Calibrating Monitors

[Bill Edmunds]

I'm using a Sharp HDTV to monitor my HD productions. How do I calibrate it? The normal rules for NTSC SD monitors don't seem to apply... or do they?

[Glenn Chan]

1- If you are sending an analog signal to the monitor, then that might need to be calibrated.

see
http://www.videouniversity.com/tvbars2.htm

*the blue gel trick doesn't work.

Digital signals don't need calibration.

2- Other aspects of the monitor may not be ideal (e.g. white point), but the monitor may or may not have controls that would let you change that.

3- Your real problem is likely that the Sharp HDTV does weird things to the signal.

Scaling, sharpening, poor deinterlacing (e.g. bob), image "enhancements", non-standard primaries, s-shaped transfer curve (inherent to LCDs, should be calibrated away), low bit depth with poor dithering, etc.

[Giroud Francois]

whatever signal you send to whatever monitor you use, calibration is needed.
while it is easy with computer (you find USB calibration tools for cheap), it is more difficult with video. The are some tools (like special DVD) to help this.
http://www.ramelectronics.net/html/V...libration.html

[Bill Ravens]

Glenn...

again you're right in theory but wrong in actual practice.

ahhh, but the display most certainly does. With the plethora of brightness, contrast, color gain, bias, etc., etc., nothing comes out of the box already set up.

And many LCD screens, these days, e.g. Samsung LCD's, come with auto settings and sensors to adjust for room light. Beware of these automated adjustments. You never really know what you're working with.

[Bill Edmunds]

Originally Posted by Bill Ravens

Glenn...

again you're right in theory but wrong in actual practice.

ahhh, but the display most certainly does. With the plethora of brightness, contrast, color gain, bias, etc., etc., nothing comes out of the box already set up.

Exactly. How do I know if I have my brightness, contrast, hue, color set properly?

[Giroud Francois]

that is the problem. you never know.
what you know that all your screens give same result (from camera monitor, to editing preview monitor, to you big LCD screen in living room.).
and prefferably.. this must be close to what you see with your eyes.
At least with such calibration you do not know if you are right, but you know that what you get at one end will stay the same at the other end.
Now if you distribute DVD directly to your customer, it is pretty much all what you can do, since you can not guess the settings your customers have.
if you work with professional equipement, it is easier, since calibration actually means something it term of physical settings. (the vectorscope summarize this almost in one view).
What you need to know, is that consumer product (especially LCD ) tends to display very strong colors, most alway with a cast into one red,green,blue primary color. (Sony screens have really hot/bright red, while others screen goes easily on blue).
My advise is to set colors low, so you will not be disappointed on how nice the blue was displayed on you computer screen and how bad it renders on your DVD.
That is for colors.
From the luminance side, you find two schools. The one in favor of bright, but has very poor dark rendering (most of batman movies are a good scale to measure), the other one is in favor of darkness, but tends to burn easily all bright pictures and gives no real black.
a simple gray scale allow to set up your screen, but that does not help on you customers screens, except if you put as goodies on your DVD a little movie with an explanation on how to set the screen properly.(i have seen that on few DVD)
for example in a very dark picture, you can hide a message that people can read only if their screen is properly set.
http://www.drycreekphoto.com/Learn/C...itor_black.htm

[Michael Jouravlev]

Calibration for cheap:

* Buy Datacolor ColorVision Spyder2express colorimeter (~$70)
* Download HCFR Colorimeter program from here, it is free: http://www.homecinema-fr.com/colorimetre/index_en.php
* Get one of the test DVDs, either burn the image provided by HCFR, or buy a GetGray disk or other TV setup disks.
* Install HCFR Colorimeter onto your laptop
* Read this thread on how to use HCFR Colorimeter program: http://www.avsforum.com/avs-vb/showthread.php?t=737550
* Connect the Spyder to the computer
* Play your test disk and log monitor output with the Spyder colorimeter

You will get several graphs, it is your job to make sense of them and to adjust your TV accordingly. It is likely that you will not have access to all needed controls from a regular TV's menu, you may need to access special service menu.

Below are couple of charts graphed with HCFR, the TV is my 50-inch Panasonic plasma. The results are not perfect but still better than original setup.

Gamma, before: http://www.jspcontrols.net/misc/panasonic/std_gamma.jpg
Gamma, after: http://www.jspcontrols.net/misc/pana...alib_gamma.jpg

Grayscale, before: http://www.jspcontrols.net/misc/pana..._grayscale.jpg
Grayscale, after: http://www.jspcontrols.net/misc/pana..._grayscale.jpg

Color temperature, before: http://www.jspcontrols.net/misc/panasonic/std_temp.jpg
Color temperature, after: http://www.jspcontrols.net/misc/pana...calib_temp.jpg

Primary and secondary colors: http://www.jspcontrols.net/misc/pana...ic_cie_709.jpg

I used "warm" standard preset as it was the closest to neutral. The TV does not have explicit gamma control, so I had to play with black/white levels to adjust gamma curve. I prefer my blacks to be discernable, not crushed, you can see it by the adjusted gamma curve.

As with most plasmas, the green is oversaturated and cyan and magenta are off.

Note, that Rec 709 (HD) and Rec 601 (SD) are different, particularly in CIE colors and in gamma profile. SD gamma is 2.5, HD gamma is 2.2.

The same SpyderExpress can be used to calibrate your computer monitor, luckily if you have reasonably new computer, monitor and OS, this process is fully automatic.

[Bill Ravens]

As a "professional", one must adhere to the standards of the industry. To try to second guess what an average user might have his/her monitor set to is whistling in the dark. To the best of my knowledge, MOST users are still viewing on NTSC compliant monitors. If this is the case, it is necessary to make your editting monitor conform to this standard, which, I guess is SMPTE. If you're working in HD and the delivery schema is TV, then SMPTE it is. If you happen to be delivering to an HDTV, things change. HDTV isn't SMPTE. I do know that there is a different version of the SMPTE color bars that are used for HDv. The color bars defined for HDv are ITU 709 compliant. Perhaps these are the best standards to use on an monitor intended for HDTV distribution. But, I'm just guessing. I don't really know.

This is where Glenn Chan could jump in and tell us what the color mapping is for HDTV. Is it RGB? And would that be RGB16-235 or RGB 0-255?

[Glenn Chan]

1- Sorry, I wasn't clear in my original post. There are different aspects of the monitor that can be calibrated... the interface (e.g. composite, S-video, component, HDMI, DVI, etc. etc.) may or may not need calibration... digital interfaces don't need calibration.

Other aspects of the monitor can be calibrated.

The monitor may also have settings that let you change certain aspects of the image processing (e.g. turn off overscan to get 1:1 pixel mapping). This will depend on the menu structures in the monitor.

2- A properly-designed LCD broadcast monitor that doesn't drift will not need the user to calibrate it.

3- The real problem with most consumer monitors is that they do wacky things to the image. They don't look the same as a reference monitor. Put them side by side with a reference broadcast monitor and they don't look the same.
They may also have limitations like limited color gamut, whatever deinterlacing circuit they have, and raised black level that you will never be able to calibrate/adjust/tweak away.

3b- It's generally easiest/best to get a broadcast monitor.
*Many old LCD broadcast monitors have some problems... they are getting better at quite a fast pace. And it looks like this might be the year where they finally surpass the Sony BVM CRTs in performance for HD.
(For SD, a CRT broadcast monitor is a much better idea IMO.)

3c- The exception to 3c might be something like the Apple Cinema Display + eCinemasys's EDP100. The EDP100 is a signal processing box that does the appropriate signal processing to get good color, deinterlacing, etc. the ACD (AFAIK). It has since been discontinued for better approaches (they rebuild the LCD panel and integrate everything into a single device). This approach doesn't work well if you have one of the newer flawed ACDs with the pink cast problem.

3d- You might be able to take a consumer display + add something to it that would make it reasonably good (like what ecinemasys did with the EDP100). I'm not familiar enough with the current options for this (e.g. Decklink) to know how well those solutions work.

[Glenn Chan]

Monitoring standards:

1- Viewing conditions:
For TV, there might be different standards here. (Not sure.)
SMPTE RP 166
ITU-R BT.710

In practice, this is not always followed.

2- Primary Chromaticities - just a fancy way of saying the exact shade of red, green, and blue. Chromaticity is an objective way of measuring the "shade" of a color. Primaries = red, green, blue

For SD, for NTSC countries (except Japan), the standard is the SMPTE C primaries.
For SD, for PAL countries (and Japan), the standard is the EBU primaries. *Sorry, I don't know the exact name of the standards documents.
For modern HD systems (i.e. not 1035i, not analog HD), the standard is the Rec. 709 primaries. ITU-R BT. Rec. 709 should be the document.

In practice, a lot of HD material is monitored on Sony BVMs which have SMPTE C phosphors. The Sony BVMs are probably the de facto standard (and they have particular shortcomings; though less than most LCDs).

In practice, the difference between the standard primaries is often glossed over. And people don't notice (so in a practical sense this omission works; it's ok).

2b- The original NTSC primaries are obsolete. No consumer display has primaries like those.

3- There is actually no standard defining some aspects of the reference monitor. The reference monitor was always assumed to be a CRT, with the CRT's natural transfer function and motion reproduction.

There are some working groups working on a standard there.

It will likely be something that emulates the CRT (but not its flaws... e.g. not that bright, resolution not that good for high frequencies).

4- Color mapping for HDTV:
Laid out in ITU-R BT. 709. It calls for Y'CbCr (with Rec. 709 luma co-efficients).
16-235 range for Y', for 8-bit formats (black at 16 Y', white at 235 Y'). 16-240 range for chroma.

[Bill Ravens]

It sounds as tho' there are some fundamental problems that still need to be worked out. For example, if an NLE displays only in RGB, it's going to misrepresent HD colors in preview, if the display is thinking in Y'CbCr, and the NLE is trying to display in RGB...well, you can see there's a problem, no? So the editor has to juggle the color space, during editting, to fool the monitor to display right colors.

But, if understand what you're saying, the best calibration one could do, right now, is to use Rec709 color bar. Provided, of course, the end user is on an HD capable monitor. Most of the world I know, is still using SMPTE specced equipment. Here's a comparison of the SMPTE SD colorbars and the SMPTE HD color bars...
http://en.wikipedia.org/wiki/Color_bars

Care should be taken because the pluge bars are significantly different from the SD version.

[Glenn Chan]

The conversion from Y'CbCr <--> R'G'B' is defined by the standards documents.

The editor should not have to juggle the color space to get the monitor to show the right thing, in a well-designed system. (By that definition, Vegas is not because you do have to juggle color spaces.)

2- I didn't think about this:
There are different flavours of Y'CbCr depending on which set of numbers they use (based off of the Rec. 601 or Rec. 709 luma co-efficients; there is also a third obsolete set).

Some TVs may let you choose which luma co-efficients are assumed when decoding Y'CbCr signals (because Y'CbCr signals either use Rec. 601 or Rec. 709 numbers). Using the wrong set of numbers will cause dramatic color shifts/inaccuracy and clipping of certain highly saturated colors. It is worth checking that your system gets the numbers right.

One way to check is to send color bars and eyeball them to see if they look correct... though that method is not 100% foolproof. It's possible to make a test pattern that makes it easier to see.

2b- Many consumer sets will always use the Rec. 601 numbers... which is wrong for a lot of HD.

But, if understand what you're saying, the best calibration one could do, right now, is to use Rec709 color bar. Provided, of course, the end user is on an HD capable monitor. Most of the world I know, is still using SMPTE specced equipment.

I don't think there is such thing as the Rec. 709 color bar?

2- Color bars should have the same R'G'B' code values in the end... their Y'CbCr values differ. If you originate them in R'G'B', then they should be correct as long as you do the right R'G'B'-->Y'CbCr conversion.
The Y'CbCr values differ depending on whether you used the Rec. 601 or Rec. 709 luma coefficients to encode.

[Bill Ravens]

here's another set of what someone is calling HD colorbars, a little more credibility than wikipedia

http://www.belle-nuit.com/testchart.html

[Glenn Chan]

The problem there is that it assumes a particular mapping from R'G'B' to Y'CbCr... e.g. 16 16 16 RGB gets mapped to 16 Y' (and neutral chroma).
A- Some programs (or rather, codecs) will map 0 0 0 RGB to 16 Y'.
B- Quicktime may apply inappropriate color management onto the still.

So those things will screw everything up.

[Giroud Francois]

electronic color chart are great, but useless if you do not get a paper version to compare with..... which is still the best way to quick calibrate a screen.