For embedded interface, it really is no problem, as long as you can restrict error messages from overlapping it. Mouse pointer doesn't effect it? I have often thought of overlaying transparent information on the window, and you have to do that with Zebra patterns, is this a worry on your technique?

>The only real, completely unavoidable work is processing the raw data from the frame grabber, and sending the raster to the GPU.

I have noticed there is upto an 266MB's back-channel capability with AGP, and have often advocated sending the RAW footage to the shaders to be processed for preview/debayering, as well as processed for storage (packing and compression etc) and sent out again to storage. How possible does this seem from where you stand?

I posted several GPU development API's (3-4) in the past in the technical thread as well.

Wayne.

<<<-- Originally posted by Rai Orz : edit:

2.) Additional with build in double removable HDD disk frame to record direct full 12Bit RAW data up to 30fps on 2 x 2.5" HDDs. (In this case Firewire is only for prewiev and setup)

-->>>

Rai,

That sounds interesting.
Could you be more specific?

@Valeriu,
Our camera head manufactor work on a future camera head designs, with internal high speed FPGA and MC, also it will be possible to change the sensor plate and adapt new or different sensors in a few days without other hardware changes. All tests work well, but the most problems are not with sensors, but with high speed data record (PC interfaces and others). So, together, we went a different way and add HDD controllers to the MC to write RAW data direct to disk (1 to max 4). At this way, a interface, like firewire or usb is only need for setup, controlling or preview, not for record, so you need no big pc. Maybe just a handheld? Max. Bit deph and frame rate are just a HDD question. The Hardware will record 8, 10 or 12Bit (packed, but uncompressed). At the moment we use HDDs with max 39MB/Sec. cont. write (so 2xHDD=78MB/Sec, 4xHDD=156MB/Sec).
It will be possible to change the sensor(plate) not only for new sensor versions, also for other sensors (CCD for Video look or CMOS for film style....). The camera case dim. can be somewhere about 70x70x100mm incl 2 x 2.5" HDDs, maybe 4.

The system in not ready to sell, and i don´t like to talk over it before, but it is at a point we can say it is possible in the very near future, so now i like to know what is on YOUR wish list and how much would you like to pay for this or that...

Is Ben still on this forum?

first of all hi everybody , i would like to know if anybody tried to build something like this:

a cmos sensor ( maybe inside an old 16 mm russian camera) recording at HD 1920*1080 @24/25 fps 10bit recording on a notebook through firewire in raw image format ( i would like to apply LUT after )

ever tried?

please let me know and have a good job

PS do you know where can i find HD cmos? Where to find raw example images?

I know this is no new ground as far as monitors for these cameras go, but they are fairly inexpensive, hope it helps somebody out.

http://www.digitalproducer.com/articles/viewarticle.jsp?id=31010

Interesting Rob.

They have a very nice 8" Touchscreen.

Obin, you might want to check it out, seems much more advanced than what you (may) have (not sure since you didn't list any models).

Leon, I don't believe nobody has answered this for you, but everything that you listed (except Firewire and 1080 (though that might be in now)) is Obin's original project of this thread, listed at the beginning, with sample HD footage ;)

We talk about many HDCMOS in this thread. Most commonly Fill factory Ibis5 from Drake and Sumix, Altasens from Silicon Imagining and Sumix, and Micron from Silicon Imaging.

Run down, Altasens is vgood, Micron 3 megapixel 1080p part good, newest versions of Ibis5 good, older version of Ibis5 and Micron 1.3 Mp not so good.

Problem is, that many camera makers produce very different quality in cameras, and most are not supported by pro camera capture aps, like we do here.

You need professional cinema capture software to shoot. For now many groups work on. Obin's SI, Drake's are closets on this list. Sumix and Rob Scott's coming. Other groups, not mentioned here ?? Should take two weeks to write capture for very fast, very component, videocamera programmer (6 months for everybody else). Tell you the truth there are some people who could probably do it is two days, after catching up on all research, but they are probably virtually one in a million). I used to lecture at university on programming, there is something like a 1000:1 difference between best and worse programmers. (Sorry, couldn't resist dropping a few jaws with that one ;)

<<<-- Originally posted by Leon Nox : first of all hi everybody , i would like to know if anybody tried to build something like this:

a cmos sensor ( maybe inside an old 16 mm russian camera) recording at HD 1920*1080 @24/25 fps 10bit recording on a notebook through firewire in raw image format ( i would like to apply LUT after )

ever tried?

please let me know and have a good job

PS do you know where can i find HD cmos? Where to find raw example images? -->>>

Rai,

In the end the industry will eventually go to 24-bit capture (there is already some special high bit mode on Altasens or Foveon, I can't remember which) to record the full range of light. But for the moment as the interfaces record 16-bit, and 16 bit camera simular to film, you should build handling support into the FPGA for future 13-16 bit modes.

At the moment I am, and long term people are very worn out by the long process, to think about things. I suggest looking through the first 4 months of posts, in the threads, and list all the all the suggestions made, and ask people which they like?

My wish list on price stands, I prefer $1000 camera and sell truckloads to TV stations and wall-mart etc. If you are interested contact me. I have been thinking of a platform for this price range, could be sold for $200AU, but people will still pay $1000AU ($800 profit). I originally was going to do another "easy" "simple" platform because too much hassle doing custom hardware, but have been let down by potential partner. But if you are already doing custom hardware instead, maybe we can bypass simple idea, for really cheap idea.

Something you may notice from performance of capture apps, is that we could even look at slow VIA processor on 12*8cm PC motherboard, real cost less than $100. With dual processor (2GHZ (AT LEAST) =2W) we start to look at editor in camera, just plug into USB2.0 USB hub (with keyboard, external drives, peripherals) etc and monitor. So, NOBODY, in the video industry can touch this, add pro lens, or DIY 35mm adaptor and old lens. As I have discussed with you and Rob many times in the past. But it is all upto what you want to do.

One thing I must note, is prosumer/video production/documentary market. Forget indie only market for cheap camera, you could see 100 times more to these other markets. Many want to be prosumer people are forced into buying $1000-, rubbish cameras, instead of $3000 rubbish camera, we can out do all, just RAW makes the difference. But for this all controls must be simple as well as manual, instant auto snap set, full auto, and follow. I can show many ways to do things, if involved. I don't often talk because it is related to other potential IP of mine. Forget all follow like old film camera, no view screen, estimate and calculate image in head rubbish, there is a reason many people can not do this and pro cameras use view finders and more convenient controls.

Now on the back-end, you have to output/record common high quality tape standards as well ;) . You go straight into the situation and say, "I have story for news in half an hour", and not say: "I need 6 hours to load in, de-bayer, colour balance, edit etc and then give it to you in a file that you can't feed into your system"). In all prosumer, low end production and doco, speed and auto is the need. Good news is that you can have cinema and quick on the same camera (as discussed before) change the rail setup, and change the control setup. We know all this already, but it needs to be done. We do not replace cameras like Drake/Altasens, just complement them for those that can't afford a Drake or Altasens (the majority of potential prosumers).


Thanks

Wayne.

Rob,

Impressive, how come none of these guys sell a 1280*1024 monitor, or 720*576, cheap (rhetorical question)?

How do people feel about using 640*480 screen, we can't get individual pixels on 1024*768 anyway, so 640*480 should show us the general lighting/color levels?


Thanks

Wayne.

Wayne,
24Bit capture: The FPGA i played with, have a lot of not used input pins. It can be reconfigured by software to adapt different sensors (fillfactory, micron, altasens). For a sensor with 12Bit at two channels, it used only 24pins (less than a quarter at all), so 24Bits should be possible, in theory. With faster HDDs, it should also be possible in practice, but i do not see the necessity, remember its RAW. 3 x 24Bit, whohww. What will you do with those thing?

You mentioned „long process“. I think differently. My fist inquiry on this thread was middle july last year. At that time i hoped i can find here a working solution for a comming movie project. Then, september we started our first own trials with camera heads from different manufacturers, because here i found no working system. As you know, it take only a few weeks till the film crew started to shoot the movie with our beta version (incl. a camera head with own hardware changes). End december we finished our camera system. Not only the camerahead + cine software, also all mechanical parts and asseccories, like case, focus follow, matebox, support rag etc. Now, we got a title story in a german (HDTV) magazine. I think, thats not a long process.
And, be sure, if i now said those new universal camera head is possible in the near future, i mean what i said.

Your wish list about price is maybe possible (numer of items). But is this the goal? Why cheaper than other companies? (SI, SUMIX) On the other hand, with those camera heart (its maybe a better name than head) its possible to build cameras from low cost home video up to high end cinema, so it will be not a good idea to sell it in all cases at the same price. More details not public...
Yes, its right, you do not need high end PCs, because capturing and record are part of the camera. Viewfinding is another thing. A cheap camera need only a small tft, at low fps. For the first version you need external solutions, so from gameboy, handheld over notebook to high end PCs, all things are imagingably, because in all cases the camera can record also without those things.

Your absolute right, the prosumer/video production/documentary market is 100, i add maybe 1000 times more to the indie market. But as cheaper a camera go, as more automatics (autofocus. etc.) is a must. This take a lot of extra development time, so i see it at the end of the list.

You mentioned the output time. The good thing is, RAW datas hold all captured informations, the bad thing is, it need a software to transform it to available formats. While transforming, you lost informations. And there are a lot of different ways how you can transform: compress, color correct, etc.. At the moment i see the best way, to put the camera HDD block out of the camera and in a spezial PC connected box (maybe USB or Firewire) and let do a PC those works. This is the way we do it also with our existing camera. You can also change the „style“ by software select and it is easy to install a new version, new better debayer, etc. on PC. To implement those function in the camera head, it is possible, but it need a lot of development time.

<<<-- Originally posted by Rai Orz : Wayne,
>RAW. 3 x 24Bit, whohww. What will you do with those thing?

Just predicting where industry will go, eventually, with new technologies. But as I said, 16 bit is enough for mement.

> You mentioned „long process“. I think differently. My fist inquiry on this thread was middle july last year. At

In terms of what we were talking about before you jioned. Camera in less than 3 months. People are simply "talked out" because too much talking and no systems out yet, nothing really to do with your camera, just the projects. So probably best to think for them and list, so they can pick which features they like.

> Your wish list about price is maybe possible (numer of items). But is this the goal? Why cheaper than other companies? (SI, SUMIX) On the other hand, with those

Marketing, if too many competitors in market, can get new market and undercut them, if you can make more profit selling to 100 times more people, and moral ideal, so the poorer people don't get left out. So we are in agreement.

>Your absolute right, the prosumer/video production/documentary market is 100, i add maybe 1000 times more to the indie market. But as cheaper a camera go, as more automatics (autofocus. etc.) is a must. This take a lot of extra development time, so i see it at the end of the list.

I have done some figuring long ago, canbe done simply (very simply). External physical controls developement hardest though. We really, also, only need one or two codecs to dump directly into TV station system (one uncompressed and one standard compressed). This also means that you can pass live feed from camera, as studio or feild camera. So not really the greatest jump. In a way, the next step for cinema camera. As we know some open FPGA compressiuon designs allready out. But maybe just use software (fast Arm, FPGA processor etc). With the increases in processing power, soon this will be possible cheaply.

I do not know if this:

http://www.isgchips.com/Templates/t_quadhdtv.htm

might be of interest to you. Found and delivered.
I hope it helps.

trying many may ways of image save...a real pain...progress in the save? yes..but slow...more later today

Monday morning I will have some more updates...a test program has been coded that tells us the specs of my hard disks and what the msec of time is we need to let the drives 'recover' between writes..this is helping a bunch and letting us tune the save for this. things are looking up ( but taking longer then I would have thought )

cheers!

weird question...what if...we could find a 35mm sized CMOS chip like the D-20 Arri camera and take pixel 'chunks' from it to make a native 1080p image with the FOV of a TRUE 35mm camera!!! this would ROCK..anyone know of a box cam with a 6-12megapixel chip that is full 35mm sized?? cameralink?

And DOF! That's a pretty cool idea actually. That and your low light sensitivity and signal to noise ratio would go up a lot (if I understand things correctly)!

I suspect these are very expensive though :(. My father has a dalsa scientific camera in one of his instruments that shoots 30fps, with a full sized chip, (not as high res as you are looking for though) but it costs approx. 10k.

Obin,

> and take pixel 'chunks' from it

Many CMOS image sensors can perform "binning", where 4 or 9, etc, pixels are combined into one pixel (albeit with higher resolution). However, with a Bayer matrix, this is much harder.

> ...we could find a 35mm sized CMOS chip like the D-20
> ....with a 6-12megapixel chip that is full 35mm sized??
> cameralink?

Then you will need a very high pixel clock, and/or more CameraLink taps, and a large sensor pixel size, all expensive items.

I got into this CameraLink world, first by using a couple of Olympus 3 megapixel cameras. I had the idea to do a stereoscopic film in timelapse. I actually found some very nice controller software from PineTree (www.pinetreecomputing.com/camctl.asp), which allowed me to shoot and download from multiple cameras, completely synchronized. One cycle took a minimum of 12 seconds.

Then I started to look at SDKs from other manufacturers. Kodak's SDK for their 14 MP DCS SLR is very nice (the chip is from FillFactory). One could actually make a 3D IMAX film with that. There was a timelapse (flat) IMAX film once made...Chronos. You could make something like that now, in 3D, for much less money.

Basler has a nice camera, the A402k (and A403k, etc., with higher pixel clocks). The chip is from Micron. 7 micron pixel, 2352x1726. But only 10 bits, and monochrome. And $10k, and higher ( www.baslerweb.com ) The sensor size is about 16mm x 12mm, getting close to 35mm cine (camera has F-mount)

Also, you will need multiple CameraLink taps. The Basler A404k can do 96 fps, with a 50 MHZ clock, using 8 CameraLink taps. These grabber cards are expensive; some of them allow for 4GB RAM, installable on the card.

This link lists many of the CMOS/CCD industrial cameras out here (Dalsa, Basler, Jai/Pulnix, Sony, etc, but not Silicon Imaging). Aaron is correct about Dalsa: nice cameras, but expensive.

http://www.opsci.com/

Without a programmable clock synthesizer, getting 24..0000. 23.9760, or 29.97 FPS, with variable ROI is also difficult.

"I have also found out that splitting the
file between two drives with IDE drives can slow things down due to the IDE
disk system that cannot handle split bus transactions. I will see if this
applies to SATA too. It might seem that you might need SCSI drives to
implement this properly. I'll have more on that later when I get a bit more
information on this."


anyone here have a firm answer to this question from my programmer?

<<<-- Originally posted by Kyle Granger :....
> ...we could find a 35mm sized CMOS chip like the D-20
> ....with a 6-12megapixel chip that is full 35mm sized??
> cameralink?
Then you will need a very high pixel clock, and/or more CameraLink taps, and a large sensor pixel size, all expensive items.--->>

No, 40 - 66Mhz pixel clock is all you need, but at multiple ports. Thats the way all high pixel sensors (or high speed sensors) work. For example:
ARRI D-20 Sensor work at only 20MHz, but with 32ports.
Panavisions QuadHD™8MegaPixel Sensor, 30fps: 37Mhz, 8 Ports
Microns MT9M413 1,3M Pixel, 500fps: 66MHz, 10 Ports
Microns PB-MV40 4M Pixel, 240fps: 66MHz, 16 Ports
FillFactorys LUPA-1300 1,3M Pixel, 450fps: 20MHz, 16 Ports

If you distibute the data to multiple HDDs inside the camera head (see my postings before), it will be not a expensive item.

<<--- Then I started to look at SDKs from other manufacturers. Kodak's SDK for their 14 MP DCS SLR is very nice (the chip is from FillFactory). One could actually make a 3D IMAX film with that. There was a timelapse (flat) IMAX film once made...Chronos. You could make something like that now, in 3D, for much less money.--->>

This nice FillFactory sensor is based on the IBIS5 technologie (see the pictures and you know what IBIS5 can do), but they dont sell it to others than Kodak, because Kodak hold exclusive rights on it. On the other hand, it have no multiple ports, so it work only at low fps.


<<---Basler has a nice camera, the A402k (and A403k, etc., with higher pixel clocks). The chip is from Micron. 7 micron pixel, 2352x1726. But only 10 bits, and monochrome. And $10k, and higher ( www.baslerweb.com ) The sensor size is about 16mm x 12mm, getting close to 35mm cine (camera has F-mount)--->>

There are other comanys, made cameras with color versions of this high speed 240fps sensor. For example:
http://www.fast-vision.com/cameras/camera40.htm

<<--- Also, you will need multiple CameraLink taps. The Basler A404k can do 96 fps, with a 50 MHZ clock, using 8 CameraLink taps. These grabber cards are expensive; some of them allow for 4GB RAM, installable on the card.-->>>

Or, you go a different way (HDDs inside camera head) like i wrote it before...

<<<-- Originally posted by Obin Olson : weird question...what if...we could find a 35mm sized CMOS chip like the D-20 Arri camera and take pixel 'chunks' from it to make a native 1080p image with the FOV of a TRUE 35mm camera!!! this would ROCK..anyone know of a box cam with a 6-12megapixel chip that is full 35mm sized?? cameralink? -->>>

People have discussed adapting still cameras to motion. In time larger and larger sensors will demand higher and higher base clocks, so 1080p binning should be possible someday, some sensor. Note that some of these cameras will only do a movie mode that is less than the base clock, this might be a restriction that might be related to the sensor, so some firmware code modification might be needed (unless it is hardwired).

Binning is not perfect. It should have higher niose and less low light than single pixels of the same size. But you should get closer to 3CCD colour accuracy as the bayer is binned.

IDE:

Cheap/nasty chipset/motherboard. I haven't heard of this one, most likely depends on the motherboard. Is this one one IDE channel, can you use two channels instead to get around it? Makeing your camera with one good motherboard should get around it. Search for the IDE standards spec/organisation. It should have some terminology for this problem, and any solution, then cross search on that to find the terms the MB manufacturers use for the solution, and serach for boards with that feature. Should be something at sites like that storage review site too, under raid schemes.


Wayne.

I am working with a test app today to find the speed and the delay we need to save with multithreading and twin disk..this should clear everything up so we can go forward with REAL data from my twindisk DFI system and complete the save part of CineLInk...

Over on the technical thread I have listed some information on the recent announcements on the Xbox2 Xenon. It appears to be much more of a super machine than first thought with one Teraflop of processing power, which would make it capable of capturing/compressing/editing a UHD (IMAX) like picture live (except we don't know interfaces yet). And software can be made to run on both the PC and the Xbox2.

http://www.dvinfo.net/conf/showthread.php?s=&postid=284833#post284833

when will that be out Wayne?

It's in the post, the holidays this year. So if anybody would like another interesting capture device project they should look at developing now. As far as cost effectiveness, and performance, the PC's is becoming dated for low cost camera capture.

Nobody will record audio on camera?

Well . . .

There's always double-system sound, which is the "pro" way of doing things, especially for a film person like yourself ;)

I wouldn't count on using the new X-Box system at all for such an operation.

Yes, sure.But these camera systems look more like a cheap one in all thing for the average budget guy.....
So I think Audio (on camera) is not a crazy thing.....
If we were heading exclusively for Top quality then we couldn't pretend to pay 10 grands for it, cause bill of materials would be more than that....
Anyway the price/performance ratio of these things beats everything available today......

Yes, lots of businesses run around with minidv codec cameras prosumer (even pd170) cameras trying to erk out a living. These systems can compete with those and beat the quality of the Eng cameras, who are another market again. People on this level might like to try a bit of amateur indie. So I am not a cinema nut (yet!) so to me it makes sense to sell to all these people and lower manufacturing costs by extra volume.

So audio on camera, and even xbox2, are good attempts to lower costs for this market. When we get around to the run and shoot software. Bill of materials (leaving out lighting and set design etc) is very low. We can expect $500-$1000 camera head for this market. If Gigbe or firewire, or USB2.0, or UWB, then we are getting close to $500 laptop territory. Now we add capture/editor, let's just imagine a freeware linux package (rather than cheap software for $200-$300), extra $99 VGA display, extra battery/solar power panel, and 35mm adaptor with old $50 lenses (with electronic controls, or CCTV lens). So the min cost is close to the lowest price for the HD1.

Which, I think, is exactly what this market is looking for.

running circles working on datasave..pain in the ass ..but it looks like we make headway with the RAID setup instead of twin disk save....jsut testing testing testing...more later...seems like things NEVER work as we wish!!

They would, on a Playstation, or an Apple, I am not going to say xbox, to many PC roots. You'll get there, just a PIB to program around PC problems.

Ideal computer: one hi-speed (1Kb/s-10+GB/s) serial bus interface through a few sockets and wireless ports, processor/co-processor/s, memory, storage, thats it, that's all that is needed (display, sound, keyboard disks, cameras can all go through serial interfaces). All this PC stuff is mostly outdated rubbish, with manufactures producing differing versions of each part of it. The way it hangs compatable is by scaling back potential and writing huge amounts of software/hardware to cover up the incompatibilities (sic). The day, hopefully, is coming.

All the Linux developers should look at the model I suggested above, and realise there job would be at least ten times easier (you can tell my Pc has been giving me trouble again, can't you ;).

re-edit:

Finally, dual processor VIA, GigE:

http://www.dvinfo.net/conf/showthread.php?s=&postid=285530#post285530

<<<-- Originally posted by Obin Olson : running circles working on datasave..pain in the ass ..but it looks like we make headway with the RAID setup instead of twin disk save....jsut testing testing testing...more later...seems like things NEVER work as we wish!! -->>>

Welcome to product development :-) On the bright side, the more problems you have to work through, the less easy it is for someone to come in and copy your work.

Are you planning to bring your camera to NAB? I'm sure people would be very interested in your progress.

RAID......mmmm.......
I hope not to hear in a couple of days that you are recording in an .AVI container....
Anyway good luck. :)

Woohhhh!! Now this is cool i am on my phone posting on this board!! Got a internet phone service!

hey guyz dont want to bother this forum, but i posted a question under alternative image methods>high resoultion wireless...

i think you guys can help me out, but i assumed you missed the post?

Obin, I do not know if this is of any help to your project but just in case it might help you...:
http://www.artray-ap.com/visionsystems.htm

Obin, do you have any screen grabs or footage you could post from your camera?

After 2617 post we demand a screen grab! Just one.

Ok a screengrab will do..let me get to work...i will post one then...the last issue we now have is the save of raw data..we are waiting on a 3rd party at the moment for a software fix of the profiler we are using...once this is done we should have a working beta of CineLink!

www.dv3productions.com/pub/dog.tif

shot with an OLD 16mm c-mount lens ;)

Hey Obin, some great lattitude and tonal range in that frame. However, I noticed a strange grid-like pattern when viewed at larger magnification.

yes grid..we have the most basic bayer filter you can have..this is why you see the grid. If your digital still camera took every pixel from the chip and gave it an RGB value it would look the same.

after we get raw save working we will have some Bayer Filter options

That's a huge file for one frame. I can see the grid, but besides that, it has very film like detail and tones. OK great...so now we demand a second image.

No, really it looks great.

found a workaround for the profiler issue..back to work..more later

Greetings all. I'm a newcomer, joined years ago but rarely kept up with DVi... but I am exceedingly pleased to come back and find amazing projects like this going on. Obin, you are da man. I have no idea who you are but your perseverence with this project is amazing. 175 pages of replies? What in heaven's name?! And this is soooo the right way to do things, too - start with a sensor, not a Nikon F and a CD/sandpaper.

I will be enthusiastically checking up on this from day to day.... the passion that a lot of folks are displaying for DIY cinema cameras (sic) is great, but here is one place where the passion has some real direction to it.

Also, Obin, if you need any help with very accurate machining (lathing or CNC milling) let me know. Also, if you have any parts that need to be physically aligned to very accurate tolerances, give me a holler - I'm a flutemaker and we have some amazingly accurate ruby-probe 3D measurement tools here in the shop (+/- 0.0001").... I can build you fixturing or whatever you need, at cost. All the equipment belongs to my father and I (old family business) so I'm open to doing whatever work you need to get this camera to the place where you want it. No idea if you'd ever need this sort of work, but I figured if nothing else you'd appreciate knowing that there's someone else willing to donate some time to this project. Our shop is very flexible so don't be afraid to email me (klimermonk -at- hotmail -dot- com) if you have something in mind.

JL

John;

thanks a bunch for the offer..i have a small cnc machine but it seems you have years of working with them on your side(I don't) I will take you up on the offer when the time comes...or atleast you could help if we want to do it at our shop? I am sure good things will come however we do it.


thanks for the positive feedback... ;)

we have some good news.

today we ran save tests that are showing we can get full 24fps 1080p datarate on our 2 drives. We have a test app that saved 4 gigs in 512kb chunks on the disks at 97MB/Sec with no problems...now it's time to write that in the code and get some real progress....

.

Hi Obin,

Congratulations! Just one thing I noticed...You should only need about 72MB/sec for 24p, 12-bits per pixel. Are you receiving one pixel in 16-bits? With the GigeLink I can pack my 10-bit data, two pixels in 3 bytes. The 10-bit data is bit-shifted by two ("normalized 12-bit").
This packing results (for me) in only moderate overhead for the display (more shifting by 4 must be done).
I'm assuming your capturing about 48MP/sec.