HOW-TO: Calibrating Display to Match HTPC Output

[jimwhite]

"On a dlp this is something akin to clipping them off at the display. (others, don't argue about BTB and such here, it's not synonomous with clipping because the information can be used/processed and affect the display for elements that are above black and clearly visible)."

You have a lot of gall to slip this in when it's convenient for you..... I'm not gonna argue the point, but note that you have caved when it's convenient...

[ChrisWiggles]

Yes, but if I lower my brightness in order to have the Below Black Bar disappear into the Digital 16 Black background of DVE title 12 chapter 13, doesn't that mean that I have automatically set my brightness at Digital 7 (where the BTB bar is encoded)?

Not quite. First we have to be clear that the other forces at work here that obscure the shadow detail is washout from brighter portions of the pattern. If you're using a DVE pattern with an 80% bar, that will washout the dark parts a lot more than with like a 20% bar. Use a low-APL (dim) pattern and it will help minimize this washout problem when setting the black level on a DLP. (I'm only recommending this specifically here, I don't hold the same advice for CRT projection necessarily)

Think of it this way, if the BTB bar (encoded at say 7) is no longer visible against a black background(digital 16), then your black level is at 16. If your black point resolves *anywhere* below 16 you will see the BTB 7 bar resolved, though it may not have the full difference between its encoded level of 7 and the background of 16. Say theoretically your black point places level 10 as the lowest black of your DLP. You will still see the BTB bar at 7, but it will look the same as 10, and every other level below ten.

Does that make sense? So if you can't see any BTB bars with, then you are at 16 or higher, on these patterns. For others: obviously if your source/processing is not passing BTB then this is a separate issue.

[cyberbri]

You shouldn't be able to "see" the BTB bar. That's why it's called "blacker than black". The reason it exists is because there may be information below "level 16 black" that helps your display process the image. But you do not want to be able to see black at level 7, because actual "blacks" that are at level 16 will no longer be black, they will be grey. As you lower your "brightness" black level slider, the black level rises, meaning that more areas are black. Say you have the black level set to digital 7. As you lower the brightness adjustment, digital 8, 9, 10, etc. will blend in with the black. You want to set your brightness/black level setting so that the 16 value is where the black point lies, and everything below that blends in with black, and so that you can still see slight noise/light in 17 and up.

Conversely, if your brightness/black level is calibrated to Digital 16, then you would have to raise that slider adjustment to show digital 15, 14, and so on until digital 7 is visible.


Back a page or two, someone kindly posted some files that show 0-25 and 230-255 on two separate images, in the form of bars for each value. I suggest you find and use those to experiment with your brightness setting. This should help you dial in 16-black.

[ChrisWiggles]

You have a lot of gall to slip this in when it's convenient for you..... I'm not gonna argue the point, but note that you have caved when it's convenient...

My statement is completely in-line with everything I've said in the past. On a DLP that has a fixed, unmoving black point, data below the black point will not be resolved. There is no float, so it won't be resolved. Note that I say "AKIN" to clipping, because that data is still present for image processing and dithering for data that can affect shadow detail that *is* being resolved. I clearly point this out when I say that it is conceptually similar, but NOT completely synonomous to clipping.

On a DLP with a moving black point to mimic CRT, it is CLEARLY not clipping, though may appear as such when viewing a fixed pattern. As the image content changes and the black point moves around (like a CRT display which is our reference) then it is obvious that clipping of data in the signal chain is a very different thing than the visible "clipping" at the actual creation of the image itself. They are not the same, and I have not 'caved' to anything, as I quite obviously pointed out in what I quoted. It's why I added the clarifier to emphasize the word akin.

[ChrisWiggles]

Conversely, if your brightness/black level is calibrated to Digital 16, then you would have to raise that slider adjustment to show digital 15, 14, and so on until digital 7 is visible.

As I pointed out, it is easily lost but if you are resolving just a couple steps below 16 (multiple steps may be necessary for your eyes to be able to resolve the change, as the point of 8-bit video is that individual steps are not clrealy resolvable by your eyes, so no visible banding) then you will see the bar at 7, but it will appear identical to whatever level you're at for black, say 10. If you have a bar at 13, a bar at 10, and a bar at 7, (pretend your eyes can see the difference just for illustration) then the 13 will look somewhere between 10 and a 16 background, but your 10 and 7 bar will look exactly the same, at 10.

[takisot]

Originally posted by ChrisWiggles
Not quite. First we have to be clear that the other forces at work here that obscure the shadow detail is washout from brighter portions of the pattern. If you're using a DVE pattern with an 80% bar, that will washout the dark parts a lot more than with like a 20% bar. Use a low-APL (dim) pattern and it will help minimize this washout problem when setting the
black level on a DLP. (I'm only recommending this specifically here, I don't hold the same advice for CRT projection necessarily)]

I do use the 20 % bar (it is the one at DVE title 12 chapter 13).

Originally posted by ChrisWiggles
Think of it this way, if the BTB bar (encoded at say 7) is no longer visible against a black background(digital 16), then your black level is at 16. If your black point resolves *anywhere* below 16 you will see the BTB 7 bar resolved, though it may not have the full difference between its encoded level of 7 and the background of 16. Say theoretically your black point places level 10 as the lowest black of your DLP. You will still see the BTB bar at 7, but it will look the same as 10, and every other level below ten.
Does that make sense? So if you can't see any BTB bars with, then you are at 16 or higher, on these patterns. For others: obviously if your source/processing is not passing BTB then this is a separate issue. [/b]

Thanks Chris!
With your explaining it now makes sense! Your sentence: “ If your black point resolves *anywhere* below 16 you will see the BTB 7 bar resolved, though it may not have the full difference between its encoded level of 7 and the background of 16” was the key for me!

So, to come back to my original question, where exactly is my black level when the dithering of micro-mirrors stops? At digital 16 or below it?

[ChrisWiggles]

Again, the simple way to recommend it is the way I do it, it's to lower your black level a bit too far, raise it until black (digital 16) *just* starts to dither, then lower it one click below that so that 16 is a solid, no-dithering black. That's what I would shoot for on a DLP. I also mention that some will lower a couple clicks below this to resolve some detail that is below 16, to include for the floating "real" black level in the source. Obviously it's easier to explain and set the former. In any case if you are within a couple steps of having 16 be your lowest black (you don't want to be above this though, you don't want to clip shadow detail) then you should have a solid image. The choice to resolve a few steps below black on a DLP with a fixed black point it your own preference.

Also note that I am biased to have the 16 without dithering since I find it distracting on a full black frame. Some would be one click higher and *just* have the beginning of dithering at 16, but I disagree.

Hope this helps!

[takisot]

Originally posted by cyberbri
You shouldn't be able to "see" the BTB bar. That's why it's called "blacker than black". The reason it exists is because there may be information below "level 16 black" that helps your display process the image. But you do not want to be able to see black at level 7, because actual "blacks" that are at level 16 will no longer be black, they will be grey. As you lower your "brightness" black level slider, the black level rises, meaning that more areas are black. Say you have the black level set to digital 7. As you lower the brightness adjustment, digital 8, 9, 10, etc. will blend in with the black. You want to set your brightness/black level setting so that the 16 value is where the black point lies, and everything below that blends in with black, and so that you can still see slight noise/light in 17 and up.

Conversely, if your brightness/black level is calibrated to Digital 16, then you would have to raise that slider adjustment to show digital 15, 14, and so on until digital 7 is visible.

Cyberbi thanks.
After Chris' post, I re-read your post and I now understand the logic behind this.

[takisot]

Originally posted by ChrisWiggles
Again, the simple way to recommend it is the way I do it, it's to lower your black level a bit too far, raise it until black (digital 16) *just* starts to dither, then lower it one click below that so that 16 is a solid, no-dithering black. That's what I would shoot for on a DLP. I also mention that some will lower a couple clicks below this to resolve some detail that is below 16, to include for the floating "real" black level in the source. Obviously it's easier to explain and set the former. In any case if you are within a couple steps of having 16 be your lowest black (you don't want to be above this though, you don't want to clip shadow detail) then you should have a solid image. The choice to resolve a few steps below black on a DLP with a fixed black point it your own preference.

Also note that I am biased to have the 16 without dithering since I find it distracting on a full black frame. Some would be one click higher and *just* have the beginning of dithering at 16, but I disagree.

Hope this helps!

You are clear as crystal!

[ChrisWiggles]

You are clear as crystal!

Yay!

That's a rare thing for me!

[maxleung]

3no, I finally tried out your test screens - works great! I really appreciate your effort - I love how it continuously plays multiple frames in each test pattern, making it dead-easy to calibrate TheaterTek (or any other VMR9 player) without having to go back and forth with the chapter controls like I have to do with DVE and (some) AVIA patterns.

I verified that my TheaterTek calibration is dead-on perfect. It is much quicker than testing my VMR9 Brightness and Contrast settings using DVE or AVIA - simple menu, one click and you're done!

I will always use this pattern if I change drivers or try out new players. Great work!

[3no]

maxleung- Glad you found it useful.

I have an updated version, attached. The release notes follow.
---------------------------
The purpose of this tool is to provide quick and easy calibration of the grayscale at Reference Black and Reference White. This is intended to be an adjunct to DVE t12/ch14 with several advantages for reference point calibration a) finer resolution around the reference levels, b) video loops so changes to your DVD player setup will show immediately without having to flip between chapters, c) bars in the “Black 15-19” screen wiggle for better visibility.

For best results, set up each element of the video digital signal path in order as follows. Warning - this is a summary of several hundred AVS Forum posts and several spirited debates that you can and should read for yourself. Good references are ChrisWiggles’ Source Settings Guide, and cyberbri's Calibrating Display to Match HTPC Output.

Here's my highly compressed summary:

- Software DVD player: adjust its video levels so an alt-prt scrn pasted into MS Paint (for example) yields bar RGB values in the screen capture that are equal to the source bar label values. {edit: more detailed explanation of this step 6 posts further down}

- Video driver: leave flat.

- Display black: adjust brightness so bar 17 is barely visible and bar 16 (reference black) is not (blends with the blacker-than-black background). Do this first with the Black 1-24 screen, then fine tune with the Black 15-19 screen.

- Display white: adjust contrast so 235 (reference white) is clearly distinguishable from the background, and as many higher peak white values to the right as you feel are appropriate after reading the posts and judging your own source material. I find that adjusting so values up to 245 are visible yields the best results on my system.

- Recheck display black and white for interactions.


This is release 2.0, April 24, 2005. New with this release:

- Added a Reference Black fine tuning screen with only bars 15-19, plus the bars wiggle for better visibility.

- Crisper bar edges due to resized screens, custom-tuned rendering, and progressive encoding. Screens now display in 4:3 aspect for the same reason. The bar edges on the DVD file’s native 720x480 image are (essentially) pixel perfect, but the DVD player’s anamorphic image resizing and desktop scaling blur the edges somewhat.

[stanger89]

Very nice, I've added them to the "Calibration" dir on my server, I'll have to try them out later.

[peter caesar]

Thanks a lot, 3no~

Another great AVSer's masterpiece!

BTW, how can I adjust my displays with your file for HDTV .ts files?
Is it similar to DVD, or should I adjust to 0 ~ 255 PC level?

[takisot]

Great work 3no!
Can't wait to test it!
Thank's a lot!

[3no]

Originally posted by peter caesar
BTW, how can I adjust my displays with your file for HDTV .ts files? Is it similar to DVD, or should I adjust to 0 ~ 255 PC level?

Originally posted by ChrisWiggles
Video from DVDs or other digital sources, follows Studio RGB standards which encodes reference black at level 16, and nominal reference white at level 235.

That being said, I have not personally worked with .ts files, so maybe someone more knowledgeable could confirm.

[takisot]

Originally posted by 3no

- Software DVD player: adjust its video levels so an alt-prt scrn pasted into MS Paint (for example) yields bar RGB values in the screen capture that are equal to the source bar label values.

-

Can you explain exactly how to achieve that? I am not familiar with this procedure.
Thanks.

[3no]

Originally posted by takisot
Can you explain exactly how to achieve that? I am not familiar with this procedure.
Thanks.

Assuming you are using VMR, the objective is to set your DVD player to neutral, i.e. the RBG levels of the image coming out of the player to the video card should be the same RGB values as the image encoded on the DVD (out=in). For example, fire up the Black 1-24 screen, make sure the DVD player is the active/foreground task (highlighted on the windows taskbar), press Alt-PrintScreen (captures the window of the active task to the clipboard), fire up MS Paint, Photoshop or equivalent, and paste the clipboard image. Assuming Paint (as everyone has it even on an otherwise stripped HTPC), select the eyedropper, click on the 16 bar, go to Colors/Edit Colors/Define Custom Colors and observe the Red, Green and Blue values, which should all be 16. If they are not, adjust your DVD player's brightness control in the appropriate direction and try again. This is easier with Photoshop as one can scan the eyedropper across the image while getting a continuous RGB reading on the Color Picker.

Do the same with the White 232-255 screen, adjusting your player's contrast control until out=in. Check all the bars on the screen, not just 235, to ensure that some expansion/recompression is not inadvertently clipping the higher values.

[maxleung]

Again, great work 3no!

Hmmm, it would be wonderful in the future to have XVID/DIVX, H.264, WMV-HD etc. versions of these same test patterns - perfect for the future when we start playing mpeg4 content with different decoders (ie. calibrate for Nero H.264, DIVX HD, XVID H.264, etc.). It would be tremendously hard work though - I wonder if an AVISynth script could be made for this kind of thing - generalized so they can be fed to any encoder for our own evil calibration purposes.

[peter caesar]

Thanks for your kind input, 3no!

[takisot]

3no, You are the Man! Thanks!

[shawn2bGoofy]

I was hoping someone can lend some thought to helping me out. I am an intermediate home theater user that gets a bit lost when talking about upconverting and downconverting. I am trying to get the best possible playback when using my pc and dvd's a sources on my hdtv.

This is what I got...any thoughts???

OS: MCE 2005
video card: G-Force FX 5700LE
Tv: Samsung rear projection CRT capable of 720p w' a DVI input
Set top box: Motorola HD DVR
Linksys Media Center Extender
Power DVD 6.0
Zoom Player
DiVx Pro


I don't know if you need more info or this is too much but I want to get the output from the pc to the tv up to 720P if that possible?

Thanks for the help,

Shawn

[cyberbri]

Your desktop resolution should be 1280x720, or any smaller setting you need to use in order to get rid of overscan (I use 1224x694). The nVidia drivers should let you do this very easily. Although you have a RP-CRT (I have a DLP), so it might take a little more work to get yours going. I'm surprised your CRT isn't 1080i. Is the native resolution 1080i or 720p?


For playback, use Zoom Player Pro 4.50, and use VMR9 Renderless (check the box for "Exclusive" mode). Use DScaler 5.0.0.6 codecs. Get the August version of FFDSHOW, and use this guide to get it set up. Then change the DScaler 5 options, output colorspace to YV12 (not YUY2) and set Deinterlacing to Force Weave. In FFDSHOW leave only Resize checked for filters, and use Lanczos for the algorithm, parameter of 2, and Luma Sharpen of .60. Try a resize value of 2x DVD resolution, or 1920x1080 if that is your display's native resolution. Then in Output Colorspace uncheck everything except YV12.

After that, make sure your display is calibrated correctly for VMR9 and you should be set to do minor tweaking and experimentation (resize values, Luma Sharpen strength, etc.). See the FFDSHOW Faq, ZP Pro 4.50 or DScaler 5 threads for questions on issues with those.

HTH

[Manchild]

cyberbri or anyone else that can provide some insight,

I use a 30" Sharp LCD HDTV (Native 1280x768) for all my video playback and as my main PC monitor. Up until now I've been using overlay without level adjustments in FFDShow and gotten glorious results. However, whenever I try to use VMR9 (I have a Radeon 9800) I always get dominating grays. I then tried to adjust output Levels to 16-235 in FFDShow and it didn't improve much. Since I use the monitor for both PC & Video am I best to just stick with the Overlay option?

Also, How do the latest DScaler 5 codecs match up with the NVidia Forceware's...have they been tested with BeyondTV and the like?

[maxleung]

Manchild, I believe this thread contains most (all?) of the information you need to deal with your problem.

But, in short, you need to recalibrate your display for use with VMR9, but I am making the assumption that your Radeon 9800 (in overlay mode) and/or DVD player are clipping levels beyond 16-235 and then "expanding" to 0-255 levels. In this situation, you will get elevated blacks in VMR9 mode, as VMR9 preserves the color information properly and places reference black at level 16, which would look grey on your LCD monitor.

(Sidenote: It is possible that your player's brightness, contrast, and saturation settings are not set correctly and damaging the image. To test for this, you would need to use the screenshot technique - pressing ALT and the PrintScreen button on your keyboard while playing a calibration pattern, and comparing to what the pattern SHOULD be. 3no's grey bar patterns are superb for testing brightness and contrast. Saturation is more difficult and requires using a color bar pattern.)

Unfortunately, I am not sure how to calibrate TV tuner output. Anyone have ideas on that?

EDIT: Meant to say that I'm assuming overlay on the Radeon 9800 is clipping and then expanding levels.

[cyberbri]

Originally posted by Manchild
cyberbri or anyone else that can provide some insight,

I use a 30" Sharp LCD HDTV (Native 1280x768) for all my video playback and as my main PC monitor. Up until now I've been using overlay without level adjustments in FFDShow and gotten glorious results. However, whenever I try to use VMR9 (I have a Radeon 9800) I always get dominating grays. I then tried to adjust output Levels to 16-235 in FFDShow and it didn't improve much. Since I use the monitor for both PC & Video am I best to just stick with the Overlay option?

Also, How do the latest DScaler 5 codecs match up with the NVidia Forceware's...have they been tested with BeyondTV and the like?

The easiest thing would be to just use Overlay. That will let you watch DVDs without doing anything.

You could also just try adjusting your "brightness" setting down to make level 16 black on the test patterns appear as the blackest black on your screen. You may find that everything looks fine. I have my TV set to VMR9 levels and still use it to do normal PC things. You should only have to worry if you use it to do professional imaging work, etc. All this does is make the 0-16 black/dark grey levels blend together basically.

I suggest the latter, as VMR9 should give you a better picture than Overlay.

[cyberbri]

GSB (Gary) has detailed a nice method of calibrating Service Menu adjustments on (Samsung) DLPs HERE by tying the contrast level to the maximum red to achieve 6500K color temp. I have not yet tried this method, but it seems to work very well.

[3no]

Originally posted by maxleung
Hmmm, it would be wonderful in the future to have XVID/DIVX, H.264, WMV-HD etc. versions of these same test patterns

maxleung-

I played around with avi's, but was disappointed in both the slop (fuzzy bar edges) of the encoders (DivX, XviD, 3ivx) and the seemingly uncontrollable colorspace remapping of the players (WMP10, DivX Player). Only TheaterTek came close, but not close enough for calibration.

FYI, I *am* using AVISynth feeding VirtualDub, which makes it easy to swap encoders, but also to control the colorspace and verify pixel perfect input to the encoder.

Attached are two DivX avi's (the ones I said aren't ready for primetime) for you to play with. I'd appreciate feedback on your results with these on your system. Each is a 60 sec clip at 5 fps, so you should set your player to loop.

[maxleung]

Thanks 3no! You're reasoning makes a lot of sense - colorspace conversions is just a big pain in the butt, among other things.

I'll try to have a look at your new patterns early this week.

[Manchild]

To whomever it may concern:

Alright...I got 3no's fabulous test patterns, loaded em up into Zoom Player with VMR9...print screened...checked in Photoshop and both the "16" bar had RGB of 16-16-16 and the white pattern had a 235 of 235-235-235. So I suppose this implies that the brightness/contrast settings in my zoom player are set appropriately. However, whenever I watch material in the DVD player it still looks washed out or "grey." The main thing that's noticeable are the "black" letterboxes that stand out.

Did I miss a step or am I doing something wrong here? It seems like no matter what I do I can't get this thing to display VMR9 properly...oh lord please save me...thanks...hehe