Anybody know what limitations the new Silicon Optix Reon-VX chip will have compared to the Realta? I assume the Reon-VX will be able to stretch the image for CIH applications using an anamorphic lens. Correct or no?

The Reon-VX has been picked up by Mitsubishi in the HC5000 and Benq in the SP-820 projectors, both due out this fall/winter.

The Reon certainly has the capability to do anamorphic stretching. However, I cant make any claims as to whether or not any products that might use Reon (or any of our chips) will implement a feature like this.

Thanks,

Andy K.
ASIC Design Engineer
Silicon Optix, Inc.

Andy, thanks for chiming in. Could you please give us a small run down on what the differences are between Reon and Realta?

Also can the Reon be passively cooled?

Another question: Are you planning to do a Realta die shrink with the aim to reduce power consumption/heat, so that VP manufacturers can get along without fans? I do like what the Realta offers, but every Realta VP (Radiance, Vantage-HD, DragonFly) uses a fan, which is really putting me off a bit.

Thanks very much!!

I'm thinking same algorithms, lower computational accuracy. Am I far from the mark?

Regarding the comparison between the two product families, I'm working on it :)

Reon can indeed be passively cooled.

I can't comment on future plans, unfortunately. :)

Thanks,

Andy K.
ASIC Design Engineer
Silicon Optix, Inc.

I can't comment on future plans, unfortunately. :)
Too bad... :)

A strange idea: If a VP manufacturer desperately wanted to get along without a fan while using the Realta, would it be a solution to use two Realtas and downclock + downvolt both of them? Would that technically work out fine? Of course it would probably be too costly...

They could do better than that, actually - we do have customer designs (for example the Denon 5910) that have a fanless Realta solution.

Thanks,

Andy K.
ASIC Design Engineer
Silicon Optix, Inc.

Thanks a bunch for your replies, Andy. I look forward to seeing a comparison between the two chips.

Richard

Andy K,

Heat pipes?

Andy K,

Heat pipes?
And/or undervolting/underclocking it. In the 5910, the Realta is only deinterlacing SD, so it should not be working too hard in there. Even scaling is performed by an alternate chip (ABT, IIRC).

And/or undervolting/underclocking it. In the 5910, the Realta is only deinterlacing SD, so it should not be working too hard in there. Even scaling is performed by an alternate chip (ABT, IIRC).

The 5910 still has plenty of ventilation slots in the top of the unit as well. Internally it must still generate a fair bit of heat.

I just wandered over to take a look at the heatsink in a 5910 - its fairly hefty but still just a regular heat sink - no heat pipe to a larger heatsink somewhere else.

Although its only deinterlacing SD we still measure that the power drain is very nearly the same. The algorithms do more operations on each pixel in SD mode so the total number of operations done is roughly the same.

Andy K.
ASIC Design Engineer
Silicon Optix, Inc.

So there is not enough power to do the SD level of processing on HD?

So there is not enough power to do the SD level of processing on HD?
Well, if you have a processor which runs the same algorithms in both SD and HD, the processor would be almost idle for SD processing. That'd not be optimal, would it? I do like the idea that a processor spends more time on each SD pixel to further optimize image quality.

However, Dale Adams already said that his ABT102 will run the very same algorithms for HD content and SD content. What he said sounded very much like HD processing per pixel would be identical to SD processing per pixel. So I guess the HD version of the ABT102 will be almost idle for SD processing.

The interesting tidbit is: If the SD version of the ABT102 is similar to the Realta, and if what I wrote above is true, then the HD version of the ABT102 should have superior image quality compared to the Realta for HD content.

Any thoughts about this, anyone?

Well, if you have a processor which runs the same algorithms in both SD and HD, the processor would be almost idle for SD processing. That'd not be optimal, would it? I do like the idea that a processor spends more time on each SD pixel to further optimize image quality.

However, Dale Adams already said that his ABT102 will run the very same algorithms for HD content and SD content. What he said sounded very much like HD processing per pixel would be identical to SD processing per pixel. So I guess the HD version of the ABT102 will be almost idle for SD processing.

The interesting tidbit is: If the SD version of the ABT102 is similar to the Realta, and if what I wrote above is true, then the HD version of the ABT102 should have superior image quality compared to the Realta for HD content.

Any thoughts about this, anyone?
It depends upon how much processing is required for each increment of PQ. Since ABT seems to specialize in deinterlacing, I would expect the HD version to be superior here as well. However, noise reduction, et al., may not have similar functionality. Realta and Gennum both have extensive NR algorithms, IIRC, so it is not surprising that the Realta would do more for SD given how intensive NR can be.

Look here for a first DragonFly user report:

http://www.avsforum.com/avs-vb/showthread.php?t=694130

Unfortunately, it's another VP with fan/noise problems. So now we have 3 Realta based VPs (TheaterSync, Vantage-HD and DragonFly), all of which have a fan. And Lumagen has already said their Realta based VP will also have a fan. I don't know why the 5910 gets along with passive cooling, probably there's a specific reason for that (maybe it's clocked lower or whatever). But if all 4 general purpose external VPs based on Realta have a fan, then that can't be just coincidence. IMHO the Realta needs another stepping or maybe a die shrink, so that it can be passively cooled. I hope Silicon Optix takes this seriously.

The DragonFly could have been a nice VP, I love the small form factor. But the fan inside kills the deal for me.

Look here for a first DragonFly user report:

http://www.avsforum.com/avs-vb/showthread.php?t=694130

Unfortunately, it's another VP with fan/noise problems. So now we have 3 Realta based VPs (TheaterSync, Vantage-HD and DragonFly), all of which have a fan. And Lumagen has already said their Realta based VP will also have a fan. I don't know why the 5910 gets along with passive cooling, probably there's a specific reason for that (maybe it's clocked lower or whatever). But if all 4 general purpose external VPs based on Realta have a fan, then that can't be just coincidence. IMHO the Realta needs another stepping or maybe a die shrink, so that it can be passively cooled. I hope Silicon Optix takes this seriously.

The DragonFly could have been a nice VP, I love the small form factor. But the fan inside kills the deal for me.
I'm pretty sure the 5910 has huge heatsinks. Not something that's really an option in a 1RU video processor.

Of course, they could always make the box bigger to fit bigger heatsinks. It's not like going for a 5-6RU case is going to affect picture quality any.

The DragonFly could have been a nice VP, I love the small form factor. But the fan inside kills the deal for me.
It's short about three HDMI inputs for me, but I agree the 1U case is very appealing.

I'm pretty sure the 5910 has huge heatsinks. Not something that's really an option in a 1RU video processor.

Of course, they could always make the box bigger to fit bigger heatsinks. It's not like going for a 5-6RU case is going to affect picture quality any.
In the Lumagen forums Jim Peterson has stated that the Radiance case will be as big as the Vision HDP Pro, so that he can put a bigger fan in. See here:

http://www.convergent-av.co.uk/forum/index.php?t=msg&th=566&start=0&rid=0

Both the RadianceXS and the RadianceXD will be in a 2U case similar to the VisionPro HDP, but with a new look to the faceplate. We need the extra height, verses the VisionHDQ, so we can use a larger fan and keep the noise level as low as possible.
So although the Radiance has a bigger case, Lumagen doesn't seem to think they can get along without a fan.

So although the Radiance has a bigger case, Lumagen doesn't seem to think they can get along without a fan.
Well yes, but there's quite a difference between 2RU and, say, a 5RU case.

Well, a 5RU case is not a case but a wardrobe... :D

Well, a 5RU case is not a case but a wardrobe... :D
Flip it upside down, put some big legs on it, get a blanket to make some curtains around the base, and you've got the makings of a nice kotatsu (http://en.wikipedia.org/wiki/Kotatsu). It'll keep you warm, and improve your picture quality.

Your kidding, but I tried this and the fan noise almost becomes inaudible. You have made a hush box without a vent pipe! The air volume under the table for cooling, the slight leakage around the bottom and the floor, and the blanky for noise reduction. QUIET! Turning it upside down because the bottom is heavier than the top. Be sure the blanky doesn't block the vents, that pathway should be free to exit into the tent made by the blanky A small table but with a top substantially deeper than the unit with the unit on the top and the blanket covering it all.A couple of pieces of rubber between the upside down top and the table to prevent vibration and cratches. An IR repeater to control it, the built in IR sensor is covered by the blanky. NO I DON'T HAVE A MOSQUITO YET SO I ACTUALLY HAVEN"T DONE THIS. And I have an equipent walk through closet so fan noise to me is never a problem for me. BUT CARLED has a BRILLIANT solution. Embroider your blanky with a BIG THANKS CARLED!

Your kidding, but I tried this and the fan noise almost becomes inaudible. You have made a hush box without a vent pipe! The air volume under the table for cooling, the slight leakage around the bottom and the floor, and the blanky for noise reduction. QUIET! Turning it upside down because the bottom is heavier than the top. Be sure the blanky doesn't block the vents, that pathway should be free to exit into the tent made by the blanky A small table but with a top substantially deeper than the unit with the unit on the top and the blanket covering it all.A couple of pieces of rubber between the upside down top and the table to prevent vibration and cratches. An IR repeater to control it, the built in IR sensor is covered by the blanky. NO I DON'T HAVE A MOSQUITO YET SO I ACTUALLY HAVEN"T DONE THIS. And I have an equipent walk through closet so fan noise to me is never a problem for me. BUT CARLED has a BRILLIANT solution. Embroider your blanky with a BIG THANKS CARLED!
Aww shucks...

not a good idea with the mosquito, folks, I put mine on a slightly hot PVR and I had to send it in for repair. I've lifted it up some more now and no problems since.

not a good idea with the mosquito, folks, I put mine on a slightly hot PVR and I had to send it in for repair. I've lifted it up some more now and no problems since.
Bet that wouldn't have happened if it had foot-high heatsinks. :D

carled,

It's the most stylish thing on my HT system... The clever use of the blue microled inside the i in algolith and the displayless brushed aluminum look usually causes people to ask WTH is that?

After all, isn't that part of the big attraction of having components like that in your system?

Ofer. No. Actually it isn't and it shouldn't be unless you are having an HT meet. In the best theaters you do not see anything but the screen and some speakers. Let the picture and sound speak for your good taste and wisdom in component selection. Put it in a nice Mid Atlantic rack in an equipment room or closet.

Suppose you don't.

Guest. OH. What are those BEEEAUUTIFUL silver boxes?

Host. Thanks, I thought you would never ask. Uh, one's a video noise reducer and the other is a video processor.

Guest. Huh. Whatz they do?

Host. Well the first silver box removes Mosquito noise and a few other types of noise from the picture.

Guest. Is that why I don't hear any noise coming from the picture?

Host. No. I said it removes various types of video noise, not audio noise. Here, I'll do some A/Bs to show you the difference with it in or out of my system.

Guest. That's OK. Uh, could we just watch the game (or movie)?

Host. And the second silver box does everything that my display does except display the picture but it does it better. Here, I'll get my demo test pattern discs and show you how it removes deinterlacing artifacts.

Guest. You mean you used it to steal artifacts from the de international lace museum?

Host. No. But that's OK. Let's just watch the game. Wanna beer?

No. Actually it isn't and it shouldn't be unless you are having an HT meet. The best theaters you do not see anything but the screen and some speakers. .Let the picture and sound speak for your good taste and wisdom in component selection.
Precisely.

LOL,

I guess it's a question of perspective. There are two philisophies - one that puts all the equpiment in the back, deeply hidden away. The second is to put the stuff right next to your display, in plain view.

That also explains people's facination with crazy tube designs for amps...

Regarding the comparison between the two product families, I'm working on it :)

Reon can indeed be passively cooled.

I can't comment on future plans, unfortunately. :)

Thanks,

Andy K.
ASIC Design Engineer
Silicon Optix, Inc.

So, where can we find the products comparison.

On the silicon optix site, the only difference in features is the 16 to 1024 tap adaptative scaling only present on Realta features...

Does the Reon VX do scaling ?

What are the exact differences between the two ? Are these commercial only :rolleyes: ?

Not a side-by-side comparison, but filed away this WSR summary (http://www.widescreenreview.com/news_detail.php?id=11421) of the Reon that outlines features. -- John

Thanks,

But there is nothing more here than the reonvx press release as found on the siliconoptix site...

What can the Realta do that the ReonVX can't ?

Why don't we see any "low price" scaler based on ReonVX ?

Jim Peterson said in June that Lumagen was thinking about a Radiance version based on this chip, but no news...

Actually, there is no 1080i capable scaler (except Lumagen Vision line) in the 1000$ to 2000$ MSRP price range...

The ReonVX seems to be best the way to produce such video processors...

Richard.

Hi,

I work at Silicon Optix,so let me offer a little bit of clarification.

Realta is a far more capable chip than the Reon VX but costs more and uses more power. Realta has the power and flexibility of the Teranex video processing architecture, which came out of Lockheed's spaceborne military reconnaisance and target identification systems of the '90s, commercialized for broadcast and post-production applications.

Reon VX has many blocks that implement in hardware selected algos that Realta can perform and the system designer can choose to use these blocks as desired. In essence, a lot of the capabilities of the Realta are made available at much less cost and with a small sacrifice in flexibility and upgradeability.

Reon VX is meant for the mainstream and is targeted for 1080p systems. Realta can support much higher resolutions. Both these parts can optionally support our eWARP2 technology which allows high-quality keystone correction, or projection on curved screens, or dynamic compensation for lens distortions, though most projection systems manufacturers are very conservative when it comes to moving optics functions to silicon.

Testing Reon VX based systems against Realta-based systems (like the different Denon DVD players) against each other on the SD HQV DVD shows Realta does better only on a few tests (the different cadences). We'll probably have to upgrade these test disks with more challenging material to show the improved capabilities that are now possible with new algos and new ways to use old algos.

Speaking of new HQV tests. we have an HD HQV DVD in beta form that was given to some reviewers at CEDIA (and yes, there will be Blu-Ray versions available as well at some point). At CEDIA, we showed this DVD running from two Toshiba HD DVDs and driving two 1080p 47" diag. LCD TVs, a new Sony Bravia and a beta Syntax 747 which uses Realta. The differences were very dramatic. I'm sure there will soon be reports on how Reon VX-based systems do on these tests as well as other HD tests starting to show up. I know of at least one major reviewer preparing to publish tests of a lot of currently available systems using this beta HD-DVD's test sequences. The Reon VX was only introduced this past summer, so give it a little time before you see new products introduced with it. Denon and Mitsubishi were surprisingly fast coming to market with Reon VX.

Hope this helps as I don't have time to participate very regularly on this forum.

Testing Reon VX based systems against Realta-based systems (like the different Denon DVD players) against each other on the SD HQV DVD shows Realta does better only on a few tests (the different cadences). We'll probably have to upgrade these test disks with more challenging material to show the improved capabilities that are now possible with new algos and new ways to use old algos.
How do the Realta and Reon-VX compare in their motion-adaptive, per-pixel deinterlace of 1080i signals? Most 1080p displays on the market today use motion-adaptive, region-based deinterlace.

The differences were very dramatic. I'm sure there will soon be reports on how Reon VX-based systems do on these tests as well as other HD tests starting to show up. I know of at least one major reviewer preparing to publish tests of a lot of currently available systems using this beta HD-DVD's test sequences. The Reon VX was only introduced this past summer, so give it a little time before you see new products introduced with it. Denon and Mitsubishi were surprisingly fast coming to market with Reon VX.
Have any other CE vendors expressed interest in using your silicon in consumer display products? Can you say whether any CE vendor expects to use the Reon-VX in a rear projection display within the next 6-12 months?

What presence does Silicon Optix expect to have at CES in January?

Thanks Oyfoo. Very interesting info!

Thanks oyfoo,

some clarification questions:

(1) From what you say it sounds like the Reon supports a subset of the algs the Realta can run. Would you mind listing which algs the Reon does not run in comparison to the Realta?

(2) You mentioned cadence tests. So the Reon only supports 3:2 and 2:2?

(3) The algs that the Reon supports - is it every bit as good image quality wise as the Realta in these algs, or are there differences in PQ, too? E.g. edge diagonal processing. Is there a difference in PQ between Realta und Reon there?

Thank you!

oyfoo,

very interesting indeed. Welcome to the forum and we hope you can participate and share more of your wisdom.

Thanks a lot oyfoo,

This is a good start of clarifications.

I'm now waiting for your response to madshi additional questions.

Richard.

Thanks oyfoo,

some clarification questions:

(1) From what you say it sounds like the Reon supports a subset of the algs the Realta can run. Would you mind listing which algs the Reon does not run in comparison to the Realta?

(2) You mentioned cadence tests. So the Reon only supports 3:2 and 2:2?

(3) The algs that the Reon supports - is it every bit as good image quality wise as the Realta in these algs, or are there differences in PQ, too? E.g. edge diagonal processing. Is there a difference in PQ between Realta und Reon there?

Thank you!

Hi Madshi,

Re your Q1: Sorry, I can't discuss this in detail.
Re your Q2: Yes
Re your Q3: Let me just say that Reon VX performs many algos just as good as Realta, and surprisingly better in a couple of them, but Realta, in general does better, and can continue to be improved because its algos can be reprogrammed, re-tweaked, and re-optimised. That's why we have "HQV Powered by Teranex" as a logo approved for Realta-based products, and plain "HQV" for Reon VX-based products. Realta does better owing to its use of the massively parallel, pixel-oriented Teranex architecture. The Teranex Xantus box probably has in excess of 3 Teraops/sec image processing capability (Realta has about 1 Teraop just within its video processor, not including a couple other processing engines). Specifically, the Reon's edge diagonal processing is just as good as Realta's standard algo load (Realta's could be improved even more, but there is a point of diminishing returns, and why sacrifice other functions just to improve processing of a line inclined just a couple more degrees closer to horizontal), and maybe a shade better. Having said that, not all of our customers elect to implement all the algos we make available to them, for reasons that baffle us sometimes.

By the way, there are a lot more PQ tests that we have than are on our released DVD (mainly because it takes a lot of time to secure releases from the original owners of the material and/or shoot equivalent material). Some of this is familiar to the readers of this forum (different clips from movies, anime sequences, broadcast video clips, for example) and others are specially designed by different organizations to stress deinterlacers/scalers/noise reducers/ For example, we have the same material in uncompressed format as well as a large selection of compressed formats so we can evaluate BAR, CNR, etc that shows up after this material, at different bandwidths is decoded by different hardware and presented to our chips.
We will release more of these tests in the future as we have been pleasantly surprised by the positive reception of the current HQV DVD. This will also come out in HD and Blu-Ray versions in the future.

Thank you!

Hi Madshi,


. Realta does better owing to its use of the massively parallel, pixel-oriented Teranex architecture. The Teranex Xantus box probably has in excess of 3 Teraops/sec image processing capability (Realta has about 1 Teraop just within its video processor, not including a couple other processing engines). Specifically, the Reon's edge diagonal processing is just as good as Realta's standard algo load (Realta's could be improved even more, but there is a point of diminishing returns, and why sacrifice other functions just to improve processing of a line inclined just a couple more degrees closer to horizontal), and maybe a shade better. Having said that, not all of our customers elect to implement all the algos we make available to them, for reasons that baffle us sometimes.



I really wish SO would manufacturer a stand alone box with all the soup and nuts to show what really could be with all the available algo and updates.

I really wish SO would manufacturer a stand alone box with all the soup and nuts to show what really could be with all the available algo and updates.
Ditto!

I really wish SO would manufacturer a stand alone box with all the soup and nuts to show what really could be with all the available algo and updates.

Hi,

We do (or at least, our Teranex division does). For a table of specs for different models (from 5 Tops to 30 Tops models) check out this link:
http://teranexlive.dimentians.com/home/platforms/VideoComputer.cfm?CFID=&CFTOKEN=3147f12cdf94203f-393A1D07-7E90-E2A3-B46D21E996A59580

I think these will run just about all the Teranex algos.
A detailed table of specs can be downloaded.
These are rack-mounted units designed for high-end users (dual 10b HD-SDI, 1.5Gbps) running the boxes 24/7 and are pretty standard equipment with most of the broadcast stations, post-production houses etc

The Teranex website also has info on some other interesting products. I like the Mini myself. It has been used to do upconversion of SD broadcast cameras to HD by slapping it onto the camera with velcro. It supports HD-SDI as well as DVI and HDMI and is the size of a small book. It's more affordable than the other Teranex boxes.

You can download a review of the Mini from "Videographer", a trade magazine for post-production house pros, written after the Mini was introduced at the most recent NAB. The review describes some of the advanced algos and controls the Mini offers, things I'm sure you guys would love, and which may eventually make it to consumer electronics products if there is a demand for them.

I don't think the Mini runs as many different algos as are available on the 2RU,3RU, and 6RU boxes, but at least it allows us to ship some very useful algos that may not otherwise be easily available at reasonable cost.

It would be great if it outputted 1080p. Last time I looked at it, it didn't.


BTW Thanlks for the info. It is indeed a pleasure to have you here.

Thanks again for the great input to the forum

Like Mark says the MINI does not output 1080P, and from some of the other forum members the Teranex boxes where not very consumer friendly( I think Alan can chime in with his experience).

It seems to me that since SO OEM's the NEC Theatersync and Digital Projection Video Processors I would like to the a Teranex or SO using those same boards but with all the Algos functioning and on going upgrades. I think Vantage, and Algolith have droped and fumbled the ball with the HQV chip. The NEC Theatersync is functional but lacks any upgrades.

Bottom line SO or Teranex should set the standard in a stand alone box with all the bells and whistles of the HQV chip and not leave it to other manufaturers implementation.

Have any other CE vendors expressed interest in using your silicon in consumer display products? Can you say whether any CE vendor expects to use the Reon-VX in a rear projection display within the next 6-12 months?

What presence does Silicon Optix expect to have at CES in January?

Hi,

Yes, we have a tremendous amount of interest from CE companies in using our silicon in consumer products. It's a good news/bad news situation as we don't have resources to support everyone, as we are a 6-yr old VC-funded private company, and have to account for how we allocate resources and spend our money. We're of course mostly interested in high-volume opportunities (ask any semi mfr and that's what they'll tell you) and name-brand customers (our board always likes blue-chip companies in our customer list, makes for nicer logos on the powerpoint presentations!).

We'll definitely be at CES and there will be a lot of new products shown using both Realta and Reon VX. I can't tell you the brand names and I can't tell you whether all our customers will have products ready to show at CES, but we expect to show more projectors (1080p as well as 720p), more DVD players (Blu-Ray as well as HD), TVs (rear projection as well as flat direct-view), and we will introduce some new products of our own as well, which I think many in this forum will appreciate. We typically have suites rather than booths in public areas at CES (cheaper and better control over who gets to see sometimes unannounced products), but there will certainly be some press coverage of some of the products. The products that will be publicly announced will certainly be shown on the show floor at the manufacturer's booths as well so everybody will get to see them.

Yes, we have a tremendous amount of interest from CE companies in using our silicon in consumer products. It's a good news/bad news situation as we don't have resources to support everyone, as we are a 6-yr old VC-funded private company, and have to account for how we allocate resources and spend our money. We're of course mostly interested in high-volume opportunities (ask any semi mfr and that's what they'll tell you) and name-brand customers (our board always likes blue-chip companies in our customer list, makes for nicer logos on the powerpoint presentations!).

We'll definitely be at CES and there will be a lot of new products shown using both Realta and Reon VX. I can't tell you the brand names and I can't tell you whether all our customers will have products ready to show at CES, but we expect to show more projectors (1080p as well as 720p), more DVD players (Blu-Ray as well as HD), TVs (rear projection as well as flat direct-view), and we will introduce some new products of our own as well, which I think many in this forum will appreciate. We typically have suites rather than booths in public areas at CES (cheaper and better control over who gets to see sometimes unannounced products), but there will certainly be some press coverage of some of the products. The products that will be publicly announced will certainly be shown on the show floor at the manufacturer's booths as well so everybody will get to see them.That is excellent news.

I'm looking forward to CES. :D

Why don't we see any "low price" scaler based on ReonVX ?

Jim Peterson said in June that Lumagen was thinking about a Radiance version based on this chip, but no news...

Actually, there is no 1080i capable scaler (except Lumagen Vision line) in the 1000$ to 2000$ MSRP price range...The ReonVX seems to be best the way to produce such video processors...Richard.

Exactly ! $3,000 for a video scaler is out of the question for me, but at ~ 1,000 for one with 1080P output based on the Reon VX may work for my set up and budget.

Why don't we see any "low price" scaler based on ReonVX ?

Jim Peterson said in June that Lumagen was thinking about a Radiance version based on this chip, but no news...

Actually, there is no 1080i capable scaler (except Lumagen Vision line) in the 1000$ to 2000$ MSRP price range...The ReonVX seems to be best the way to produce such video processors...Richard.

Exactly ! $3,000 for a video scaler is out of the question for me, but at ~ 1,000 for a video processor with 1080P output based on the Reon VX may work for my set up and budget. I have a sony60" SXRD 1080P rear projection TV (A2000) and a panasonic AE 900 720P projector and both would benefit from state-of-the-art video processing.

We will release more of these tests in the future as we have been pleasantly surprised by the positive reception of the current HQV DVD. This will also come out in HD and Blu-Ray versions in the future.

Some new test results using our beta HD HQV disk:

http://www.hometheatermag.com/hookmeup/1106hook/

This may be a stupid question, but Realta and Reon use 4 field motion adaptive deinterlacing and the DVDO VP50 claims 5 field. What is the difference between the two? Does 5 field versus 4 field make any difference? Which one of these high-end beasts, Calabre Vantage or DVDO VP50 is going to produce the best and clearest PQ for fast moving action sports and fast horizonal panning scenes for 1080i, 720P and SD source material? Why?

This may be a stupid question, but Realta and Reon use 4 field motion adaptive deinterlacing and the DVDO VP50 claims 5 field. What is the difference between the two? Does 5 field versus 4 field make any difference? Which one of these high-end beasts, Calabre Vantage or DVDO VP50 is going to produce the best and clearest PQ for fast moving action sports and fast horizonal panning scenes for 1080i, 720P and SD source material? Why?
Really, 4-fields is all you need to confirm stillness. Field 1 is a match with field 3, and field 2 is a match with field 4.

Just 2 fields is not enough, because the fields consist of different scanlines--they will never match even if they are from the same film frame or a non-moving image, and the only way to detect non-motion is with a computationally difficult, and unreliable, measurement of the comb factor (this is particularly difficult on a per-pixel basis).

3 fields would almost work, but might occasionally produce false detection of stillness (and therefore combing artifacts) in the case where there is motion lasting for 1 field (specifically, field 2, which is not matched with any other field, as opposed to fields 1 and 3, which are matched with each other). This could happen with a video overlay that lasts for 1 field. If the motion detector only relied on 3 fields to detect stillness, then 1 field overlays on non-moving images would always comb.

4 fields allows true stillness detection because every pixel is confirmed identical to a field before or after it. There is no way for motion to "sneak through" and be detected as non-motion.

As far as I can tell, the only advantage of 5-field motion adaptive detection is in the case of source material which has repeated-frame errors. If a single frame is repeated, this would cause a 4-field match, but not 5-field. I'm not sure how common this is in material across the board, but I do own a DVD that has repeated frame errors and it looks decidedly better on the VP50 than with 4-field motion detecting deinterlacers. The reason is that the repeated fields don't comb. You still perceive a slight judder (unavoidable as the error is in the source) but it doesn't comb. So in this scenario, the 5-field motion detection is really showing its advantage.

The big disadvantage of 5-field detection is that there is now one extra field of bobbing whenever motion comes to a stop, because the lookahead delay buffer in the VP50 is less than 5 fields. This is particularly noticeable on subtitles, overlays, and moving cursors, but is also noticeable just on ordinary video. In my opinion, a video processor's delay should always be long enough to look ahead the full number of fields for which motion is being detected. This is only problematic with video games, and that's what the game modes are for.

The VP50 also has a quirk where the look-behind buffer is not large enough! Non-moving objects bob just before they start moving. This can be easily confirmed by displaying a menu with sharp edges (text) or a cursor--when it disappears, the bobbing in the menu's text is noticeably visible for an instant before it disappears. So although 5-field motion detection does have an advantage in a certain case, the motion detection algorithm in the VP50 really isn't as good as it could be.

I haven't (yet) been able to test any HQV devices, so I don't know how its motion detection performs in these types of situations.

4 fields allows true stillness detection because every pixel is confirmed identical to a field before or after it. There is no way for motion to "sneak through" and be detected as non-motion.
Would that it were so. That would make life lots easier.

But it's not reality. It's very, very possible for there to be motion which is such that a comparison across 2 field periods indicates no motion while there is still motion between fields which are 1 field period apart.

An easy example case to consider is one where there is a regular spatial feature with a rate of motion which causes it to repeat or overlap itself every 2 field periods. In this example, all 2-field comparisons will yield no motion, but there would in fact be significant presence of interlace motion artifacts in every combination (weave) of 2 adjacent fields.

Simply comparing 2 field-pairs spaced 2 field periods apart just doesn't cut it.

- Dale Adams

What I meant to say, is that 4 fields allows stillness detection because each pixel can be confirmed identical to the next or previous corresponding even/odd field (so not actually the next or previous field).

So:


Field 1: Field 2: Field 3: Field 4:

xxxxxx ------ xxxxxx ------
------ xxxxxx ------ xxxxxx
xxxxxx ------ xxxxxx ------
------ xxxxxx ------ xxxxxx
xxxxxx ------ xxxxxx ------

Fields 1 and 3 can be directly compared, and fields 2 and 4 can be directly compared. But, for example, field 2 cannot be directly compared to fields 1 or 3 because the pixels are on different scanlines (the comb factor can be measured, but that's unreliable).

But if fields 1 and 3 match, and fields 2 and 4 match, then it is known that: Fields 1 and 2 can be safely weaved (for the matching pixels) because they are from the same frame. Fields 2 and 3 can be safely weaved, and fields 3 and 4, for the same reason.

But say we have this:


Field 1: Field 2: Field 3:

xxxxxx ------ xxxxxx
------ xxxxxx ------
xxxxxx ------ xxxxxx
------ xxxxxx ------
xxxxxx ------ xxxxxx

Let's say that fields 1 and 3 match. Then in most cases, field 2 could be "guessed" to be from the same frame (for the pixels above and below the matching pixels from fields 1 and 3), but not always. If an overlay appears for only field 2, then by weaving fields 1 and 2, and fields 2 and 3, you would get combing.

With 4 fields, you can be sure that this is not happening, because field 2 matches field 4.

You can still get combing with 4-field motion detection if two fields that are not from the same point in time occur repeatedly in the source. However, this is a source error--it doesn't happen with any normal pulldown. On a native interlaced TV, you'd see flicker (and a sort of combing effect). 5-field motion detection allows a single frame to be repeated erroneously and not comb.

An effect that can also cause combing is for something to flicker at a 30Hz rate. I've seen this happen in video games. This is intended to be seen, as, indeed, flashing at 30Hz. However, a motion detector will detect it as non-moving and what will be seen on a progressive-scan display is a still, combed, image. The only way to alleviate this is to integrate comb detection into the motion detection algorithm. The Greedy (High Motion) plugin for DScaler does this, but few other modern deinterlacers do. Furthermore, it is generally regarded as better not to do this--many deinterlacing test patterns use horizontal stripes aligned to scanlines (even fields are all white, odd fields are all black), and a "pass" result is for it to be seen as still stripes, and a "fail" result is flickering all white and black.

An easy example case to consider is one where there is a regular spatial feature with a rate of motion which causes it to repeat or overlap itself every 2 field periods. In this example, all 2-field comparisons will yield no motion, but there would in fact be significant presence of interlace motion artifacts in every combination (weave) of 2 adjacent fields.
This seems like much the same case as my example. Say there is a horizontal pan with a repeating pattern:

ABCDEFGHABCDEFGHABCDEFGHABCDEFGH
Say the camera pans to the right:


Field 1: ABCDEFGHABCDEFGHABCDEFGH
Field 2: EFGHABCDEFGHABCDEFGHABCD
Field 3: ABCDEFGHABCDEFGHABCDEFGH
Field 4: EFGHABCDEFGHABCDEFGHABCD

Yes, this will cause combing artifacts. But if the pattern continues:


Field 5: ABCDEFGHABCDEFGHABCDEFGH
Field 6: EFGHABCDEFGHABCDEFGHABCD

Then even a 5-field motion detector will detect fields 1-5, or 2-6, as nonmoving, and combing will occur. But if the repeating pattern lasts only for 4 fields, then the 5-field motion detector makes the difference, resulting in no combing, vs. brief combing for the 4-field motion detector. If the pattern continues forever, then it seems the only way to prevent combing is if the motion detector considers comb measurements in determining the presence of motion.

Did I get your example correct? :)

Therefore, if my analysis is correct, then 4 fields is enough to distinguish motion from non-motion, except in the case of temporal aliasing (cyclic motion aligned to 30Hz). Brief periods of temporal aliasing can be alleviated by requiring more fields to detect stillness, but 5 fields isn't a magic number; it could be 6, or 7, and the more fields required, the longer the duration of temporal aliasing that would not result in combing. Perhaps 5 makes sense because single instances of repeated frames are not detected as stillness. Maybe, testing on a wide range of material, 5 fields yields significantly better quality than 4, but the returns diminish after 5 (Dale, you'd obviously be in a better position to know that. :)) But once you go beyond 4, the temporal aliasing case is the only advantage. And the big disadvantage is the increased delay, or increased bobbing in many situations when weaving would be the best thing to do, if the delay is shorter than the number of fields being tested for.

Perhaps an ideal solution is, indeed, to integrate advanced comb detection into the motion detecting algorithm. Which changes the character of the algorithm into more of motion/coherence detection, rather than just motion detection. (This also means you're going to fail the horizontal stripe test pattern, but if the comb detection is accurate enough, that's ok.)

I checked the spec sheet for the mini on the Silicon Optix web site and it now list 1080p 60 as the maximum output resolution.

Mark:

Did they add dual HD-SDI output to handle 1080P60?

I have no idea. Go to the Silicon Optix web sight and get to the mini and click on the spec sheet.

Just checked no dual HD SDI output so no 1080P60 output via HD-SDI, in fact no output resolution listing 1080P60 by the mini, just a note that the optional DVI and HDMI outputs have a maximum output resolution of 1080p60.

I assume that it will do 1080p 60 via the optional DVI or HDMI outputs. Rather misleading if the unit won't but the optional out puts will pass it.

Hi Madshi,

Re your Q1: Sorry, I can't discuss this in detail.
Re your Q2: Yes
Re your Q3: Let me just say that Reon VX performs many algos just as good as Realta, and surprisingly better in a couple of them, but Realta, in general does better, and can continue to be improved because its algos can be reprogrammed, re-tweaked, and re-optimised. That's why we have "HQV Powered by Teranex" as a logo approved for Realta-based products, and plain "HQV" for Reon VX-based products. Realta does better owing to its use of the massively parallel, pixel-oriented Teranex architecture. The Teranex Xantus box probably has in excess of 3 Teraops/sec image processing capability (Realta has about 1 Teraop just within its video processor, not including a couple other processing engines). Specifically, the Reon's edge diagonal processing is just as good as Realta's standard algo load (Realta's could be improved even more, but there is a point of diminishing returns, and why sacrifice other functions just to improve processing of a line inclined just a couple more degrees closer to horizontal), and maybe a shade better. Having said that, not all of our customers elect to implement all the algos we make available to them, for reasons that baffle us sometimes.

By the way, there are a lot more PQ tests that we have than are on our released DVD (mainly because it takes a lot of time to secure releases from the original owners of the material and/or shoot equivalent material). Some of this is familiar to the readers of this forum (different clips from movies, anime sequences, broadcast video clips, for example) and others are specially designed by different organizations to stress deinterlacers/scalers/noise reducers/ For example, we have the same material in uncompressed format as well as a large selection of compressed formats so we can evaluate BAR, CNR, etc that shows up after this material, at different bandwidths is decoded by different hardware and presented to our chips.
We will release more of these tests in the future as we have been pleasantly surprised by the positive reception of the current HQV DVD. This will also come out in HD and Blu-Ray versions in the future.

Oyfoo,

I recently demo'd the new Denon 2930 and 3930 DVD players. I wasn't able to do an actual A-B comparison because I didn't have both players at the same time. However, for film-based Region 1 DVDs, the deinterlacing and scaling seems to be very, very similar with both players which is to say EXCELLENT. What, if any, differences or improvements in deinterlacing and scaling can the Realta provide over the Reon for typical film-based DVD material? Are they almost identical in this regard? Thanks.

Dale and Jeff- thanks for the good information and pros/cons of the 5 vs 4 field deinterlacing. I see now why fast horizonal pans are difficult and look blurry on your average progressive scan DVD player and why either the Calabri Vantage or DVDO VP50 will do a significantly better job upconverting DVDs and SD signals with regard to fast moving objects, horizontal pans and 3:2 pulldown, etc. But what if most of what I watch critically is HD 720P and 1080i from a satellite box that only needs to be upconverted to 1080P native resolution of my TV. It's a no -brainer now that for standard definition and DVD that these algorithms will enhance the picture quality, but how about a signal that's already high-def (720P or 1080i) that needs only to be converted to 1080P?

Aren't the extra frames (3:2 pulldown) and anti-combing techniques (4 or 5 field motion adaptive deinterlacing) already applied to these signals prior to broadcasting and therefore there isn't much additional benefit of better conversion to 1080P (assuming no down-rezing of the signal by the TV)? Is the answer different for 720P vs 1080i?

Aren't the extra frames (3:2 pulldown) and anti-combing techniques (4 or 5 field motion adaptive deinterlacing) already applied to these signals prior to broadcasting and therefore there isn't much additional benefit of better conversion to 1080P (assuming no down-rezing of the signal by the TV)? Is the answer different for 720P vs 1080i?
These all fall under deinterlacing, and yes, they are needed for 1080i, but not 720p. You don't want your satellite box doing any conversion from the original decoding, because the video processor (at least one from DVDO or based on the HQV algorithms, or some of the other latest competing processors) will do a better job.

The ability to perform the advanced deinterlacing on HD video was the big new feature of the VP50 vs. the VP30 with the ABT102 card, which could only do it for SD, and used simple field upscaling on 1080i sources. Realta-based devices, such as the Vantage HD, can already do advanced 1080i deinterlacing.

Ideally a 1080i signal arrives as 3:2 pulldown (if it is film sourced), and although it is possible to perfectly reassemble this into the original film frames, this is what the deinterlacing after your satellite box must do--it can't be done before broadcasting because it is broadcast in interlaced form. 720p is progressive, but this is only ideal for video-sourced material--1080i is used for film because it yields the ultimately higher resolution, if your video processor or TV can perform inverse 3:2 pulldown.

Jeff - You indicate that Ideally a 1080i signal arrives as 3:2 pulldown (if it is film sourced), .....

So I get two different 1080i signals from satillite? 1080i movies requiring reverse 3:2 pulldown (e.g. HBO) and 1080i video (e.g. HDNet original programming) not requiring reverse 3:2 pulldown. Why couldn't the process of converting film to 1080i just include a first step of conversion to 1080P (with reverse 3:2 pulldown applied) and then convert these frames of 1080P to interlaced format (i.e. 1080i without the need for reverse pulldown). Unless I'm missing something, then all that would be required to convert accurately to 1080P would be proper deinterlacing (i.e. no 3:2 reverse pulldown required)?

I still see a need for high-end deinterlacing to get best picture, but I don't get why true HD 1080i film-based material needs 3:2 pulldown if the creation of the digital master is first done in 1080P. Are ou saying that even if I have an HD DVD or BlueRay player that outputs 1080i, then I'll still need to do 3:2 pulldown to get accurate picture for any film based material? Much thanks for your help.

I just looked up the mini and it indicates the highest output resolution is 1080i/~60. For my application I need maximum output resolution of 1080P/~60. This is probably because this item was geared toward broadcasters who only output maximum 1080i due to bandwidth considerations.

Oyfoo- do you think that it's possible to add this capability to the mini through a software patch for us home theater forum guys? I seems the chipset should easily handle the 1080i to 1080p noise reduction, Pixel motion, 3:2 reverse pulldown, then I'm fairly certain the functionality of the mini would work for me.

I especially like the split screen mode that allows you to view the noise reduction processed video on one side and the unprocessed signal on the other. Sweet.

Jeff - You indicate that

So I get two different 1080i signals from satillite? 1080i movies requiring reverse 3:2 pulldown (e.g. HBO) and 1080i video (e.g. HDNet original programming) not requiring reverse 3:2 pulldown. Why couldn't the process of converting film to 1080i just include a first step of conversion to 1080P (with reverse 3:2 pulldown applied) and then convert these frames of 1080P to interlaced format (i.e. 1080i without the need for reverse pulldown). Unless I'm missing something, then all that would be required to convert accurately to 1080P would be proper deinterlacing (i.e. no 3:2 reverse pulldown required)?

I still see a need for high-end deinterlacing to get best picture, but I don't get why true HD 1080i film-based material needs 3:2 pulldown if the creation of the digital master is first done in 1080P. Are ou saying that even if I have an HD DVD or BlueRay player that outputs 1080i, then I'll still need to do 3:2 pulldown to get accurate picture for any film based material? Much thanks for your help.
Is your question can flags be placed in the 1080i signal so no image analysis is required for 1080p conversion? Theoretically, yes. Which is what the HD-DVD format supports, and what future 1080p HD-DVD players will support. (I think BD supports it differently, storing as native 1080p meaning it comes out of the decoder as progressive, but the result is similar.) However, then your satelite box would have to switch modes depending on whether flagged content is being sent or not, which is not something you would likely find in a satelite box. Keep in mind this might have to be frequent, for example, at every commercial break--in short, not going to happen. Furthermore I wouldn't put too much reliance on broadcasters properly flagging the content.

Other than possible flagging, 3:2 pulldown is a 2-way reversible operation. If you re-interlace 1080p which has 3 full frames/2 full frames alternating, then you have the original 1080i source. So reinterlacing a progressive stream that was created via inverse 3:2 pulldown doesn't actually do anything useful.

So yes, if your HD-DVD/BD player outputs 1080i, you need a good deinterlacer to get proper 1080p due to the image analysis required. Depending on the source, a "lesser" deinterlacer might do ok, also. This depends on the frequency of bad edits, the amount of video noise in the image, the amount of vertical detail in the image, among other factors.

Oyfoo,

I recently demo'd the new Denon 2930 and 3930 DVD players. However, for film-based Region 1 DVDs, the deinterlacing and scaling seems to be very, very similar with both players which is to say EXCELLENT. What, if any, differences or improvements in deinterlacing and scaling can the Realta provide over the Reon for typical film-based DVD material? Are they almost identical in this regard? Thanks.

Realta and Reon VX are pretty equal in their deinterlacing and scaling performance. Realta can generate output resolutions all the way out to 2560x1600 (looks spectacular when scaling 1080i material), and Reon VX only goes out to 1080p.


Oyfoo- do you think that it's possible to add this capability to the mini through a software patch for us home theater forum guys? I seems the chipset should easily handle the 1080i to 1080p noise reduction, Pixel motion, 3:2 reverse pulldown, then I'm fairly certain the functionality of the mini would work for me..

I believe that the 1.5Gbps Serial Digital Output is limited to 1080i. The optional DVI and HDMI outputs will support 1080p. No s/w patch needed. Call Teranex, I'm sure they'd be glad to sell you one if you do a good Ted Turner or Rupert Murdoch imitation! Just kidding.

I believe that the 1.5Gbps Serial Digital Output is limited to 1080i. The optional DVI and HDMI outputs will support 1080p. No s/w patch needed. Call Teranex, I'm sure they'd be glad to sell you one if you do a good Ted Turner or Rupert Murdoch imitation! Just kidding.

That's good news and an excellent solution if "supporting 1080p output from optional DVI or HDMI output" means that all the functionality of the unit (scaling, Pixel Motion deinterlacing, noise reduction, 3;2 pulldown, etc) will extend to work on a 1080i input signal. Thank you for replying back! Unfortuntely, the Mini is actually more expensive (3-5K?) than other solutions so this is not cost effective solution.

UPDATE: Silicon Optix sales just got back to me and the Mini up down cross converter does not output 1080P and there are no near-term plans for this to be added. Max output resoluton is 720P.

It appears that no video processors based on the Reon chip were announced at CES. This is especially disappointing in light of the fact that lorelevitt found the Reon in his Toshiba XA2 HD DVD player to be better than the Gennum VXP in his Crystalio II. I really thought that we would see a mid-priced Reon based video processor announced at CES.

In light of the overheating issues that they have with the Dragonfly, I was really hoping that Algolith would come out with a Reon based processor they can be passively cooled. They could have the Dragonfly and Mosquito as their flagship de-interlacer, scaler and compression artifact reducer products and the Flea and a Reon based processor as their "budget" solutions.

Even better would be a product that combined the Reon and the Flea in one box for say $1500. I don't think that price point is unreasonable since both the Flea and Reon are sold as "plug and play" solutions that have few options to tweak. I know the "limited" functionality would detract some people here from getting it, but the only other product with a de-interlacer, scaler and compression artifact reducer in one box is the upcoming Lumagen which is $4000.

Is anyone at Algolith listening? :)

I think that SO is focusing on integration partners (projectors, displays and player manufacturers). This makes sense because the VP market is pretty small.

I am reading all this and trying to make sure the Denon DVD players and/or AVRs that use Realta based processing do indeed support 1080/24p. It sounds like it does but not obvious. A nod would help.

This is an excellent thread.I currently have the Sammy BDP5000 with it's Reon and will be looking at new new Marantz blu-ray(Realta).Wonder if any difference in performance-BD and upscaling DVD.I know quality in the Marantz will be better.

I am reading all this and trying to make sure the Denon DVD players and/or AVRs that use Realta based processing do indeed support 1080/24p. It sounds like it does but not obvious. A nod would help.

The Realta chip in the Denon DVD-3800BDCI can apply all picture adjustments to a 1080p/24 signal from a Blu-ray disc. The Realta chip Denon AVP-A1 pre/pro cannot perform any picture adjustments to a 1080p/24 signal. I don't know how the other Denon AVR's work.

I hope this helps.

Mark

The Realta chip in the Denon DVD-3800BDCI can apply all picture adjustments to a 1080p/24 signal from a Blu-ray disc. The Realta chip Denon AVP-A1 pre/pro cannot perform any picture adjustments to a 1080p/24 signal. I don't know how the other Denon AVR's work.

I hope this helps.

Mark

It may have helped him 2 months ago when he asked. :)

It may have helped him 2 months ago when he asked. :)

Time must really fly on your calendar. The question was asked two weeks ago.:)

Time must really fly on your calendar. The question was asked two weeks ago.:)

Or maybe your calendar is broken? The question was asked 03-07-08.

Or maybe your calendar is broken? The question was asked 03-07-08.

Sorry, you are correct. I looked at the date at the bottom of the post and that date does not go with that post. My mistake.:o

It may have helped him 2 months ago when he asked. :)

Better late than never:confused:

Mark

I'm a bit of newbie...please be patient with my dumb questions.
I have a Sony XBR4 which I can not stand and am this close to ordering an Olevia 747i because I've read nothing but great things about it's picture quality due Olevia's inclusion of the Realta SoC. However, I'm a bit hesitant to purchase the set because WOTS is Olevia could go under any minute.
I see that there is a Vizio plasma coming out next month which features the Reon chip. Should this TV be able to handle standard def TV as well as the Olevia despite the fact it "only" uses a Reon as opposed to a Realta?