Gary and Ofer (and other Beta Testers) - What you have is the final version of the ABT102. If you would like to make comments, then go right ahead.
Following is a repost from AVforums.com (the British equivalent of this AVSforums.com) where I posted my structured comments about the final version of ABT102.

I've received a number of requests to repost it over here, so here is it.

I've taken this opportunity to make it less PAL/50Hz specific; and not overuse the phase “simply awesome” too. ;)

Introduction

This is a review of Anchor Bay Technology’s Precision Deinterlacing Card for iScan VP30, i.e. the ABT102. This new daughterboard adds high quality SD deinterlacing to VP30, using part of Dale Adams’ new DL algorithms.

My daughterboard is beta 5, but Josh Allen says that’s final version. The firmware is the same current 1.07 (28-Apr) beta that is available for download from DVDO.com (http://www.dvdo.com).

Installation

The retail version of ABT102 will come with five items: Precision Deinterlacing Card, Stand-offs (2), Software CD-ROM, ABT Reference DVD, manual (which may be printout or PDF on CD-ROM).

After unpacking the ABT102 the next step is to simply remove the top from VP30, which is achieved by removing nine screws, and then locate the Expansion Module Connector socket for the ABT102. Next to the socket are two holes which fit the stand-offs, these are simply pushed in; they are very similar to SDI module ones. Once stand-offs are in place the ABT102 card can be mounted, it fits into socket and both stand-offs go through the two holes in the card. One has to assure a tight fit, but not pressing so hard as to break anything. Once the top cover is reinstalled, the firmware must be updated (from CD-ROM) for ABT102 to operate; update involves the usual RS232 malarkey.

All in all, a pretty easy upgrade that even the most technologically challenged should be able to perform with no problems.

New Features

Once ABT102 and new software are installed the info screen is now headed “VP30 – ABT102” and under the “Input Adjust” menu the “Film Mode” menu is replaced with a new one entitled “Deinterlacing”. This new menu is where the ABT102 is controlled. There are three main areas (my categories, not DVDOs) that are: ‘Auto’ and ‘Film Bias Mode’ and ‘Video Mode’ that are for general use; ‘2:2 Even’ and ‘2:2 Odd’ which are specific to PAL film content; ‘Game Mode 1’ and ‘Game Mode 2’ which are for game consoles.

From DVDO.com website:

New modes available under ‘Input Adjust —>Deinterlacing’ are:
Auto – This mode is the default.
‘Auto’ represents the best balance between automatic detection of film and video sources, bad edit detection, and identification of mixed-mode sources. This mode should be used when the content may be a mix of film and video content or you are not sure.
Film Bias Mode – This mode is intended for use on content that is known to be film-based.
Video Mode - This mode is intended for use on content that is known to be video-based.
2:2 Even – This mode should be used when the user knows that the source is high-quality 2:2 pulldown (i.e. film-based content played back in a country with a 50Hz video standard) and wants to avoid any loss of cadence lock while watching that source. This mode weaves two adjacent fields together starting with an even field and combining it with the following odd field. This will provide a higher quality overall signal than the ‘Auto’ or ‘Film Mode’ settings, providing that the source really is 2:2 pulldown and does not have bad edits. Only one of the ‘2:2’ Deinterlacing settings is correct for any given source and the correct mode can be chosen by simply trying both of them and selecting the one which does not result in combing artifacts.
2:2 Odd – This mode is very similar to ‘2:2 Even’ except that that this weaves two adjacent fields together starting with an odd field and combining it with the following even field.
Game Mode 1 - This mode is intended for use with game consoles (like those from Sony, Microsoft and Nintendo).This mode gives you minimal latency with edge-adaptive processing. The total amount of delay with source-locked output mode set on the VP30 is about half a frame of delay. Unlocked frame rates will increase this delay.
Game Mode 2 - This mode is intended for use with game consoles (like those from Sony, Microsoft and Nintendo). This mode gives you minimal latency with both motion and edge-adaptive processing. The total amount of delay with source-locked output mode set on the VP30 is about one and a half frames of delay. Unlocked frame rates will increase this delay.

Deinterlacing film content

The new arbitrary cadence detection for 3:2 pulldown and 2:2 pulldown is truly excellent and very rarely misses. In ‘Auto’ it is extremely hard to find anything (even with very difficult PAL 2:2 pulldown) that does not lock.

However, if you do have a PAL film source that doesn’t lock, you can simply use either ‘2:2 Even’ or ‘2:2 Odd’ to ensure solid lock throughout a movie.

Take difficult content such as “Star Trek: Insurrection” (beginning of movie), or PAL “Bridget Jones: Edge of Reason” (Coca-Cola sign saying “GO BRIDGET GO”), or PAL “Day After Tomorrow” (opening), or any of the dozens of movies R1 and R2 that I’ve run through and not a bad lock or moiré in sight in ‘Auto’ mode.

Yep, that good.

Deinterlacing video content

Improving the video deinterlacing is the raison d’être of ABT102, and to say that it’s in a different league to SiI504 is understatement of the year. To put it simply, video deinterlacing is astonishing -- ABT102 makes broadcast SD 576i video content look like 720p, and 480i video content looks just as good (caveat: although I’ve only seen DVD based 480i video and not broadcast).

ABT102 deinterlaced video has the picture quality of film, but with the smooth motion of video.

With broadcast SD digital television high bit-rate channels look stunning (as do DVDs), but amazingly even low bit-rate channels look superb; I’ve gone and looked at some really terrible PQ channels, but they are still pleasingly rendered. Broadcast SD television has never looked so good on a progressive display; I seek out video material just to see how great it looks: music videos, even commercials, are so much better than before.

To test ABT102 fully I even viewed sport! Soccer looked fantastic, great detail (crowds and nets), no jaggies on diagonal lines on pitch, no combing, very good PQ.

Cartoons are particularly splendid, ‘The Simpsons’ has never looked as good nor ‘Southpark’ or any other animated material I’ve tried. The Edge Adaptive Processing seems to work especially well with the solid colours and black lines of animation.

The Edge Adaptive Processing (gets rid of jaggies by smoothing edges) does equally well with non-animated video. ABT102 blows Faroudja DCDi out of the water – and I have yet to see Gennum VXP and Realta HQV but I don’t think they’re going to be significantly different to ABT102 on SD material.

As to cadence locks, combing is extremely rare with ABT102’s arbitrary cadence detection; especially as it can deinterlace different parts of the screen as film/video. This means that difficult challenges such as video-titles-over-film are not even an issue. Get used to credits and titles being super-smooth with no combing; vertically or horizontally scrolling text looks smooth and sharp and completely judder free, no matter if on solid color background or over film/video.

Most of the time it looks like video content just got a free upgrade to 720p.

Deinterlacing video game content

I haven’t has as much chance to test this with my GameCube as I would like, however, it all works as expected and has no noticeable delay. Period.

If you are a game player and need a video processor, this is the one for you.

Tests

Specific tests that I haven’t already discussed above…

HQV Benchmark DVD
HQV disc is completely and utterly beaten by ABT102.
Color Bar : perfect pass
Jaggies Pattern 1 : excellent, no jaggies even at low angles
Jaggies Pattern 2 : excellent, laughs in the face of this test
Flag : excellent
Detail : excellent, no detail lost at all
Film Detail : excellent, no moiré
Assorted Cadences : excellent, didn't miss one
Mixed 3:2 with titles : excellent, smooth video text over film

The ABT102 doesn't have noise reduction built in so:
Noise Reduction : fail
Motion Adaptative Noise Reduction : fail
However, both NR tests look great ... but I do not believe that cleaning up analogue broadcast artifacts and mosquito noise are appropriate test for deinterlacing.

Digital Video Essentials
Test pattern that contains video mixed with film. ABT102 laughs at it.

Difficult DVDs
I’ve tried all the ones that people suggested to me on AVforums.com and I can get my hand on, and the ones I’ve played through ABT102 haven’t had any problems.

Summary
ABT102 could not have passed the tests with more flying colors, it really is superb.

Conclusion

I have yet to see Gennum VXP and Realta HQV but I am willing to bet that ABT102 gives them a run for money, at fraction of price.

This upgrade is simply outstanding: Film content is improved because of awesome cadence detection.

Video content is in different league because of cadence detection and diagonal edge anti-aliasing.

Game content is supported like never before with two modes. The ABT102 transforms the iScan VP30 into the best-in-class SD deinterlacer, for very little money. You’re going to have to spend thousands of dollars more to see comparable results from other manufacturers. Could not be more highly recommended.

:) :) :)

StooMonster

I want! :D

Great review, Stoo. Much appreciated. :)

Wow....I am impressed. Thank you for posting this. The Sil504 was the only reason I hadn't been looking at the VP30 for my future scaler needs. This definately changes things.

Wow twice!!

Thanx StooMonster for a great review. :) :)

So let me see if I understand the implications of this new board:

The only weakness of the VP30 was in its video deinterlacing, right?
And with this new board the video deinterlacing is at least up HQV or Gennum quality, right?
So a VP30 equipped with this board would all around be equal or better than any of the current HQV or Gennum units, but at a much lower price point, right?

What is the MSRP of an ABT102 equipped VP30?

Bob,

As a fellow Ruby owner and a VP30 owner the VP30 w/ ABT102 should really help SD video and film big time since the internal deinterlacer/scaler for SD in the Ruby is fairly weak. The MSRP is $1999 for the VP30 and currently $199 for the ABT102 (until May 15th then $499).

Now the only weakness for the Vp30 is the HD deinterlacing. Can't wait for ABT's HD offerings.

Thanks, Andy!

Currently I own a Vantage-HD, but I still have HDMI/HDCP -> DVI connectivity issues, so I am beginning to look for alternatives. This new addition to the VP30 has naturally caught my attention. ;)

Thanks, Andy!

Currently I own a Vantage-HD, but I still have HDMI/HDCP -> DVI connectivity issues, so I am beginning to look for alternatives. This new addition to the VP30 has naturally caught my attention. ;)

remember this only works on SD inputs, the next dvdo scaler will also do it for HD. The vantage also does it for HD. How much you will miss the new deinterlacer for HD is personal.

Daniel.

Thanks, Andy!

Currently I own a Vantage-HD, but I still have HDMI/HDCP -> DVI connectivity issues, so I am beginning to look for alternatives. This new addition to the VP30 has naturally caught my attention. ;)

As much as I love my VP30 if your problems are HDMI/HDCP, you should know that others have reported HDMI audio issues. I have not heard much re: HDMI video issues though. Just thought you should know. Great thing about DVDO is you could get a 30 day trial.

On video mode material, is there a sense that the image is bobbing up and down, or is it rock solid and stable. This is one of the great failings of DCDi, so hopefully Dale has been able to correct for it.

The only weakness of the VP30 was in its video deinterlacing, right?


I'd also add that its user interface is a bit clunky, compared to a Lumagen or your Lexicon MC-12. If you order the ABT102 before May 15, you can get it for $199, otherwise it's $499.

--Andre

The SiI504 also has problems with getting false positives with 2:2 material, so the advantages aren't limited to video.

How does the VP30 handle 1080i? If I wanted to deinterlace and scale 1080i to 720p, does the ABT102 help at all?

ABT102 is only for SD. The next major DVDO release is slated to be HD-oriented (i.e., beyond bob/weave).

How does the VP30 handle 1080i? If I wanted to deinterlace and scale 1080i to 720p, does the ABT102 help at all?


As I understand it, the VP30 will scale 1080i to 720p.
However, the ABT102 will de-interlace SD only (not 1080i).
But I could be wrong.

As I understand it, the VP30 will scale 1080i to 720p.
However, the ABT102 will de-interlace SD only (not 1080i).
But I could be wrong.

The vp30 does work for HD signals but the way it does it is far from perfect. The end result will be pleasing for many but there are other products that will do it alot better. The 102 card brings the important area of SD signals into a new realm and leaves the HD area as it was before good but not perfect. Perfect deinterlacing and scaling on HD comes at a price that is higher than the vp30 is today. In the near future dvdo will release a new scaler based on the same concepts it now uses for SD but extended to HD how much extra that will cost is unsure but these new inovations will be used.

The DL102 can be seen as a showcase for what they can do but its limited in its deinterlacing to SD only because of the design of the vp30.

As a user of several of their products i also would liked to see the dl102 card do HD but i am happy for the upgrade they do to current users and when the HD version is released we can expect a good upgrade path from dvdo if history is any guide.

The parent company of dvdo is http://www.anchorbaytech.com/ it also has alot of information on the things they design (and then showcase in their dvdo products) and that others can license.


Daniel.

^^ Probably been asked a billion times now, but what is the really ballpark timeframe estimate on this "next DVDO product?" ;)

^^ Probably been asked a billion times now, but what is the really ballpark timeframe estimate on this "next DVDO product?" ;)

doubt anyone who really knows will tell you :), we do know that they have shown working boards for HD for all and more of the things found on the dl102 at tradeshows.

My guess is we will see products announcements near or at the bigger shows in september but remember they have a great upgrade path with a proven track record so investing in a vp30+dl and then a trade in might be closer to the price of this new model
than you think. Ask dvdo if you want if they will do a nice upgrade path again unlike timeframe and specs they might respond to that one ....

Daniel.

^^ Probably been asked a billion times now, but what is the really ballpark timeframe estimate on this "next DVDO product?" ;)
Rumours say announcement at Cedia 2006 (mid September) with shipping to follow right afterwards.

Danielo and madshi: That's fine by me. Certainly no one wants them to rush something like this, but I, too, have heard "Keep an eye out on CEDIA 2006" more than once. ;)

I do know that DVDO/ABT has an EXCELLENT upgrade path and all that. :)

Why are we guessing September? BECAUSE that is when the BIG show (CEDIA) is. Exhibiting at a show is a hugh expense. A company does get to present new products to its dealers and gets a lot of free press. Soos you want to maximize things by announcing at the show etc etc. Not 30 days later. Delivery later is OK Soos we are guessing product sometime thereafter. Could it be months later? Sure could. But the dynamics dictate showing at least a prototype and saying it will be on the shelves soon.

The parent company of dvdo is http://www.anchorbaytech.com/ it also has alot of information on the things they design (and then showcase in their dvdo products) and that others can license.
So "VRS" will be their competetor against "HQV" and "VXP". Hurray for three letter abbreviations.

From the ABT homepage:

Anchor Bay will be rolling out other VRS solutions, such as detail enhancement and MPEG noise reduction, in both their semiconductor and DVDO brand systems throughout 2006

Hopefully this will be included in the "next DVDO product"!

Hopefully this will be included in the "next DVDO product"!
One would assume so. Big FPGAs cost a fortune, so the sooner they get the stuff onto a application specific chip, the better it'll be for their bottom line.

Unless I am missing something, the linked press release dated 5/6/06 while mentioning edge enhancement and noise reduction as roll outs for later this year, makes no mention of improved (motion adaptive) deinterlacing of 1080i this year.

Unless I am missing something, the linked press release dated 5/6/06 while mentioning edge enhancement and noise reduction as roll outs for later this year, no mention is made of improved (motion adaptive) deinterlacing of 1080i this year.
Maybe. But the WideScreenReview VP30 article said this:

"Anchor Bay Technologies plans to introduce its own 1080i inverse-telecine and motion-adaptive deinterlacing in a new DVDO video processor later this year."

Surely WSR wouldn't claim such a thing without confirmation from DVDO/ABT?

Would like to add that ABT's VRS "Evaluation & Optimization DVD" is superb.

Who made it?
Stacey Spears and Don Munsil and team have done a fantastic job with this DVD, big pat on the back / high five (dpending on your cultural preference) from here!

What's on it?
DVD is in four sections: Static Paterns, Motion Paterns, Demo Loop, About.

Static Paterns contains: Picture Controls, Geometry, Gray Scale, Half Patterns. All of these contain a number of test patterns to help calibrate a video processor.

Motion Paterns contains: Chroma Upsampling, Source Adaptive, Edge Adaptive, Mixed Mode, Montage. All of these sections contain a number of different items to test the capbilities of a video processor.

How is it?
Not only are the test patterns extremely well thought out, no redundancy or pointless duplications, they are very well made and of the highest quality. Futhermore, the menu system is the best structured of any calibration / benchmark DVD period.

This will be the new benchmark DVD, it's got better and more comprehensive tests than the HQV one and it's really easy to use.

Top marks (again) to DVDO team at ABT.

:) :) :)

StooMonster

Static Paterns contains: Picture Controls, Geometry, Gray Scale, Half Patterns. All of these contain a number of test patterns to help calibrate a video processor.

I could not think of a better name than "half patterns". These map to the patterns in the VP30 where half of the image is transparent. The idea is you would pull up that pattern on the disc and then go to it on the VP30. This will allow you to verify color space conversion, as an example. The bars may not line up exactly due to pixel cropping in a player, but the colors should match.

I need to explain the bad edit tests located under source adaptive. It contains the following bad edits:

# ...AABBBCCDDD-BBBCCDDDAA... (...3-2-3-|3-2...)
# ...AABBBCCDDD-DDAABBBCCD... (...3-2-3-|2-2-3-2...)
#
# ...BCCDDDAABB-BBBCCDDDAA... (...3-2-2-|3-2...) *
# ...BCCDDDAABB-AABBBCCDDD... (...3-2-2-|2-3-2...) *
# ...BCCDDDAABB-DDAABBBCCD... (...3-2-2-|2-2-3-2...)
#
# ...BBBCCDDDAA-AABBBCCDDD... (...3-2-|2-3-2...)
# ...BBBCCDDDAA-DDAABBBCCD... (...3-2-|2-2-3-2...)

These are 7 of the 25 possible bad edits with 2-3. These are the only 7 that do not contain a single field. The purpose of this test is to see if a processor can track across an edit w/o dropping to video.

The first time you see the text, this means the next edit, or loop, will contain that edit. At that point the text color will invert but remain the same. This gives you time to read what the edit was, incase it dropped out of film lock. The next edit, however, will actually be clean. So you have bad, clean, bad, clean, etc...

If a processor drops lock, it should only occur on the first of each. If it occurs on the clean edits too, that is even worse. :)

The montage was edited by hand to ensure that almost every cut is a bad edit. The cross fades are usually 2-3 to 2-2, 2-2 to 2-3 or 2-3 to 2-3 where one 2-3 is out of phase with the other.

The montage is more enjoyable with audio. :)

I could not think of a better name than "half patterns". These map to the patterns in the VP30 where half of the image is transparent. The idea is you would pull up that pattern on the disc and then go to it on the VP30. This will allow you to verify color space conversion, as an example. The bars may not line up exactly due to pixel cropping in a player, but the colors should match.
These are a great idea, couldn't believe I'd forgot to mention them in my post above.

edited by hand
DVD certainly oozes quality and shows the hard work to get it just right; as per above, nice job!

StooMonster

hi sspears: do you plan to have some form of simple documentation for the dvd ?

(For the moment I'm trying to understand the chroma ICP clips.... )

Would like to add that ABT's VRS "Evaluation & Optimization DVD" is superb.

Who made it?
Stacey Spears and Don Munsil and team have done a fantastic job with this DVD, big pat on the back / high five (dpending on your cultural preference) from here!

What's on it?
DVD is in four sections: Static Paterns, Motion Paterns, Demo Loop, About.

Static Paterns contains: Picture Controls, Geometry, Gray Scale, Half Patterns. All of these contain a number of test patterns to help calibrate a video processor.

Motion Paterns contains: Chroma Upsampling, Source Adaptive, Edge Adaptive, Mixed Mode, Montage. All of these sections contain a number of different items to test the capbilities of a video processor.

How is it?
Not only are the test patterns extremely well thought out, no redundancy or pointless duplications, they are very well made and of the highest quality. Futhermore, the menu system is the best structured of any calibration / benchmark DVD period.

This will be the new benchmark DVD, it's got better and more comprehensive tests than the HQV one and it's really easy to use.

Top marks (again) to DVDO team at ABT.

:) :) :)

StooMonster

Maybe I missed it, but: Where, when, and how can the rest of us get this DVD you refer to?

The DVD is included with the purchase of the ABT102.

I hope the dvd will be offered separately as well, around say $15 would be great. I'd pick one up.

There are multiple edge adaptive clips. These show a processors ability to deal with jaggies.

Aside from the synthetic clip, there is ropes, ship and hockey. This clips show was sets apart the ABT102 apart from the HQV and the Gennum.

In ships there are a pair of horizontal and vertical ropes that intersect. The Gennum and HQV shows jaggies across the ropes. The ABT102 only shows jaggies at the intersection of the ropes.

With hockey you should look at the markers on the ice. Again, the ABT102 maintains solid lines while the HQV and Gennum shows jaggies. The top of the glass is also another good location to look at.

The Gennum does the worst on jaggies. Take a look at the far right side of the ship. You want to look at the side of the ship.

The montage may actually be a bit boring if you watch it on the ABT102. You should take a look at it on something else, like the Toshiba HD DVD player when it is outputting 480p. It combs on almost every cut. There are also many markers in the montage that show jaggies when not in film mode. The coffee cups, the car hood, the round parts of the open piano, the lights on the motor cycle. You want to look at scene changes and those objects. The Gennum does unusually bad on the motorcycle near the end. There is an isolated portion of the image that combs on every cut.

The Denon 5910 does well on the cadence patterns, but the NEC Theatersync does poorly. It only locks on the 2-3 pattern on the high detail tests. It does fine on the diner clips. These are both using the HQV but the results are much different. The 5910 does a lot better than the NEC as well as the Yamaha with HQV.

The tracking ability of the ABT102 is something I like a lot. Play the high detail play all clip. This goes through each cadence pattern. It will repeate each one twice and then go on to the next. The AB102 drops on each 2-2 loop, but never drops after that.

hi sspears: do you plan to have some form of simple documentation for the dvd ?

(For the moment I'm trying to understand the chroma ICP clips.... )

I believe Josh and co are working on some docs. We did not have time to get them done prior to the disc shipping. The goal was to have on-screen examples like the HQV disc, but we ran out of time.

For a refresher on ICP and CUE, please read: http://www.hometheaterhifi.com/volume_8_2/dvd-benchmark-special-report-chroma-bug-4-2001.html

The chroma bug tests are both DVD player and processor tests. First you want to disable the CUE correction feature of the VP30 when running these tests. All DVD players suffer from ICP, some worse than others. DVD players should be able to pass the 2-2 and 2-3 tests. These tests should tell you if they suffer and how severe. There are four rows of squares. The 2nd row can be ignored. That is really there for processors to detect a cadence. You see processors only look at the luma channel when doing cadence detection. To show the chroma bug you want to remove luma. The chroma lines in motion, should not flicker and should be clean and straight. By straight I mean not jagged. :) If you see jaggies in the bottom two rows, you have the bug. Again, 2-2 and 2-3 should be clean while ICP will have jaggies.

On the processor side, you are testing to see if it has a chroma filter to mask thee artifacts. Once you have done the above, set CUE correction to on or auto. The bars should clean up some. If you have an RP56 or RP81, the ICP cleans up very well because it is minor. If you have a Pioneer, it does not clean up as well because the ICP is so severe.

I am interested in some feedback on the image cropping pattern. This is located under static patterns -> geometry. This is a bit different than AVIA. AVIA is 0 based while ours is 1 based. You simply go to the first number that is missing the end and that is how many are cropped.

The HD version is much more interesting because it includes test for accurate gamma, linear light scaling and 1:1 mapping.

There are four rows of squares. The 2nd row can be ignored. That is really there for processors to detect a cadence. You see processors only look at the luma channel when doing cadence detection. To show the chroma bug you want to remove luma. The chroma lines in motion, should not flicker and should be clean and straight. By straight I mean not jagged. :) If you see jaggies in the bottom two rows, you have the bug. Again, 2-2 and 2-3 should be clean while ICP will have jaggies.

Thanks! This is exactly what I'm looking for, i.e. how to identify


If you have a Pioneer, it does not clean up as well because the ICP is so severe.

I can confirm this... :(


The montage may actually be a bit boring if you watch it on the ABT102

I was abt to ask what to see on this since there is really nothing "unusual" to see while using the ABT102. :)

I should be able to have an opportunity to try this disk on a VantageHD. Will look at the jaggy clips.

I am interested in some feedback on the image cropping pattern. This is located under static patterns -> geometry. This is a bit different than AVIA. AVIA is 0 based while ours is 1 based. You simply go to the first number that is missing the end and that is how many are cropped.

The HD version is much more interesting because it includes test for accurate gamma, linear light scaling and 1:1 mapping.
HD version?! By this, do you mean that there will be a HD-DVD or Blu-Ray (or both :D ) edition of this test disc. If so, you just became my new best friend!

Dicey

Don and I are planning an HD test disc. We wish to cover a lot of area, mostly for 3rd parties to have difficult content to work with for product development. It will have the normal calibration patterns plus much more. No ETA at this time as we are still trying to scope the work and continue to build tools to create more tests.

Stacey,

Keep us TS mongers in mind...

HW and SW install in 25 minutes; very impressive results on SD cable TV. Haven't tried DVD yet - wife has TV proramme she wants to watch.

Question: does anyone know what is the purpose ( or future use) of the six pins at the top of the card? It looks like it is designed to accept an adapter of some sort.

I haven't seen the card yet, but if I had to guess, it's probably the interface to reprogram the ROM. BTW, some more information on the ABT102 can be found here, including the FPGA model used and the ROM used:

http://www.anchorbaytech.com/products/ABT102chip.html

--Andre

Andre,



Thanx for the link...what struck me, after reading the specs, was the outputting of 20-bit 4:2:2 YCbCr.

Can I assume this means that an 8-bit 4:2:2 YCbCr signal from a DVD player is processed into a 20-bit 4:2:2 YCbCr signal and sent out to the display?

Either that or it's a typo and should read "10-bit 4:2:2 YCbCr" ;)

StooMonster

The output has 10-bit resolution. The "20-bit" part of the spec just means that the luma and chroma are not multiplexed together into a single stream.

- Dale Adams

There are multiple edge adaptive clips. These show a processors ability to deal with jaggies.

Aside from the synthetic clip, there is ropes, ship and hockey. This clips show was sets apart the ABT102 apart from the HQV and the Gennum.

In ships there are a pair of horizontal and vertical ropes that intersect. The Gennum and HQV shows jaggies across the ropes. The ABT102 only shows jaggies at the intersection of the ropes.

With hockey you should look at the markers on the ice. Again, the ABT102 maintains solid lines while the HQV and Gennum shows jaggies. The top of the glass is also another good location to look at.

The Gennum does the worst on jaggies. Take a look at the far right side of the ship. You want to look at the side of the ship.

The montage may actually be a bit boring if you watch it on the ABT102. You should take a look at it on something else, like the Toshiba HD DVD player when it is outputting 480p. It combs on almost every cut. There are also many markers in the montage that show jaggies when not in film mode. The coffee cups, the car hood, the round parts of the open piano, the lights on the motor cycle. You want to look at scene changes and those objects. The Gennum does unusually bad on the motorcycle near the end. There is an isolated portion of the image that combs on every cut.

The Denon 5910 does well on the cadence patterns, but the NEC Theatersync does poorly. It only locks on the 2-3 pattern on the high detail tests. It does fine on the diner clips. These are both using the HQV but the results are much different. The 5910 does a lot better than the NEC as well as the Yamaha with HQV.

The tracking ability of the ABT102 is something I like a lot. Play the high detail play all clip. This goes through each cadence pattern. It will repeate each one twice and then go on to the next. The AB102 drops on each 2-2 loop, but never drops after that.

Hey Stacey

I remember the Gennum not doing as well with the HQV jaggies pattern but quite good with the hockey one. The boat was kind of hit or miss with everything, and almost a pick your poison as everything dropped out a little. What I liked about the Gennum is the fact that it does diagonal line processing with HD as well, as demonstrated when you came over that day. I haven't had the chance to test that with the HQV yet and the ABT doesn't do HD unfortunately but that will change soon with their new stuff.

Gennum had issues with extreme corners or end if I remember right, which is why the hockey scene probably looked good but not the HQV angles.

Kris,

you wrote:

The HQV solution still performed a bit better with mixed cadences, including those commonly found in foreign animation. It would also lock onto high detail images with mixed cadences a bit faster.
In what way did the HQV perform better with mixed cadences - apart from locking on a bit faster? Or was the fast locking on the only advantage? Your use of "also" sounds to me as if there was something else (apart from the faster locking on) where the HQV performed better.

Thank you! :)

Kris and Stacey,

did you try this on HQV yet? It would be very interesting to compare results with ABT102...

Noobs question...would this add on (ABT102) make the PQ close to HD (broadcast or HD-a1/x1)?

Noobs question...would this add on (ABT102) make the PQ close to HD (broadcast or HD-a1/x1)?

The ABT102 is just a deinterlacer. It converts 480i to 480p. The scaling from 480p to any higher resolution is still done by the same chip in the VP30 that was doing it before.

Thanks. Would you say that the VP30 output PQ is close to HD as advertised?

Kris,

you wrote:


In what way did the HQV perform better with mixed cadences - apart from locking on a bit faster? Or was the fast locking on the only advantage? Your use of "also" sounds to me as if there was something else (apart from the faster locking on) where the HQV performed better.

Thank you! :)

As I update the benchmark the differences will show. The HQV in the Denon did a bit better with some of the high detail test patterns that Stacey generated with mixed cadences. At some point I would like to generate a competition of sorts between the ABT-103, National Semiconducter, HQV, Reon, and VXP chips since they are all next generation chips. But I don't know when (or if) this could be arranged based on my travel schedule. I would use newer material such as the HQV and VRS test discs, some home brewed stuff and some of our older benchmark tests.

As I update the benchmark the differences will show. The HQV in the Denon did a bit better with some of the high detail test patterns that Stacey generated with mixed cadences. At some point I would like to generate a competition of sorts between the ABT-103, National Semiconducter, HQV, Reon, and VXP chips since they are all next generation chips. But I don't know when (or if) this could be arranged based on my travel schedule. I would use newer material such as the HQV and VRS test discs, some home brewed stuff and some of our older benchmark tests.
That sounds like a very nice plan to me. Nobody did this before for VP deinterlacing capability. I think it would really help.

Btw, what is ABT-103? Is that the name of the HD version of the ABT102? Or is it a newer revision of the SD version?

Also never heard of Reon yet! And also I didn't know that National Semiconducter had a next gen chip available. Seems I learn something new every day... :)

Thanks. Would you say that the VP30 output PQ is close to HD as advertised?

I would refer you to the VP30 thread for information on that.

The short answer is that SD is not HD. You can improve SD performance in a number of ways including better deinterlacing and scaling, but it will never be a match for true HD. The difference in real picture detail is very significant and cannot be artificially added.

That said, watching DVDs upscaled via the VP30 on a large screen can be very impressive.

Btw, what is ABT-103?

I believe Kris meant ABT102, given that there is no ABT-103, and that he is talking about using the VRS and HQV DVDs for testing.

The ABT-103 is a prototype bloody mary made with clam juice, grated horse radish, and kentucky moonshine, hold the tomato juice.

I think we need a deinterlacer/scaler special olympics. I suggest we hold it in Rio. BTW steroids are allowed if applied to any gated chip.

Mark,

Thats some funny poop!

Watched my first video last night with the abt102 installed. Overall very impressed with the quality of the picture. One strange symptom I did notice watching SD satilite is when there was what I guess in a error in the transmission (not that uncommon with their service) the abt really seems to have a hard time adjusting or ignoring the error. It's what I can best describe as a large color band in the effected area where as before without the abt102 you just had a small fragment on the screen. I'm not sure exactly how to describe as it's something new I've never experienced before.

Anyone else run into this?

I meant the 102, sorry about that! Typed too fast I guess!

I completely agree. It almost never breaks up, DVDO did a fantastic job on this unit.

Perhaps a dumb question: Is there somewhere in the VP30 user menus than can confirm that the ABT102 is installed and working?

Firmware 1.07 works with or without the ABT102. To confirm that the chip is actually seated and installed properly, are there certain user functions that would be grayed out if the chip weren't connected and the scaler was still using the Sil504?

Perhaps a dumb question: Is there somewhere in the VP30 user menus than can confirm that the ABT102 is installed and working?

Firmware 1.07 works with or without the ABT102. To confirm that the chip is actually seated and installed properly, are there certain user functions that would be grayed out if the chip weren't connected and the scaler was still using the Sil504?

The Info page will have ABT102 on it and the extra options under deinterlacing (I think) will be un-greyed, i.e. game mode etc.

So I'm told. :rolleyes:

It definitely says ABT102 on the top of the info screen.

It definitely says ABT102 on the top of the info screen.

But does it say that with this firmware even if the chip isn't installed?

dunno that, Joshua, but you will definitely have the whole gamut of deinterlacing choices if the 102 is properly installed. (and if it is not functioning properly, in my case the VP30 would not handle or deinterlace 480i at all.) But I installed V1.07 first - the deinterlacing menu option was there, but you couldn't choose anything. After the ABT was installed, and finally functioning, with an interlaced input, you can go to the input adjust (?), go to Deinterlacing - which is at the top, and choose one of the many options mentioned before.

But does it say that with this firmware even if the chip isn't installed?

The info screen will only say 'ABT102' at the top if the ABT102 is installed.

Thanks for the confirm. Just making sure.

I'm still wondering whether to call this new combo the VP30+ or the VP40, though.

I remember the Gennum not doing as well with the HQV jaggies pattern but quite good with the hockey one.

No, it performs much like the HQV on the Hockey. Take a look at the markers on the ice, they are jagged on the HQV and Gennum. They are clean on DCDi and the ABT102.

did you try this on HQV yet? It would be very interesting to compare results with ABT102...

I have tested multiple HQVs, Gennum, National and FLI2310. Not all HQVs are equal. The 5910 is the best performing HQV. The NEC and Yamaha perform a lot worse than the 5910. They will only lock on 2-2 and 2-3 high detail clips. They drop on the others. They also drop on the bad edits test. I tried both force film and auto.

I have tested multiple HQVs, Gennum, National and FLI2310. Not all HQVs are equal. The 5910 is the best performing HQV. The NEC and Yamaha perform a lot worse than the 5910. They will only lock on 2-2 and 2-3 high detail clips. They drop on the others. They also drop on the bad edits test. I tried both force film and auto.

I tried on VHD. For the high detail clips it locks on all, although not as quickly. All of them locked at the point when the car reaches the middle yellow seats. Does this mean anything ?

althogh for the synthetic wedges it fails on all of them. only the ABT102 maintains all the stright lines of the wedge except the tips, for all cadences. (also just for the kick tried on a lumagen hdp and no surprise it obviously failed everything)

Perhaps I'm doing something wrong: I don't notice any difference in the picture quality before and after installing the ABT102. I'm testing with a DVD player at 480i using the component input and while playing one of my DVDs.

Does the 102 card provide better resolution overall or does it only help with displaying scenes that contain motion of some sort, in better resloution?

The DVD that they send along with the 102 card does show the demo loop quite well and it seems the resolution is true 720P. However, while playing one of my dvds, I don't notice any difference. Please tell me that I'm doing something wrong. Yes, I did verify that it's installed properly - ABT102 appears next to the v1.07 in the info screen and I've confirmed the additonal options found under deinterlacing.

...Rohit

RNaval, you need to differ between video and film content. Most DVDs are film content and the SIL504 was already quite good in NTSC film. The ABT102 will only add some better "bad edit" handling on top of that and more robust cadence detection. But apart from that the difference for NTSC film content shouldn't be big. The major improvement can be seen in video content and in content with strange cadences like Anime.

I tried on VHD. For the high detail clips it locks on all, although not as quickly. All of them locked at the point when the car reaches the middle yellow seats. Does this mean anything ?

althogh for the synthetic wedges it fails on all of them. only the ABT102 maintains all the stright lines of the wedge except the tips, for all cadences. (also just for the kick tried on a lumagen hdp and no surprise it obviously failed everything)


Actually it locks on the vertical wedge but fails on the horizontal wedges. :p

Oliver

Thanks for that explanation. So, is the diner scene on the demo dvd, Video content? If so, it is impressive.

I have tested multiple HQVs, Gennum, National and FLI2310. Not all HQVs are equal. The 5910 is the best performing HQV. The NEC and Yamaha perform a lot worse than the 5910. They will only lock on 2-2 and 2-3 high detail clips. They drop on the others. They also drop on the bad edits test. I tried both force film and auto.

Stacey, would you happen to have the firmware version numbers of the HQV scalers? Sometimes the latest version is not the best.

What's a good example of a popular DVD with video content?

Many performance art DVDs (concerts, etc.) and many TV show DVDs are video-sourced. Sports and documentary DVDs can be video-based as well. Most mainstream movies are film-based, and aren't too challenging for most video processors.

--Andre

rohit - if you have it, try Eagles Hell Freezes Over; look at the diagonals - guitar strings, boom mikes. I'm also waiting for the Yankees to be on again - over ESPN SD, the base lines look terrible - like stairsteps. I am expecting the ABT to do a much better job of rendering them.

Most anime (especially older programs) requires good video deinterlacing.

Most anime (especially older programs) requires good video deinterlacing.

Most Anime requirse odd cadence detection vs. good video deinterlacing. The 6-4, 5-5 and 8-7 cadence detection will cover a lot of your Anime.

Most anime (especially older programs) requires good video deinterlacing.
Stacey has it right, I think. With older deinterlacers which did not detect the odd anime cadences, the video deinterlacing performance was indeed the important thing. However, if the deinterlacer can detect and lock to the anime cadences then it isn't operating in video mode.

- Dale Adams

What's a good example of a popular DVD with video content?

I could be wrong, but I think some of the early Smallville seasons on DVD are in video (i'm judging by the jaggies I see in the opening montage).

Perhaps I'm doing something wrong: I don't notice any difference in the picture quality before and after installing the ABT102. I'm testing with a DVD player at 480i using the component input and while playing one of my DVDs.

Does the 102 card provide better resolution overall or does it only help with displaying scenes that contain motion of some sort, in better resloution?

The DVD that they send along with the 102 card does show the demo loop quite well and it seems the resolution is true 720P. However, while playing one of my dvds, I don't notice any difference. Please tell me that I'm doing something wrong. Yes, I did verify that it's installed properly - ABT102 appears next to the v1.07 in the info screen and I've confirmed the additonal options found under deinterlacing.

...Rohit

What kind of DVD player do you have and what is the connection to the VP30? Are you sure the DVD player is in fact outputting 480i? Hit info on the VP30 remote to verify that is what is being seen.

I do not have much experience with the ABT102. BUT the old Sil504 chip it replaced does a very good job of deinterlacing FILM (480i to 480p). The Sil504 does not do as good of job deinterlacing video (480i to 480p), ok let me restate that, as a video deinterlacer it sucks.. The big improvement accomplished by the ABT 102 is with respect to improving the crappy video deinterlacing of the Sil 504. Watch some SD TV and marvel at how much better the video deinterlacing is. Film, I doubt, is improved very much. The scaling in the VP30 is not touched by the new chip.

It also improves on diagonal line handling as well, correct?

Sounds nice, might be worth it for an upgrade. Any improvements with LaserDisc? I see at least a couple of errors in just about every movie with the 504.

It also improves on diagonal line handling as well, correct?
Yes, but this feature is only needed/useful/used for video content. It just doesn't make any sense for film content.

Many of us were big SiI504 fans for film deinterlacing, however it should be noted that this chip was very much aimed at the 'NTSC' market and not the 'PAL' market, where it's performance (2:2) was not quite as good as when used in 'NTSC'. 'PAL' users will therefore see many additional benefits in addition to the better deinterlacing. This new solution covers ALL the bases (PAL, NTSC, multiple cadence detection, film and video) for SD where many of the 'older' solutions did not.

Can you give some examples of what is in video an how can you tell it is video?

Thanks

I could've been sitting here with my ABT-102 now, had it not been for the slow postal service in this country. So close, still so far away... :(

Anyway, that's not why I called.

Lately I've been thinking about this deinterlacing stuff, and taken more note of it. And found that I'm quite happy with the performance of the VP30 as-is (PAL SD material from my digital cable box).
The only times I really notice jaggies is in soccer games. The white lines there have a tendency to become jagged. There are a very few other situations as well, but really not many.

And - is it just me, or did the deinterlacing get a tad better when going from firmware 1.00 to 1.07, even without the ABT-102 card? Maybe it's just placebo, or wishful thinking, or perhaps I set a few settings slightly differently than before?

I can't really see much stair stepping, combing or jaggies now at all. Like right now, there was a guy on TV with a pinstriped suit (tan suit w/black stripes), and none of the lines were jagged or had steps.

Most all diagonal lines or edges that I see look nice and smooth.

So in light of this; will the ABT-102 card improve the overall image quality even though I don't really see the problems of deinterlacing now? Except in the mentioned case of soccer games and a few other situations?
Does it do anything else that would cause improved image quality? Since people are raving about 'almost HD quality', I mean.

Yes, but this feature is only needed/useful/used for video content. It just doesn't make any sense for film content.

Doh. I need to continue learning :)

I'm not sure if I'm seeing the right thing... but here it goes: It seems for poorly mastered video DVDs (both NTSC and PAL), the ABT102 seems to introduce a little bit more mosquito noise. Without the card I do remember seeing the noise also. With the ABT102 it just seems a bit more. Does any one see something similar ?

With a clean source, the image has improved quite a bit with the video having a 3D look (used to be quite flat).

(maybe it's just because the noise was previously masked in the deinterlacing .. ? )

I agree that the ABT102 seems to react to noise differently than the Sil504. I have a recorded broadcast that's very noisy, and the noise becomes more prominent and pulses more with ABT102 than the Sil504. I wonder if it's better to apply noise reduction before or after deinterlacing in this case (eg. using a Mosquito). Also, I wonder how noise affects the deinterlacing, since the random noise may make it more difficult to figure out correlations between fields.

--Andre

I'm not sure if I'm seeing the right thing... but here it goes: It seems for poorly mastered video DVDs (both NTSC and PAL), the ABT102 seems to introduce a little bit more mosquito noise. Without the card I do remember seeing the noise also. With the ABT102 it just seems a bit more. Does any one see something similar ?

With a clean source, the image has improved quite a bit with the video having a 3D look (used to be quite flat).

(maybe it's just because the noise was previously masked in the deinterlacing .. ? )

Did you by any change check the sharpness control per input ? The working seems to have changed to me. I first had it to 3 for video but now have reduced it.

Daniel.

So is DVD considered film or video? I am ignorant on the film vs video thing. Can someone educate me?

So is DVD considered film or video?

It can be either. It depends on how the footage was recorded.

--Andre

I agree that the ABT102 seems to react to noise differently than the Sil504. I have a recorded broadcast that's very noisy, and the noise becomes more prominent and pulses more with ABT102 than the Sil504
What's actually happening is that the ABT102 responds to motion differently than the SiI504. Since noise of sufficient amplitude can be regarded as motion, you may well see a difference with a noisy source.

If you closely examined the behavior of the 504, you'd see that its video deinterlacing does not respond well to certain types of low-level motion. If you could freeze the image out of the 504 you would see that in many cases it is missing much of this low-level motion and has significant combing artifacts in its output. The ABT102 is much more robust in this regard and produces a deinterlaced signal with much less combing - i.e., it does a much better job of removing interlace motion artifacts than the 504. The downside of this, however, is that it reacts differently to noise and its motion-detection logic will trigger in many places where the 504's will not. This can cause flicker or pulsing with the right type of source material (typically high contrast edges or very high detail). It's all a big tradeoff, really. The ABT102 is currently programmed with a bias towards removing more motion artifacts, and that's probably resulting in what you're seeing.

I wonder if it's better to apply noise reduction before or after deinterlacing in this case (eg. using a Mosquito). Also, I wonder how noise affects the deinterlacing, since the random noise may make it more difficult to figure out correlations between fields.
I believe that the Mosquito is supposed to operate on the least processed form of the signal. For an interlaced signal this would be before deinterlacing.

From the perspective of the deinterlacer, it's often, but not always, better to do noise reduction before deinterlacing. For video deinterlacing, the removal of noise before the deinterlacing process should result in a more stable image as the noise which can trigger motion-detection logic has been removed or reduced. For film cadence detection, it's more of a mixed bag. Noise reduction can result in both better and worse correlation between fields in different situations. It depends a lot on the nature of the noise and the capabilities of the noise reduction hardware.

- Dale Adams

Dale. thanks for the info. You are correct. The Mosquito works best processing 480i, then out into the VP30.

Most DVDs are film. Films (movies) are Film. Rock concerts, DVDs of TV shows etc, are video.

Broadcast SD and HD of sports etc are Video. The great advantage of the VP30 with the ABT102 is for SD video. Sports. Yea. Get rid of the noise fro compression DTV video SD with the Mosquito first. Assuming the VP40 or 50, whatever, handles 1080i videos well as the VP30 with ABT102 handles 480i video, I will be one happy camper.

Assuming the VP40 or 50, whatever, handle 1080i videos well as the VP30 with ABT102 handles 480i video, and I will be one happy camper.
Dale already said in an older thread that the HD deinterlacer will run the very same algorithms as the SD deinterlacer, just for HD, of course... :)

The Mosquito adds some pulsating artifacts too, particularly with static images.

The mosquito certainly needs the 480i source because then it gets the macroblock sizes correctly (feeding it the output of the processor may cause it to incorrectly detect macroblock edges and lost it's "edge").

I would think that MPEG noise reduction in the process would only serve to confuse the ABT102, however, wouldn't it? If the processing is too strong, it could misidentify combing artifacts.

So is DVD considered film or video? I am ignorant on the film vs video thing. Can someone educate me?
There are two types of 'sources' for interlaced signals (e.g. DVD, 1080i HDTV):

'Film sources' require "pulldown" to resequence frames. 'Film sources' are movies and high quality TV dramas.

'Video sources' require removal of combing over fields. 'Video sources' are news, sports, chat shows, and most studio based stuff.

Therefore DVD could be either: it will be 'film source' if it's a movie, but will be 'video source' if it's sports highlights.

StooMonster

... and u also have video materials produced/mixed with film/CGI based "background", for which some parts of it need to be in video mode, some parts of it need to be in film or other cadence tracking mode. The ABT102 is excellent at this.

... although I still have some clips even the ABT102 combs severely... (for example: a closeup of a cat on a completely black background, I see the whiskers combing everywhere.)

... although I still have some clips even the ABT102 combs severely... (for example: a closeup of a cat on a completely black background, I see the whiskers combing everywhere.)
Is this a clip I can get access to? E.g., is it available on DVD?

- Dale Adams

The downside of this, however, is that it reacts differently to noise and its motion-detection logic will trigger in many places where the 504's will not. This can cause flicker or pulsing with the right type of source material (typically high contrast edges or very high detail).

- Dale Adams

Thank-you, Dale. I have the ABT currently installed and functioning - finally got to watch the Yankees on ESPN-SD last night. Superb job on removing the jaggies on the basepaths - they had been horrible before, the VP30 +ABT made them perfectly, smoothly diagonal on my crt. HOWEVER, on closeups of the batter, waiting at the plate with little or no motion, I was seeing the pulsing you describe above. I was afraid my 980 was going south - glad to know it is just an artifact. I couldn't figure it out, since on wide shots, pans, and commercials there was no evidence of pulsing (as if there was a toggle on the brightness control, from one level to another, at about 1(?) second intervals,) but on the players face, with out of focus crowd in the background, the strange variation in black levels was there.

Thanks for putting my mind at ease. And while the game did not look as good as the HD feed (went to a friend's house to check it out on a small 52" hd set,) it looked about as good as I expect SD will look. Good job.

wow Dale I was hoping u would notice this post and take a look at it. :)

Pls dun laugh, this DVD is kid's stuff. It's called "Baby Shakespeare", from the "Baby Einstein" series. Chapter 5, it shows a closeup of a real cat. The ABT102 combs real bad here. Chapter 4, there is one part it shows assembling of a wooden butterfly toy. The ABT102 combs quite badly here as well. (Each chapter is very short. These are for babies.)

I found the amazon link
http://www.amazon.com/gp/product/B00005YUPP/qid=1148306919/sr=1-1/ref=sr_1_1/103-2304259-4975857?s=dvd&v=glance&n=130

Thank you for your help!

Dale,

Thanks for the explanation.

--Andre

Pls dun laugh, this DVD is kid's stuff. It's called "Baby Shakespeare", from the "Baby Einstein" series. Chapter 5, it shows a closeup of a real cat. The ABT102 combs real bad here. Chapter 4, there is one part it shows assembling of a wooden butterfly toy. The ABT102 combs quite badly here as well. (Each chapter is very short. These are for babies.)
Perhaps it's something to do with the cat being in slow motion? The artefacts could be encoded into the DVD.

StooMonster Junior has these DVDs, although way too old now, but she's a genius so I wonder if they really work?

Anyway, they are authored on Macs I think, and are mostly video and computer generated effects. c722's post reminded me I had these, so I put them in and there are a few combing incidents, but as I said above I think they could be due to encoding. Dale ... it's over to you.

StooMonster

The disc is ordered. I just have to wait 'til it comes. Too bad there's no way to just download the silly thing. . . :) I'll post results when I get the disc.

- Dale Adams

The disc is ordered. I just have to wait 'til it comes. Too bad there's no way to just download the silly thing. . . :) I'll post results when I get the disc.

wow what can I say ? One more time reminds me that I did the right thing of sticking with DVDO instead of the other guy. :)


Perhaps it's something to do with the cat being in slow motion? The artefacts could be encoded into the DVD...

well for the cat I can make it not comb through some other cheap deinterlacer so I guess it's probably a simple video mode. Maybe because of the background the ABT102 thinks it's more complicated than that. :)


Anyway, they are authored on Macs I think, and are mostly video and computer generated effects

Yes . This material is a mix of everything, video, CGI, film. Real gd to show a video deinterlacer. And it's very clean so deinterlacer has gd samples. Plus this is done by Disney so you have *all* sorts of chroma errors.

Btw just to be clear: the video performance of the ABT102 is excellent. I'm just very curious to see what's going behind for those exceptional cases.

Plus this is done by Disney so you have *all* sorts of chroma errors.
I have an original version from Julie Aigner-Clark's Baby Einstein Company, before she sold her business to Disney in November 2001 for $25M.

So I don't think you can blame Disney for *all* the chroma errors. ;)

StooMonster

I have an original version from Julie Aigner-Clark's Baby Einstein Company, before she sold her business to Disney in November 2001 for $25M.

oh is that so ? I didn't know that.. :)

btw:

StooMonster Junior has these DVDs, although way too old now, but she's a genius so I wonder if they really work?


Good to hear that ! Lots of ppl told me it works so I bought the whole lot. Hopefully it works the same way for her also. :)

ok end of OT...

Just installed my ABT102, and am very happy with it. Much clearer image, and any hint of jaggies etc. are gone, as far as I can see. The image has more of that 'looking through a window' quality to it than it did before the upgrade.

A question though:

Am I right in thinking that it's a good idea for me to set it to 'Video mode' for my digital SD cable box (which of course is mostly video material), and to 'Film bias mode' for my DVD player input?

If a movie then is suddenly shown on TV, the ABT102 should still automatically fix that, even though it's set to 'Video mode' for the cable box, right? And vice versa? I'm asking mostly to confirm what I think is the case.

Just installed my ABT102, and am very happy with it. Much clearer image, and any hint of jaggies etc. are gone, as far as I can see. The image has more of that 'looking through a window' quality to it than it did before the upgrade.

A question though:

Am I right in thinking that it's a good idea for me to set it to 'Video mode' for my digital SD cable box (which of course is mostly video material), and to 'Film bias mode' for my DVD player input?

If a movie then is suddenly shown on TV, the ABT102 should still automatically fix that, even though it's set to 'Video mode' for the cable box, right? And vice versa? I'm asking mostly to confirm what I think is the case.

The best setting for all content is 'Auto'. If the 'Auto' mode is having issues, then you may want to change to the proper mode, 'Film Bias' or 'Video', based on the content that you are watching.

Oh, OK - thank you for the reply. I'll set it back to Auto for all inputs then. :)

wow Dale I was hoping u would notice this post and take a look at it. :)

Pls dun laugh, this DVD is kid's stuff. It's called "Baby Shakespeare", from the "Baby Einstein" series. Chapter 5, it shows a closeup of a real cat. The ABT102 combs real bad here. Chapter 4, there is one part it shows assembling of a wooden butterfly toy. The ABT102 combs quite badly here as well. (Each chapter is very short. These are for babies.)
I received the disc and took a look at the 2 scenes you mention.

On the chapter 5 scene with the cat, I don't see any combing at all. The only thing I notice is that the freeze frame of the cat at the start and end of the scene looks to have a problem similar to the butterfly puzzle scene (see below). It could also just be that the still frame image was generated (poorly) from a single field so that you see a lot of jaggies in the cat's whiskers. As soon as the cat starts moving, though, the scene looks fine. What deinterlacing mode are you running the ABT102 in?

The wooden butterfly scene in chapter 4 is a problem with the source (i.e., it's encoded that way on the disc). It looks to me like the fields are misaligned (possibly reversed). When there's motion in this scene the motion-adaptive processing in the ABT102 smooths out the image since it's generating that portion of the picture from just one field. However, where there's no motion then adjacent fields are combined and you see the misalignment as jagged edges. The only way you can make something like this look good is to force the ABT102 to use only a single field to generate the output image (since any combination of 2 fields will give you the jaggies) using Game Mode 1.

I notice quite a bit of this on this disc. The scene with the apples right before the wooden butterfly is this this way, too. It's more noticeable on the butterfly scene due to the static background and the sharp edge transition on the butterfly puzzle. Not everything's like this of course - the shapes puzzle scene after the butterfly scene is fine, for instance.

- Dale Adams

Thanks for the investigation.


On the chapter 5 scene with the cat, I don't see any combing at all. The only thing I notice is that the freeze frame of the cat at the start and end of the scene looks to have a problem similar to the butterfly puzzle scene (see below). It could also just be that the still frame image was generated (poorly) from a single field so that you see a lot of jaggies in the cat's whiskers. As soon as the cat starts moving, though, the scene looks fine. What deinterlacing mode are you running the ABT102 in?


In Auto it combs a lot. Forcing it to video is much much better. The player is a pio 59, dun know if it's related. I cannot force it to Video for the whole disc as it will comb almost every minute. (That's why I'm impressed by the ABT102 since this material is a mix of everthing.) Do you think you can add a Auto with Video Bias ?



The wooden butterfly scene in chapter 4 is a problem with the source (i.e., it's encoded that way on the disc). It looks to me like the fields are misaligned (possibly reversed)....

I notice quite a bit of this on this disc. The scene with the apples right before the wooden butterfly is this this way, too.

If this is the case, the entire series are full of this. I saw a lot. On the same disk of the cat, for every "camera" stop where it introduces the next subject, it's obvious the fields are reversed ( e.g. the blue colors around the "camera" lens )

In Auto it combs a lot. Forcing it to video is much much better. The player is a pio 59, dun know if it's related. I cannot force it to Video for the whole disc as it will comb almost every minute. (That's why I'm impressed by the ABT102 since this material is a mix of everthing.) Do you think you can add a Auto with Video Bias ?
I'm not sure I follow your comment about it combing a lot when you force it to video. In video it should never comb. Combing is a result of merging two fields which are from different points in time. In video mode it never does this for the whole frame, and will only do it on a pixel by pixel basis if there's no motion in the image. If there's no motion and it's combing, then there's a problem with the source not with the deinterlacer. I also never even saw it comb in auto mode, although admittedly I did not watch the entire disc. However, I did look at about 10 minutes of the disc and never saw any combing type of artifact that was not in the source material itself.

For the record, what I saw wasn't strictly what I would call combing. There were jagged edges on objects and a slight offset between fields, but it was not at all the same as when I forced the deinterlacer to combine 2 fields which should not be merged. There were very severe and obvious large-scale combing artifacts when I did this. That's not what I saw in auto or video modes, though. For my testing I used both S-Video and 480i HDMI signals from two different DVD players (although neither was a Pioneer DV-59AVi) - a Sony and an Oppo.

To repeat, if you're seeing combing with this source in video mode, it's a problem with the source - i.e., the combing is in the source itself. (Either that or your DVD player is doing something very, very odd. What type of connection are you using from the player to the VP30?) The only way I know of to get around this type of problem is to force the ABT102 to use only a single field to construct the output image (as any combination of two fields for the problem sequences will result in jagged artifacts) - i.e., Game Mode 1. The ABT102 literally cannot comb in this mode as the output is based on only a single field.

- Dale Adams

In video it should never comb. Combing is a result of merging two fields which are from different points in time.

Unless the src itself is wrong. In the case of this (abt a dozen dvds from this series) I find it better to leave it at Auto instead of Video, because in Auto it tries to track diff part of the screen. For example, on the same disc u have now, one of the last chapter shows a "sleeping" toy dragon, and the background is a moving moon. This moving moon must be from some real time-lapse shoot of the actual moon. For this scene, Auto does it flawlessly. in Video, the moving moon has "jaggies" which looks like combing. I dun know how Video can introduce jaggies to the moon while Auto doesn't. (Is it because Auto correctly tracts the cadence of the moon ?) On the other hand, for the cat scene, Video is excellent, all the whiskers have no "breaks". But in Auto u see quite a lot of breaks, even when it moves.


There were jagged edges on objects and a slight offset between fields, but it was not at all the same as when I forced the deinterlacer to combine 2 fields which should not be merged. There were very severe and obvious large-scale combing artifacts when I did this.

Yes this is the best description of the image. "offset between fields". It's not the massive combing error when the deinterlacer is doing something wrong (of which I've seen from a poor deinterlacer from another cheap dvdp)

c722,

how are you connected to the VP30, via component?

are you sure the pio isn't deinterlacing too? Pioneer's deinterlacer is quite poor, in terms of combing, and you might be inadvertently using it in progressive mode.

I tried both component and HDMI, in 480i. The VP30's info screen also verified the input is 480i...I tried on both output, a 720p HDMI and 852x480 RGBHV, both doing 1:1 (checked via VP30's checker pattern... although the analogue one has some issues but that's a separate problem)

can the pio be doing 480i->480p->480i ?.....hope not...

Some players deinterlace and then interlace again to achieve 480i output. However, I can't believe the Pioneer would do that! c722, do you have a different DVD player to try it with?

Just managed to confirm that the ABT102 seems to do the following with 480i 60Hz or 576i 50Hz video :

1. When graphics or still picture is put on screen, I see a huge amount of edge chroma noise (ICP/CUE) on the edges of text, and whatever image is on the screen goes really aliased ( like what you see digital zooming in on a pic )

2. As soon as credits start rolling, the ABT102 locks on and the edge chroma noise disappears and the still image that was on the screen goes smooth and clear again.

I've also seen the same behaviour on a broadcast MPEG2 SDTV (576i 50Hz) stream output from my STB via component.

I could believe that my Pioneer DV-733/47Ai CUE is so bad that the VP30 can't fix it completely, but the blocky image still problem was a surprise. And to see the same thing on a broadcast MPEG2 stream also makes me wonder.

I now have a DVD that can repeatably do it ( US Region1 480i 60Hz Thomas Tank Engine : Percy's Chocolate Crunch Track 9 ).

Unless the src itself is wrong.
In that case the deinterlacer isn't going to fix the problem unless you manually put it in some special processing mode (like I suggested with the ABT102's game mode). Discussing this sort of thing with relation to deinterlacer performance is moot, unless it's intended to identify source material which is just blatantly faulty.

In the case of this (abt a dozen dvds from this series) I find it better to leave it at Auto instead of Video, because in Auto it tries to track diff part of the screen. For example, on the same disc u have now, one of the last chapter shows a "sleeping" toy dragon, and the background is a moving moon. This moving moon must be from some real time-lapse shoot of the actual moon. For this scene, Auto does it flawlessly. in Video, the moving moon has "jaggies" which looks like combing. I dun know how Video can introduce jaggies to the moon while Auto doesn't. (Is it because Auto correctly tracts the cadence of the moon ?)
I don't see anything like this in either auto or video modes. I see no jaggies on the moon in any processing mode (outside of the forced 2:2 modes). There's no hint of a cadence that the ABT102 could track, and video and film modes perform identically.

On the other hand, for the cat scene, Video is excellent, all the whiskers have no "breaks". But in Auto u see quite a lot of breaks, even when it moves.
Again, I see no evidence of any of this. I see no combing problems in auto mode.

I'm wondering if we're both looking at the same thing. Singapore is a PAL country. Are you looking at the same R1 disc I am, or at a different version? Are you viewing a 50 Hz source or a 60 Hz source?

- Dale Adams

1. When graphics or still picture is put on screen, I see a huge amount of edge chroma noise (ICP/CUE) on the edges of text, and whatever image is on the screen goes really aliased ( like what you see digital zooming in on a pic )

2. As soon as credits start rolling, the ABT102 locks on and the edge chroma noise disappears and the still image that was on the screen goes smooth and clear again.
This sounds a lot like the 'Baby Einstein' disc c722 mentioned. There are numerous examples on that disc where a still frame appears to be generated from a single field, resulting in lots of jagged edges and blocky aliasing artifacts. I haven't noted the chroma issues there, though.

I now have a DVD that can repeatably do it ( US Region1 480i 60Hz Thomas Tank Engine : Percy's Chocolate Crunch Track 9 ).
I'll look for this disc to see if I can reproduce what you're seeing. I suspect it's a problem with the source, but you never know for sure until you look. Does the VP30 with SiI504 do this as well?

- Dale Adams

In that case the deinterlacer isn't going to fix the problem unless you manually put it in some special processing mode (like I suggested with the ABT102's game mode). Discussing this sort of thing with relation to deinterlacer performance is moot, unless it's intended to identify source material which is just blatantly faulty.

Yes I just wanted to know what's happening. As u said the most obvious is the butterfly scene. That's one thing I saw before the ABT102, and was hoping the new DL can remove it. Now I know it's plain bad in the source so I'm happy.

For the moon scene, I just retried. The "jaggies" of the moon is not the significant type. The small jaggies are near the top of the moon, i.e. if u look very closely that part of the circle is not very smooth. For a video I guess any pan will definitely leave gaps for the deinterlacer to fill (pretty much like diagonal processing), so this should be normal. Am I right ?


I'm wondering if we're both looking at the same thing.

... should be. I bought them from US.

Actually I was beginning to think if it's related to the scaling. If the deinterlacer is mixing the wrong fields, the result will be massive. I can see that kind of significant combing if I use the pio deinterlacer and set it in the wrong mode. The ABT102 doesn't have any of this. My plasma's resolution is quite low. To scale a small (moving) circle nicely on it may not even be possible.

Dale,

Can you explain what is happening when using DVE and viewing the Snell and Wilcox Zone Plates. The center concentric circle that moves around the center of the display distorts, the circles become unstable, it is almost like the circle aliases, like there is some type of undersampling. Is this caused by the source material?

Thanks.

I am interested in some feedback on the image cropping pattern. This is located under static patterns -> geometry. This is a bit different than AVIA. AVIA is 0 based while ours is 1 based. You simply go to the first number that is missing the end and that is how many are cropped.
Ok, some feedback.

I have the Oppo 971H with the Pixel Magic SDI card and ABT102 de-interlacing plug-in adapator.

The image cropping is limited to 4 lines at the top of the screen/image. I underscanned the image and checked that the line offset was zero to be sure I was not inadvertently cropping it. Left, right and bottom show zero cropping. The bottom of the image, although not cropped, is raised by about the same number of pixels, suggesting that I have an offset in the data fed into or coming out of the PM board. However, when looking at the DVI ouput, there is NO cropping on any side. The PM board is fed from the same data lines as the DVI board is. That leaves the question as to why there is a difference between the 2 output formats. I do know that Oppo had some difficulty in getting the MPEG decoder to work satisfactorily with the Faroudj chipset, but I don't think that that explains it. The PM board is in essence just a simple parallel to serial converter, so it should not be dropping lines. Perhaps Dale or someone experienced in the design of SDI could aswer this one.

Edit: When I force the Oppo to display PAL, the cropping and bottom offset disappear on SDI. The ABT102 disc is in NTSC. On DVI, there is no change in cropping when toggling between NTSC and PAL (there is no cropping on either). This suggests that the ABT102 is doing the cropping, but only at 60Hz (NTSC).

On the CUE front, after having turned it OFF on the VP30, I could not detect any using the criteria as you described it in one of your previous posts.

Overall, it would appear from the examplary results I am getting using both the ABT102 and DVE test discs, that the combination of the PM SDI card and the Oppo 971H represent a very good solution to getting the absolute maximum out of SD DVD. I would not have said that only just a week ago, before I installed the ABT102, as PAL was totally useless using the Sil504 de-interlacing chipset on the VP30. Having used the ABT102 NTSC disc and my DVE PAL disc has tested both PAL and NTSC operation on my entire system. My display can natively lock onto both 50 and 60Hz and it is also set up to its native resolution of 1280 x 768. See my PP for complete equipment specs/models.

Can you explain what is happening when using DVE and viewing the Snell and Wilcox Zone Plates. The center concentric circle that moves around the center of the display distorts, the circles become unstable, it is almost like the circle aliases, like there is some type of undersampling. Is this caused by the source material?
There are 3 versions of the S&W test pattern. One is 3:2 pulldown, one is (supposed to be) 2:2 pulldown, and the other is video (i.e., motion between every field). When there is a repeating cadence in the source, such as 3:2 pulldown, the deinterlacer should detect that cadence, reconstruct the original film frame, and give you an image where the moving zone plate (which is what the concentric circle pattern is called) looks smooth.

When the deinterlacer cannot (or does not) do this, then it must use its motion-adaptive mode to deinterlace the image. The problem with a test pattern like this is that the multiple angles and very high frequencies in part of the image typically result in jagged edges or moire type patterns for parts of the image that are moving. This is what should happen on the video version of the pattern (the one called 'field rate', I think). The 3:2 version should be rendered with smooth looking circles, although you will typically see a brief instance of the more jagged pattern when motion first starts up after the zone plate stops moving, as the deinterlacer has to reacquire the 3:2 repetition cadence.

The 2:2 version (called 'frame rate', I think) is more problematic. Ideally, this should be handled the same way the 3:2 pattern is. However, I believe there is a mastering problem on the disc. I cannot detect any 2 adjacent fields that come from the same original frame (except when the zone plate is not moving). There zone plate in each adjacent field is in a different position, so the deinterlacer cannot reconstruct a perfect frame. Consequently, you see artifacts in this version of the pattern as well.

- Dale Adams

Hello. I just finished installing the ABT card and the sofware.

I'm not very happy.

Has anyone noticed with the new card- watching the opening "crawl" of SW:Episode 3, that the yellow text seems to be shimmering, flickering, moving, etc.? I don't see this happening with any of my dvd players, nor did it happen with the vp30 before the upgrade.

I was hoping the software would fix the audio problems I've been having, but it's made them worse. I got loud buzzing noises and had to bypass the vp30 for audio altogether.

This is quickly becoming the most disappointing product I've ever owned. Everyone is raving about the product and the customer service. While they've answered some of my questions, many have been ignored in emails.

Is anyone experiencing problems like this? :mad:

Hello. I just finished installing the ABT card and the sofware.

I'm not very happy.

Has anyone noticed with the new card- watching the opening "crawl" of SW:Episode 3, that the yellow text seems to be shimmering, flickering, moving, etc.?

I was hoping the software would fix the qudio problems I've been having, but it's made them worse. I got loud buzzing noises and had to bypass the vp30 for audio altogether.

This is quickly becoming the most disappointing product I've ever owned.

Is anyone experiencing problems like this? :mad:
Yes, I know what you mean. believe me. Oddly, my audio problems have diminished a bit - no longer any dropouts on PCM, but DD is still misbehaving and some DD streams don't come though at all now. Didn't think that the ABT102 card would have anything do do with this.

The only other new, or shall I say odd thing, I have seen with the ABT102, was a scene where a prop aeroplane was taxing down a runway. The view was from behind. There was a complete circle of severe "combing" - or maybe missing fields - within the prop's outline. I guess it would have been a three bladed prop spinning at about 1000 rpm. The ABT would have interpreted the spinning prop blades as very fast movement object(s). Perhaps Dale could tell us more about this. I can't recall seeing this sort of effect on the old SiI504.

Thanks for your reply!!

I'm considering asking DVDO if I can just have my money back, though I doubt they'd do that since I bought in April.

Thanks for your reply!!

I'm considering asking DVDO if I can just have my money back, though I doubt they'd do that since I bought in April.
Just hang in there a little longer if you can. When it's all working it's good - bloody good in fact. The individual bits work well - it's just that they need to get all the bits to work together well.

The only other new, or shall I say odd thing, I have seen with the ABT102, was a scene where a prop aeroplane was taxing down a runway. The view was from behind. There was a complete circle of severe "combing" - or maybe missing fields - within the prop's outline. I guess it would have been a three bladed prop spinning at about 1000 rpm. The ABT would have interpreted the spinning prop blades as very fast movement object(s). Perhaps Dale could tell us more about this.
I'd have to see the scene in question to make a meaningful comment.

- Dale Adams

If we do not use the VP30 for audio delay, what is the delay of the unit if scaling and deinterlacing is used? Is the delay constant?

Oliver

If we do not use the VP30 for audio delay, what is the delay of the unit if scaling and deinterlacing is used? Is the delay constant?
If you're running with the output frame rate set to be 1:1 locked to the input, then there will be a constant delay of ~3.5 frame periods. Running unlocked increases that by about 1 frame period, although that will vary a bit, typically running between 4 and 4.5 fame periods.

This is also assuming you're not running with the ABT102 deinterlacer set to either of the game modes, as these are specifically designed to reduce the video processing delay. Subtract 2 field periods if you're using game mode 2, and 3 field periods if you're using game mode 1.

- Dale Adams

I'd have to see the scene in question to make a meaningful comment.

- Dale Adams
That won't be possible in this instance Dale - it was a news item. I realise that this makes the source of "dubious" nature/pedigree and I wouldn't have commented on this except for the fact that I've seen similar behaviour with other very fast moving objects within or moving through a scene. I'll keep a close eye on it and record it if I can catch it. Thanks.

If we do not use the VP30 for audio delay, what is the delay of the unit if scaling and deinterlacing is used? Is the delay constant?

Oliver

If you go to the 'Input Adjust->AV LipSync' adjustment and move the slider all the way to the left you will see how much delay the iScan is putting on the audio signal so that audio and video are synced up. For example, if this value is -55. the iScan is putting 55 milliseconds of delay on the audio. '0' on this slider bar is when audio and video are in sync (ie the audio has already been delayed). This delay will remain constant as long as the input signal, output resolution and output frame rate do not change.

If you're running with the output frame rate set to be 1:1 locked to the input, then there will be a constant delay of ~3.5 frame periods. Running unlocked increases that by about 1 frame period, although that will vary a bit, typically running between 4 and 4.5 fame periods.

This is also assuming you're not running with the ABT102 deinterlacer set to either of the game modes, as these are specifically designed to reduce the video processing delay. Subtract 2 field periods if you're using game mode 2, and 3 field periods if you're using game mode 1.

- Dale Adams

Thanks. This helps alot. Actually is there a benefit/disadvantage of locking input and output frame rate? Also I have been searching for pronto codes that allow me to use a single button to directly access the memory modes on the VP30. Are they available? Was hoping to make a marco so that my family members need only press a button for TV/DVD watching.

Oliver

Also I have been searching for pronto codes that allow me to use a single button to directly access the memory modes on the VP30. Are they available?

Which 'memory modes' are you referring to?

Dale, do you think you might get the chance to look at those Star Wars laserdiscs we discussed some time back? Unfortunately, the ABT102 didn't do anything to help the shimmery image.

Which 'memory modes' are you referring to?

I believe the correct name is display profiles. There are 4. The Pronto codes and the Universal Remote Codes does not have this.

Oliver

I believe the correct name is display profiles. There are 4. The Pronto codes and the Universal Remote Codes does not have this.

Oliver

Display Profiles

Profile 1:
0000 006c 001b 0000 0064 0064 0016 0015 0016 0015 0016 0015 0016 0015 0016 0015 0016 0041 0016 0041 0016 0041 0016 0041 0016 0015 0016 0015 0016 0015 0016 0041 0016 0041 0016 0015 0016 0015 0016 0041 0016 0015 0016 0015 0016 0015 0016 0041 0016 0015 0016 0015 0016 0015 0044 0044 0016 0001


Profile 2:
0000 006c 001b 0000 0064 0064 0016 0015 0016 0015 0016 0015 0016 0015 0016 0015 0016 0041 0016 0041 0016 0041 0016 0015 0016 0041 0016 0015 0016 0015 0016 0041 0016 0041 0016 0015 0016 0015 0016 0015 0016 0041 0016 0015 0016 0015 0016 0041 0016 0015 0016 0015 0016 0015 0044 0044 0016 0001

Profile 3:
0000 006c 001b 0000 0064 0064 0016 0015 0016 0015 0016 0015 0016 0015 0016 0015 0016 0041 0016 0041 0016 0041 0016 0041 0016 0041 0016 0015 0016 0015 0016 0041 0016 0041 0016 0015 0016 0015 0016 0041 0016 0041 0016 0015 0016 0015 0016 0015 0016 0041 0016 0015 0016 0015 0044 0044 0016 0001

Profile 4:
0000 006c 001b 0000 0064 0064 0016 0015 0016 0015 0016 0015 0016 0015 0016 0015 0016 0041 0016 0041 0016 0041 0016 0015 0016 0015 0016 0041 0016 0015 0016 0041 0016 0041 0016 0015 0016 0015 0016 0015 0016 0015 0016 0041 0016 0015 0016 0041 0016 0015 0016 0015 0016 0015 0044 0044 0016 0001

User:
0000 006c 001b 0000 0064 0064 0016 0015 0016 0015 0016 0015 0016 0015 0016 0015 0016 0041 0016 0041 0016 0041 0016 0041 0016 0015 0016 0041 0016 0015 0016 0041 0016 0041 0016 0015 0016 0015 0016 0041 0016 0015 0016 0041 0016 0015 0016 0041 0016 0015 0016 0015 0016 0015 0044 0044 0016 0001

Josh, are these codes also usable for HD+?

That won't be possible in this instance Dale - it was a news item. I realise that this makes the source of "dubious" nature/pedigree and I wouldn't have commented on this except for the fact that I've seen similar behaviour with other very fast moving objects within or moving through a scene. I'll keep a close eye on it and record it if I can catch it.
The spinning prop blades on the plane could have been a case where there was a repetitive motion of an object with regularly-spaced features, where the rate of motion and the feature spacing was such that the features were at exactly the same location from frame to frame. The deinterlacer looks for differences between fields to determine the presence of motion, and so would see no differences in this local area and therefore see no motion.

The SiI504 should actually be much more susceptible to this than the ABT102, as the motion detection in the 102 is more reliable than that in the 504. You can also see this type of problem with other deinterlacers such as those from Faroudja and Silicon Optix. I have a couple of test clips which demonstrate this very well.

Again, I'd have to see the particular clip to know for sure whether or not what you saw would be a case of this. The other behavior you describe with fast moving objects might be something different, although I'd have to see an example to be sure.

- Dale Adams

Actually is there a benefit/disadvantage of locking input and output frame rate?
If the output rate is not locked to the input rate, then there will be dropped or repeated frames. Depending on the difference between the two rates and the type of program material you're looking at, this may be unnoticeable or may cause objectionable stutter or jerkiness in moving objects.

- Dale Adams

Dale, do you think you might get the chance to look at those Star Wars laserdiscs we discussed some time back? Unfortunately, the ABT102 didn't do anything to help the shimmery image.
I did get a chance to look at this and noticed the effect/problem you mentioned. Some of it looks to be aliasing in the original source image, but there are definitely cases where the deinterlacer is dropping in and out of lock on the 3:2 cadence. Unfortunately, the development system I use to analyze this type of thing only accepts an HDMI or DVI input so I can't use it to determine just what the problem is. (Yeah, very stupid, I know. That's what I have to work with, though. . . )

- Dale Adams

Unfortunately, the development system I use to analyze this type of thing only accepts an HDMI or DVI input so I can't use it to determine just what the problem is. (Yeah, very stupid, I know. That's what I have to work with, though. . . )
What you need is a device with a very good video decoder/comb fiter and ADC connected to an HDMI transmission chip. You can then us it for your testing, and sell them for vastly inflated sums to us rabid laserdisc fanboys. :D

Yup a real market for this device but probably not long term!!

We need a SAI mod kit for LD players that can be fitted by monkeys and bypasses all dodgy Composite->S-Video->Composite internals in any player! :rolleyes:

Think I'll give up and replace what left over LDs I have with the HDDVD/BD versions when/if they appear. ;)

I did get a chance to look at this and noticed the effect/problem you mentioned. Some of it looks to be aliasing in the original source image, but there are definitely cases where the deinterlacer is dropping in and out of lock on the 3:2 cadence. Unfortunately, the development system I use to analyze this type of thing only accepts an HDMI or DVI input so I can't use it to determine just what the problem is. (Yeah, very stupid, I know. That's what I have to work with, though. . . )

Is there any hope for this? I'm seeing similar problems with other laserdiscs as well. I was hoping the ABT102 would help to clean up analog sources like this.

What you need is a device with a very good video decoder/comb fiter and ADC connected to an HDMI transmission chip. You can then us it for your testing, and sell them for vastly inflated sums to us rabid laserdisc fanboys. :D

Isn't that what the VP30 is supposed to be? Maybe one that could have a 480i "passthrough" mode if you wanted 480i from its HDMI.

--Andre

Dale,

Thank you for the explanation of the DVE Snell and Wilcox Moving Zone Plate, I better understand what is happening now. Also, thank you for taking the time to answer my question with a thoughtful response, and you did it on the Holiday.

Regards

Isn't that what the VP30 is supposed to be? Maybe one that could have a 480i "passthrough" mode if you wanted 480i from its HDMI.

--Andre
The VP30's video decoding/comb filtering chipset could be better...

Has anyone else noticed the problem I've reported with the SW Ep. 3 dvd opening crawl?

It only happens with the vp30 and ABT card. My dvd players handle it just fine. This seems like a big step backward.

I emailed DVDO-no reply, as usual.

Has anyone else noticed the problem I've reported with the SW Ep. 3 dvd opening crawl?

It only happens with the vp30 and ABT card. My dvd players handle it just fine. This seems like a big step backward.
I'm looking at the issue. I generally don't comment on things here unless I have something substantive to say. However, I'd suggest that if you're that upset about how the ABT102 handles one 20 second video clip (assuming there aren't other, similar issues you've noticed), that you return it to ABT for a refund.

- Dale Adams

Has anyone else noticed the problem I've reported with the SW Ep. 3 dvd opening crawl?

It only happens with the vp30 and ABT card. My dvd players handle it just fine. This seems like a big step backward.




What type of DVD player do you have?

The VP30's video decoding/comb filtering chipset could be better...

I can't disagree with this. I'm still shocked that my Panny 7th gen plasma has a better comb filter than both the Lumagen HDP and the VP30.

--Andre

I have a few. I most recently tested it on the Denon 3910 (without the vp30) and the text looked perfect-no shimmering or wiggling. I've also played the same scene with no trouble in my Denon 2900 and Sony player.

A big deal has been made about the ABT card's performance on the demo footage of the testing disc. But who cares if it handles demo material and test patterns well if it can't handle the opening credits of a brand new, well-produced transfer like Episode 3?

Does anyone feel like watching test patterns after a long day at work?

The Silicon Image deinterlacer seemed to do a better job.

I've asked ABT to address this, but no reply so far. I'll give them a chance to respond. What choice do I have?

The opening scene of the latest Staw Wars dvd isn't just 20 seconds of video. As a well-produced, recent dvd it's somewhat of a benchmark. It means that there will most certainly be other problems with other discs.

Why would a $2500 outboard deinterlacer/processor give me VERY noticeable artifacts on a mainstream movie that don't appear with any of my dvd players?

The opening scene of the latest Staw Wars dvd isn't just 20 seconds of video. As a well-produced, recent dvd it's somewhat of a benchmark. It means that there will most certainly be other problems with other discs.

Why would a $2500 outboard deinterlacer/processor give me VERY noticeable artifacts on a mainstream movie that don't appear with any of my dvd players?

Just return the card/vp30. Why even bother if you're unhappy. Besides the "benchmark" scrawl at the opening of the movie how does the rest of the movie look? Bad? Worst than your dvd players then just return it. If it looks so bad why keep the vp30 for so long? Surely if you were pleased with the VP30 w/o the abt 102 you can at least return the card since its only been out for 3 weeks or so. No product or processor is perfect. Just pick your poison. If DVDO emailed you back that they were looking into it would you hold on waiting for several more months? Don't like it now, return it now. Buy it again later or buy something else. This stuff/hobby should make you happy.

I have been throwing all sorts at the ABT102 and I must admit I am astonished by what it does in comparison to the competition. It is one classy bit of engineering. Well done guys.

Has anyone else looked at this dvd with the new card? That's the first thing I want to know.

But beyond that, how could such a state of the art product not handle something so simple- the opening credits of a movie? It could not have been designed this way intentionally. Maybe I got a bad copy of the film?

I've only had a little bit of time to look at dvd's since I installed it. They seem to look pretty much the same as they did through the SilCon (pretty much the same as my $250 Sony upconverting dvd player).

I've read all the rave reviews about this product, including the ABT card. I want to believe it.

I hope that DVDO's attitidue is a little better than, "if you really like Star Wars, return the card."

I want to check the other Star Wars discs through the ABT and see if it has the same problem. Maybe the one sequence was rendered in a strange way and that's what gives the ABT trouble. Still, it doesn't happen with the other deinterlacers I've tried.

Thanks for your comments. This is a very cool forum.

The opening scene of the latest Staw Wars dvd isn't just 20 seconds of video. As a well-produced, recent dvd it's somewhat of a benchmark. It means that there will most certainly be other problems with other discs.

Why would a $2500 outboard deinterlacer/processor give me VERY noticeable artifacts on a mainstream movie that don't appear with any of my dvd players?

I had SW EPII on last night and the opening crawl was fine. I don't know if it's the same as Ep III or not. I have a Marantz DV9600 putting out 480i by HDMI (4:2:2) to the VP-30 with the ABT102. The VP-30 goes HDMI (4:2:2) to a samsung HLR 5078, calibrated. I'll compare Ep III with Ep II tonight. So far, I'm quite pleased.

Thanks! Please post your findings. I look forward to it. I'll check Ep. 2.

Can someone explain this 4:2:2 stuff to me? That was a setting I was never sure of on the vp30. I have no such controls on my dvd players. I didn't get any sense of it from the manual, either. I've also been trying to find out how I should set the 50hz/60hz lock settings. The vp30 manual doesn't explain what this is and none of my dvd player manuals say anything about it, either. Maybe these settings are negatively effecting my results. I've asked dvdo about these but never got an answer. Maybe I'll try again, but if you guys could explain it to me, I'd be grateful.

Thanks!

hifichip76 - What kind of display are you using? Where in the world do you live (i.e. do you have PAL or NTSC where you live)? Do you watch DVDs from different regions?

I just tried US R1 SW I,II,III. II is the smoothest without even the slightest hint of "shivering". SW I is also very gd, smooth but there is a tiny hint of "shivering". Something must be wrong on III. There are some obvious artifacts just as u descibed. In particular the word "Episode III" itself and the 1st paragraph are not stable.. The above is under "Auto". With Film Bias for III it improves, at least the 1st paragraph no longer flickers, but the work "Episode III" still has troubles.

Input is pio 59 HDMI YCbCr444 480i, output RGB 1280x720 59.94 locked HDMI DLP FP.

Hope this is something fixable via firmware.

Has anyone else looked at this dvd with the new card? That's the first thing I want to know.

But beyond that, how could such a state of the art product not handle something so simple- the opening credits of a movie? It could not have been designed this way intentionally. Maybe I got a bad copy of the film?

I've only had a little bit of time to look at dvd's since I installed it. They seem to look pretty much the same as they did through the SilCon (pretty much the same as my $250 Sony upconverting dvd player).

I've read all the rave reviews about this product, including the ABT card. I want to believe it.

I hope that DVDO's attitidue is a little better than, "if you really like Star Wars, return the card."

I want to check the other Star Wars discs through the ABT and see if it has the same problem. Maybe the one sequence was rendered in a strange way and that's what gives the ABT trouble. Still, it doesn't happen with the other deinterlacers I've tried.

Thanks for your comments. This is a very cool forum.

What I know but is still hard to pin down is this. The ABT does sometime seem to introduce noise. I have been trying to figure if it does that on bad transfers or good transfers. I have also seen this shimmering artifact but it comes and goes. In general it has been a a hit and miss. It is defintely less sharp and a little noiser then the Lumagen HDP that I unfortuntely sold to get this. Cant complain about the deinterlacing powess esp on video though.

Oliver

I hope that DVDO's attitidue is a little better than, "if you really like Star Wars, return the card."
No official spokesman from ABT has made such a comment. I post here on my own time and my views are my own, not the company's.

- Dale Adams

Can someone explain this 4:2:2 stuff to me? That was a setting I was never sure of on the vp30.
4:2:2 applies to a YCbCr signal. It means that the two chroma components, Cb and Cr, are undersampled relative to the luma signal, Y. In a 4:2:2 signal the chroma is horizontally sampled at only half the rate that the luma is, which means that you have only half as much signal bandwidth for chroma. Vertically the signals are all sampled at the same rate.

A DVD is actually encoded with a 4:2:0 scheme, which means that chroma is subsampled at half the luma rate in both the vertical and horizontal directions. Virtually all MPEG decoders upsample the chroma in the 4:2:0 signal on the DVD to a 4:2:2 signal for output. This is one of the reasons why many want a 4:2:2 signal from the DVD player - i.e., it's the closest you can get to an unprocessed signal straight from the MPEG decoder. Internally the VP30 is a 4:2:2 processor.

I've also been trying to find out how I should set the 50hz/60hz lock settings. The vp30 manual doesn't explain what this is and none of my dvd player manuals say anything about it, either. Maybe these settings are negatively effecting my results.
Here's a link to the DVDO website about this: http://www.dvdo.com/faq/faq_isvp30.php See the last question/answer in the list.

- Dale Adams

I've only had a little bit of time to look at dvd's since I installed it. They seem to look pretty much the same as they did through the SilCon (pretty much the same as my $250 Sony upconverting dvd player).
Look, there is film content (hollywood movies originally recorded on film in 24fps progressively) and then there is video content (tv sports and shows, originally recorded in 60fps interlaced). These are two totally different things which run through totally different algorithms in a deinterlacing chip.

DVDs are for the biggest part film content. And the SiI504 already did a good job with NTSC film content, just as some progressive DVD players do. The new ABT102 can not improve so much in this area, because the SiI504 was already good here. So it is to be expected that you don't notice much of a difference with DVDs. When watching film content, the ABT102 is better than the SiI504 only when the source material is problematic (bad encoding). In this situation the ABT102 can improve image quality over SiI504, although it can not always do magic.

Where the ABT102 is two or three steps ahead of the SiI504 is with video content (tv sports and shows). There you should be able to see a very noticable improvement and that's what all the people are raving about. This is also what the test DVD (which is shipping with the ABT102) is mainly targetting.

I hope that DVDO's attitidue is a little better than, "if you really like Star Wars, return the card."
I kind of understand your frustration. You bought something which you hoped would magically improve all your DVDs and now some look the same while the DVD you seem to like most looks even worse.

But please be careful with what you say. You need to be aware that we consumers are in a very good situation with DVDO here, because several important DVDO guys are actively helping out on these forums, doing their best to help us. They don't have to do this. We shouldn't shy them away from the forums, so please be kind. Of course you may complain, but please be polite when doing so.

FWIW, Dale (who created the ABT102) is checking the problem you're describing and he will surely report on what problems he found with the Star Wars DVD. Now be honest: Where else can you report a problem with your consumer electronics device and you get feedback from the very engineer who created the electronics a few hours after you posted your complaint? So please be kind, have a bit of patience and give Dale the time to find out what's wrong. You have a much better chance to get help from him if you post your comments in a more friendly way!

Is there any hope for this? I'm seeing similar problems with other laserdiscs as well. I was hoping the ABT102 would help to clean up analog sources like this.
It's hard to say at this point. The fact that I don't have the ability in my development system to analyze what's actually happening with LDs means I have to guess what's going on. That's never a good thing - it's always better to base things on hard data.

Still, that doesn't mean there's no hope. (There can always be a new hope. :D ) I'm in the middle of developing some performance improvements now which may well fix the problem you're seeing, but I won't know for a while yet (probably a few days).

- Dale Adams

I'm in the middle of developing some performance improvements now which may well fix the problem you're seeing, but I won't know for a while yet (probably a few days).
Can you update the ABT102 algorithms on the VP30 add in boards through a firmware upgrade? I thought that wouldn't be possible?

Can you update the ABT102 algorithms on the VP30 add in boards through a firmware upgrade? I thought that wouldn't be possible?
It's possible, but you need a special ROM programmer to do it. In other words, it's not something the end user of the VP30 can do. If ABT does decide to upgrade the ABT102 at some point, it would require either a board swap or sending in the ABT102 card to ABT to be reprogrammed. I don't believe any decision has been made yet about how we might go about doing such a thing (probably because it hasn't been needed yet).

- Dale Adams

Wow! Thanks for all the replies. It's good to know that someone else found the same artifacts.

I'd like to make it clear that I didn't know that anyone from ABT/DVDO was on this forum. THat's great! I thought this was just another forum, like Sound and Vision (my usual forum), but a particularly good one.


I meant no insult or disrespect to anyone at ABT. I wouldn't have bought this product if I didn't think they were good. I simply want to feel that it was a good investment and get the enjoyment I paid for. I appreciate you guys looking into our problems and questions.

It really is wonderful that people from ABT frequent this forum. No, I don't know of any other company who does that. I'd heard their customer service was excellent, which was part of why I bought the vp30.

Someone asked me what kind of diplay I have. It's a Sony 50' LCD RPTV with one HDMI input (which is why it's disappointing that I can't get the vp30's auto switching to work). I know it's not state of the art, but it's pretty good.

If it helps, I mainly use the vp30 with the component outs of my old Denon2900 or with the HDMI 480i output of my Sony player (975 or whatever the model is). My cable box seems to have trouble with the VP30 (I get random black boxes and sometimes an error message saying that my display is not compatible, etc.). I've never found out why that is. Maybe it's a problem with the cable box.


I've also wanted to know exactly what the vp30 does when it gets an input signal at the same resolution as its putputting. In other words, if its output is set to 1080i and I feed it a 1080i signal (from HD cable or from an upconverting dvd player,etc.), what does the vp30 "do" to the signal? Does it just pass it through (aside from any picture quality adjustments)? I've wondered this, especially as HD and Blu-Ray are on their way.

Thanks again.

Still, that doesn't mean there's no hope. (There can always be a new hope. :D ) I'm in the middle of developing some performance improvements now which may well fix the problem you're seeing, but I won't know for a while yet (probably a few days).

- Dale Adams

You should not have done this Dale now you know its our duty to bug Josh until they find some way to update our 102 cards we simple can't live without these performance tweaks...... you just created the demand .... :D

Daniel.

You need to be aware that we consumers are in a very good situation with DVDO here, because several important DVDO guys are actively helping out on these forums, doing their best to help us. They don't have to do this. We shouldn't shy them away from the forums, so please be kind.


You couldn't be more right. ;)
This relationship between DVDO and AVS is special...
I don't know of anything like this elsewhere for HT folks.
This interaction sealed the deal for me...I will continue to purchase DVDO products in the future. :)

Well, I'd have to agree with that. Sorry I didn't know.

BY the way, how do I know if my dvd player is a 50hz source or a 60hz source? I read the dvdo FAQ again, and it still doesn't make any sense to me. There's nothing in the dvd player manual about 50hz or 60hz.

You couldn't be more right. ;)
This relationship between DVDO and AVS is special...
I don't know of anything like this elsewhere for HT folks.
This interaction sealed the deal for me...I will continue to purchase DVDO products in the future. :)

AMEN. If there was an electronics Hall of Fame, it is possible that we have several future inductees in this forum alone. Not too bad - can you say that of any other forum? Like you, if it weren't for the gents at DVDO that hang here, I might have given up on the VP30 (as, say, Gary was close to doing ;) ) - instead, I hung in there thru V1.07 and the ABT, and now have a pretty nice product. Which can most likely be swapped for a VP70, when I get a projector that warrants such a processor.

Plus - I liked my old Iscan Pro SOOOO much, I would have bought almost anything from DVDO. That was one sweet unit. Didn't do much, but did it D@@@ well. :D

BY the way, how do I know if my dvd player is a 50hz source or a 60hz source? I read the dvdo FAQ again, and it still doesn't make any sense to me. There's nothing in the dvd player manual about 50hz or 60hz.
Most DVD players only produce one or the other. Typically, if you live in a country with a 60 Hz video standard like NTSC, then the player you're using will produce a 60 Hz output. The same is true if your local video standard is a 50 Hz one like PAL or SECAM - i.e., the player's output will be 50 Hz.

There are, however, players which will output either. You probably don't have one of these, as they're usually only available through nonstandard channels.

Another thing to look at is the DVDs you're using. There should be some fine print on the back of the package which states the region code (e.g., USA is region 1) and video standard used.

- Dale Adams

Yes, I'm R1 (New York) USA, so I guess it's 60hz. I do also have a multi-region player that I haven't really used much.

Thanks.

You can press the "info" key on your VP30 remote. It should tell you whether the VP30 receives 50Hz or 60Hz from your DVD player. At least my iScan HD+ does.

If it helps, I mainly use the vp30 with the component outs of my old Denon2900

FWIW, I run a Denon 2900 w/SDI into the VP30/ABT102 then out (RGBHV) to a Sony G70. I am unable to observe the artifact you described on EPIII.

Has anyone else looked at this dvd with the new card? That's the first thing I want to know.
I see this problem to some degree on at least 5 of the Star Wars DVDs. The severity of the problem varies quite a bit, with Episode V being the best of the bunch.

But beyond that, how could such a state of the art product not handle something so simple- the opening credits of a movie? It could not have been designed this way intentionally.
The problem with the opening text crawl of Star Wars and the ABT102 is that it its triggering the detection logic in the 102 for cross-fades between out-of-phase 3:2 pulldown scenes. The ABT102 tracks cadence in a number of different ways. Ideally, all of these should agree as to the nature of the cadence in any given scene. However, the motion in this sequence is such that it intermittently causes differences in two of the cadence tracking mechanisms, and this difference looks exactly like what you get with a cross-fade between two out-of-phase 3:2 sequences.

The SiI504 does not have any such cross-fade tracking capability, which is why it does not drop out of lock on this. But then it also combs on 3:2 cross-fades, too. TANSTAAFL, I guess. ;)

I've developed a fix for this, but it requires logic changes to the ABT102. It may also require that the VP30 be run in film bias mode (although I'm still looking at a workaround for this). For what it's worth, this is the first clip I've run into that exhibits this particular behavior outside of a real cross-fade. The problem with needing a logic change to the 102 is that the ROM on the ABT102 board which contains the FPGA code is not user programmable. I'm still thinking about possible ways to make this work with the existing version of the 102, but I can't make any promises at this point.

Anyway, that's what's happening. I'd be interested in hearing about any other instances similar to this.

- Dale Adams

A double cross fade, eh? :p

You never cease to amaze me in how you work these things out, Dale.

Dale - don't want to stop you from your research, and playing with the card, but sheesh - we can always wait for the next version of Star Wars - maybe just drop a line to Lucas, and he can release a new special edition that is edited better. :D

...snip....

TANSTAAFL, I guess. ;)

....snip....


The Moon is a Harsh Mistress by Robert Heinlein. Great book :)

and he can release a new special edition that is edited better

I don't know if an editor that good has been born yet. :)

Episode V being the best of the bunch.
Truer words have never been spoken. :D

All I can say is thank you ABT, my picture now looks great being fed from my BUD(10' Sat 4dtv), it was horrible with the internal proccessor in the VP30, now with the 102, it is like it should be.

A very happy customer (hopefully the HDMI handshaking issue gets solved soon on my Sony DHG-HDD250).

I'm confused, is the problem localized in LD versions of SW or in the DVD versions, or both?

I'm confused, is the problem localized in LD versions of SW or in the DVD versions, or both?
One of the LDs, and a mastering error was confusing the ABT102.

One of the LDs, and a mastering error was confusing the ABT102.
I believe the cross fading stuff Dale was talking about was directed to the problems hifichip76 reported with a Star Wars DVD.

I hope DVDO's next gen video processor will allow algorithm upgrades of the ABT102-HD through a "simple" firmware upgrade... :)

I believe the cross fading stuff Dale was talking about was directed to the problems hifichip76 reported with a Star Wars DVD.
Oh, DVD was it... hehe...

I guess Ofer wasn't the only one who got confused...

I believe the cross fading stuff Dale was talking about was directed to the problems hifichip76 reported with a Star Wars DVD.
Correct.

- Dale Adams

Since there was/is so much buzz over opening credits, I dug out my copy of the Episode 3 DVD (region 1 - USA). The "shimmer/wiggle" is there on the words, as they crawl. Uh, so what? I had watched the movie twice, which was quite enough, without noticing or caring. I'm tempted to go back and quite hifichip's smart remark that "no one wants to watch demo clips all day." (pp) While I saw what he was complaining about, I had to go back and watch it several times to be sure it was a problem. And then watched again with different deinterlacing settings. hifi - how is this different than watching clips?

fwiw, there was an imperfection, denon 1600 -> VP30 -> Runco 980 Ultra. Looked better to me with di set to "film." But now it occurs to me - I need to go home tonite, and try the Denon running 480p, bypassing the ABT altogether - see if there is any difference. (I'm kidding- it didn't bother me before, and isn't worth bothering with now. Yankees are on tonite also :D)

Yup, spend all night watching clips. :)

OTOH, between watching the Star Wars intro crawl, watched Yankees win on ESPN-SD. ABT worked like a charm - SD sports (with diagonal base paths) aren't going to look a lot better. Works for me, Dale. Thanks.

There is no such thing as a perfect deinterlacer. Period. But the ABT102 seems awfully close to perfection by providing a robust and balanced approach to the task. As such, it is the very best SD deinterlacer on the market today IMO. What is there not to like? It has better low angle processing than VXP and HQV and beats them all (including Faroudja) hands down on cadence detection. Now lets hope the HD version of their deinterlacer is every bit as good! :)

Dale, thanks so much for taking my questions seriously and looking into the issue on the SW discs. Last night, I looked at some of the other SW discs. They all seemed to have the problem to some degree, but I really only found it bothersome on Ep. 3. I understand that there's something odd about that scene. I also watched the scene using three other deinterlacers I have, and the problem only occurs with the ABT. WHy are they able to handle it so well, even though they're not as good in general? Is it because the ABT is so sophisticated and has so many ways to detect, that it's also easier for it to get thrown by something so unusual?

Thank you for your detailed explaination of the problem.

I don't mean to be picky. Most such glitches wouldn't bother me, especially if it was just a random movie scene. But I am a pretty big SW fan and have been looking forward to watching them through the ABT since I first ordered it.

I hope you can find a solution for it and I really do appreciate your efforts in any case.

Kevin

Just managed to confirm that the ABT102 seems to do the following with 480i 60Hz or 576i 50Hz video :

1. When graphics or still picture is put on screen, I see a huge amount of edge chroma noise (ICP/CUE) on the edges of text, and whatever image is on the screen goes really aliased ( like what you see digital zooming in on a pic )

2. As soon as credits start rolling, the ABT102 locks on and the edge chroma noise disappears and the still image that was on the screen goes smooth and clear again.

I've also seen the same behaviour on a broadcast MPEG2 SDTV (576i 50Hz) stream output from my STB via component.

I could believe that my Pioneer DV-733/47Ai CUE is so bad that the VP30 can't fix it completely, but the blocky image still problem was a surprise. And to see the same thing on a broadcast MPEG2 stream also makes me wonder.

I now have a DVD that can repeatably do it ( US Region1 480i 60Hz Thomas Tank Engine : Percy's Chocolate Crunch Track 9 ).
I took a look at chapter 9 on this disc. Everything you're seeing (well, almost everything) is in the source material on the disc. Here's what's happening:

1) On the still frame at the start of the chapter, the aliasing is all in the source. It looks like a combination of 2 factors. The main one is that the background image appears to have been created from a single field with simple line replication. There looks to be just a bit of ICP in this as well (this is the one thing not directly encoded in the source material), which makes it a little less obvious by imparting some visible line structure into the image.

The vast majority of the noise on the text is cross-luma artifacts (sometimes called dot-crawl, hanging dots, or chroma-crawl), not ICP or CUE. This is an artifact caused by the high frequencies around sharp color transitions bleeding into the luma component, and it occurs when the combined luma and chroma in a composite video signal is imperfectly separated. This disc was mastered from a composite source using what appears to be a mediocre Y/C separator. The cross-luma artifacts are encoded right into the video stream stored on the DVD. You can see these artifacts around any part of the image with very saturated colors (e.g., the yellow letter "1" and the red pinstriping on the train); they appear as a moving checkerboard pattern.

2) There's an interesting transformation which occurs between the still frame and the start of the rolling credits. The background image slowly morphs from the line-replicated image to a smooth looking image over the course of ~8 frames. It's not a hard switch. The scrolling credits are now moving so the ABT102 deinterlacing the moving text based on a single field of data. This reduces the visibility of the cross-luma artifacts around the text (which are still there).


I observed all this in several configurations:

A) Both an S-Video and an HDMI 480i signal from an Oppo 970 to a VP30 with an ABT102.

B) An HDMI 480i signal from an Oppo 480i to a VP30 with an SiI504.

C) Looking at the decoded data from the DVD on my computer frame-by-frame. This was done by extracting the chapter in question, decoding the MPEG 2 data, and examining the resulting sequence of frames in both QuickTime and frame-by-frame as single images in PhotoShop.

It is not the ABT102 (or any other deinterlacer) creating the problems you're seeing on this disc. They're encoded directly on the disc and would be visible with almost any video processor. (You might be able to get rid of some of the Y/C separation artifacts by sending a composite signal from the DVD player to a processor with a 3D Y/C separator in the video decoding front-end.)

As this is the only source material you've specifically called out that I can look at, I don't know if you're seeing the same type of behavior from your other sources as well. It's certainly possible, and probably likely. I saw the aliasing type of artifact on one other DVD source noted in another post in this thread - "Baby Einstein". (What is it about kid's shows that they all seem to have this lousy mastering?) Cross-luma artifacts are, unfortunately, all too common on TV shows transferred to DVD. You may be seeing some Y/C separation artifacts from analog broadcast sources if you feed a composite signal to the VP30, as it has an adaptive 2D Y/C separator which will not be as effective at eliminating these artifacts from still images as a 3D filter. You mentioned that you're using a digital STB, though, so it's more likely a head-end problem.

- Dale Adams

Last night, I looked at some of the other SW discs. They all seemed to have the problem to some degree, but I really only found it bothersome on Ep. 3. I understand that there's something odd about that scene. I also watched the scene using three other deinterlacers I have, and the problem only occurs with the ABT. WHy are they able to handle it so well, even though they're not as good in general? Is it because the ABT is so sophisticated and has so many ways to detect, that it's also easier for it to get thrown by something so unusual?
The ABT102 was designed to detect a number of problem cases which trip up many other deinterlacers. Examples of this are cross fades between out-of-phase 3:2 pulldown signals, cross-fades between 3:2 and 2:2 pulldown, multiple source types on the screen at one time, difficult or unusual cadences, etc. An older design like the SiI504 often does not handle this type of material very well.

However, it's been my experience that no detection method is bulletproof, and there will always be some number of 'false positives' for every valid detection. Ideally, these are few and far between, especially in proportion to the number of valid cases, but there will be some instances where a false detection is made. This may be a flaw in the basic algorithm (e.g., perhaps it's not selective enough) or just that one statistical case which hits all the right trigger points. This particular clip actually sets off a number of different detection mechanisms, so it's more likely a combination of a problem with the algorithm sensitivity and the exact source content of this one scene.

I hope you can find a solution for it and I really do appreciate your efforts in any case.
I do now have a revision of the ABT102 which does not trip up on this scene (or any of the other Star Wars openings). However, this currently requires that the ABT102 be run in a mode which, while good for well-mastered DVD material, is not what you'd want to use for all sources. I really like to avoid these types of special modes (although they are needed for certain very problematic sources) and have a general purpose operating mode which works well for everything. I'm not there yet, but I'm still working on it. If you do happen to notice any other sources with similar problems please let me know.

- Dale Adams

In my SW ep3 pal edition region 2, when outputting the deinterlacer in 2:2 odd (the way i watch the whole movie) the problem completly vanishes, in film bias mode or auto the problem is barely noticable; i hope it helps.

Regards

I do now have a revision of the ABT102 which does not trip up on this scene (or any of the other Star Wars openings).

Dale, just curious if you think this could be the same problem happening on those laserdiscs, or if that's probably something different going on?

So Dale is the Oppo 970 a good interlaced source for the vp30/ABT102?

So Dale is the Oppo 970 a good interlaced source for the vp30/ABT102?

I've ordered one just to use with my VP30 .....

How does it compare to a DVD player with SDI?

Many thanks!

Marcelo

Wow- thanks, Dale. So, what is the fix? Will it be something I can download?

I watched clips from other dvd's last night through the ABT and the picture was excellent. King Kong and Batman Begins look very close to HD!

I still can't get the auto switching to work and my cable box goes nuts whenever I hook its HDMI up to the vp30 (fine for a while, then starts shutting on and off by itself, maybe it's something to do with the HDMI handshake...can't find any settings that help this yet). But I guess questions like that belong in another thread.

Thanks again for all this help!

I've ordered one just to use with my VP30 .....

How does it compare to a DVD player with SDI?

Many thanks!

Marcelo


Me too! :)
Arrives Fedex today! :cool:

I ordered one yesterday, but they're back-ordered til the middle of June. I'm curious as to how it will work the ABT compared to my Sony.

I've had my board installed for about a week and have watched 8 or 9 movies, some from the oppo 971 (DVI out), some from a SDI modded player (SDI out) and two HD movies (1080i HDMI out). I can not see any differences using an AE900U, but after reading this thread I now understand that movies were not the real target of this add in board. I have 3 more weeks to decide to keep it thanks to the great return policy at DVDO. Yes, I upgraded the firmware and run everything at their default settings.

Dave

davidcrowe - If you are using the DVI output of the Oppo 971, you are sending the VP30 a signal which is not deinterlaced by the ABT102. 1080i is not deinterlaced by the ABT102 either, only 480 and 576i are processed (deinterlaced) by the ABT102. The SDI player is the only source, from what you mention, that is using the ABT102's superior deinterlacing.

I have a Marantz DV 9600 feeding a 480i via HDMI to the ABT 102 equipped VP30 and the picture is very impressive. Looks better then much of the compressed HDTV crap that DTV is sending us these days.

The more DVDs I watch, the more I like the 102. I just did an A/B with Mask of Zorro between the super bit dvd version and an HD version that I recorded on my H10-250. The dvd is very close. Very impressive

I have the 102 with a 1080i RPTV and an old sony dvd/vcr combo unit (component video). I see lots of jaggies and moire patterns. I plan on buying the oppo 970hd when it is available again. I am wandering if it is the bad dvd player or the 1080i RPTV that is giving the bad results. I love the detail I am getting even with the dvd player. I see the jaggies on Pixar movies like Incredibles and Monsters Inc. I also saw it on Serenity tonight.

DR

dvreid - Have you tried 540p output on the VP30 to see if you get the same results?

davidcrowe - If you are using the DVI output of the Oppo 971, you are sending the VP30 a signal which is not deinterlaced by the ABT102. 1080i is not deinterlaced by the ABT102 either, only 480 and 576i are processed (deinterlaced) by the ABT102. The SDI player is the only source, from what you mention, that is using the ABT102's superior deinterlacing.

Josh,

Yep, just posted the other sources for reference. They all look the same for DVDs. I have the oppo 970 on order for another 480i source. It may be a limitation of the 900u. I have have not tried the vp-30 with the vph-1292. Perhaps a more capable projector would show more of an improvement.

dvreid - Have you tried 540p output on the VP30 to see if you get the same results?

Josh,

I had tried it before and had not seen it get better. I just started Serenity again using 540p output and there are jaggies easily spotted in the first scene of the movie.

Does anyone else have Serenity with a 1080i RPTV to give me hope it is the player?

Thanks,
DR

dvreid,

Are sure your RPTV is not ALSO scaling the signal? Are you getting 1:1 pixel mapping? Check the Test Patterns of the Disk that came with the 102 or at least check the patterns from the VP30.

I recently installed the ABT102 and am very happy with the results. However, the upgrade itch is back, and I'm wondering if I can get an even *better* picture if I add SDI to the mix... or should I just get a new DVD player (currently I use a Denon 3800) that does 480i over HDMI?

MM

I have a Denon 3800.
I recommend the Oppo 970HD for 480i HDMI.
You will be amazed at the difference.

dvreid,

Are sure your RPTV is not ALSO scaling the signal? Are you getting 1:1 pixel mapping? Check the Test Patterns of the Disk that came with the 102 or at least check the patterns from the VP30.

CRT RPTVs are not fixed pixel displays like Plasmas or LCDs. It only accepts 480i/480p/540p/1080i.

I watched Ice Age on FX today and did not see any jaggies so I think it is definitely the dvd player. I will have to be patient for the oppo to come available.

DR

So, what is the fix? Will it be something I can download?
Probably not, although I can't say for sure yet. The current fix would require that the ROM on the ABT102 board be reprogrammed, and that can't be done without a special ROM programmer. I'm still looking at ways to remedy the problem without requiring a new version of the ABT102, but I can't promise anything. If a new 102 version is needed, then there would have to be a board swap or the card would have to be sent to ABT or someone else to be reprogrammed. (But note that there is no current plan or process to do this, but that's at least partly because it hasn't been required to date.)

- Dale Adams

I had tried it before and had not seen it get better. I just started Serenity again using 540p output and there are jaggies easily spotted in the first scene of the movie.

Does anyone else have Serenity with a 1080i RPTV to give me hope it is the player?
I cannot replicate this problem on my system. The most probable explanations I can come up with are that there is some sort of DVD player problem or that you're seeing interlace motion artifacts on your display (as it is interlaced, after all). However, if you haven't noticed such artifacts before, then that's probably not the issue. I don't notice this when I look at the 1080i output from the VP30. The only display I have which will show a 1080i image in its native interlaced format may not be large enough or have sufficient resolution for me to see such a problem, though, so I can't be sure.

- Dale Adams

Dale, sorry if I asked something here that's not so relevant to the ABT102 card. Scaling. The other nite we had an opportunity to see the VP30 with 2 other competing products. The 480i/576i deinterlacing and diagonal processing of the ABT102 is excellent. It's clearly well ahead of the others. However the scaling of the Vp30 is something not too exciting. Not that it's not gd. But compared with the other 2, it just seemed to be a little soft, not as sharp. In fact from memory it seems my old HD+ somehow was sharper. (I could be very wrong on this though, since that was more than 6 months since I sold the HD+). All the scalers were doing a 1:1 on a 720p DLP FP, so it shouldn't be a calibration issue on the PJ.

Just wondering, is there any plan to further improve the scaling ? Will the upcoming NLS in some way improve on the scaling also ?

Thanks

I've got a bunch of questions and suggestions in regard to the ABT-102. I don't own the VP30 yet but I will probably shortly order one with the ABT-102 card (probably regardless of the answers to these questions, as it appears to be the best SD deinterlacer right now).

Dale, I definitely appreciate you as a developer taking the time to answer questions and take feedback. Hence, these questions are primarily aimed at you. :) And also, they are aimed at any current owners of the product, or possibly even competitors (such as HQV), insofar as the questions apply to deinterlacing in general.


Inverse Telecine algorithms

I haven't tested either the ABT-102 or any HQV-based product but from reading behavior reports and feature specs online, it appears that each product primarily relies on a different technique for inverse telecine:

It appears that the HQV primarily uses field-matching, based on measuring the amounts of comb between two fields (perhaps usually or always on a per-pixel basis).
It appears that the ABT-102 primarily uses cadence detection (perhaps with comb detection (or just match detection on a small-region basis) as a backup to catch out-of-cadence video elements, or mismatching cross-fades)

My evidence for this:

This text from HQV press material: "With HQV processing, there is never any confusion about cadence. Instead of trying to match the incoming video against known patterns, HQV processing simply identifies complete frames as they come in. HQV processing is able to identify all known cadences, no matter how uncommon, and it can also detect cadences that have not yet been invented."
A post in this thread by c722 on 2006-05-18, 22:40, says that the HQV (in the Vantage HD) fails on synthetic wedge 2:2 and 2:3 patterns (and then a reply by oliverlim says "it" fails on the horizontal wedges but locks on the vertical wedges, "it" being ambiguous but probably referring to the HQV in the VHD). As a horizontal wedge is a synthetic pattern that will send a comb-detection algorithm off the chart, this is perfectly consistent with the theory that the HQV uses field-matching based on comb detection.
The same post says that the ABT-102 passes all wedge tests. This is consistent with cadence detection without regard to measuring the comb factor.
The ABT-102 seems unable to perform inverse telecine in game mode, even mode 2, where 3 fields of look-ahead buffer ought to be enough if field-matching were used.
The HQV claims per-pixel detection of film vs. video. The ABT-102 claims only "multiple source type within a frame" (If I recall correctly something I read over a year ago, doesn't the VXP do something similar? Decide based on larger regions, rather than pixels?) This is to be expected if the comb detection is less sophisticated in the ABT-102.
Although the ABT-102 claims to be able to detect mismatched 3:2 crossfades and 3:2 with 2:2, it can't detect mismatched crossfades between 2:2 and 2:2! (based on careful reading of this line: "Detection of 2:2 to/from 3:2 crossfades and out of phase 3:2 crossfades")
And this text from HQV material, highlighing the comb-detecting capabilities: "HQV processing represents the most advanced de-interlacing technique available: a true pixel-based motion-adaptive approach. With HQV processing, motion is identified at the pixel level rather than the frame level. While it is mathematically impossible to avoid discarding pixels in motion during de-interlacing, HQV processing is careful to discard only the pixels that would cause combing artifacts. Everything else is displayed with full resolution."

If I am correct in my theory that the ABT-102 relies primarily on cadence detection rather than field matching, then it is definitely excellent that detection of cross-fades of mis-matched cadences (perhaps using a strict measurement of the "match factor" between the duplicate fields in 3:2, or a more fail-safe, basic comb detection?) was added as a feature. Cross-fades are not the only place you find a "3:2 signature" where inverse telecine should not be performed! There is another pattern--one that exists in a lot of older anime throughout the program! See these screen shots:

http://www.animemusicvideos.org/guides/avtech/videogetb2a.html

Go to the section labeled "Temporal Scanline Blending" and see the two screenshots.

This is very common in older anime and is a process whereby material which is originally 24fps has its cadence destroyed using a telecine machine that blends frames together so each field (not frame, but field) on the DVD consists of a blend of two film frames--but worse--the amount from one film frame vs. the next in a single field changes vertically within the field! I've seen more than one 3:2 deinterlacer falsely comb on this because it entered film mode, often when there is only a small bit of motion on screen (typically a character's mouth moving). Again, specific logic to detect "false 3:2 signatures" is essential, and I'm glad to see the ABT-102 has it.

It seems to me that having an excellent comb detector is essential to making a very good deinterlacer. Relying most closely on the comb detection, rather than solely motion detection and cadence detection, allows the following features to be added or much improved. It seems the ABT-102 already does these things to some extent but perhaps not as much as it could:


Safer detection of 3:2
Especially, safer detection of 2:2--without pixel-level comb detection, any 2:2 inverse telecine algorithm is wide open to falsely combing on mismatched fades (because there are no duplicate fields as for 3:2), and moving video material on top of the 2:2 film content
Making the film/video decision per pixel rather than limiting the distinction to broader regions (especially for 2:2)
Entering film mode sooner without a huge lookahead buffer (very important for game modes!)
Weaving pixels when there is less than a 5 or 4-field match (again, matters most in game modes)
Weaving fields into full frames when there is no identifiable cadence at all, or when there are just isolated field matches not part of a larger cadence
A very common case of the previous point is low frame rate animation, which may have a cadence, but cannot be identified (at least not until a pan occurs). Standalone field matching works better than cadence detection in this case.

All the above improvements can happen because you can safely weave fields while omitting the few pixels that appear to comb (because they might be video overlay content), of which there will be some in any frame even if they don't really comb (because the comb detector has to be "safe").

Although in some cases, relying primarily on a comb detector would result in more pixels bobbing when they could actually be weaved, it actually results overall in more fields being correctly weaved because relying on a comb detector is the only way to identify that they belong to the same frame. Also, it results in less pixels falsely weaving.

Cadence detection is sort of a hack to identify which fields belong to the same frame. Very effective, in most cases...but combing is what the eye sees when it's done wrong, and comb detection is also the only way to reliably detect many cases, so why not base the algorithm on comb detection? It's more difficult computationally, but it seems it would result in the best output.

It seems to me that the best architecture for a deinterlacer would have a "brickwall comb detection filter" at nearly the last stage of the decision to bob or weave, using a very effective comb detection algorithm. Pixels that have been decided as "to be weaved" that are detected as combing would instead be bobbed with only two exceptions:

If it was part of a sequence of 4 or more still fields (with regard to that pixel)
If it is part of a solid cadence--"solid" meaning no possibility of a false cadence signature, and also that the cadence extends forward and backward enough fields to ensure that the cadence isn't ending with a bad edit (so a look-ahead buffer is needed here)

And then, as device options, either of these two exceptions could be independently disabled. In particular, I see the ability to disable the stillness exception as useful for content with temporal aliasing (the helicopter blades mentioned earlier in this thread), and objects blinking at 30Hz, which is common in games.

So...does this all make sense about cadence detection vs. field matching? Comments? Am I totally off-base here? :)

BTW, an interesting bit of trivia here. The racetrack test pattern that is on the HQV disc, and that HTHF uses in their benchmark, is not just an ordinary 3:2 test. The test was specifically picked by HTHF because it contained comb signatures, to be sure that video processors would lock onto the cadence anyway despite the appearance of combing. They think it's a flaw if it doesn't. I'm not so sure. If the comb detector is very good, and passes, say, the racetrack 3:2 test, but not a horizontal wedge 3:2 test, I think that's a defensible design decision. FWIW, I'd be curious if the HQV could successfully inverse telecine this clip if "odd cadence" versions of it were created, or "no cadence" versions (constantly changing patterns).


Per-pixel bi-directional weave

The HQV claims 4-field per-pixel motion adaptive processing, and the ABT-102 the same for 5 fields. But can they make the decision to weave forward in time (next field) or backward in time (previous field) on a per-pixel, or only per-frame, basis?

Here is the test pattern for this: Have some full-screen video content, to ensure the processor is not in film mode. Then change to a still scene. Then, the left half (or some portion) should change to a new still scene, and exactly one field later, the right half (or other portion, not overlapping with the first portion) should change to a new still scene. To pass, every pixel must be weaved. In the field in between the two scene changes, the pixels on the left half must weave with the previous field, and the pixels on the right half must weave with the next field.

However, I haven't heard of such a test pattern existing.

Some real cases where this matters:

Subtitles or other overlay content appearing on top of video or film (the subtitles should weave from the very first field that they appear to the very last before disappearing, if they are motionless)
Moving cursors on DVD menus when there is other motion on-screen
The strange type of anime telecine described above, where it may be possible (or common) for there to be some pixels on-screen from one film frame and other pixels from another frame

For another explanation of this, see:

http://forum.doom9.org/showthread.php?t=82264

If the weave decision is not even per-frame, but can only go backward, then any scene change from one still scene to another will bob, even though it shouldn't need to. I would guess that the HQV and ABT-102 at least do per-frame weave direction decisions, if not per-pixel, because otherwise what would be the point of HQV's "Truth in Marketing" paragraph? (On the other hand, the paragraph only says "In addition to the two fields being analyzed in the current frame, the two previous fields are required in order to determine which pixels are in motion.", when it ought to say "and the three next fields as well", so that's why I'm asking this here.)


Multiple film cadences in a frame

Quick question, for both the ABT-102 and HQV: Are they capable of performing inverse telecine on two different cadences in the same frame? Or only distinguishing between a single cadence and video in a frame?

In other words, I know they can both handle 60i scrolling text on top of 3:2 film--but can they handle a split screen of 3:2 on the left side, and a different cadence 3:2 on the right side? What about 3:2 overlay patterns in front of different cadence 3:2 footage? (I've seen this at least twice in anime, can point to some discs if anyone is interested but other than one, I will have to re-rent them to confirm and get the time index on those discs.)


Game modes

For game mode 1, why can't motion-adaptive processing be used to weave backward in time? I can see how the extra delay of game mode 2 is needed if weaving forward in time therefore occurs, but I don't see the need for the extra delay to just weave with the previous field, if 4 or 5 fields have passed with no motion around a particular pixel.

For game mode 2, it doesn't look like it is supported, but I would just like to voice support for the possibility of inverse telecine in a lower-latency mode, even if it wasn't as reliable as in the full-delay (non-game) mode. (And of course, it should err on the side of bobbing even when weaving might be possible, to prevent falsely weaving when it shouldn't.) The reason is that many games run in 30fps--essentially, 2:2 pulldown. Also, they switch cadence (between even and odd 2:2) every so often so choosing the 2:2 even or 2:2 odd modes isn't possible. Without this feature, then in 30fps games, half the vertical resolution is lost during motion when in game mode 2 (and all the time in game mode 1).


Old game mode

Suggestion: Create an "old game mode" for old game consoles that have half vertical resolution. This allows perfect progressive reconstruction even during motion.

Old game consoles that have 240 lines of vertical resolution at 60fps use a 1:1 mapping of in-game horizontal lines to lines in the current field. Therefore, the proper way to deinterlace if the source is known to be such an "old game" is to double the lines in each field down for one frame, and up for the other, rather than using interpolation.

As an additional processing step, because the source is known to be "old game", then blending can be used to reduce the blocky look of the large pixels. This should be performed after deinterlacing.

The only place I've seen this feature is in DScaler--the "old game" deinterlacing mode.



Bob stability

This is hard to measure with an "objective" test and it also requires observation in motion (not just a screenshot). But basically, when bobbing video content in motion (usually low motion), does it look "smooth"--or, instead, are there artifacts such as sudden changes in interpolation direction or flickering dots?

I've seen DCDi and it's pretty good. Some DScaler and AviSynth plugins that perform diagonal interpolation are not so good, though. Do owners of HQV devices and the VP30 with ABT-102 have comments comparing the bob stability appearance to other devices?


2D vs. 3D Y/C separation

I noticed in the VP30 specs and in this discussion that there is no 3D Y/C separation (temporal), but only 2D (spatial). However, it is "super-adaptive". Interestingly, this could actually be a superior approach to take over 3D, but I'm not sure of this.

One problem with only using a temporal solution is that nothing is done about dot crawl on parts of the scene in motion. But dot crawl can be very visible even in motion. Furthermore, its sudden removal when stillness occurs can actually draw attention to itself and be an artifact itself.

Another problem is that unless the look-ahead delay buffer is made even larger, several fields have to pass when motion stops for the dot crawl to disappear.

A final problem is that some DVDs that are sourced from composite materials have already passed through a temporal Y/C separator, and the only way to improve these further is with a spatial separator.

How about this architecture? Use only a spatial Y/C separator to process the image. But use a temporal Y/C separator to calibrate the spatial separator! This avoids all the artifacts and the delay of a 3D separator. The success of this, however, depends on it being possible to use an algorithm such that calibration values remain fairly constant throughout the program--an assumption I'm not sure is valid. (Also, the spatial separator should be designed to work as well as possible without useful input from the temporal calibrator, as some DVDs have already filtered out the temporal dot crawl.)


Super-advanced features


No processor is going to detect that a 1080p signal is really a bobbed 1080i signal, debob it, and re-deinterlace it using inverse telecine.


Not yet, anyway. :D


As long as we can dream of ultra-advanced features, here are a few requests. :)

Have a small buffer prior to that used for deinterlacing. Re-arrange fields that appear to be out of order. This is an infrequent source error, but it seems to happen with enough regularity across a wide range of DVDs that it could be worth correcting for in a product.
Have a buffer of frames after deinterlacing. If an odd cadence exists, reclock the output to a smooth cadence. For example, 2:2:2:4 would be changed to 3:2. AviSynth plugins already do this.
When performing the above, have an option for whether to allow blending. So if the blending option is enabled, then 3:2 material on DVDs would get reclocked so that every fifth output frame is a blend of two film frames. Or, when outputting to a 72Hz display, segments of 60i material would be blended and therefore smoother.
When a field consists of a blend of two frames, detect this and extract the original frames. This can be done by examining the next and previous several fields and a lot of computation. :) (This should be done prior to deinterlacing. Then, the reclocking feature above would ensure a smooth output.)
The previous feature, but support blends for which the composition varies vertically within the field (lots of anime would benefit from this; see earlier in the post).

Implement the above features and you're on your way to having a "magic device" that corrects every known video problem. :)

Most Anime requirse odd cadence detection vs. good video deinterlacing. The 6-4, 5-5 and 8-7 cadence detection will cover a lot of your Anime.
Stacey has it right, I think. With older deinterlacers which did not detect the odd anime cadences, the video deinterlacing performance was indeed the important thing. However, if the deinterlacer can detect and lock to the anime cadences then it isn't operating in video mode.

- Dale Adams
I beg to differ. In fact, I'm not quite sure where the idea came from that 8:7 is the "anime" cadence. I've observed a large number of anime DVDs, and from what I've seen, when anime has a cadence, it's usually 3:2, but 2:2 is common in new anime that is digitally produced, as is 2:2:2:4. However, 2:2:2:4 segments are usually short and not the entire program. Less frequently, I do see segments with strange cadences, or just as often, no cadence or a constantly changing cadence--again, usually isolated scenes.

Now, don't anyone tell me "anime has a low frame rate--they don't actually draw 24 frames per second". The important point is that it's filmed at 24 frames per second. The pans are 24fps, and film grain, artifacts, or errors, or even wobbling of the film projector machine that you see in a few shows, are clearly clocked to 24fps.

Unless I'm missing something, isn't 8:7 "cadence detection" of limited utility anyway? Once all the frames in the cadence pattern last longer than 4 fields (or 5 with the ABT-102), it seems most of the pixels are going to be weaved anyway. (Yes, film mode will help a little, though, in terms of weaving pixels that might have bobbed otherwise due to noise.)

The reasons anime is often such a deinterlacing challenge are in fact the very things that the latest generation of processors, such as the HQV and ABT-102, is addressing:

Many anime titles are transferred from film to video by blending frames (see my previous post for an explanation of this).
Much anime has frequent cadence bad edits even when it does have a cadence.
Lots of anime in the last 10 years has video titles or computer effects added, out of cadence with the film.
Sometimes 30p or 60i fades are used on 3:2 content.
Many watch it subtitled, which is usually not synced to the 3:2 cadence (unless your DVD player already does this, but then, you aren't relying on your VP to deinterlace).
The DVD production companies releasing anime in the US often add 60i scrolling credits on top of the 3:2 film material.
Anime, with its sharp lines, really benefits from diagonal interpolation more than live-action material.
Although this isn't relevant to video processors, it's worth noting that few anime DVDs are flagged as progressive, even when they actually have a valid cadence.

I think some may take a look at the feature sets of these new deinterlacers and think "what's the point?" Especially of things like cross-fade detection, scrolling titles on film, bad edit detection, and nonstandard cadence detection.

One reason is that these happen all the time on anime DVDs. These types of problems I think are frequent on many DVDs that are not film-based movies. Anime really does benefit from the entire gamut of features that the odd test cases on these benchmark DVDs address, not just the nonstandard cadence detection.

Inverse Telecine algorithms

[/list]
My evidence for this:
[list]
This text from HQV press material: "With HQV processing, there is never any confusion about cadence. Instead of trying to match the incoming video against known patterns, HQV processing simply identifies complete frames as they come in. HQV processing is able to identify all known cadences, no matter how uncommon, and it can also detect cadences that have not yet been invented."
A post in this thread by c722 on 2006-05-18, 22:40, says that the HQV (in the Vantage HD) fails on synthetic wedge 2:2 and 2:3 patterns (and then a reply by oliverlim says "it" fails on the horizontal wedges but locks on the vertical wedges, "it" being ambiguous but probably referring to the HQV in the VHD). As a horizontal wedge is a synthetic pattern that will send a comb-detection algorithm off the chart, this is perfectly consistent with the theory that the HQV uses field-matching based on comb detection.
The same post says that the ABT-102 passes all wedge tests. This is consistent with cadence detection without regard to measuring the comb factor.
The ABT-102 seems unable to perform inverse telecine in game mode, even mode 2, where 3 fields of look-ahead buffer ought to be enough if field-matching were used.


The "it" refers to the Lumagen HDP. For some reason the Vantage fails the synthetic wedges both horizontal and vertical. But I know Kris has actaully tried the ABT102 test disc on the Denon 5910 and I am sure he would have mentioned it if he saw this. I cannot imagine a HQV doing worse then a SII504 for this test. So I guess the next vantage firmware which is suppose to improve deinterlacing results will nail this test.



Bob stability

This is hard to measure with an "objective" test and it also requires observation in motion (not just a screenshot). But basically, when bobbing video content in motion (usually low motion), does it look "smooth"--or, instead, are there artifacts such as sudden changes in interpolation direction or flickering dots?

I've seen DCDi and it's pretty good. Some DScaler and AviSynth plugins that perform diagonal interpolation are not so good, though. Do owners of HQV devices and the VP30 with ABT-102 have comments comparing the bob stability appearance to other devices?



I had a plasma enhancer once and the bobbing of video on the ABT102 is very good or as good and defintely not worse then the DCDi. The thing I do notice at times is that the ABT102 does introduce some noise during video deinterlacing as compared to the HDP.




2D vs. 3D Y/C separation

I noticed in the VP30 specs and in this discussion that there is no 3D Y/C separation (temporal), but only 2D (spatial). However, it is "super-adaptive". Interestingly, this could actually be a superior approach to take over 3D, but I'm not sure of this.

One problem with only using a temporal solution is that nothing is done about dot crawl on parts of the scene in motion. But dot crawl can be very visible even in motion. Furthermore, its sudden removal when stillness occurs can actually draw attention to itself and be an artifact itself.

Another problem is that unless the look-ahead delay buffer is made even larger, several fields have to pass when motion stops for the dot crawl to disappear.

A final problem is that some DVDs that are sourced from composite materials have already passed through a temporal Y/C separator, and the only way to improve these further is with a spatial separator.

How about this architecture? Use only a spatial Y/C separator to process the image. But use a temporal Y/C separator to calibrate the spatial separator! This avoids all the artifacts and the delay of a 3D separator. The success of this, however, depends on it being possible to use an algorithm such that calibration values remain fairly constant throughout the program--an assumption I'm not sure is valid. (Also, the spatial separator should be designed to work as well as possible without useful input from the temporal calibrator, as some DVDs have already filtered out the temporal dot crawl.)


The 2D Super Adaptive is not too bad. But it is still no where close to what is available in for instance a Pioneer Plasma or Panasonic Plasma which has a 3D comb filter. For composite inputs, the 2D filter on the VP30 still results in dot crawls in the lines on news tickers and some "color image smearing" between letters with bright contrast like logos etc. The 3D filters in many of the current Plasma does not have these issues. It does better then the HDP on DVDs source from composite sources as I rarely see any cross colour rainbows on early series of Smallville or X-files.

Oliver

An interestin read. I think there's a few errors in there, but I'll let Dale field those. In the mean time, let's speak comb filters.

2D vs. 3D Y/C separation

I noticed in the VP30 specs and in this discussion that there is no 3D Y/C separation (temporal), but only 2D (spatial). However, it is "super-adaptive". Interestingly, this could actually be a superior approach to take over 3D, but I'm not sure of this.

One problem with only using a temporal solution is that nothing is done about dot crawl on parts of the scene in motion. But dot crawl can be very visible even in motion. Furthermore, its sudden removal when stillness occurs can actually draw attention to itself and be an artifact itself.

Another problem is that unless the look-ahead delay buffer is made even larger, several fields have to pass when motion stops for the dot crawl to disappear.

A final problem is that some DVDs that are sourced from composite materials have already passed through a temporal Y/C separator, and the only way to improve these further is with a spatial separator.

How about this architecture? Use only a spatial Y/C separator to process the image. But use a temporal Y/C separator to calibrate the spatial separator! This avoids all the artifacts and the delay of a 3D separator. The success of this, however, depends on it being possible to use an algorithm such that calibration values remain fairly constant throughout the program--an assumption I'm not sure is valid. (Also, the spatial separator should be designed to work as well as possible without useful input from the temporal calibrator, as some DVDs have already filtered out the temporal dot crawl.)
The iScans, unless they've changed things weithout me noticing, use a Philips SAA71xx series (http://www.semiconductors.philips.com/pip/SAA7118E_V1.html) chip, which is pretty venerable and clean, without being the last word in picture quality. It has an adaptive 4/2 line 2D comb filter.

You're correct that a 2D filter can be superior to a purely 3D approach. Which is why few if any filter circuits use a purely 3D approach. Instead you get 3D adaptive filters that use 3D with static images and 2D with moving ones. It's even possible to design a motion adaptive filter (along the same lines as a deinterlacer) which uses purely spacial filtering along moving parts of an image and 3D for backgrounds and static objects.

The folks at ABT certainly have the skills to write a comb filter that'll make the Philips chips seem pretty average, and the reason they don't, I guess, is that it doesn't make good business sense when you compare the price of bulk buying mass produced semiconducters compared to spending time writing algorithms in house, and the implementing it in a FPGA.

It's a shame, though, as when you spend ~$2000 on a video processor, you do sort of expect better comb filtering than you'd get from a broken, 10 year old, $5-on-ebay SVHS deck.

Dale, you're a legend !

Thanks for having a look at that disc for me and giving an expert view of the problem.

I have 2 DVD players that are both oldish Pioneers so I wasn't sure if I was looking at ICP/CUE or bad source.

Now if you could only tell me who will win the World Cp.. just don't tell big_marcelo :)

Dale, you're a legend !

Thanks for having a look at that disc for me and giving an expert view of the problem.

I have 2 DVD players that are both oldish Pioneers so I wasn't sure if I was looking at ICP/CUE or bad source.

Now if you could only tell me who will win the World Cp.. just don't tell big_marcelo :)
Andy, if the VP30 could make Brazil a guaranteed 6th time winner...... I'd buy a dozen of them myself!!!!! :)

I hope you enjoy the World Cup on that big 50" NEC XR5 of yours..... I still haven't decided between the pio pro 42" or the NEC 50" ....

The "it" refers to the Lumagen HDP. For some reason the Vantage fails the synthetic wedges both horizontal and vertical. But I know Kris has actaully tried the ABT102 test disc on the Denon 5910 and I am sure he would have mentioned it if he saw this. I cannot imagine a HQV doing worse then a SII504 for this test. So I guess the next vantage firmware which is suppose to improve deinterlacing results will nail this test.
Oh, I see. I re-read the original 3 posts very carefully, so I see how the chain of replies went. :)

Hmm...I wonder then if the Vantage failing on the wedges is still for the same reason, though...

And although I'm strongly leaning toward ordering a VP30 with ABT-102, one reason for my lingering doubt is on whether I should wait and see on that announced Vantage firmware update for mid-June...

An interestin read. I think there's a few errors in there, but I'll let Dale field those.
Yep...I wouldn't be surprised. :) I speculated a lot and went out on a few limbs. I hope I am on the right track.

BTW, I didn't pull that distinction between cadence detection and field matching out of the blue. It didn't occur to me to perform inverse telence any other way than cadence detection, until I found that some of the AviSynth plugins do it by field matching. They take a very interestingly different approach than most set-top processors do. The Telecide and Decimate pair of plugins is one example. Telecide attempts to match neighboring fields in an interlaced input, and then re-arranges them (keeping its output at the same data rate, still "interlaced") to be on aligned frames. Then, Decimate simply grabs these frames, removes duplicates, and outputs at 24p. There is no cadence tracking at all in the process--only examining neighboring fields. Also, therefore, "for free", you get 2:2:2:4 reclocking to a smooth 24fps. The data rate is never expanded to 60p, so if there's video content, it has to be blended (ugh).

The other notable example is TDeint (which I happened to link to in my first post). This also has the distinction of being one of the only two AviSynth "deinterlacing" plugins that I know of that outputs 60p, and that performs inverse telecine (the other being Greedy (High Motion), which is also a DScaler plugin). (The key words being "that I know of"--I may have missed some, but I did some searching.) The sole means of inverse telecine in TDeint, like Telecide, is an attempt to match each field either to its previous or next neighbor. The one with lower comb factor "wins", and also, the image is broken up into a number of squares (16*16 pixels, configurable), and if the measured comb factor is too high in any single one of them, the field will not be matched. So, by examining only three fields at a time and no cadence tracking, you get inverse telecine of all cadences "for free"! It is pretty reliable, too, but occasionally combs on field-blended anime, or just at the start of 60i scrolling credits on top of film. I might also add that TDeint has both a pretty good bob (a choice of two main algorithms, with variations), and bi-directional weave, making it by far the best PC software deinterlacer that I could find, and on par with the Faroudja chips. (Where it can't compete with the HQV or ABT-102 is that it does not perform video and film deinterlacing in the same frame, and the bob still needs work.) Unfortunately, it is far too CPU-intensive to work in real-time, so it's not really an option for either watching DVDs or playing games. (And no, I'm not going to make ripping and encoding a DVD a necessary step in order to watch it. :) I tried that a few times, got really annoying quick when you just want to watch it. That's why I'm buying a video processor.)

Well, the above is straying from the original topic somewhat, but I just wanted to provide the motivation for my first long post, and elaborate a little on where I got the cadence detection vs. field matching idea. I know TDeint achieves great results from field matching and zero cadence detection, so I was speculating that HQV used a similar algorithm which is probably a departure from most deinterlacing chips.

The folks at ABT certainly have the skills to write a comb filter that'll make the Philips chips seem pretty average, and the reason they don't, I guess, is that it doesn't make good business sense when you compare the price of bulk buying mass produced semiconducters compared to spending time writing algorithms in house, and the implementing it in a FPGA.

It's a shame, though, as when you spend ~$2000 on a video processor, you do sort of expect better comb filtering than you'd get from a broken, 10 year old, $5-on-ebay SVHS deck.
Yeah, you got it. Although I'm considering buying the VP30 right now, I'm looking forward to improvements in areas like Y/C separation in their future products, and things like my wish list in my first post. :) But deinterlacing is what's most important to me and it's looking like the ABT-102 is the best product for that.

I noticed that my capture card, which I currently use with DScaler, has the same chip (SAA7134). When I test the "adaptive comb filter" (the hardware one, not any DScaler filter) on the composite input, it greatly helps eliminate noise in the signal, which can be easily seen even when a DVD is paused. It doesn't really do much for dot crawl encoded on the DVD itself. The option is also available with S-Video, and I see the video blink when I disable and enable the option so I know a change is occurring but I can't tell any difference with S-Video. I hope in the VP30 it actually does improve all signals (even HDMI/SDI in), and isn't just a filter on the composite connector.

Ah...one possible big issue with the VP30 that I just now noticed...my monitor is 1920*1200...the VP30 can output 1920*1080, which is 1:1 mapping for 16:9 content. But for 4:3 content, can the VP30 output 1600*1200? It isn't clear from the manual. It's not a preset, but can any resolution be entered and if so, what is the maximum? (FWIW, 1600*1200 is fewer pixels than 1920*1080.) Otherwise, I'll have to use the scaler in the monitor. I'm not sure how good it is, but it looked good when I played a video on my computer (over a DVI connection) in non-native resolution. It's a gm1601 chip, if that tells anything. I found a brochure for it but it didn't say the number of taps in the scaler. The Vantage HD does support 1920*1200...so now I'll have to decide...get the VP30 now...get the Vantage HD...or wait an undetermined amount of time for DVDO's product announcement for their HD deinterlacer based on the ABT-102 algorithms...decisions...

Again, I want to thank the folks at DVDO for reading this thread...I hope my long post wasn't annoying, but it was sort of a braindump of various deinterlacing ideas I've been thinking about for a long time. Hopefully it is useful to some, and also I wanted to grab the chance to post while I knew DVDO/ABT was paying attention. :)

The best comb filter I've ever seen was Victor Steinberg's (formerly of Snell and Wilcox) FPGA implementation based around the Philips SAA7136. He uses the SAA7136 in a special mode where the raw CVBS samples are output instead of YCbCr samples.

I think this is Victor's website, but there's no mention of the 3D Y/C product (or IP).

http://www.videoq.com/

Ron

jeff,

My eyes got blurry after an hour into reading your post... You should really try to use shorter posts as you probably lost most people after the first few pages...

Combing is hard to detect. There are several techniques for doing this, but lots of things look like combing and aren't really combing and vice versa.

From my original reading of the HQV info, I believe HQV uses several overlapping techniques that are running in parallel. A combing detector and detail detector are supposed to decide, on a per frame basis (I don't think it works on a per-block basis, but I could be wrong) which algorithm gives maximum detail with minimal combing.

ABT's game mode is intended to provide quick feedback (deinterlacing can cause critical delays which could reduce the player's response time).

About the analog inputs, I do believe the VP30 can improve on those. I spoke to Dale about this and I believe he passed it on. Composite Y/C separation is slowly losing its importance with time, but I agree that it should be done as best as possible on a processor of this caliber. I think component inputs could also be improved and I hope the next processor in the series will achieve this.

Hi everyone,

What an informative thread, definitely learned a lot reading through all these pages..I am about to get the ABT102 card myself.

Could anyone help and kindly tell me if I need to write down all my inputs profile/picture control/output profile settings before I do the upgrade?? I mean, the firmware needs to be updated, would it reset everything??

Thanks a lot.

Many people have found it necessary to do a settings reset of their VP30 if they already have 1.07 and install the ABT102. If you don't have 1.07 installed, then installing it is going to erase your settings. So in either case, write down your settings.

I found this post by Jim Petersen from Lumagen very helpful in explaining how cadence detection is done:

http://archive.avsforum.com/avs-vb/showthread.php?postid=3470579#post3470579

--Andre

Many people have found it necessary to do a settings reset of their VP30 if they already have 1.07 and install the ABT102. If you don't have 1.07 installed, then installing it is going to erase your settings. So in either case, write down your settings.

I found this post by Jim Petersen from Lumagen very helpful in explaining how cadence detection is done:

http://archive.avsforum.com/avs-vb/showthread.php?postid=3470579#post3470579

--Andre

I see. I shall write down my settings before bringing the VP30 yo the shop then....thanks a trillion for your quick advise.

We have a work sheet on our website to help with this process. It is available here:

http://www.dvdo.com/faq/faq_pro_man.php

Thanks again, Josh Zyber for contributing this.

We have a work sheet on our website to help with this process. It is available here:

http://www.dvdo.com/faq/faq_pro_man.php

Thanks again, Josh Zyber for contributing this.

Wow, thanks. This is great cos' i was just thinking of making one myself. Just great!

If you don't mind, I would like to ask one more question.. Regarding 480i video game processing using Game 1 or Game 2 mode, other than the resolved frame delay issue, should I expect any general visual upgrade, say smoother picture, less jaggies, more details on a 36" CRT HDTV...compared with Sil504?? Unlike many people here, I am really new to all these terms of combing, cadences, etc...I do see artifacts for 480i games currently with VP30, but I don't think I know what they are, if they belong to the source or if they are due to deinterlacing problem.

In anycase, I owned previous iscans before, and the VP30 is the best, but I just want to know if the ABT102 would really make 480i video games LOOK better..

Thanks in advance!! Cheers

I found this post by Jim Petersen from Lumagen very helpful in explaining how cadence detection is done:

http://archive.avsforum.com/avs-vb/showthread.php?postid=3470579#post3470579

--Andre
Thanks very much for this link...it was exactly the type of information I was looking for!

Great thread, good info. Couple of questions ..... Im looking at the new VP20.

#1. Does the interaction of the ABT102 with the VP30 as described in this thread, apply to using an ABT102 with a VP20?
#2. It is not stated anywhere, including on the DVDO site....Can I assume that the VP30 and the VP20 do not deinterlace at all until you install the ABT102?

Thanks again for all the good info

#1. Does the interaction of the ABT102 with the VP30 as described in this thread, apply to using an ABT102 with a VP20?
#2. It is not stated anywhere, including on the DVDO site....Can I assume that the VP30 and the VP20 do not deinterlace at all until you install the ABT102?


#1: As far as we know, but only the DVDO folks will be able to answer this definitively.

#2: For both, if the ABT102 isn't installed, the Sil504 is used for deinterlacing. It's very good for NTSC films, not so good for PAL films, and fairly terrible for video.


--Andre

#1. Does the interaction of the ABT102 with the VP30 as described in this thread, apply to using an ABT102 with a VP20?

Absolutely

#2. It is not stated anywhere, including on the DVDO site....Can I assume that the VP30 and the VP20 do not deinterlace at all until you install the ABT102?

Both the VP20 and VP30 use the SiI504 (Silicon Image) as their base deinterlacers. The ABT102 is much better at deinterlacing video and sources with mixed cadences.

I love it. The Sil504 is "fairly terrible" for deinterlacing 480i video. LOL

I just finished running all the S&M (fitting name for this torture test, LOL) disk on a variety of processors. There's simply no comparison between SIL504 and ABT102. They're in completely different leagues.