AVS Forum banner
Status
Not open for further replies.
1 - 12 of 12 Posts

·
Registered
Joined
·
3,522 Posts
Discussion Starter · #1 ·
Pardon my status as a creator of video test patterns, but I'd like to air a concern I have regarding use of downconversion from 1080P sources to produce other resolution test patterns. Doing so can produce results which differ significantly from material synthesized directly in the target resolution. The problems are particulary accute when source and output resolutions are not multiples of each other. This situation is present if a 1080P signal (1920 x 1080) is converted to 480P (720 x 480). Because non-integer relationships are involved, some problems arise. Things which can be synthesized directly at full resolution in 720 x 480 cannot always be represented as a 1920 x 1080 source. In other words, one can create test signal features directly in 720 x 480 which cannot be exactly recreated through downconversion from a 1080P source. Downconversion from 1080P is not necessarily the best way to create multiple test signal formats.


For this reason I always numerically synthesize test signals directly in the target resolution. Doing so allows direct control of image content, allows flatness of frequency response out to full nyquist limits, and maintains greater geometric monotonicity. Conversion across non-integer size factors introduces image distortions and limitations which I prefer to completely avoid through generation directly in the target resolution.
 

·
Registered
Joined
·
1,512 Posts
Do I sense a comment on DVE? :D


What you state makes sense to me. It's interresting to think, however, that a good number of DVDs are now being created from HD masters. Would you expect a DVD which is based on a transfer to 720x480 directly might show improvements in PQ?


Can you comment any on Avia Pro, and how that is created/mastered?


-Jon
 

·
Registered
Joined
·
3,522 Posts
Discussion Starter · #3 ·
I generated the AVIA Pro patterns directly in 480P (720 x 480) 24 fps 2/3 pull-down. The result is that the material is full amplitude all the way out to nyquist (when appropriate) and is high enough resolution to verify pixel for pixel mapping. There is advantage in avoiding downconversion. If we didn't generate in native resolution there are quite a few test features which would not have been possible. Multiburst and Multipulse signals in AVIA Pro are flat in response and in phase all the way out to 6.75 MHz for luma and 3.38 MHz for chroma. I couldn't have done that without native resolution synthesis. That wide bandwidth means that response losses seen on a scope will be due to the playback equipment being tested, not a mix of things intrinsic to the patterns themselves + the equipment.


You mention feature film DVD's being downconverted from HD. That's not a problematic because filtering down the bandwidth is desirable in that case. Otherwise you'd see line flicker and other artifacts. On the other hand, you don't want test signals themselves to be the factor which limits response. The test signals are used to check playback and display limits so you want them to exceed what the display system can normally handle. Otherwise, things like frequency losses or group delay problems can't be as easily teased out as being due to the equipment or a limit in the test signal. Feature film and test patterns differ in goal. One is meant to look good. The other is meant to stress the limits of equipment so the equipment can be improved (and make regular material look good).


Sometimes test material is too stressful. For instance, the sharpness pattern in the original AVIA is extremely demanding. Too demanding in fact. Very few display systems can show that pattern without any ringing. The transitions exceed what can be put over the air. Because of that, I've also added filtered tests in AVIA Pro (and S&V Home Theater Tuneup) which allow lesser, but still compliant display systems to respond appropriately. Providing multiple versions of tests allows nuances to be examined. It also means a lot of test patterns are needed. It took 5,000 test patterns/variations to achieve the flexibility I wanted for AVIA Pro. Casual users might not understand the power we'll be delivering, but experts will be able to examine more things than they've ever been allowed to using a test DVD.


I couldn't do it with the same accuracy if we generated in 720P or 1080P and then downconverted. We actually already have in house, the ability to generate in other resolutions. That means not only scaling but completely refomratting the content to match the desired resolution and then numerically synthesize directly in the target resolution. When we do create PAL or HD, they will be generated in the native resolution for each format. It takes more time and expense, but I believe that approach is better than creating a single set of masters from which to derive other signals. Not everybody saw the importance of those last bits of image detail in test patterns and there was pressure to build AVIA Pro as HD and then downconvert to multiple formats. I wasn't willing to take that compromise.
 

·
Registered
Joined
·
1,512 Posts
Guy,


There's a lot of good info in your post. Thanks for taking the time to respond.


-Jon
 

·
Registered
Joined
·
1,540 Posts
Thanks for the Info on the Sharpness test, i've had a hard time trying to tame the halos on my display! Versions of the sharpness test filtered at different frequencies (closer to the limit of typical dvds) would be great.
 

·
Hello, World!
Joined
·
2,523 Posts
Quote:
Originally posted by Guy Kuo
Casual users might not understand the power we'll be delivering, but experts will be able to examine more things than they've ever been allowed to using a test DVD.
This expert casual user is looking forward to getting hold of AVIA Pro once available :D


Guy, one of the biggest criticisms of DVE right now is the lack of documentation. Will AVIA Pro contain documentation (as the original AVIA did?) with the test patterns? And will it be designed, again like AVIA was, to give rapid access to test patterns and easy switching between them?


Mark.
 

·
Registered
Joined
·
3,522 Posts
Discussion Starter · #7 ·
The manual is currently creating a horrible case of carpal tunnel syndrome. Yes, there is a printed manual. There is too much new material which has been seen before so we HAVE to explain it.


You'll be very pleased with navigation I think. There is still heirarchial menu access to the patterns so you don't have to memorize, just point and click to get to patterns. I've arranged them so the menus are more logically behaving than in AVIA. Commonly used groupings are together. You even get selectable triplets and quadruplets of patterns which loop amongst themselves. For instance, you select a quadruplet with the high and low intensity gray windows desired and you are then put into a pattern loop which advances between black, low, middle and hi windows with a single press of your remote. Makes jumping between your desired targeting points during grayscale calibration a snap.


Sometimes there are too many patterns to put in a single menu. The windows which are in 2.5 IRE increments are indexed in the menu as every other pattern, but you can go forward and backwards in the sequence at will. (Of course there are other sequences like every 5 IRE, 10 IRE, Automated by 2.5, automated by 5, automated by 10) The automated sequences are for use with Colorfacts for handsfree data gathering.


There are also subsets of patterns which act as suites suitable for display calibration. I've culled together a suite for fixed panel and another one for CRT's. The suites concentrate on the things appropriate to the type of display.


Finally, there is always title and chapter number navigation for those who realy like memorizing numbers.


re the sharpness test - That's already been done for the disc. Even the window patterns now have three types of edge transition severities to check for ringing. Many patterns have multiple use features built in. Staying with just the windows you'l always have on screen indication of which window is being displayed, black level indicators and white clipping indicators WHILE you are using the window patterns.
 

·
Registered
Joined
·
1,540 Posts
Will there be any specific deinterlacing tests (video, 2:2, 3:2, flagged, unflagged, bad edit, time to change from video to film and from film to video, PAL, NTSC deinterlacing etc)? How about mpeg decoder tests that could identify the chroma upsampling error and whether a chroma filter is being used in either film or video mode?
 

·
Registered
Joined
·
3,522 Posts
Discussion Starter · #10 ·
Quote:
Originally posted by cybersoga

Will there be any specific deinterlacing tests (video, 2:2, 3:2, flagged, unflagged, bad edit, time to change from video to film and from film to video, PAL, NTSC deinterlacing etc)? How about mpeg decoder tests that could identify the chroma upsampling error and whether a chroma filter is being used in either film or video mode? [/B

]



Deinterlacing testing is provided in a test sequence which runs through all 25 2/3 pull down film to film, 5 2/3 pull down film to video, and 5 video to 2/3 pull down film edit transitions. That covers both good and bad edits since we've basically run the edits points across the 35 types of transitions. The motion transitions sequence is not flagged. As such it will severely test the ability of equipment to actually detect and sync to 2/3 cadence. The test material also has some noise which tests how well the interlacer can ignore minor noise and still not lose lock. You'll be able to count how many frames it takes for a deinterlacer to recover after each type of transition without using special instrumentation. It is however, in NTSC and does not cover 2/2 pulldown transitions. There is a specific moving zone plate for looking at 2/2 pull down at multiple motion rates and directions but that does not include edit transitions.



There are multiple vertical and horizontal chroma sweeps which are specially built for visual as well as instrumented response testing. The moving zone plate patterns (in 2/3 2/2 and video motion) also have chroma plates which make checking for chroma filtering very simple. Visually oriented chroma tests in AVIA Pro uniquely restrict chroma signal polarity so they are easier to see. Some specific chroma upscale error test patterns are provided, but currently only for progressive material being incorrectly upscaled as interlaced.


New high resolution Bowtie patterns now allow luma and chroma timing tests with double the normal time resolution were also built. The bowties in AVIA Pro are RGB legal. Prior bowties tests were unable to survive conversion to and from RGB color space. The ones in AVIA Pro are designed to survive such colorspace conversions.
 

·
Registered
Joined
·
1,540 Posts
I'm drooling along with Mark_H! Sounds great, even if I can't afford it (i'm assuming as it's a Pro product it's not aimed at consumers?) this should give us (and more importantly hardware reviewers) a good idea at how good a DVD player is at deinterlacing and chroma upsampling among other things, so we can make more informed purchasing decitions, and hardware manufacturers can aim to make dvd players that pass the tests so it's all good.
 

·
Registered
Joined
·
4,525 Posts
This really should be something. Guy was showing some material at CEDIA that was unlike anything I had seen before.


I can't wait to see what else is in store...
 
1 - 12 of 12 Posts
Status
Not open for further replies.
Top