I think we have a real new problem and it seems to be affecting Gennum VXP based scalers.
Silicon optix scalers are somewhat of a standard in
the Broadcast industry and their Benchmark test
material is considered a true "benchmark" for video
processor manufacturers. Gennum VXP processors do
not pass the "Film Resolution Loss Test" off the NEW
Silicon Optix HQV HD Benchmark Evaluation materual
from either a Blu-Ray player or HD-DVD player when fed
with the SMPTE 133 1080i testpattern.
I had a suspicion all along why HD content (HD-DVD &
BluRay) suddenly seemed to have a soft "touch" when I
discontinued to use my previous scaler (Silicon Optix
Whithorse aka NEC Theatersync) and started to do 1080i
to 1080p deinterlacing a Gennum VXP processor
(e.g. Anthem AVM-50 or JVC RS-1 projector).
The specialized (24p) SMPTE 133 test
pattern painfully reveals the problem. I did an ISF
calibration of an RS1 last week and could not quite
figure out what was going on with the de-interlacing.
Now with the HQV Benchmark disc this suddenly makes
sense. The Gennum engineers better get back to the
drawing board to sort that problem out quickly and
pass the fix on to the OEM manufacturers.

Helmut

What's the test? How is pass/fail measured? What percent or whatever is the failure? A loss of high frequency above what?

Mark,

The test pattern is a typical frequency test pattern which contains single pixel vertical information and a rotating bar. The entire pattern also pans around. If the test passes you can see the single pixel on/off lines - if it fails, these are bobbed into solid blocks of white or black.

Cheers,

Mark

The test pattern is a typical frequency test pattern which contains single pixel vertical information and a rotating bar. The entire pattern also pans around.Is it a film mode or video mode test? I.e., is there a repeating cadence like 3:2 pulldown used or is there motion between every field pair?

- Dale Adams

Both separate film and video tests are on the disc.

Mark

I believe to remember that a Gennum guy said that the Gennum fails on some 1080i60 test patterns, but not so on real life material. It's some time ago, though, so I'm not fully sure my memory is still intact... :eek:

I believe to remember that a Gennum guy said that the Gennum fails on some 1080i60 test patterns, but not so on real life material. It's some time ago, though, so I'm not fully sure my memory is still intact... :eek:


I guess Melgon wouldn't notice a softer pic in movies then.

So, there may actually be a problem with the Gennum (or its firmware).

Interestingly, while it fails the film resolution test, it seems to process 3/2 pulldown correctly and it passes the video resolution test.

Don't know what's going on here.

Is it a film mode or video mode test? I.e., is there a repeating cadence like 3:2 pulldown used or is there motion between every field pair?

- Dale Adams

For the "Film Resolution Loss Test", it's the panning of a SMPTE chart. Base on documentation it's supposed to be a straight 3:2 cadence. "The SMPTE chart is recorded 24p and transfered 60i". It failed this.

However, right after this there is another "Film Resolution Test - Stadium", which is supposed to be a "real life" 3:2 sequence, and it passed this. (The HQV doc said if a processor failed the one above, it will fail this also. Well it doesn't for Gennum)

So whatever the "Gennum guy" said may be correct: "it failed some 1080i60 test patterns", but pass on "real life" samples. :rolleyes:

I think the Gennum is probably doing processing by regions and is having trouble identifying the 3:2 cadence of the SMPTE chart at the corner, and is doing video mode there. Just a guess.

Helmut posted the response that he got from Gennum here: http://www.avsforum.com/avs-vb/showthread.php?p=10819148

One can always create a test pattern or sequence which can be
demonstrated
to break someone else's processing!

While what he says is true in general, wasn't it specifically the SMPTE 133 1080i testpattern that the Gennum failed?

That's hardly a testpattern "someone" created to show that another video processor fails.

While what he says is true in general, wasn't it specifically the SMPTE 133 1080i testpattern that the Gennum failed?

That's hardly a testpattern "someone" created to show that another video processor fails.


I'm currently looking at the SMPTE 133 1920x1080 test pat from the DVE HD DVD @ 1080i into a Gennum based PJ. At each corner and also in the centre of the screen there are groups of 12 squares that contain horizonal and vertical lines at different freqs. Non of the squares is solid black or white and each group has squares that contrain 1 pixel wide alternating lines.

Should I expect to see something different ?

Dazzer

That suggests your scaler is working for video based material. The HQV test discs also contain the same patterns but in film mode which requires a different technique to deinterlace correctly.

That suggests your scaler is working for video based material. The HQV test discs also contain the same patterns but in film mode which requires a different technique to deinterlace correctly.

Thanks Mark.

Dazzer

A moving stack of single pixel (horizontal) lines is a real torture test.

A de-interlacer that puts great faith in cadence matching should pass such a film de-interlacing test. Once every 5 fields it will size up matching current field and the field two beyond which are the first and third of a threesome. The de-interlacer will lock into film mode. The other four situations are: second if a threesome and first of the next pair, third of the threesome and second of the next pair, first of a pair and first of the next threesome, and second of a pair and second of the next threesome. Because the stack test pattern is moving, all of these other four situations do not correlate (a sameness test does not recognize it as stationary subject matter) and a motion adaptive de-interlacer in video mode will and should bob them.

Even-even or odd-odd video mode correlation (a field and the field two beyond) will also match up once every five fields If it assumes the subject is stationary through those three fields (including the one in the middle that is) the de-interlacer in video mode will not bob out the detail some of the time.

Odd-even or even-odd video mode correlation (two consecutive fields) will never match up.

Finally, a de-interlacer that, after the other processing is done, specifically looks for combing which a stack of alternating one pixel high horizontal lines resembles, will bob away this detail. Processors that handle film mode with superimposed live video are likely to run into stumbling blocks here.

Even pixel by pixel de-interlacers are not immune from these problems.

Video hints: http://members.aol.com/ajaynejr/viddoubl.htm

is there already a HDDVD or BluRay Disc of this HQV test disc ?
If not I would only perform tests with a DVD player that can truly output 480i.
If you use a HDDVD or BluRay unit that upconverts the 480i to 1080i you are also testing this upconversion not being able to tell who is responsible for the problem.
If you have a DVD player with the Gennum chip installed then I may be corrected.

Spatz,

The HD-DVD and Blu-Ray versions are available to order.

http://www.avsforum.com/avs-vb/showthread.php?t=855219

Mark

That suggests your scaler is working for video based material. The HQV test discs also contain the same patterns but in film mode which requires a different technique to deinterlace correctly.

Mark

If the 1080p test pat is encoded at 24p does this not mean it will be treated as film ?

Quote from DVE -

"High definition content is provided in both 1080p and 720p. All of the 1080p content is at the 24 frame rate while 720p materials are either 24 frames or 60 frames per second."

Dazzer

Dazzer,

Hmmm, good catch, I didn't spot that note... so the 1080 patterns seem to be film mode after all... and the 720p patterns could be used to test video processing.

Mark

"High definition content is provided in both 1080p and 720p. All of the 1080p content is at the 24 frame rate while 720p materials are either 24 frames or 60 frames per second."

"frames per second", meaning progressive source. "video" deinterlace is for 60i interlaced source.

regards,

Li On

www.pixelmagicsystems.com

"frames per second", meaning progressive source. "video" deinterlace is for 60i interlaced source.

regards,

Li On

www.pixelmagicsystems.com

Li On

How does this apply to DVE HD DVD and HQV HD DVD tests in question ?

Dazzer

Li On

How does this apply to DVE HD DVD and HQV HD DVD tests in question ?It applies because the tests are 1080i, not 1080p. There is no need to process a 1080p carrier so there is no point in testing that.

The HD HQV disk doesn't really tell you much about how your display will do with Blu-ray movies, unless your display is limited to 1080i input. Rather, it is a test to show how well your display or display processor does with other sources, namely high-definition television content.

Among other things, the HD HQV disk tests whether a video processor can 1) correctly perform motion-adaptive deinterlace on 1080i60 video sources, and 2) can correctly perform film-deinterlace -- otherwise known as inverse telecine -- on a 1080p24 source within a 1080i60 carrier. Basically, it tests whether your display processor can tell the difference between video (such as sports) and film sources (such as series and movies) on television, and deinterlace each appropriately.

Take the example of HBO. HBO sends its series and movies as 1080p24 within a 1080i60 carrier. The HD HQV test disk will tell you whether your current display or video processor is able to detect the 1080p24 source within HBO's 1080i60 signal and display it without detail or resolution loss. The overwhelming majority of video processors in consumer displays cannot.

Most newer displays treat everything as a 1080i60 video source, bobbing and weaving as appropriate (i.e. guessing) to fill in the missing information necessary to create a 1080p signal. This is the right approach for 1080i60 video, but it is the wrong approach for 1080p24 content, where all the information for the progressive image already exists in the signal. Bobbing and weaving a 1080p24 source throws away resolution and picture detail, yet 99% of consumer displays do it [or worse].

Take the example of HBO. HBO sends its series and movies as 1080p24 within a 1080i60 carrier. The HD HQV test disk will tell you whether your current display or video processor is able to detect the 1080p24 source within HBO's 1080i60 signal and display it without detail or resolution loss. The overwhelming majority of video processors in consumer displays cannot.

Most newer displays treat everything as a 1080i60 video source, bobbing and weaving as appropriate (i.e. guessing) to fill in the missing information necessary to create a 1080p signal. This is the right approach for 1080i60 video, but it is the wrong approach for 1080p24 content, where all the information for the progressive image already exists in the signal. Bobbing and weaving a 1080p24 source throws away resolution and picture detail, yet 99% of consumer displays do it [or worse].

I assume the same applies to HD DVD film material output at 1080i/60 ??

Dazzer

I assume the same applies to HD DVD film material output at 1080i/60 ??Yes.

The HD HQV disk doesn't really tell you much about how your display will do with Blu-ray movies, unless your display is limited to 1080i input. Rather, it is a test to show how well your display or display processor does with other sources, namely high-definition television content.

I fully understand that. The disc is for testing display device (or video processor) deinterlace performance when feed with a 1080i signal from a HD player.

But some seems to compare the HD player self 1080p output quality which really does not apply if the disc itself is encoded in progressive mode because in that case the HD player can access the disc native format and simply output the correct progressive frame as is and no deinterlacing is performed inside the HD player.

regards,

Li On

www.pixelmagicsystems.com

I fully understand that. The disc is for testing display device (or video processor) deinterlace performance when feed with a 1080i signal from a HD player.

But some seems to compare the HD player self 1080p output quality which really does not apply if the disc itself is encoded in progressive mode because in that case the HD player can access the disc native format and simply output the correct progressive frame as is and no deinterlacing is performed inside the HD player.

regards,

Li On

www.pixelmagicsystems.com

Except that most HD players right now *are* interlacing internally and then deinterlacing for 1080p output. Which means the HQV discs *are* relevant tests for these players...

Mark

I fully understand that. The disc is for testing display device (or video processor) deinterlace performance when feed with a 1080i signal from a HD player.

But some seems to compare the HD player self 1080p output quality which really does not apply if the disc itself is encoded in progressive mode because in that case the HD player can access the disc native format and simply output the correct progressive frame as is and no deinterlacing is performed inside the HD player.

regards,

Li On

www.pixelmagicsystems.com

Li On

In my test I'm using a Tosh A1 that outputs a 1080i signal. That said I believe the current Tosh XA2/XE1 uses the Reon to deinterlace the 1080i signal first to generate 1080p i.e. not direct from disc.

According to another thread (http://www.avsforum.com/avs-vb/printthread.php?t=832465) the XA2 using Reon shows moire artefact in the HD DVD MI3 Vatican wall scene at 1080p on a broad range of displays. The Gennum based display doesn't suffer from this moire artefact when fed the same disc at 1080i.

Dazzer