AVS › AVS Forum › Display Devices › Display Calibration › HD MPEG-2 Test Patterns
New Posts  All Forums:Forum Nav:

HD MPEG-2 Test Patterns - Page 3

post #61 of 290
7/7/06 at 3:49am
I have a "weird" feature request!

All these great patterns are in 1080i format. I'd want some patterns in 1088i format, especially a 1088i overscan pattern. As we know there are lots of 1080i MPEG2 programmings are actually in 1088i native format due to the limitation of some MPEG2 encoders with the 16x16 pixels block size of the MPEG2 format.

I'd like to know the behavior of a display, video processor, MPEG2 decoder how to handle these common 1088i source. A ideal MPEG2 decoder (like the DScaler5 MPEG2 decoder!) should simply drop the buttom 8 lines of pixel and output in 1080i. A worksaround solution should be output the full 1088i image and let a downstream video processor to crop the buttom 8 waste lines. The worse processing is to decode the full 1088i, then rescale to 1080i output which the native 1080i image will be destroyed forever.

regards,

Li On
Reply
Reply

AVS Top Picks

post #62 of 290
7/9/06 at 3:18am
Thread Starter 
Quote:
Originally Posted by Li On View Post

I have a "weird" feature request!

All these great patterns are in 1080i format. I'd want some patterns in 1088i format, especially a 1088i overscan pattern. As we know there are lots of 1080i MPEG2 programmings are actually in 1088i native format due to the limitation of some MPEG2 encoders with the 16x16 pixels block size of the MPEG2 format.

I'd like to know the behavior of a display, video processor, MPEG2 decoder how to handle these common 1088i source. A ideal MPEG2 decoder (like the DScaler5 MPEG2 decoder!) should simply drop the buttom 8 lines of pixel and output in 1080i. A worksaround solution should be output the full 1088i image and let a downstream video processor to crop the buttom 8 waste lines. The worse processing is to decode the full 1088i, then rescale to 1080i output which the native 1080i image will be destroyed forever.

regards,

Li On

All 1080i bitstreams are actually 1088 lines. It's not an encoder limitation, it's how how the MPEG-2 specification works. For interlaced resolutions, the actual vertical pixel height must be divisible by 32 (1088/32 = 34).

All the test patterns on my website are 1088 pixels high. However, the sequence header vertical_size parameter for all the patterns is set to 1080. The last 8 lines on all the patterns are padded to black (Y = 16, Cb = 128, Cr = 128) as opposed to many encoders that pad to grey (Y = 128, Cb = 128, Cr = 128).

However, as a decoder test, I'll create some patterns with the last 8 lines padded to something besides black, and I'll also create some patterns with the sequence header vertical_size set to 1088.

I'll post them sometime this week.

Ron
Reply
Reply
post #63 of 290
7/9/06 at 4:29am
These patterns are great.

Have you considered a 720P overscan pattern?

I have not tried this, but I suppose we can burn these to DVD-R for playback on a Toshiba HD-A1. That would be interesting.

- Rich
Reply
Reply
post #64 of 290
7/9/06 at 7:54pm
Wow! Thanks Ron!

Now I understand! I know there is that MPEG2 header 1080/1088 indicator. And there is a small software that change a stream header from 1088 to 1080. But the header change tool never work for me on a HTPC. WinDVD/Nvidia MPEG2 decoder always show the bottom 8 gray lines while DScaler5 MPEG2 decoder never show the waste pixel.

So ALL MPEG2 1080i content are in fact in 1088i native format. Some with 1080i header and some with 1088i header. Some with the bottom 8 lines in black and some in gray/white/whatever.

I'd like to see these test "overscan" patterns:
- 1080i header, bottom 8 lines in black (already in the current test patterns)
- 1080i header, bottom 8 lines in gray
- 1088i header, bottom 8 lines in gray
- 1088i header, bottom 8 lines in black

Btw, is it possible to actual encode normal image in that 8 bottom lines? If possible, I'd like to see a "overscan" pattern which include the bottom 8 lines as part of the actually image, such as label the bottom 8 lines as 0 to -8 pixels!

Thanks in advance!

regards,

Li On
Reply
Reply
post #65 of 290
7/11/06 at 1:21am
I changed your fix1088 tool to convert 1080 to 1088! (sorry for messing with your program! Now I can test your patterns in 1088 mode.

regards,

Li On
Reply
Reply
post #66 of 290
7/11/06 at 11:15pm
Thread Starter 
Quote:
Originally Posted by Li On View Post

I changed your fix1088 tool to convert 1080 to 1088! (sorry for messing with your program! Now I can test your patterns in 1088 mode.

regards,

Li On

fix1088 is Jacob Balazer's tool, not mine. I've added a couple of bitstreams for your testing. They are:

http://www.w6rz.net/crop1080red.zip

http://www.w6rz.net/crop1088red.zip

They are encodings of the overscan pattern with the last 8 lines (1080 - 1088) padded to red (so that it's very obvious). For my hardware encoder, it's quite difficult to put anything complex into the last 8 lines due to the memory format (it's not in raster format, but in a special "tile" format). Also, it's impossible to set the last 8 lines with my DVS HD-SDI server (it only takes 1920x1080 YCbCr files).

Ron
Reply
Reply
post #67 of 290
7/11/06 at 11:25pm
Thread Starter 
Quote:
Originally Posted by RichB View Post

These patterns are great.

Have you considered a 720P overscan pattern?

I have not tried this, but I suppose we can burn these to DVD-R for playback on a Toshiba HD-A1. That would be interesting.

- Rich

720p overscan pattern is doable. Give me a few days. As for a DVD-R for the Toshiba, see this thread:

http://www.avsforum.com/avs-vb/showthread.php?t=693441

I think the guy is using my patterns, but for some strange reason, he doesn't know. I'd check it myself, but I can't ********** from work and my home internet connection is too slow.

Ron
Reply
Reply
post #68 of 290
7/12/06 at 5:15am
Quote:
Originally Posted by dr1394 View Post

720p overscan pattern is doable. Give me a few days. As for a DVD-R for the Toshiba, see this thread:

http://www.avsforum.com/avs-vb/showthread.php?t=693441

I think the guy is using my patterns, but for some strange reason, he doesn't know. I'd check it myself, but I can't ********** from work and my home internet connection is too slow.

Ron


Thanks. I will give that a shot.

- Rich
Reply
Reply
post #69 of 290
7/13/06 at 6:58pm
Thanks for the new patterns. And there are a few new requests too!

I'm trying to find out if a MPEG2 decoder outputs different HDTV MPEG2 formats in native resolution. The most common 1920x1080i is well covered with the "horzpix1.ts" and "vertpix1.ts" single pixel on/off patterns.

I'd like to see a pair of these patterns for 1440x1080i and 1280x1080i too. Or a single pixel on/off "checkerboard" pattern (like this WMV patterns in this post) to check both horizontal and vertical resolution.

I know there are the "burst1280.ts", "burst1440.ts" but I don't really know which part to look for for single pixel resolution output checking.

Thanks in advance.

regards,

Li On
Reply
Reply
post #70 of 290
7/15/06 at 10:38am
Might already be in the 'package' online, but if not perhaps a judder test pattern might be useful. Noticed this thread within another forum. Haven't linked my computer to my HD RPTV yet, but if all these goodies keep developing I'll have to try them. -- John
Reply
Reply
post #71 of 290
7/17/06 at 10:24pm
Thread Starter 
Quote:
Originally Posted by John Mason View Post

Might already be in the 'package' online, but if not perhaps a judder test pattern might be useful. Noticed this thread within another forum. Haven't linked my computer to my HD RPTV yet, but if all these goodies keep developing I'll have to try them. -- John

Actually, I do have some "judder" test patterns that I developed for my co-workers in the MPEG-2 decoder group.

The zip file:

http://www.w6rz.net/judder.zip

contains four test streams.

1) interlacejudder.ts - the moving bar advances 16 pixels horizontally each frame. However, since it's interlaced each field of the bar moves 8 pixels per field.

2) pulldownjudder.ts - the moving bar advances 16 pixels horizontally each frame. However, 3:2 pulldown has been applied.

3) progessivejudder.ts - the moving bar advances 16 pixels horizontally each frame. This is a 1080p@29.97 video bitstream.

4) filmjudder.ts - the moving bar advances 16 pixels horizontally each frame. This is a 1080p@23.976 video bitstream.

Ron
Reply
Reply
post #72 of 290
7/18/06 at 8:01pm
Thread Starter 
Quote:
Originally Posted by RichB View Post


Have you considered a 720P overscan pattern?

- Rich

The 720p overscan pattern is ready.

http://www.w6rz.net/overscancrop720.zip

Ron
Reply
Reply
post #73 of 290
7/18/06 at 9:11pm
Quote:
Originally Posted by dr1394 View Post

The 720p overscan pattern is ready.

http://www.w6rz.net/overscancrop720.zip

Ron

Great! I will try it out tomottow.

Thanks,

Rich
Reply
Reply
post #74 of 290
7/19/06 at 10:59am
If I may ask, do any of these test patterns have the ability to test whether or not 1080i is being weaved and not bobbed?

Thank you either way.
Reply
Reply
post #75 of 290
7/19/06 at 2:45pm
Thread Starter 
Quote:
Originally Posted by dallas27 View Post

If I may ask, do any of these test patterns have the ability to test whether or not 1080i is being weaved and not bobbed?

Thank you either way.

Yes. The specific patterns are:

http://www.w6rz.net/vertrez1080.zip

http://www.w6rz.net/vertrez720.zip

http://www.w6rz.net/vertrezmotion.zip

http://www.w6rz.net/vertrezivtc.zip

See this thread for more information:

http://www.avsforum.com/avs-vb/showthread.php?t=608670

Ron
Reply
Reply
post #76 of 290
11/21/06 at 8:20pm
dr1394,

Thanks for the patterns, they are great. I was wondering if you could add a 10/20 second countdown to your site? This will allow easy latency tests along with all of the others you currently offer.

Thanks,

- Brian
Reply
Reply
post #77 of 290
11/22/06 at 9:24pm
Ron - I have a question about your hdtestpatterns.zip that has the .iso in it. The web site says its an HD DVD ISO. Can this be burned to a standard DVD-R or DVD+R or do you need a special burner and/or special DVD media? Assuming it can be burned to a standard DVD, when played back in a HD DVD drive, does it really output at 1080i or does it somehow get downrezed in the process (such as burning it to a DVD)? Will this be available for BluRay as well? Thanks!
Reply
Reply
post #78 of 290
11/23/06 at 3:32am
Thread Starter 
Quote:
Originally Posted by padrino121 View Post

dr1394,

Thanks for the patterns, they are great. I was wondering if you could add a 10/20 second countdown to your site? This will allow easy latency tests along with all of the others you currently offer.

Thanks,

- Brian

I'm not exactly sure what you are asking for. What latency are you trying to determine or measure? The judder test patterns have a frame number counting in the upper right hand corner.

http://www.w6rz.net/judder.zip

Something like that?

Ron
Reply
Reply
post #79 of 290
11/23/06 at 3:42am
Thread Starter 
Quote:
Originally Posted by lovingdvd View Post

Ron - I have a question about your hdtestpatterns.zip that has the .iso in it. The web site says its an HD DVD ISO. Can this be burned to a standard DVD-R or DVD+R or do you need a special burner and/or special DVD media? Assuming it can be burned to a standard DVD, when played back in a HD DVD drive, does it really output at 1080i or does it somehow get downrezed in the process (such as burning it to a DVD)? Will this be available for BluRay as well? Thanks!

It can be burned to a regular DVD-R or DVD+R with any burner and will play in HD on the Toshiba HD-DVD players. See this big thread on the topic of HD-DVD authoring on red-laser DVD.

http://www.avsforum.com/avs-vb/showthread.php?t=705146

The original thread for the .iso file is here.

http://www.avsforum.com/avs-vb/showthread.php?t=707769

I started hosting the file since it was deleted from the original freebie hosting site. I also verified that the patterns were not disturbed by the Ulead MF5 authoring program. The results are here.

http://www.avsforum.com/avs-vb/showt...&&#post8169385

Ron
Reply
Reply
post #80 of 290
11/23/06 at 5:25pm
Thanks. So do I understand correctly than that there is no resolution loss in the images? In other words if there is a 1920x1080 cross hatch pattern that is just 1 pixel wide, when I play these back I should receive these as still just one pixel wide? Likewise am I correct to assume that the full fields and window patterns have been checked to make sure there is no contamination?

Lastly, would you please consider adding 5 IRE full fields and/or windows (5, 15, 25 etc)? These are very helpful for my calibrations. Particularly useful is a 5 IRE field/window as I am able to accurately read 5 IRE and like to calibrate it as close to D65 as possible. Thanks!
Reply
Reply
post #81 of 290
11/25/06 at 6:20am
Quote:
Originally Posted by dr1394 View Post

I'm not exactly sure what you are asking for. What latency are you trying to determine or measure? The judder test patterns have a frame number counting in the upper right hand corner.

Something like that?

Ron

Something like that might work, I need to check it out. I have a need to measure end to end latency in a system that I can only feed from a source and observe the output of. I was thinking something that counted down from like 10 seconds to 0 would suffice, albeit a bit rough. I liked seconds because it's much easier to see it just from observation then watching something as fast as the frame count but the better solution would be to have both.

I'll take a look at judder, do you have any suggestions that might better solve my problem?
Reply
Reply
post #82 of 290
11/25/06 at 5:53pm
Thread Starter 
Quote:
Originally Posted by padrino121 View Post

Something like that might work, I need to check it out. I have a need to measure end to end latency in a system that I can only feed from a source and observe the output of. I was thinking something that counted down from like 10 seconds to 0 would suffice, albeit a bit rough. I liked seconds because it's much easier to see it just from observation then watching something as fast as the frame count but the better solution would be to have both.

I'll take a look at judder, do you have any suggestions that might better solve my problem?

Take a look at the A/V sync pattern (it's at the very end of the web page). When I was working on low-delay MPEG-2 encoders for two-way videoconferencing, that's what I used to use.

With a dual-trace oscilloscope, you can easily measure the delay with millisecond accuracy. Connect the source video or audio to one channel and the delayed output to the other channel. Set the scope for "chop" rather than "alternate", since the sweep time will be quite low (100 ms per division). Trigger the scope trace on the channel connected to the source on the rising edge of the audio tone or the rising edge of the video bars. If you use video, you'll have to set the trigger level above the sync pulse (that is, somewhere between 10 and 70 IRE), so that it actually triggers on the color bars pulse.

You should then be able to see the delay of the tone/bars on the second channel. At 100 ms per division, the tick marks will be 20 milliseconds. Typical delays are 100 to 500 ms, so you'll get a pretty good picture of the delay.

When I was doing the low-delay work, the holy grail was 100 ms one-way (200 ms round trip). That's about the most two people in a conversation can tolerate without inadvertently talking over each other.

The pattern on the website has 1 second tone/video pulses. Depending on how long your delay is and how accurate you need to measure it, you might want a shorter duration pulse. Let me know, I can generate that pattern very easily.

Ron
Reply
Reply
post #83 of 290
11/26/06 at 2:18am
Quote:
Originally Posted by dr1394 View Post

Take a look at the A/V sync pattern (it's at the very end of the web page). When I was working on low-delay MPEG-2 encoders for two-way videoconferencing, that's what I used to use.

With a dual-trace oscilloscope, you can easily measure the delay with millisecond accuracy. Connect the source video or audio to one channel and the delayed output to the other channel. Set the scope for "chop" rather than "alternate", since the sweep time will be quite low (100 ms per division). Trigger the scope trace on the channel connected to the source on the rising edge of the audio tone or the rising edge of the video bars. If you use video, you'll have to set the trigger level above the sync pulse (that is, somewhere between 10 and 70 IRE), so that it actually triggers on the color bars pulse.

You should then be able to see the delay of the tone/bars on the second channel. At 100 ms per division, the tick marks will be 20 milliseconds. Typical delays are 100 to 500 ms, so you'll get a pretty good picture of the delay.

When I was doing the low-delay work, the holy grail was 100 ms one-way (200 ms round trip). That's about the most two people in a conversation can tolerate without inadvertently talking over each other.

The pattern on the website has 1 second tone/video pulses. Depending on how long your delay is and how accurate you need to measure it, you might want a shorter duration pulse. Let me know, I can generate that pattern very easily.

Thanks for the good information. When I used to do video work I would measure the latency at the application level using timing in the video stream, I never knew how to use an oscilloscope to measure the latency so this is very useful.

In this case I cannot attach anything to the output stream, just view it on the system that's built to play it so an instrument isn't really a possibility at the moment. The end to end latency may be a number of seconds however I want to be able to measure it to a resolution of less then a second so I was thinking a visual clue indicating seconds (countdown was my initial thought) and one indicating less then that (maybe something like 1/4 second, or ticking frame count per second) would handle it. I haven't looked at judder yet so I need to check it out and see if that will work.

Brian
Reply
Reply
post #84 of 290
11/30/06 at 11:31pm
Could anyone point me in the right direction? I've searched thread after thread and I've come up short.

I want to create basic grayscale and color patterns playable on my DVD player to calibrate my TV. I have Photoshop, but don't know how to start. I hope someone can help me out.
Reply
Reply
post #85 of 290
12/19/06 at 1:20am
Thread Starter 
I've added a new pattern. It's a combined luma and chroma frequency sweep.

http://www.w6rz.net/sweeps.zip

Ron
Reply
Reply
post #86 of 290
12/29/06 at 12:08pm
Thanks for this great resource. I had no problems burning the test DVD, but I have no idea how to use it. I have done a few basic calibrations using AVIA, but these patterns are a little different. So could someone please explain how you use these patterns for a plasma screen, or at least point me in the right direction to look?
Thanks
Reply
Reply
post #87 of 290
1/2/07 at 8:53pm
Yes, it would be nice it the "special patterns" had a brief explanation on how they work or what to look for.

In my case, what do I look for in the judder and Y/C delay patterns?
Reply
Reply
post #88 of 290
1/2/07 at 10:13pm
Quote:
In my case, what do I look for in the judder and Y/C delay patterns?

Judder... and Y/C delay...
Reply
Reply
post #89 of 290
1/2/07 at 10:33pm
Thread Starter 
Quote:
Originally Posted by HDholic View Post

Yes, it would be nice it the "special patterns" had a brief explanation on how they work or what to look for.

In my case, what do I look for in the judder and Y/C delay patterns?

The judder patterns serve two purposes:

1) Shows how smoothly a software decoder can render frames. Any hiccup, pause, stutter or jump is easily detected by watching the horizontal movement of the bar. It should be smooth as silk, since the movement is exactly the same number of pixels in each frame.

2) Deinterlacing functionality. The interlacejudder.ts and pulldownjudder.ts clips have field based motion, which should be eliminated by a deinterlacer. That is, the left and right hand edges of the bar should not be fuzzy.

The Y/C delay pattern is used to estimate (by eye) how much Y/C delay your decoder/display chain may be suffering. The middle row of bars have perfect Y/C alignment, while other rows are offset 1, 2 or 3 pixels in each direction. If the middle row looks to be aligned properly, you're in good shape. If one of the other rows seems better aligned, you're off by 1, 2 or 3 pixels.

Ron
Reply
Reply
post #90 of 290
1/2/07 at 10:47pm
Thanks dr1394. Using VLC media player to test on my PC LCD (1440x900 native res.) the white box edges where blurry on the judder test. Blue and green Y/C delay seemed to be about 1 pixel off.

Do you have or could a pattern to test motion blur on LCD HDTVs be made for that purpose? Also, is it possible to test how many lines of res. a display is processing at any input resolution?
Reply
Reply
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Display Calibration

AVS Top Picks

AVS › AVS Forum › Display Devices › Display Calibration › HD MPEG-2 Test Patterns