Silicon Optix Announces the Realta Chip -- Bringing Teranex's $60k HQV to the Home - Page 3  

Well, we are still going to build our Post Production and Broadcast boxes and blades, so nothing has changed. Our performance of these devices have already been upgraded to handled high-resolution and more processing functions as well.

We have many customers that we are supporting with these products.

The Realta is targeted more for the Homer Entertainment markets, where the Teranex boxes have the I/O for the professional segments.

Dennis

dcrespo,

That's great!

Looks like you can perfect the algorithms on the professional segments and the jump to the consumer segment will be that much quicker.

Are you licensing the algorithms seperately from the chip? i.e., can a company purchase the chip and only some of the algorithms, but not all of them?

Ofer,

You hit the nail right on the head! And it's a point so crucial that I've got to reiterate:

High-end post-production and broadcast require the BEST possible image quality and Teranex has delivered just that; our algorithms have been refined over 100,000+ hours of content verification by the 'Golden Eyes' of Hollywood, broadcast, etc. and are being continuously verified and perfected before being programmed into the Realta.

"So Teranex will no longer be manufacturing stand alone video processors, unlike Faroudja whose chip was sold almost exclusively for DVD players and really limited its appearence in competing video processor boxes."

I dunno... DCDi is in more and more players, like A/V receivers, as well as some video-processor boxes.

I think a big difference with Teranex is that the pro / broadcast market was / is using Teranex and will keep doing so for a long while. If it's cheaper, that's great news, but not necessary.

This chip -- with the ability to go into products at $1500 and up (DVDO / Lumagen-priced scalers? DVD players? upscaling A/V receivers?) -- will obviously get the technology into 100x as many hands.

Now, when the next generation of the product comes out, it can go compete for the $10-20 scaler-chip market and democratize great video for all eternity!

Mark

At the party and at the Silicon Optix booth there were $2K or so consumer video processing boxes which used the last generation Silicon Optix chips. I got sheets and prices on those boxes. My question to them was whether they were going to make boxes with the new chip as well. Depending on who I asked I got different answers. Some said they wanted to sell chips not make boxes. Others said a box from them would be available within 6 months or so. At the booth, the sales guys said they would have a replacement box but it was about 1 year out. The Algolith guys at the party said they would have a combined SiliconOptix/Algolith box out about the end of the 1st Q 2005 (about 6 months from now) for about $3.5K. Hopefully we can start an Algolith thread with Algolith staff participation were we can lobby for features or added cost options such as SDI and HDSDI in and the use of BNC connectors instead of RCAs.

Ofer. Nothing to be embarrased about. I am in the business and my plane RT ticket to Indy was only a little over $100. If the Cedia show was in the middle east (and even if travel was safe over there) I still wouldn't have gone given the cost of the airfare and the travel time necessary. Are you considering going to the phone show in Las Vegas this January? :)

Dcrespo:

One of the features, and to some extent the very high cost of Teranex video processing was the enormous amount of computing horserpower packed into them, almost like a super computer, does this new Realta Chip require less computing power than say the Xantus?

Lon

Mark,

I think the overall costs were around $2K, and since I couldn't merge it with any business trip - it would have to be taken from my own pockets (& I'm a real cheapskate...). I have no idea what my travel itinerary will be this upcoming year.

LJG,

I think you misunderstand. The Realta does not require any computing power - it is a processor all of its own. It is not like the FLI2X00, where you still need a main processor to drive OSD and other features. It is a very very powerful processor and it runs Terranex's algorithms for video processing (in fact very similar ones to the Xantus).

Basically, Terranex could rebuild new boxes that have 10x the comping capacity if they design them around Realta chips...

From what I understand, the Realta can also be the processor that runs the entire show + does OSD and other neat stuff.

Ofer:

So no super computer needed, WOW.

LJG,

Exactly, the Realta is the super computer...

"PQ to the masses has nothing to do with it."

Not to go OT, but this patronizing [over]generalization gets old. A good portion of "regular" people I know never considered RPTV because ''it's not as clear" as their direct view TV's, and I agree. It hasn't been til relatively recently that RPTV has gotten competitive with tube sets, and I think that's why sales have skyrocketed.

Of course DVD and HD are necessary to give the PQ at the bigger sizes, but my poiunt remains.

To some people, PQ does make a difference. However, one must define PQ. If overly processed sharp and high temperature pictures (brighter) mean PQ than I guess the masses want that. If a dim picture that contains the HF detail means PQ then I would still maintain to the masses (not the video nerds here (you and me)) PQ means little. Look how well early plasmas sold and those LCD flat panel with all that contouring or whatever. Picture quality of plasmas is getting better and better and the manufacturers are traveling all over the country to teach their dealers how to sell a better quality picture. Noah you live in an ivory castle. Make it brighter and sharper and the masses will buy it. Maybe 15% of the market knows what good PQ is and would choose it. But even if the rest, the masses, saw it and knew it was indeed better, they still would buy the brighter and sharper because that's what they want. The job of the market is to fill a need and if there isn't a need or desire, to create it first. That is why ISF runs all those ads. They are trying to create a need for the services of ISFers.

Hi Guys,

I gotta admit I can't believe this is already happening (see my other thread here) and will be invading our Plasma displays (etc.) in coming months! :D

You'll notice from my thread in the Plasma forum, that I came to realize the dire need for such technology in today's plethora of input sources - especially for digital displays such as Plasma screens which only work well when fed cristal clear signals preferably just the way they like it (native rate):
http://www.avsforum.com/avs-vb/showt...hreadid=444985

So it looks like we're entering a new, better than ever, democratization of Broadcast/Hi-End technology. I'm happier than ever for the hundreds of laserdiscs of old movies not out on any better media. I may even start hunting down rare VHS tapes of classics not even available on Laserdisc, as the Realta chip may well make them enjoyable even on a Plasma display!

Thanks Silicon Optix - you've got things figured. :cool:

Hiya Mark,

I don't know if you recall, back in the days when the Teranex threads were lively here at the AVS forum, I had tended to make such purist statements myself. Actually, without going so far as lending to "Teranex bashing" my position was that if you totally process, and entirely recalculate, a given image, what you output is NOT the original image, but an ersatz of that image, a synthetic fabrication which, much like a cartoon or special effects, simulates reality but does not remain faithful to it.

In the audio realm, much has been understood about this: many folks prefer a given type of sound that pleases them, and don't care much about whether it is faithful to the original source which was recorded - to the dismay of those who think that Hi-Fi means High Fidelity to the original event.

In the video realm, much is the same as you pointed out eloquently above. However, the Hi-Vi community seems less developed with many folks prefering eye-boggling blasts of oversaturated color sometimes accompanied by butt-kicker vibrations. :eek:

The CASE I am however willing to make for the advent of a low priced widely distributed Realta chip is the following:

When sources are so poor that viewing enjoyment is reduced, then it would be nicer to have a watchable "false" picture processed by its Teranex algorithms than to switch program material.

Sure, it won't be as REAL as a non processed source without cybernetic intervention. Definitely, sizeable chunks of its picture will not be truthful to the original, carrying even the risk of betraying the original source in ways which go against its artistic intent (imagine Hitchcock's dark scenes artificially brightened for clearer viewing of the scary silhouettes = loss of suspense).

This is why, just like for automatic suspension control for sportscars, it is PARAMOUNT that we be able to switch it OFF whenever we want! Otherwise, it wouldn't be a Porsche if it's tail won't spin out in a hairpin, and it wouldn't be true CINEMA if you toy around aimlessly with the Director's artistic intent.

Another reason why it might be necessary to require that it be able to be disactivated, is that using a specific subcarrier code detection system, these chips could be pre-programmed to change or interfere with given images that "whoever" may not want us to see - their technlogical roots are, after all, military - let that signify whatever you choose to interpret it as meaning.

But for cases, like in my thread above, where you are inputting poor quality files off of the internet, or trying to view ageing demagnetized video tapes, then it might just turn the trick, making a case for occasional purpose built FAKE REALITY.

Cheers,

Brett. Very nice post. However, I was not bashing Silicon Optix at all. I was praising them. I have always wanted a Teranex and now with the upcoming Dragonfly from Algolith, I will be able to have Teranex quality deinterlacing and scaling with mosquito noise reduction from Algolith all for under $3500! I suspect, but do not know, that the Realta's 1080i deinterlacing will be superior to the the DXP chip's 1080i deinterlacing. From what I've heard, the DXP chip is quite a bit less expensive than the Realta and thus I think it willll show up in more mass market products. Me, I'll buy stuff with the Realta chip, if it is in fact what I think it is.

I have to admit I was bushing a little bit, but the bashing was of Noah, a long time AVSer with who I have cross posted with before, and one who I have the utmost respect for. I think a little friendly kidding of each other with no malice is OK. I am sure my friend Noah understands my ivory tower crack.

All kidding aside, one makes a lot of friends here at AVS and ocassionally at shows and elsewhere we get to meet and become more than posting friends on the internet.

The Realta chip and the HQV chip are the same. The powerful Realta chip brings 'Hollywood Quality Video' using well-proven Teranex algos to your home, thus the HQV name. Only this chip and the Teranex chips can run the Teranex algos. It takes a LOT of horsepower to run per-pixel processing and the teraops/sec needed to do this.

By the way I work for Teranex, now a business unit of Silicon Optix.


Ray

Hi Ray,

Welcome to the AVS forum, the best place in the universe for Home Theater. :) You're sure to get interested in many of the topics and discussions here, folks in this forum are a great crowd to mix and mingle with - loads of new ideas and tons of professional insight.

By the way, do you have, or do you plan to have a home theater of your own? Does it include a hot-rodded souped-up Teranex? What display?

Regarding your job at Silicon Optix, I imagine you are not free to disclose much about future market prospects. But have you guys had contact with any Plasma panel manufacturers? (no need to specify which if you're not allowed to)

Also, is Silicon Optix still working with the military to further improve the image calculation capabilities of the more recent secret "new series" of processors which is still kept outside of any civilian implementations? This would be interesting for us to know because it would mean that as soon as you are so authorized, you could then roll out future technologies to us folks in the mainstream, for our Home Theater uses and possibly even for video editing of our home movies?

Thanks both to you - and to your entire team! :cool:

This is true for any processor. For example, unless you watch 480i60 or 1080i60 on an interlaced display, you are not seeing the original info. Here I am talking about true video material.

If you resize an image, you are not looking at the original. So, anyone watching a plasma, dlp or LCD is not seeing the original.

You could argue that what you watch on DVD is not true since the source it came from was 10-bit 4:2:2 at one point and is now 8-bit 4:2:0. The interpolation filter your DVD player will use is most likely not an inverse of the low pass filter used to get you to 4:2:0.

Hi Stacey,

I guess you got me "Smack on the head of the nail" there.

Maybe I should have done a better job of phrasing that. For example, one could mention a subjective tolerance level at which you do not see a disturbingly excessive discrepancy between the original image (or master recording) and the displayed "fabricated" picture.

To make it clear, if you are watching a documentary on the Louvre museum and they show you the Mona Lisa della Joconda painting by Leonardo da Vinci, it should look more like what you would expect instead of bearing a closer resemblance to a Disney rendition of a Van Gogh painting. ;)

But humor aside, just like with audio fidelity, what is like the original versus what is "fake" will depend upon each individual's sensitivity, education of one's expectations (knowing what to look for), and priorities. What worries me is which trade-offs might be made to "clean-up" a picture which has noise in it. For example, the Realta substitute, for a noisy gradually graded colored field, a noise-free clean but unicolored field? Some folks might not mind, but others might feel they've lost in the trade-off.

This is why I stated above, that in its implementations, an on-board Realta chip would need to be DISCONNECTABLE for those rare or frequent occasions when a viewer wishes to process less rather than more.

JMHO

Mark,

"If a dim picture that contains the HF detail means PQ then I would still maintain to the masses (not the video nerds here (you and me))..."

I like a bright picture with HF detail a lot more. It adds to the illusion of 3-D and gives a sense of lifelike vividness.

But you're speaking in outdated CRT terms; there's no relation between brightness and resolution with digital displays (well, maybe a little with LC devices).

"Noah you live in an ivory castle. Make it brighter and sharper and the masses will buy it."

I guess you overestimated me:)

Sorry, everyone, for going OT.

I can't say much other than we would be fools if we didn't talk to plasma reps. One inportant thing to remember about the Realta HQV; the HD output (digital or analog) can be custom scaled to the native res of any display it is installed in - 1024x768, 1600x1024, 2048x1536, in p, SF, i, you name it. NR algos can be tailored for the display as well. There are many handles in the chip so the manufacturer can add as much control as (they think) the consumer can grasp.

I hope they choose 'normal', 'advanced', and 'professional' menu settings myself.

As for the military, Teranex continues to be the interface for that. Yes, the military still uses our technology and new developments there would eventually 'trickle down'. In many cases though, our technology has trickled up due to advances in our lab. Our Image Restore (see website) has a lot of processing wizardry in it.

My HT? Not much. Two Kappa 9a spkrs (remember them? I love 'em), generic surrounds in a 7.1 config, a Sunfire Signature True Sub (love that too) Pioneer VSX-49TX, and a RCA F38310 direct view 16x9 38" set w/HD tuner and HD DirecTV receiver built in. That's all my wife will let me buy for now. I do have my ultimate dream theater though, it definitely revolves around a Teranex and a native 1920x1080p display. DLP or D-ILA 3 panel, whichever is cheaper at the time!

To follow up on Ray's point, if you do the processing right... you don't need to add any noise reduction and/or detail enhancement. It's only when the processing is sub-par that you need to apply detail enhancement to make the image sharper... (Brett's point).

With more powerful 10-bit image processing chips you'll find the images will look a lot more natural!

A lot of our customers don't even turn on the detail ehancement or noise reduction and we don't turn them on either when we do our demos unless the customer asks to see it.

There is an extensive desire for good PQ, but it has proven difficult to be translated into measurable need. It's much easier to "hear" the consumer's clammoring for lower prices than it is to hear them complaining about bad PQ.

Other than judging brightness/sharpness, consumers just don't know what to look at/for when gauging the PQ of a display. If they were to become educated about the various elements that constitute good PQ, if they were given the analytical tools with which they could intelligently judge PQ, then we would hear much more critique of bad PQ; and we would more easily hear the desire for good PQ.

If the PQ discussion was to be brought out into the mainstream then I suspect that there would be far fewer consumers making purchases based only on the brightness/sharpness metrics...

As Mark says, "one must define PQ." Defining PQ for the mainstream is an uphill battle but it sure would make for a sweet victory!

It's very much like the distinction between pornography and erotica.

It seems to me that HQV has some significant implications for home theater system design that haven't yet been discussed on this thread. IIRC, the Teranex processing has about 10 frames of delay, which is 1/3 second in the context of 30 fps. This absolutely REQUIRES audio lipsync delay compensation; there's no way around it. (It is also on the borderline of making user controls seem to respond sluggishly, but that ergomic issue is another discussion.)

So...the lipsync issue implies that a stand-alone scaler makes less sense than ever. The scaler belongs in the receiver or preamp, which will switch both audio and video and which will automatically manage lipsync without the user's ever having to think about it.

But wait...there's more! I think Immersive has exactly the right idea in putting an optical drive in the receiver as well. Suddenly, all of the problems with copy protection, HDMI, etc., simply vanish. The receiver can play DVD, CD, DVD-A, SACD, etc, without any hassles. It might still need a couple of HDMI inputs for one's satellite receiver or whatnot, but the big systems problem regarding optical media becomes solved in a graceful way. The optical drive should probably be a plug-in assembly with integrated decoding electronics so that it could be readily upgraded for Blu-Ray or HD DVD. (This type of configuration is already as common as dirt in the world of PCs.)

To protect the investment of the early adopters whose displays do not include HDMI inputs, it would be wise to try to force Hollywood to rescind its droolingly stupid, non-technical-lawer-driven prohibition against upsampling DVDs and outputting the results in analog component and DVI form -- otherwise, this idea would be stillborn, even though anyone can build an HTPC right now that does exactly what Hollywood has forbidden for stand-alone DVD players. This might well take legislation, action by the FTC, or even a class action lawsuit on behalf of said early adopters. It would be up to Hollywood to prove that it had a compelling reason (other than sheer technical incompetence) in meddling with the consumer electronics industry in such an anti-consumer manner.

In short, I think that introducing advanced video processing with delay that destroys lipsync unless proper audio delay is added has major implications that shake up current home theater paradigms by causing a ripple effect in almost every part of the system.

Hi Robert,

You're exactly right on the latency issues when doing image processing to avoid audio-to-video lip sync error. This is something that we encounter in the broadcast environment all the time. It's annoying!

One frame of lip sync error is generally accepted. Two or more frames of lip sync error is not acceptable. Consumer electronics manufacturers do not automatically correct for the lip sync error because:
(1) it is very expensive to do;
(2) Audio and video are not always processed in the same box so it is often not always possible to correct.

The reason for this is that human perception is conditioned to receive audible cues after visual cues but not the other way round - i.e. hearing a door close before you see it close is not a natural condition.

While it is true that high end home theatre systems do have the capability to correct for the lip sync error, the range of correction is likely limited to one video frame (40 msec) or so. In addition, the average consumer does not have the technical expertise to set this up themselves. Usually it is the home theatre system installer that would set the system up correcting for the audio to video timing error.

Consumer products such as TVs and projectors definitely do not have the same audio correction capabilities as a high end home theatre systems.

Gennum's VXP processor guarantees one frame of latency when the input and output frame rates are the same!

I have to disagree with the "very expensive" part. Once the audio has been digitized, delaying it by 1/3-second requires about 768 kilobytes of RAM assuming six channels (5.1), 96 kHz sample frequency, and 32-bit floating point words. This is only required if we assume that the receiver has provisions for 6 channels of uncompressed domain. Otherwise, the Dolby or DTS-encoded stream can be delayed before the decoder, which reduces the memory requirement very substantially. In any event, this amount of RAM is not expensive at all if implemented in a receiver that has a digital architecture, as every mid to high-end receiver now does.

OUCH !

I had been hoping to be able to find the Realta chip in an external scaler, one similar to the Cristalio Plasma Enhancer Pro which one could connect to a variety of displays or projectors.

No such luck I guess unless there's a simple solution for the lip-synch issue of about eight frames if the 1/3 of a second delay required for Teranex quality processing is correct. My modded Proton preamp has this feature, but for its rear speakers. I guess I could just wire the system in Stereo and use the rear inputs/outputs as main channels.
:confused:

Who's the first one on the deck gearing up to sell a cheap high-quality add-on adjustable audio delay circuit which we can hook into our audio signal path? I'm sure that someone here will rise to the challenge.
:cool:

And DVDO has an inexpensive, relatively, scaler that delays audio.

And several reasonably priced AVR's from major name brands do it too.

Actually both the Realta chip and the Teranex professional system has a 4 frame delay on average for a 480i-1080i up. The absloute max delay with ANY Teranex product is 6 frames and it's only one conversion, something real obscure like 1080p25 to 576p23.98sf (Slow PAL).

And by the way, both Teranex and the Realta HQV has a built-in audio delay to match video. The Teranex has full AC3, Dolby E (professional), and DTS delay support (2 digital channels) with embedded, analog, and AES audio - all 8 channels. The Realta has similar specs geared towards high-end consumer use.

As for the display, my thought is that you take the optical out of the display's Toslink connector and back into the receiver. There you go, no delay, no lip sync problems.

There is a very good reason for the processing delay, one is the amazing quality you get with the Teranex/Silicon Optix algos. With per-pixel processing, when EVERY SINGLE BIT of information is being attended to, nothing being discarded, truncated, or rolled off, it simply takes time to obtain perfection. Personally I think it's well worth it from my experience. Take the HQV, watch the 5th Element DVD with a DPI HD-35 DLP projector on a 30 foot screen at 1080p60 and you WILL believe!

After almost 4 1/2 years with Teranex seeing this stuff all the time, I'm still amazed at what I see.