Sunday, June 19, 2005

Tweaking Gamma on My Samsung DLP, Part I

In HDTV Quirks, Part II, Gamma I tried to explain why, hidden in their service menus usually, today's "digital" HDTVs offer a number of gamma choices.

Now I'll go into my experiences fiddling with service-menu GAMMA on my Samsung 61" DLP-based rear-projection HDTV monitor.

My Samsung DLP-based RPTV apparently features some six GAMMA choices, numbered 0-5. The factory setting is GAMMA 4. Right now, I'm experimenting with GAMMA 5 on the DVI input from my hi-def cable box-cum-DVR.

When I'm referring to the Samsung's internal GAMMA numbers, I'll put GAMMA in all-caps to distinguish it from "true gamma" and its values.

These 0-5 GAMMA numbers have nothing to do with mathematical, "true gamma" exponents per se. They're just arbitrary numbers, indexes into locations in the ROM firmware of the TV, I suppose. Nor do higher GAMMA numbers necessarily mean higher, more contrasty gamma. For example, GAMMA 3 obviously yields much lower "true gamma" than GAMMA 2. To my eye, GAMMA 3 probably implements a "true gamma" equal to 1.0 — for when gamma is being set by external gear, I assume.

Nor is it clear that the other GAMMA numbers implement simple
(output luminance) = (input voltage)gamma

functions. When you think about it, digital signal processing can be done by means of lookup tables, rather than math functions with exponents. Take a digital value of input voltage for a single pixel, look it up in a table, and find the output luminance value. If you plot all the generated output values against all the possible input values on log-log axes, maybe you get not a line but a snake.

If the user changes the brightness or contrast setting, that may completely change the "shape" of the lookup table's snake — i.e., the snake may writhe for you.

In my very limited experimentation with Samsung's GAMMA choices so far, I find the different GAMMA numbers (excepting GAMMA 3) give subtly different looks to the picture. The differences are so subtle that I can't really decide if, say, GAMMA 1 is really any different from GAMMA 0 — that type of thing.

The situation is aggravated by the fact that, each time I change GAMMA, I can't always remember exactly what the previous picture looked like. It's too bad there isn't a split-screen approach to comparing two different GAMMA settings.

For that matter, it's too bad GAMMA can't be adjusted by means of user-accessible menus, just like contrast and brightness.

Still and all, I seem to be able to hit pause on the DVR, so that I have a constant image to judge by, and when I toggle among the GAMMA numbers I think I can see subjectively meaningful differences in things like how brightly and with how much color a face appears in an overall-dark scene.

That is, in the lower IRE ranges where I'd expect to see the biggest differences, I do see such differences. With the factory GAMMA 4 setting, such images seem to look "flat." Switching to GAMMA 5 increases the apparent denisty or solidity of a darker, low-IRE image, making it look less flat.

Or so it seems in my early experimentation. More later.


After experimenting with GAMMA 5, I decided it wasn't exactly right. It did make mostly dark, low-IRE scenes look less flat than GAMMA 4, but in scenes with a mixture of bright areas and dark, the dark portions still didn't have enough "shadow detail."

So I switched to GAMMA 2. GAMMA 2 appears to implement a lower "true gamma" value than either GAMMA 4 or GAMMA 5, meaning small changes in input signal levels in fairly dark areas of the scene produce more sizable changes in output luminance. This brings details out of the shadows more easily, à la Goldilocks #2, the image on the right above.

The downside is that the image has less "snap" or "punch" overall. What I think is going on here is that snap or punch sells TVs. It wows friends who drop in and glance casually at your TV screen. But for extended viewing, you're better off sacrificing snap for shadow detail. Lowering gamma does that, opening up the shadows in a satisfying way.

I also note that changes to true gamma (via switched GAMMA values in the service menu) seem to mandate altering user-menu brightness and color settings. When I was using GAMMA 5, I felt I had to boost brightness from what had been "correct" with GAMMA 4. But when I switched to GAMMA 2, the "correct" brightness went right back down to where it had been before, and it was color that had to be boosted.

These details of interactions between service-menu GAMMA and the two user-menu settings are probably specific to my particular model of TV. But, as a general rule, I'd say you can expect there to be such interactions.

My Samsung does not make it easy to deal with said interactions, by the way. When you are in the service menu, you don't have access to the user settings — and vice versa. When you go into the service menu, furthermore, the user-menu settings automatically switch over to the TV's so-called Dynamic mode, bypassing whatever Custom settings you may have in force.

Interestingly, it turns out that the default Dynamic mode seems to furnish exactly the right user settings for GAMMA 2, whereas Dynamic mode, with its boosted color level, isn't quite the thing for GAMMA 4 or GAMMA 5.

That makes me think Samsung may have optimized the TV for GAMMA 2/Dynamic mode, but changed to GAMMA 4 for in-store display and the wowing of friends.


Another update: I grew dissatisfied with GAMMA 2, for reasons that are hard to put in words, and so I tried GAMMA 0. GAMMA 0 seems to offer a lower "true gamma" value than any of the other GAMMA settings on the Samsung DLP (with the exception of GAMMA 3, which I think has a "true gamma" of 1.0).

Using GAMMA 0, I watched the standard-definition cablecast of Alfred Hitchcock's classic, Vertigo, on TCM last night. This is one of my favorite films; I've watched it countless times. I think I know what it's "supposed to" look like — and that's exactly what it did look like with GAMMA 0. The dark scenes, such as the opening rooftop chase, were dark but accessible to the eye, and the bright, almost washed-out look of the scenes in which James Stewart first begins tailing Kim Novak had, yes, a bright, almost washed-out look. The colors of the flowers in the flower shop where Novak goes to buy her "Carlotta" nosegay were rich ... but not too rich.

Note: I had to boost the user brightness setting from 50 in the default Dynamic mode to 60 to get things looking just right. Again, changes to gamma tend to necessitate changes to brightness, or color, or both.


Yet another update: yesterday I went to the movies for the first time in a while, and what I saw on the screen appeared to have much higher gamma than I've been aiming for at home. (I saw Judi Dench and Maggie Smith in Ladies in Lavender, a bit too much on the feminine, emotionally nuanced side for testosterone-poisoned me ... but still a pretty good movie.)

The image on the screen had a lot of dynamic range. The bright stuff was very bright, so much so that a cut to a super-bright scene after an ultra-dark one hurt my eyes, and the dark scenes were really, really dark. In scenes setting backlit darker elements against directly lit bright ones — for example, faces incast shadows against a bright sky — I had to interrogate the barely lit elements with my eyes to pick up their subtle gradations of shadow detail.

If the cinematic "system gamma" or "end-to-end exponent" (the "inverse power function" of the camera negative times the intrinsic gamma value of the print) had not been so high, this would have been unnecessary. Shadow detail would have popped right out at me.

I sat there thinking my Samsung's contrast ratio probably is incapable of giving me that broad a dynamic range between 100-IRE or "peak" video white and the TV's 0-IRE, minimum-light-output "black level." Given that fact of life, still and all I was strongly reminded of the gamma curve I had witnessed when using the service menu's GAMMA 5 setting.

I also noted that many of the movie's skin tones were typically much deeper — more highly saturated, redder — than I had been targeting at home. When Judi Dench, that excellent actress, went all emotional, her face got as red as a red-hot poker. GAMMA 5 provides exactly such rich skin tones.

So I've switched back to GAMMA 5 ... which makes the TV's dynamic range no different, but compresses shadows and intensifies colors. It gives Hitchcock's Vertigo a different, but not unpleasant, look, not unlike that of Ladies in Lavender in the theater.


Let me note yet again that changing "true" gamma (via service-menu GAMMA) typically requires changing the TV's user brightness and color controls afterward. I have a DVD, Avia Guide to Home Theater, which makes that a snap. It's designed to help you set up your video and audio environments to something approximating perfection. At some length, it teaches you everything you need to know about home-theater fundamentals ... but I typically skip all that jazz and go right to the basic test patterns.

These patterns let you adjust contrast, brightness, sharpness, color, and tint. The last two adjustments involve matching the intensity of special patches of color when viewed through a supplied blue filter. On my Samsung, though, the tint control is disabled for hi-def inputs ... no biggie, since the test pattern shows the TV's default tint setting as spot-on, anyhow.

Contrast is set using a pattern that places a moving 98-IRE vertical stripe against a 100-IRE background. Many TVs (but not my Samsung) "crush" such subtle differences in high output brightness when the contrast control is set too high, making it impossible to pick out the stripe. In such cases, you simply back contrast down until the stripe reappears. In the case of my Samsung, however, white detail is never crushed, even with the contrast setting maxed out at 100.

Brightness is set with a moving black stripe at (I think) 2 IRE against a 0-IRE backdrop. You lower the brightness control until the stripe vanishes and then raise it slowly until the stripe barely re-emerges. I find this adjustment gives me roughly the same result as putting a letterboxed 4:3 image on the screen and matching the horizontal black bars above and below the framed image to the vertical black bars at the sides of my 16:9 screen.

Sharpness is set using a standard resolution pattern with groups of alternating white and black lines grouped by increasingly narrow widths. You're supposed to adjust the sharpness control so you can distinguish lines of the finest possible pitch ... but without adding any false "haloes" or "ringing" to the picture. Only problem is, the sharpness control on my Samsung inexplicable seems to have no demonstrable effect whatever!

Color and tint are set using a standard "color bars" pattern, slightly modified to juxtapose extra patches of flickering, complementary colors where they can do the most good. When the color and tint controls are properly set (if the tint adjustment is even available), the flashing patches blend into the background when you look at them through the supplied blue filter.

The main drawback to all this is that it really calibrates the TV only with respect to the DVD player and its associated input to the TV. My Bravo D2 DVD player, by V Inc., has its own brightness, contrast, and color adjustments, adding a measure of confusion to the whole deal.

So what I've done is set up the TV/Bravo D2 interface using Avia, and then play my DVD of Vertigo on the Bravo. Meanwhile, I cue up a scene of the version of Vertigo I recorded on my cable DVR: the shot of the flower shop interion after Scottie opens the door works quite well. I pause the playback on this scene, then switch to the DVD, pausing it (hopefully) at the very same frame. Toggling between the two inputs to the TV, I adjust contrast, brightness, and color on the cable-box input to match that of the DVD input.

Not surprisingly, this gets me a pretty good match. Even if the Vertigo that came into my home via analog cable has a lot more noise and video "trash," it looks just the same as the DVD otherwise. Also unsurprisingly, the actual settings of the relevant controls on the TV are not exactly the same ... close, but different. Different signal sources can be expected to need different settings — which is why I am dubious about people who set up one video source using Avia and then slavishly copy those settings to other sources.

No comments: