can see from the graph that the orange line is too light when the picture is dark (10%
to 40% white) and too dark when the picture is bright (60% to 90% white). This sort of
response gives images a milky, dull appearance that can’t be removed no matter how
bright you make the picture. Of course, there are degrees of severity. This example would
be fairly noticeable. But this isn’t the only type of gamma problem manufacturers
leave in their displays -- it’s just one of the errors I’ve seen so far. For
example, if the gamma curve was a little darker for the dark steps and a little brighter
for the lighter grayscale steps, most people wouldn’t find that as objectionable as
the opposite performance indicated by the orange curve in Figure 2.
Figure 3
Few product reviews offer all the details about gamma
performance. But if you can find the information, it’s certainly something to
investigate when you’re considering a purchase. The review may plot gamma for each
grayscale step from 10% to 90%, giving you a line graph like Figure 3. A straight
horizontal line at 2.2, 2.25, or 2.3 would be ideal for most current displays. The blue
line in Figure 3 represents that ideal performance. In fact, the blue line in Figure 3
shows what the blue 2.2 gamma curves in Figures 1 and 2 look like when plotted this way.
The orange line in Figure 3 matches the orange luminance curve in Figure 2. It’s
lower than the 2.2 line at the darker steps and higher than 2.2 at the lighter steps. The
gray line represents a TV that has very poorly controlled gamma with a different value of
gamma for many steps. Both the orange and gray lines below have an average gamma of 2.2,
so if a review gives just a single number for gamma you might think the display is fine.
The truth, as you can see, is that the gamma performance may be quite poor. The more
deviation from the straight horizontal line, the more likely it is you’ll be able to
see some sort of problem with the display’s luminance. Note that there is no gamma
value for 0% or 100%. Those points are used to calculate gamma for the other steps, but,
to put it in simple terms (the issue is too complex for this column), they don’t have
a gamma of their own.
It’s my experience that no matter how good the TV is
in other ways, if the gamma is not well controlled and fairly close to 2.2 or 2.3 for
newer displays (circa 2009), the picture will be less satisfying and perhaps even annoying
compared to the same display with good gamma. Poor gamma performance is often more
noticeable than color errors in primary or complementary colors. As of the summer of 2009,
depressingly few video displays have gamma controls and nice uniform gamma measurements
for each grayscale step. But it’s worth seeking out video displays that offer some
control over gamma, especially if you’ll be having your video display calibrated or
you plan to teach yourself calibration.
. . . Doug Blackburn
db@hometheatersound.com |
