eLCOS Technology Thread

Hi all,

At CES we were happy to demonstrate our new 1920x1080 panels. It seemed that we generated some interest there and were happy overall with the response at the show.

As I find time I will post information about the panels and more specifically the set up that we showed at CES.

For a start, feel free to check out our website here.

I include again a link to the Kolin press release here.

Thanks for your interest - more as I find the time.

Cheers!
DAve.

__________________

Disclaimer: These comments are all my opinion and do not represent any official stance from eLCOS. I am just an engineer - sometimes I think that the marketing guys would prefer it if I didn't talk to anyone...!

The 1080p solution that we demonstrated at CES uses our 0.7 inch LCoS imagers. It is a three panel system with a controller for each channel. The controller board accepts standard DVI and maps one to one on each panel for each colour. So the controller board must be fed 1920 x 1080 progressively over DVI. We have operated two modes - one being 60 (or 59.94) fps one to one and the second was 24 fps displayed at 72 fps. We were using 59.94 fps at CES.

The three panels are slightly different and are tuned to give the best response to the red, green or blue wavelength. The pixels on each panel are approx. 8.1 microns square. We are using a digital modulation to generate the grey scale as opposed to driving an analogue level, so in that sense we are kinda like DLP. However we do get some averaging effect / filtering as we switch faster than the response time of the liquid crystal. This allows us to use relatively low voltage silicon to keep the cost down. Our goal is not so much to compete with Sony SXRD and JVC D-ILA in the $20k+ projector market, more to offer 1080p resolutions at the $5k level for RPTV - so we are aiming for material that is relatively easy and cheap to make in mass quantities.

There are other advantages to using digital modulations with respect to taking care of non-linearity in the LC response. These things can be taken care of in the digital domain.

Feel free to ask questions. I will answer as I find time - assuming it is not sensitive information

Cheers!
DAve.

__________________

Disclaimer: These comments are all my opinion and do not represent any official stance from eLCOS. I am just an engineer - sometimes I think that the marketing guys would prefer it if I didn't talk to anyone...!

Am I understanding correctly that the Kolin HDTV will accept 1080p input through DVI?

That will actually largely be up to Kolin. They will add a front end board with all of the inputs and they will feed our controller board 1080p.

There is definitely an issue with 1080p 60 fps input at the front end board. Mature scaling technology for the most part is limited to 75 MPixels at the input so that will handle 1080i or 720p (or 1080p up to 30 fps). For 1080p at 60 fps you need 148.5 MPixels per second - too much for the input to the scaler.

But, you say, who needs to scale it if it is 1080p already??? The scaler is typically where the On Screen Display and such is added. So bypassing it for 1080p is not likely to be offered.

All that said, the next gen scalers will all handle 150 MPixel per sec or even higher and also offer better HD deinterlacing and so on... so 1080p input will be a reality at some stage.

Cheers!
DAve.
__________________

Disclaimer: These comments are all my opinion and do not represent any official stance from eLCOS. I am just an engineer - sometimes I think that the marketing guys would prefer it if I didn't talk to anyone...!

Thanks. Things have sure changed since I got my Captain Midnight Secret Decoder Ring from a can of Ovaltine. In fact they now make it in China and charge 25 times as much.

This is making me nervous. Low-voltage, low-cost silicon for the chips?
Hasn't the Hitachi nightmare taught us anything?
"Cheap" chips fail prematurely.
The Toshiba 1080p 3-chip LCoS has already come and gone.
This eLCoS project is definitely not breaking new ground.
Toshiba gave us 6.2 million pixels at around a $6k street price
and is now stuck eating the cost of that failure.
The sets worked GREAT... for 8 months. Then some large percentage
of them failed catastrophically due to chip deterioration.
These chips need to be robust. Being adjacent to a projector bulb is a
pretty harsh environment.

I guess cheap was an unfortunate choice of words - think inexpensive.

It is unfortunate that Hitachi and Toshiba were unable to produce reliable product. That said, comments like this are unwarranted.

It is our ground breaking technology that allows low voltage designs and the low voltage designs that allow for low cost and very reliable silicon wafers.

Cheers!
DAve.
__________________

Disclaimer: These comments are all my opinion and do not represent any official stance from eLCOS. I am just an engineer - sometimes I think that the marketing guys would prefer it if I didn't talk to anyone...!

Agreed. I will certainly apologize for the tone.
Upon re-reading my post from yesterday, I can certainly
see how it could come across as harsh.
That wasn't really my intent.
And there may have been some frustration over the
whole state of this market in general that was leaking through.
It wasn't directed toward you or your company.
It's just that so many LCoS projects have failed to date,
and 1920x1080 fixed-pixel displays have been shown
for years and years at trade shows, yet still manage to
have essentially zero consumer availability.

If it's any consolation, if eLCoS does deliver a 6.2 million pixel set,
and hits their target price and more importantly their vague availability date,
it's possible (and in fact LIKELY) that I will be quite keen to give eLCoS my $$$.

Phloyd: Isn't the plan to have OEMs sell these sets, not eLCOS?

I was asked in the Intel thread to post about the difference between analogue and digital LCoS.

The analogue side is what JVC and others have been doing for some time and there is good information out there about this technology.



This image comes from the Sony press release about SXRD, and does a good job of showing how an LCoS cell is layered.

Those technical amongst you will recognise the architecture - it looks a lot like a DRAM cell! And it works along the same lines.

Essentially a voltage is written to the pixel and stored on the capacitor - this in turn causes the LC to align according to the voltage.

However, the voltage on the capacitor will decrease over time, so they need to understand three things - the voltage they want to write, the rate of decay and the time between pixel refresh. So they need to write a higher voltage so that the average voltage on the LC gives the grey scale level that they desire.

So, the faster they update the screen, the better off they are. The slower that they write, the higher the drive voltage needs to be in order to be the correct average voltage.

Ok, now let's imagine for a second that white is represented by voltage Vw and that the higher the voltage, the whiter the LC will get (this is not always the case, but it is easier to think of it this way).

So, for argument, a full white can be achieved by Vw = 4.5 volts or greater across the LC. So for an analogue system to write a white pixel, they need to write say 5.5 volts to the pixel so that as the voltage decays, the average is going to be 4.5 volts. Remember this part when we start talking about the digital system...

Cheers!
DAve.
__________________

Disclaimer: These comments are all my opinion and do not represent any official stance from eLCOS. I am just an engineer - sometimes I think that the marketing guys would prefer it if I didn't talk to anyone...!

In a digital LCoS system, pulse width modulation is used to achieve the grey levels instead of trying to drive an analogue voltage. So we have some similarities to DLP in that sense.

The basic lay up of our cell is much like the analogue cell except that our cell only knows how to drive to black or white - no analogue levels between are possible. It uses an SRAM type cell so that the 0 or 1 will stick and stay until it is written again. For white, the drive will be to white over the whole frame period and for black, it will be black for the whole frame period. As a first approximation, half black and half white gives you a mid grey. I am sure that there are many places that you can find out about PWM if you want to know more about it.

One essential advantage comes where we talk about white and black. As we noted above, the analogue system will need to overdrive white to get white as the average. In the digital system, the drive level for white is the actual white level voltage (in our example, 4.5v) - so we can get away with lower voltage silicon.

The second essential advantage is flexibility when all things are not equal. With a digital modulation we can tailor our PWM to take care of nonlinearities in the LC response and other systems. We can do things in inexpensive digital ways.

The third advantage is that we only have to have two analogue levels set accurately - those being Vblack and Vwhite. Everything else is generated in the modulation. Analogue systems need to track the exact response of the panel across voltage ranges and accurately drive the correct analogue voltage so that the average drive to the LC is the correct level - this is obviously considerably more variables to deal with.

Lastly the digital system is probably easier to scale. As we use less voltage to drive, the transistors can get smaller. Digital will no doubt lead analogue in pixel density. We are currently more dense than the Sony SXRD - they have 9 um pixels, ours are 8.1 um on the 1080p panel.

So... I hope this helps explain the differences... and I hope it doesn't raise more questions than it answers....

Cheers!
DAve.
__________________

Disclaimer: These comments are all my opinion and do not represent any official stance from eLCOS. I am just an engineer - sometimes I think that the marketing guys would prefer it if I didn't talk to anyone...!

Not to worry - I later saw elsewhere that you had one of the Toshiba sets and I am sure it was sad to have to return it.

I think LCoS has a lot of potential and it is exciting that Intel seem to be following a strategy that is very close to ours - to me that is a sign that we are on the right track and I hope that we can remain ahead of Intel in the race to a real product.

Rogo: yes - currently we have no plans to make TV's - our customers will make TV's. It is possible that we will create a reference design for our not so technical customers - in this case I will try to ensure that there are a good range of digital inputs to take advantage of a fully digital system.

Cheers!
DAve.
__________________

Disclaimer: These comments are all my opinion and do not represent any official stance from eLCOS. I am just an engineer - sometimes I think that the marketing guys would prefer it if I didn't talk to anyone...!

Ok, one last thing and then I will shut up for a while

The unit we demonstrated at CES was a three panel engine in a 60" RPTV case. The whole system was essentially the same as the 720p system that we also had on show, except that we used the 1080p imagers instead. Obviously there was some difference in the control files but the optics engine and controller are both compatible with the new imagers.

We were running source that we licenced through CES. I wish that we had the writes to some more interesting material, but those are the breaks I guess. The source was provided on DVHS and was all 1080i. We played back from a JVC 30000U via component output into a KDS Leeza which scaled to 1080p. I tried to get a directly digital conenction using 1394 to PC or STB but there were either motion effects (I think due to incorrect field order) or DVI sync problems, so the best result was to use component. We definitely lost some sharpness and colour depth with the analogue step. KDS provided excellent support (it helped that they were just up the hall ) in getting the best picture with the equipment available.

Other than the analogue video step, the other major thing for us to contend with was a large amount of ambient light and reflections on the screen. It is hard to show decent contrast in such an environment.

Still, overall we were pretty happy to be there with such a tight schedule to get a decent image on our set. Thanks to those who came by to see us.

Cheers!
DAve.

Phloyd:

Thanks for the information. I've got a couple of questions for you.

First of all, according to your website, the lcos imager has a contrast ratio of 400:1. This seems quite low compared to TI's claim of 3000:1 for the HD2+. I guess my question is am I correct in assuming we can make a direct comparison between the CRs of 2 different technologies, or is it like comparing apples to oranges? How will the final CRs that viewers see coming out of their TVs compare the CRs of the imagers?

Secondly, in one of the earlier posts, you mentioned that the liquid crystal response time wasn't fast enough to keep up with the switching time of the digital modulation. Does that mean that we can not expect 256 distinct levels of gray from a 60 fps source?

Sorry, one more question.

You mentioned that your low-cost solution resembles SRAM in that it holds a charge, as opposed to the high-cost competion which more closely resembles DRAM.

Admitedly, it's been a long time since my electronics courses at university, but I seem to remember that SRAM was quite a bit more complex (and costly) then DRAM, and required 5 times as many transistors. Is this no longer the case?

Hi padlock,

I don't know too much about the contrast ratio comparisons. If you ask Runco about it, they quote three different CR's based on how they were actually measured. It seems that the CR of a chip and the CR of a system based on the chip can differ greatly. I think that the 400:1 you saw is most likely a system CR, where there DLP may be the chip contrast ratio. In any case, I think that our true contrast performance is in the same ball park as the current DLP RTPV's from Samsung, just based on looking at them...

The LC response is slower that the modulation, but that results in it acting as a smoothing filter, partially holding an average value based on the modulation. We can achieve similar or better colour performance to other technologies - in fact I think that colour fidelity is one of our strong points.

SRAM is more complex than DRAM, but there is still plenty of room under each pixel for our pixel cell. The overall cost is largely based on the materials and process used as well as the size of the final wafer. Our cost savings come from the wafer being relatively small (0.7") and the process being a fairly standard silicon that is easy to fab - it is not the bleeding edge of silicon geometry by any means...

__________________
Disclaimer: These comments are all my opinion and do not represent any official stance from eLCOS. I am just an engineer - sometimes I think that the marketing guys would prefer it if I didn't talk to anyone...!

Dave,

Welcome to AVS and the RPTV forum. I am sure this will be a popular thread. I am one of the people that has great hopes for the LCoS technology, though so far its been struggling. Maybe your twist on the chip will make it successful.


Question for Phloyd and the more I think about it it should be for the light engine maker, but maybe you can give it a shot.
How will the alignment/convergence of the 3 panels be handled? Some 3 panel fixed pixel display have had issues in this area.

And on a related note, why don't people make the resolution 1080/1100P, and then if the three LCOS panels are out of alignment, you can shift the image on the panels around to align it via "software". You only have to use 1080, the extra 20 are just around the edges for image shifting.

Thanks. I enjoy talking about the technology.

That said, we are still at a phase where many things need to be kept under wraps, so I may not offer much detail in certain areas.

You may notice that I have added a disclaimer to my posts - this is just to emphasise that I am not posting as an official source of information from our company. I am just trying increase peoples' awareness and help them be better informed about the LCoS technology.

The short answer to your question is that the images are combined optically in the light engine, so we are somewhat at the mercy of the light engine guys.

And to answer Axel's question, people do this.

Cheers!
DAve.

Wow so much great information!

In case any of you did not see it in my post in the Cineos thread here is the link to the pictures I took at the eLCoS booth at CES while talking to Dave (there are Cineos pictures in there as well). http://www.aecustom.com/ces04/lcos.html

Dave I just want to thank you for spending so much time discussing this technology. You are amongst a group of people who want to see your product be a success so bad we can taste it (and imagine it after what we saw with Toshiba).

Hopefully this is the beginning of good things to come and I can tell you from the short time I have spent on AVS your involvement and feedback with the community that will be using your products is the most invaluable research you will ever get!

Good Luck to eLCoS...this is going to be fun!

Phloyd, this is a lot of good information. Thank you for sharing it with us.

Because of overscan a panel does not have to be bigger than 1080x1920
to correct for alignment errors .. these will be very small, and will
be covered up by the bezel if they move the images around.

Doug

Doug: None the less our panels are slightly bigger than 1920x1080.

TetsujinWave: No problem - it is good to get the information out there.

Peter: Thanks for adding the link to the pictures. I like the basketball picture.

We have some pix that were taken at the show too - if I can track them down I may post one or two. I don't want to be seen as advertising, so let me know if you think it would be a bad idea to post pictures.

I guess you picked up on my enthusiasm for what we are doing - we are definitely excited about what can be achieved. We definitely understand the benefits of being able to talk about all this stuff in a nice, relaxed and informal discussion.

Anyways, I have presented about everything I can think of so if you guys have any questions or want clarification about anything, feel free to ask. I will offer what information I can.

Cheers!
DAve.

Please do not encourage overscan. Overscan is more of an abomination on mankind than interlace.

Every pixel is designed to be seen -- from (0,0) to (1079,1919).

I currently have one of the Toshiba LCOS sets. Sadly, it is going to be shipped back to Toshiba sometime over the next couple of weeks.

I absolutely loved the picture...smooth and film like. It kills me to return the set.

The big problem appears to be longevity of the LCOS panels. After about 6 months of use, my set started to develop a tendency to temporarily retain static images. For example, I like to watch baseball and usually there is a static score board on the screen. If after watching for a while, I switch channels, you could see a sort of ghostly afterimage of the scoreboard. A similar behavior would happen if I was watching a 4:3 SD show for some period of time then changed to a 16:9 HD channel. The 4:3 area in the center of the screen had a decidedly redder appearence. Eventually this would fade then no problem. As time went on, the afterimages became more pronounced and it took longer to fade.

Another common problem fellow Toshiba LCOS owners encountered were the dreaded "blue flares" at the bottom of the set. I first started to see these after about 7.5 months. They changed from day to day, sometimes there, sometimes not. I also experienced large irregularly shaped blue "blobs" on the left side of the screen. They were not as intense as the blue flares but were more like blue discolorations, that changed shape and size over time.

Hence the return of a non-fixable set (no parts!)

My question is: have these new panels gone through long-term use testing? Any signs of degredation? Over 6-months of use?

(I'd be happy to be a beta tester of a prototype)

One thing to remember about LCOS longevity is that many of the JVC projectors have been in use for 4 years or so. They are used in the simulation market quite extensively and are popular there. There is no way they are suffering the same maladies as the Toshiba RPTV has, or else they'd have long since been ripped out.

Mark

Feel free to ask questions. I will answer as I find time - assuming it is not sensitive information

Dave,

I figure this you can answer one of the questions that has been bugging a lot of on AVS for some time - assuming it isn't top secret or anything.

lie-cose?

el-cose?

el-cos?

el-sea-oh-ess?

He he ...

We tend to say el-see-oh-ess, el-cose or el-cos.

Also some people say our name with an extra "e" and some don't - e-el-cos or el-cos.

Nobody I know of says lie-cos(e).

The largest vote I think goes with el-cos...

Cheers!
DAve.

The image retension issue is seen in LCoS when a DC bias is allowed into the control drive. In normal operation the LC should only ever see an AC signal and so image retention is typically not a problem.

I am really interested to hear more, or particularly see pictures of the racetrack and blue issues. It seems very odd to me that issues can show up in blue but not red and green.

Naturally we will be doing extensive reliability testing on our parts. As Rogo notes, JVC have had LCoS in the field for quite some time, and from the friendly chat I had with them at CES it seems that they are very pleased with how the LC's have aged. So I don't think it is necessarily an issue with LCoS itself, but just something to do with how Hitachi and Toshiba built their systems and panels.

Of course, unless someone gives us some hard information, this is all speculation.

Cheers!
DAve.