White Level - Can't see moving bars

Greetings

Look to the DVD player and if it has image settings that could be set wrong.

regards

Something is clipping. As Michael suggests, it is likely the source. What is the source?

If this is regular Avia, this is also a more severe problem, because BOTH of the bars in Avia are below reference white and should ALWAYS be seen.

Avia II adds an additional white bar above reference white. Generally, I would say that this also should remain visible, but some people for reasons of maximizing contrast on digital displays may choose to clip peak whites partially or totally.

Greetings

Maximize contrast at the cost of losing detail and adding discoloration into the whites too. That's what John Q Public likes to see.

Now tell me why my image looks soft again?

Regards

Greetings

Just make sure you didn't pause the video at white level. i did the same mistake at the beginning.

regards,
seelan

If your DVD player has it's own contrast adjustment, you may need to alter the setting. One of my Denon DVD players was clipping white at the default player contrast setting.

Most modern digital displays can't have the white level set with ANY test pattern because they show every shade of white even beyond 100% white which is represented by a digital 235 for all 3 colors (red, green, and blue).

LCDs don't clip no matter how bright the backlight or Contrast control are set. So you won't get a color shift (but the color could have errors all the time.

That said, it is POSSIBLE your backlight is flooding the LCD panel with so much like, it is washing out top-end detail. So you should start by turning the Backlight control down as far as possible. Then see if the white bars are visible.

I haven't seen one of these Samsung LCDs yet, but there is a creeping problem with newer and newer LCD panels... they are so freakin' bright and the control for backlight is so limited, that you can't set the white level low enough for comfortable viewing in a dark room... which is really the whole point of home theater.

Now, back to white level... just about all the "white level" adjustment patterns on test/setup discs are geared to CRT displays and the instructions for using the patterns don't apply to digital displays. Even though consumer digital video is limited to digital values between 16 and 235, digital video works on a 0-255 range and there are few digital displays that do anything to eliminate visible detail in whites that are encoded at 236-255. So almost all digital displays will display all or most whiter-than-white values. And that means 99% white, 98% white, 97% white, etc. are also displayed and are visible.

It can be difficult to see these whiter-than-white and near-white steps when the display is in torch mode (maximum backlight, high Contrast) because there's just so much light your iris doesn't contract enough to be able to distinguish the white detail that is on the screen.

And as far as "setting white level" with a test/setup disc on modern digital displays... there's very little you can do by eye if the display is revealing all the steps OK. This is making the Contrast control useless for anything except setting the peak white level which requires a meter. You display a 100% white window pattern (though on an LCD it doesn't matter if you use a 100% white field), set backlight to minimum (usually, though occasionally there might be a display where the minimum setting is too low, not very common though) and use the Contrast control to get 30 fL or thereabouts. This is a comfortable luminance (brightness) level for viewing the display in a dark room. If you need more brightness for a room with lights on or daylight, you can crank up the backlight which is usually all that's needed for bright-room viewing... but a contrast tweak may help on some brands/models also.

Lacking a meter, it's hard to know how close you are to 30 fL for the peak white level, but I find if you adjust Contrast so that a 100% white window pattern doesn't make you squint and so a 100% white field (full screen) is bright but doesn't feel overpowering, you are probably fairly close to where you need to be (these evaluations need to be done in a dark room).

I hope this info helps you get some good results. Let us know.

Huh? Sure they do.

Greetings

I can show you some Westie LCD flat panel sets in the THX class that clip the heck out of white when the contrast is set higher than 60%.

To get contrast set right, the patterns should have white all the way to 255 ish ... not just 235. There is additional image detail there that just enhances the image. The difference between a man in a white T shirt and a white shirt with all the subtle gradations.

Regards

S-Video? DUDE are you stuck in the 1980s????

So now you're on component - welcome to the 1990s! LOL!

I think it's time to join the new millenium! HDMI cables are under $6 each at www.monoprice.com --- of course if you don't have any source components with HDMI outputs you'll need to take care of that first

That effect may LOOK like clipping, but it's not clipping. LCDs don't emit light, they filter light to produce different colors. The backlight provides all the light. Once an LCD pixel is completely transparent, the pixels for the other 2 colors would have to get more transparent than 100% transparent so they pass more light than the "weak" color, but that doesn't happen to LCDs. The luminance histogram with red, green and blue graphed separately shows that LCDs don't clip.

If the manufacturer lets the Contrast control crush any color or colors, that's a crush problem, not a clipping problem. That means for your Westinghouse example, let's say that a 60 Contrast setting produces a pixels that are 100% transparent when you send a digital 255, that means if you send digital 254, each pixel is a little less transparent than it was when receiving a digital 255. Now, say you increase the Contrast control to 61... now both the 254 level and the 255 level produce the same level of transparency (100% transparent). Then move to a Contrast setting of 62 and now you have digital values of 255, 254, and 253 all being 100% transparent. That's crushing whites not clipping because there are no colors getting brighter (more transparent) at settings of 61, 62 or higher, you're actually making lower digital values reach peak white (100% transparent) before they should.

I forgot the Westinghouse panels did that - so you could very well get some use out of a white PLUGE for setting one of those (or others that do something similar).

I've read and heard experts claim digital values of 236-255 are in consumer video all the time, but other experts say it never happens these days. It is certainly no big trick to limit digital values to 16-235 anymore... there didn't used to be tools back in the 1980s (for digital audio), but these days it's a simple thing. But you have to analyze the entire movie (if that's what you are digitizing) first to find the darkest and lightest points and assign those to 16 and 235 so everything else you digitize falls in between. You can crush whites just fine without it being a loss of 236-255 data. Just make all pixels equally transparent for all values from 225-235 and you lose the detail in white shirts without higher bits ever being involved. Personally, until I see actual measurements of digital values present in consumer digital video sources, I have to say I don't know if 236-255 are ever present in DVDs or high-def discs on any kind of routine basis.

I learned my lesson, though, I should never use never when talking about video displays, there's always going to be something out there that does something unnecessary/unexpected.

Greetings

Clip crush ... that's all semantics. Why not crush clip ...? Both cases you lose detail. One side because you are supposed to ... the other side ... not really.

You have too much faith in TV makers, Doug. They still do plenty wrong ... often on purpose.

MAjor brand players like the Samsungs and the Panasonics and the Sonys and the Pioneers and the Sharps are now regularly shipping products (I think LG and Toshiba are here too) that have a hard time clipping/crushing or discoloring. That's a good thing.

The second tier players are not quite there yet ...

The knowledge trickles down slowly ... but given where we have come from in the last 10 years ... it's rather amazing.

I would not have thunk this in 1998 when I bought my first HDTV RP set.

Regards

Doug, Mr.D who works in video production says:

http://www.avsforum.com/avs-vb/showt...0#post13406530


http://www.avsforum.com/avs-vb/showt...0#post13776980



After taking all of this into consideration, I recalibrated my display protecting my grayscale and gamma all the way out to 255 and it's seemed to make a subtle but noticable improvement (XBR2 SXRD). Of course I've sacrificed quite a bit of contrast ratio to do so.

Greetings

Either you want it set up right ... or you don't.

Also remember that the human eye actually has less CR than the TV in most cases. High contrast ratio numbers that give you a headache to watch ... do not result in good images.

If you can live with discolored whites and soft images on the bright end lacking in detail ... by all means pump the contrast back up there. The scary thing is ... where do you stop then ...? If killing 5%-10% of the top end detail is okay to you ... why stop there? Go for 20% ... and get even bigger CR numbers ...

The train destined to image accuracy would have long since left the station.

Regards

Do any of the popular calibration discs have "whiter than white" window patterns? I know that Avia II does have window patterns above 100 (IRE in the Avia world) - I think 102.5, 105, and 107.5. It seems like it would be useful to include those and for software to allow folks to optionally do a 12 or 23 (0 - 100, 105, 109 percent) grayscale measure in order to check for RGB tracking above 235.

cheers,


--tom

Greetings

The AVS HD DVD and Blu test discs have all the white you could ask for.

DVE on the HD side gives you one block of WTW. Not enough really to do it right.

The Monster ISF disc gives you the man in the White shirt pattern ... and that is a full white test pattern.

Other good white contrast patterns to look for are white gradation patterns. The ones that smoothly start from black on one end and end up at white on the other end. If the white is at 255 ... then turning up the contrast higher will cause ... (you just have to lookat it yourself. The effect is interesting.)

Stacey Spears created this parting RED SEA test pattern as I call it for his HD test disc. It is really nice and easy to use ... but unfortunately, it is not for sale.

Regards

The AVS HD disc has a 105% and a 109%(254) white window pattern that you can use to check your grayscale above 100%. I mainly just make sure that red hadn't been clipped at 109% since my SXRD runs out of red at higher contrast levels.

Thanks Michael and sperron - unfortunately, no Blu-Ray player yet. So, I've only been working with GetGray and Avia/Avia II. Interestingly on my Samsung LCD, I've got contrast at 98/100 with the backlight at 1. This actually improved RGB tracking at the lower end where there clearly is a blue rise - the higher the contrast setting the more that is pushed a bit right to left thus improving things at 20%-30% without causing more color error at the high end (according to the i1D) or dropping gamma. Getgray includes above 235 on its ramp patterns as well as a max full field white which are what I've used to visually look for any visible color shifting above 235. With my eyes though, grayscale errors are more noticeable in the low end - above a certain luminance level, the error has to be fairly large for me to perceive it.

cheers,


--tom

I agree, this would be a good question to address to the HCFR guys. 236-254 is a visible part of the picture and it should be accounted for.

Edit: And I did address it in the HCFR 2.0 thread.

DVE Professional has window patterns ranging from interface level 001 to 254.

Below reference black: 001, 005, 010, 012, 014.
Above reference white: 240, 250, 254

Of course, it's pretty expensive only to get these patterns.

Well... yes and no... it depends on how they get the high contrast ratio. If it's a display that doesn't have black levels as dark as Pioneer Kuro displays and the high contrast ratio is achieved by extremely high peak white levels... that's torture. But
But I know what you are saying... today's typical black levels (around .02fL) with a contrast ratio higher than 1500:1 or so are headache-inducing. One of today's typical displays with .02 fL black levels would have to output 600 fL to achieve the 30,000:1 contrast ratios being advertised - it's really silly. 600 fL! That's HUGE. It's much easier to get high contrast ratios by making blacks blacker - but it's harder for today's display technologies to get blacker.

I'm looking forward to infinity contrast ratios - as long as you have full control over the peak white level. If the display produces true black at 0 fL, the contrast ratio will be infinity for every/any peak white level... even if the peak white level is1 fL the contrast ratio would still be infinity when the black level is really 0 fL. Once everybody can achieve 0 fL blacks will finally be able to do away with contrast ratio all together. At that point, the only measure needed will be peak white level.

It's the engineer in me - using the right description/formula/data can mean the difference between success and disaster. A damp road and black ice can look the same, but if you treat both the same way, you're headed for big trouble!

You make no sense with this. It absolutely is clipping. When you run you out of output and can reach that through the adjustment range, that's clipping. Those values are eliminated. It can either be a hard clip across all three colors, or a more subtle clip that affects only one or two at a time, causing first a colorshift, then the clipping/obliteration of detail.

Your distinction about LCDs being transmissive is not at all relevant to whether the device is clipping. The values are clipped off, that's just as straightfoward as can be. They're gone. They're not crushed or compressed, they're clipped.


No, crushing is where the delta between levels decreases and obscures their visibility. What we're talking about here is clipping, not crushing. These values are completely excoriated from the image. And it's exactly what it looks like. You crank white level through the roof and at a certain point the detail at the brightest portion of the luma range vanishes because it is being clipped off. And it isn't any kind of soft-clipping either. It looks terrible.


No, that's called clipping. All the values above a certain point are simply clipped off the video content altogether. They all get clipped back to the lowest value reproducible, in effect.

I'm not sure where this impression comes from, but this is most certainly not the case. Standard displays don't even come close to matching the capabilities of the human eye.

I know this is something that has been taught for some time by the ISF, and that's quite unfortunate, but it is not at all true. Darrin Perrigo has spent a great deal of time combating this misunderstanding and confusion, which it appears arises from a misapplication and misunderstanding of CSF limitations.

Yes. Avia II has many patterns that go both below black and above white. Among the most useful of these is the "2.5IRE" black window pattern, and the xxdeep ramps pattern (also extremely useful to observe banding and clipping on a digital). It also has a steps pattern from 0-255, and 1-254 which may be of interest. Many of the other patterns also have moving black and white bars which include a bar that is below black, and above white, respectively. And also there's small black and white crossing ramps/checkers there too which are helpful.

DVE has the crossed ramps pattern which goes above and below, the dots mark reference white and black.

Get grays ramps go all the way too.

Your description of Clipping in the 3rd sentence is not Clipping, it's just the maximum luminance of the display in question. When you increase Contrast, at some point the display won't get any brighter - that's normal/typical. If it happens for all 3 colors at the same time, it is not clipping and may or may not cause crush depending on whether the display is still responding in a linear fashion. Clipping is an luminance limiting issue affecting one (or sometimes 2) colors before the other color(s) stopped increasing output level.

For example, you could start with a low Contrast control setting and all 3 colors would reach the same peak luminance level. Increase the Contrast control up to a certain point, and usually 1 color (occasionally 2 colors) would no longer be able to achieve the same luminance level. But this happens only in plasma and CRT displays, not LCD or DLP. It's like having a 3-channel audio amplifier with 2 channels having 100 watts, and 1 channel having 50 watts... the 50 watt channel will clip before the other channels.

In an LCD, 100% white is always the level of the backlight minus the transmission losses through the LCD panels. All the pixels can do is reach 100% transparent. If the pixels are 100% transparent when they SHOULD be 95% transparent, that's not clipping, that's just bad design and it will cause crush in the highlights. It causes a complete loss of detail without a color shift.

Crushed blacks happen when multiple digital values produce the same or very similar luminance level. For example, crushed blacks happen if digital values 20, 21, 22, 23, 24, 25 & 26 all produce the same dark gray shade instead of slightly different luminance levels. This is a simplified example, 27, 28, 29, 30, 31, 32 might all produce a luminance level 1% higher than the previous group of steps, etc. Crushed whites happen if 220-235 all show the same luminance level. Crushed whites happen without a color shift. Crushed blacks happen without a color shift.

Clipping is completely different and does not produce the same luminance level for all 3 colors in PDPs and CRTs as it's just about impossible for the red phosphors, green phosphors, and blue phosphors to all peter out at the same luminance level. Calibrators have to set a peak level that keeps all 3 colors within a linear operating range to avoid clipping.

Example - let's say a PDP has the red gamma curve flatlined at 95% so red never achieves the same luminance level as blue and green. ANd blue and green continue producing the light they need to produce to reach 100%. The 90% white step (assuming a reasonably well-calibrated display) will be neutral because all 3 colors are able to achieve the same luminance level. But at 100%, red luminace will trail green and blue so 100% white will have a cyan color shift. (this assumes the red problem is not fixable with the Gain control of course)

An LCD (like the Westinghouse in the example) that has the Contrast control set well beyond 60 causing all pixels in all colors to be 100% transparent for digital values from 220-255 to detail will be lost, but there will be no color shift because there is no clipping. For clipping to exist in an LCD, one panel would have to be less transparent than the other 2 panels so that it could not achieve the same luminace level for any given backlight level. Let's say the Red panel behaved just like the green and blue panels up to 95%, but then the red panel never got any more transparent, but the blue and green panels would continue to get more transparent all the way to 100% transparent. That would be clipping in an LCD, but that never happens in the real world. As has been pointed out by Michael, most LCDs DO NOT clip - if you run into it, more than likely it will be in lower-tier products.


You are making my own arguement with this statemet. This is what happens for highlights and shadows when there is crush.


Not true. Clipping simply means 1 color (maybe 2 colors in rare cases) can't produce a luminance level equal to the other colors. You will still have the steps in 2 colors, but one color stops getting brighter at some point. You end up with no steps in that 1 color and a color shift because the other 2 colors continue to get brighter.



Only for 1 color at a time (2 colors in rare cases), the colors that continue to get brighter are NOT limited and still produce gradation steps as digital values increase.


Displays that clip won't all cut off at the same point (it would be a miracle if they did, at least) - one color will give up first, another color will give up next, and the "strongest" color will give up last. Clipping causes color shifts because it doesn't happen equally in every channel at the same time.

Crush is a condition where all the channels stay together, but they are "wrong" in that there's not enough luminance difference between digital levels and steps will blend together without a luminance shifts for each digital value.

Crush can happen in shadows or highlights.

Clipping can only happen at the bright end of the spectrum when the pixels produce the light and one color doesn't achieve the same peak luminance as the other colors (and can't be brought back in line with the Gain control).

Transmissive pixels can't clip - every 100% transparent pixel will pass the same amount of light in a transmissive display like LCD.

I can't help if these terms have been used interchangeably over the years - but they are not the same thing. It's like wheel and rim - people use them interchangeably all the time but they are NOT the same thing from an engineering/technical point of view.

I would have to think that SXRD is a transmissive technology similar to LCD. When turning up contrast it is possible for one primary in the grayscale to stop increasing while the others continue to go up. The behavior causes a color-shift in the near whites similar to your multiple descriptions of clipping. As far as I can tell you seem to be stating that the behavior I described is not technically defined clipping due to the process by which different displays produce grays. From a practical standpoint for an end-user calibration the distinction seems rather trivial to me, because regardless of display technology the end-result can still be an unintended color-shift in the grayscale when turning up contrast as one primary can no longer match the others.