HD MPEG-2 Test Patterns

[dr1394]

I've hosted some HD MPEG-2 test patterns on my website.

http://www.w6rz.net

They are HD MPEG-2 Transport Streams at the standard 19.39 Mbps ATSC bitrate. The first patterns are 601 and 709 color bars, and some resolution frequency bursts.

I've already had a request for IRE window patterns, so they'll be available soon. Any requests or suggestions are welcome.

Ron

[Li On]

Some thought off the mind:

- 1:1 single pixel on/off pattern
- sharpness
- Y/C delay (horizontal/vertical)
- overscan, pixel crop
- BTB, WTW, ramp map
- color field (R/G/B/Cyan/Magenta/white/black etc)
- geometry, linearity, convergence
- motion judder test
- 3/2, 2/2 pulldown
- video jaggies
....

regards,

Li On

[trbarry]

Ron -

Very cool! Thank you.

- Tom

[Bob Sorel]

Ron...excellent!!

There is a huge shortage of HD patterns to use for calibration and your willingness to host and even create such patterns will be a very valuable addition to those of us who would like to calibrate our HD setups (me... ).

I appreciate any and all patterns that you will contribute, but here is my wish list. I originally posted this with the intentions of putting together a DIY DVD calibration disc, but the same patterns are equally valuable for HD, and they are much harder to get in either 601 or 709 color space, and nonexistent AFAIK in .ts format.

Here's what I figure:

DVE reverse ramps - to check for BTB and WTW and to get a general idea of the grayscale calibration - precalibration procedure
DVE color bars (horizontal on white background) - to check and adjust color decoding
AVIA PRO secondary color checkerboard - to check and adjust hue/tint
75% red window
75% green window
75% blue window
75% cyan window
75% magenta window
75% yellow window
full field black
10% white window
20% white window
30% white window
40% white window
50% whilte window
60% white window
70% white window
80% white window
90% white window
100% white window
75% white (for tristimulus chart)

and then I went on a rant as to how I would recommend setting up a calibration disc:

If I were to organize a disc, I would make it incredibly simple to navigate. and put all of the "fluff" elsewhere. Or, at least make it very simple to get to the "professional" patterns right from the main menu and then make it easy to navigate. Avia comes pretty close, but still has its problems - too many submenus. Put the "meat and potatoes" patterns in one section, and then put all of the extra and/or specialized patterns in another.

That is, there should be a "calibration" menu that is further subdivided by categories like:

1. Basic - including color bars and patterns for setting brightness and contrast, as well as check for the ability to pass BTB and WTW. Just include the BEST patterns to get the job done, not EVERY pattern you can imagine. Put the alternative patterns elsewhere.
2. Color - 75% windows in all 6 colors
3. Grayscale - 100% black, 100% white fields, and the 10% to 100% windows in 10% increments (all *labeled* on screen to avoid mistakes, and all verified to be color correct)
4. Crosshatch - for CRT convergence and geometry

You get the idea. Within each subcategory, all patterns should be chaptered and set to infinitely loop to avoid timing out if the calibrator is delayed. By using chapters, the calibrator can just hit the "previous" or "next" button on the remote to switch from one to another - a lot faster than going back to ANY menu. And since the patterns are *labeled* on screen, there is no chance of putting up the wrong pattern. Then put all of the specialized and/or alternative patterns in another area and organize those similarly for equal ease of use. There is no reason, for example, to have 20 different sets of color bars in a section when you only need one or two to get the job done. That just forces needless navigation.

The amount of patterns necessary to do even a full professional level calibration is not all that many, especially if the patterns are selected judiciously. Putting together a disc using .ts format will be incredibly simple.

Ron, I'm not suggesting that you do all of this work. This is just a list of *ideal* patterns that are normally needed for professional level calibration. But you asked, so I didn't feel shy about responding...

Thanks again! This is a wonderful effort on behalf of the forum!

[Ursa]

Bob - I have created full-field patterns for gamma calculations (16 - 236, using 236 for reference white for convenience to eliminate rounding issues - the 10% steps are then 22 code values apart). I also have 100% full-field color patterns as well, but 75% would be pretty easy as well (given my white value at 236). I used 100% to reduce by one the number of measurements to take.

Personally, I am curious about folks that are super anal on accuracy to the thousandth of a decimal in other areas, but then introduce way more error by using inherently analog test patterns on digital displays while still treating them as whole-value integers.

Ron - Did you author the patterns via a converted bitmap or directly in an MPEG container? If you used a bitmap, or other static graphic format, how did you author differently between Rec601 and 709? The code value ranges are the same (16 - 235), and the resolution is the same in your patterns, so I'm guessing you authored directly in YUV2/YV12? BTW, Thank You! You and 3no are really helping out your fellow hobbyists!

Later,
Bill

[Bob Sorel]

Bob - I have created full-field patterns for gamma calculations (16 - 236, using 236 for reference white for convenience to eliminate rounding issues - the 10% steps are then 22 code values apart). I also have 100% full-field color patterns as well, but 75% would be pretty easy as well (given my white value at 236). I used 100% to reduce by one the number of measurements to take.

Bill, full field patterns and 100% colors work fine for me... Jeff (UMR) recommended changing the list to windowed patterns and 75% colors basically due to the difficulty in plasmas to display these images, thus making the reduced patterns more universal in their usefulness. DLP owners won't have any problems, though...

To anyone creating these patterns: I highly recommend labeling the pattern on the pattern itself to avoid any possibily of mistake on the part of the user. Especially with the grayscale windows or full field patterns, it would be very easy to grab the wrong unlabeled pattern and throw off your gamma curve measurement, for example. Hopefully most people would know the difference between cyan and magenta, but you never know...hehe.

[dr1394]

Originally Posted by Ursa

Ron - Did you author the patterns via a converted bitmap or directly in an MPEG container? If you used a bitmap, or other static graphic format, how did you author differently between Rec601 and 709? The code value ranges are the same (16 - 235), and the resolution is the same in your patterns, so I'm guessing you authored directly in YUV2/YV12? BTW, Thank You! You and 3no are really helping out your fellow hobbyists!

Later,
Bill

All the current patterns have been "authored" in 4:2:2 YCbCr. To feed my hardware encoder, I load the YCbCr file onto a DVS HD-SDI server. However, I do have a program to convert from 4:4:4 RBG to 4:2:2 YCbCr (with either 601 or 709 matrix), so I can do bitmaps (or any format that Photoshop can handle). In fact, I have an overscan pattern that I made from a GIF found on this site:

http://www.tigerdave.com/images/test...x1080_grid.gif

I haven't put it on the website because it looks wacky in 1080i since he used alternating lines to create the color shading. Also, I was going to get his permission.

Ron

[Li On]

Hmm... that "overscan" pattern is a bit.. plain! I'd like some maker around the edge to indicate overscan range.

regards,

Li On

[Ursa]

Originally Posted by dr1394

All the current patterns have been "authored" in 4:2:2 YCbCr. To feed my hardware encoder, I load the YCbCr file onto a DVS HD-SDI server. However, I do have a program to convert from 4:4:4 RBG to 4:2:2 YCbCr (with either 601 or 709 matrix), so I can do bitmaps (or any format that Photoshop can handle). In fact, I have an overscan pattern that I made from a GIF found on this site:

http://www.tigerdave.com/images/test...x1080_grid.gif

I haven't put it on the website because it looks wacky in 1080i since he used alternating lines to create the color shading. Also, I was going to get his permission.

Ron

Ron - Definitely some cool toys! Thanks!

[jflatt]

I just put together a DirectShow source filter that will create the 1x1 black/white checkerboard pattern. It can provide a very clean image to the attached Overlay, VMR7, VMR9, whatever. I think it would be nice to get a bunch of various patterns in it and roll it up into a useful application. Attached is the source and a VS.NET 2003 project.

[dr1394]

The 0 thru 100 IRE window patterns are now available on www.w6rz.net, along with IRE bars and split bars (10 IRE steps). Sorry Bob, no labels (for now).

Luma values used for IRE patterns are:

0 IRE = 16
10 IRE = 38
20 IRE = 60
30 IRE = 82
40 IRE = 104
50 IRE = 126
60 IRE = 147
70 IRE = 169
80 IRE = 191
90 IRE = 213
100 IRE = 235

You may need to hit reload to see the new links.

Ron

[dr1394]

Added ramp (luma levels 1 - 254) with -5 to 105 IRE bars pattern.

Ron

[kschmit2]

These are all excellent so far.
Thank you very much for that effort.

[Mac The Knife]

Sorry to complain about a freebee, but could someone with a lot of bandwidth volunteer to rehost these patterns?

DR1394's web site has such a low bandwidth limit that I gave up and canceled my first two attempts to download.

[Diode1]

Originally Posted by Mac The Knife

Sorry to complain about a freebee, but could someone with a lot of bandwidth volunteer to rehost these patterns?

DR1394's web site has such a low bandwidth limit that I gave up and canceled my first two attempts to download.

I am on comcast cable and the bandwidth via 2x d-loads @ a time via http://www.w6rz.net/ has been really good.

[RichB]

This is a great.

I think it might help downloaders if there was a single zip or at least 1 zip per section.

-- Rich

[jmeer]

It looks like they zip from about 70mb to about 2mb

[dr1394]

Originally Posted by RichB

This is a great.

I think it might help downloaders if there was a single zip or at least 1 zip per section.

-- Rich

Doh! Thanks for jogging my feeble brain cell(s). As jmeer notes, the files are very compressable. This is because the IRE window patterns are almost all stuffing bits. It's 1 Mbps video and 384 kbps (silent) audio in a 19.39 Mbps TS. I'm going to change everything over to zip files later today.

As for bandwidth of the website, I just did a test download here, and it sustained 280 kbytes/sec, or about 2.3 Mbps.

Ron

[dr1394]

Okay, all changed over to zip files along with a zip file containing all the patterns.

Ron

[Bob Sorel]

Great work, Ron!

And the zipped files are much easier to download. It was taking me about 4 to 6 minutes to dl the separate .ts files, but I was able to download allpatterns.zip in about 9 minutes.

I'm anxious to check out the new grayscale patterns...

[kny3twalker]

just wanted to say thanks

[Kroot]

Thank you Ron for test patterns!

[benpatient]

wonderful. thanks!

[Mac The Knife]

Originally Posted by dr1394

...
As for bandwidth of the website, I just did a test download here, and it sustained 280 kbytes/sec, or about 2.3 Mbps.

Ron

Well, that's strange, because I was only getting about 20KB/s from your site whereas I usually get between 150 and 200KB/s from other sites.

I guess I'll try again after you zip everything up.

[RichB]

I was playing with the color pattern bars and there appear to be two black bars.

I assumed the brighter one was above video black ( greater than 16) and the other was below black ( less than 16). So I set black so that the darker one just disappeared.

Is that correct?

-- Rich

[Bob Sorel]

If my readings are correct, then the black background is at video level black (16), while the below black bar is 7 and the above black bar is 25. I mean, you could use that pattern, but a much better option would be to use the pattern labeled "ramp.ts", as you can use the same pattern for several purposes:

1. To evaluate the overall "grayness" of the grayscale.
2. To get an overall feel for the gamma curve.
3. To see if you are passing BTB/WTW information.
4. To adjust BOTH contrast and brightness. By using this pattern, you will see how the 2 controls interact, and cab quickly go back and forth to make adjustments until your ramps look as good as possible. The bars aligned with the dots represent video black (0% - 16) and video white (100% - 235).

I love this particular pattern!

[dr1394]

Added crosshatch with circles pattern.

Ron

[Li On]

Thanks!

For the crosshatch w/circles pattern, is there a edge outline at 4 sides? Using VLC player I only get the top edge line. That's why I want a good pattern to check edge pixel crop or overscan effect.

Are the patterns in 1080 or 1088 lines?

regards,

Li On

[dr1394]

Originally Posted by Li On

Thanks!

For the crosshatch w/circles pattern, is there a edge outline at 4 sides? Using VLC player I only get the top edge line. That's why I want a good pattern to check edge pixel crop or overscan effect.

Are the patterns in 1080 or 1088 lines?

regards,

Li On

There is an edge outline (actually 2 lines) on the top and left hand side of the crosshatch pattern. I thought about adding outlines on the bottom and right side, but a real overscan/pixel crop pattern is in the works.

All patterns are 1080 lines, since the DVS HD-SDI file server I'm using only supports 1920x1080 uncompressed images. Also, the video bitstream sequence header is set to 1920x1080. However, the HD MPEG-2 encoder that I'm using pads lines 1081 thru 1088 with black instead of the usual grey.

Tomorrow's pattern, Y/C delay.

Ron

[dr1394]

Added Y/C delay pattern. Here's the 0 delay red bar in MPEG-2 encoded/decoded 4:2:0 on top and the original 4:2:2 bitmap on the bottom.



Ron