<<<-- Originally posted by Obin Olson : Wayne, seems to me that Micron is doing the best lowcost chips..I have the Silicon Imaging camera and am going to start some tests! foveon is 4fps I think and the fillfactory chips suck ass in image quality the rockwell chips are great but cost alot -->>>

Yeah, I'm really disappointed that Foveon hasn't come out with something suitable for HD. Their 10 megapixel (actually 2268 x 1512 x 3 layers) sensor could be perfect if it had a decent frame rate. The diagonal size is 25mm which is getting close to 35mm.

I'm curious -- How much more do the Rockwell/AltaSans chips cost vs. the Micron chips?

That's what Foveon has on their website at the moment, but who knows what might be coming. I'll make some private enquiries.

What is this Altsens (Rockwells??) chip and how does it compare in cost. I still hold to my 80/90% auguement, but not 50%. There must be more manufacturers out there, though I would still like a pro opinion on the perforamnce of all the chips. I'm walking aroud like a sensor with an inkblot on it at the moment, in respect to their true performance.

Rob,
The chip cost is the tip of the iceberg. The data rate from the Micron is 60Mpix/sec, the Rockwell is 150Mpix/sec. I suspect most people will use the Micron in 8 bit mode, the Rockwell in 12 bit (unpacked so 16 bit) mode. The system costs for the acquistion, wider PCI bus and disk bandwidth will chew up much more than the camera costs.

A person could buy a Rockwell/Altasens based camera and use a 1280x720 window initially to keep the cost down I suppose......

On sensors, I agree with Obin - Micron is the leader right now in low cost, low noise chips. Fill Factory is full frame shutter. It is like saying that a mustang is better than a humvee because it accelerates better. Low noise, high sensitivity is not what Fill Factory does.

Sorry, I am proprietorising cost sensitive issues down the line. There is a miminal cost for a miniminal (or desired) resolution/datarate, that we accept and is why we are willing to go three chip if cheap enough. When it comes to splitting hairs on chip perforamnce/cost, I'd rather go with the one that is much cheaper, or much better, and get much better value either way, just pragmatic.

So Steve, for a cinematic sensor you would chose the Micron that was suggested in the Viper thread, over all others, a week ago? I must say I think my upper limit for a sensor would be $1K single chip (divide by three for three chip).

Still wish I could read those techncial specs properly.

<<<-- Originally posted by Wayne Morellini : What is this Altsens (Rockwells??) chip and how does it compare in cost. -->>>

http://www.altasens.com/products.html

Aspect ratio - 16x9
Resolution - 1936 x 1086
Size - 2/3"
Up to 60 fps progressive (for the 3560 model)
Depth - 12 bits (Bayer)

It's the chip used in the Kinetta camera (http://www.kinetta.com) I haven't been able to determine the price range yet.

Steve -- Thanks for the info. You're right about the chip cost. What it boils down to me, though, is I really want a 10- or 12-bit 1080p solution that will provide 40+ fps. I'm really trying to determine the cost of an entire system that meets this critera and then decide what direction to go. That's why I was asking about the various frame grabbers and upcoming cameras from your company. Perhaps I should wait 8-16 weeks and see? :-)

It would be a fine thing to wait 8-12 weeks for an announcement. I'll do it here. You will be the first to know. I don't have exact numbers but the Altasens is something like 20x to 50x the cost of the Micron.

I think there are three groups here:
Low cost, very good quality - the Micron 1280x720 @24-30 (or 60)fps single chip will be fine.

Medium cost, excellent quality - Altasens, single chip

Hey, commercial cameras with recorders are $100K+, so $20K is a great deal - 3 chip, the Altasens.

Almost all of the discussions in Alternative Imaging Methods can be viewed this way. Each requires a different camera and system.

my thougths
 

well,

this is extremelly well explained, i mean the three main (and different) needs what Steve Nordhauser just noted here before.

personally - i will start with the cheapest solution and therefore learn (i hope! :)) something - and not spend tons of $$$. especially if there is also solution for slomo.

just a thought.

by the way - Steve, thank you for your help. i suggest you officially for the first DVinfo community medal! :))

filip

Hmmmm

Okay, I guess I'm a little confused. For some reason, I thought it was between medium 720p at around 3-5k and high quality at around 10k. Not sure 20k will work, but then if that covers the whole NLE as well, I guess not a bad wrap. If not, can't do 20k for just a camera. but beggars can't be choosers, I guess. Either way, thanks for everything, we love ya steve.

Laurence,
I was talking system prices to build one system. The first level - low cost, very good quality would be a micron camera, 720p, capture, PCI-32 computer with a two drive serial ATA RAID. Maybe $4K plus recording software.

The second level would be a PCI-X computer with at least a 4 drive RAID. I don't have real prices yet but around $8K-$10K.

The third level involves recording three 150Mpix/sec (300MB/sec for uncompressed raw data) simultaneously. Plus 3x the camera cost of #2 plus a prism. I took a hipshot at $20K. The results should be as good as any commercial HD camera.

Steve,

So are these real-world cameras that your company is offering right now, or are they cameras that are coming to market?

For instance, the Micron 720p camera or the Altasens 1080p camera, were are those currently in the production/release cycle?

Also is there anywhere we can see moving samples of these chips? Still shots are nice, but they don't tell the whole story, especially when you're wanting to evaluate motion characteristics.

Thanks.

The Micron 720p is available now in camera link and gigabit ethernet our model SI-1300.

Obin just received an SI-1300-RGB-CL. Maybe he can provide some independent feedback. If he wants, I will host a few video clips if he doesn't have a good host.

Other cameras will be announced in the next 8-12 weeks that will be of interest to this group. For specifics on our cameras, please contact me directly.
steve@siliconimaging.com

Hi Everyone -

Steve, thank you for all your help. The end result seems to be that for an investment of $20k, we'd have a camera with modular sturdy design and the abilty to upgrade chips as they become available. The repair would be as simple as removing a faulty sound card on a pc; the output would be equal to michael manns current movie shot on a viper, and this camera would easily handle any of the demands for tv broadcast. It amounts to about a month's rental of a viper , but leaves us with a camera we own and are able to upgade. In the upgrade, we'd come here and sell the old chip for someone else who is building a camera ...


this is very very good news

obin
I found a great site on the bolex16 camera
http://homepage.newschool.edu/~schlemoj/film_courses/bolex.html

Richard let me tell you what is very very good news! Steves single cmos 1300 camera!!!!!!!!!!!!!!!! I got some test images today from this unit by holding a film lens up infront of the camera and I can already tell it's a good chip - very good- can't wait to get the lens from Steve and get some real footage from this beast! I will setup a lit scene in our studio and shoot food/people/objects and also go outside and shoot sunlit stuff for all you boardpeople to take a look at....I will say that cameralink stuff is no childs play but I think that once I get the hang of all the software options things will be good )))))))))))WE NEED SOMEONE TO CODE A SIMPLE VERSION OF CAPTURE SOFTWARE FOR CAMERA LINK((((((((((((

anyone ???

obin, that is great news . I have built a agus35 with a piece of ground glass from a optical company . the glass is 1500 grit
It works perfect on mini dv. It might be high enough quality for your new camera.

man that would be awesome if it is...can you send me info on the stuff?

oolsen1@ec.rr.com

Obin,
I don't know too much about Windows environement, but here is a a link for some video capture software:
http://www.alternaware.com/

thanks for the post, I contacted Jeff via email asking if he would care to talk about code writing on this project....lets hope he digs the idea !

Hi

The 20* to 50* price of the Altasns over the Micro, is what I was getting at elsewhere (but if it is still below $1K single chip, it is good).

The problem with the thread is that the Summix is aimed at a very cheap rate, and once you go above 10K there are several planned commercial pro alternatives (the Kinnetta he most famouse). So really I think the sub 10K stuff is the place still to be developed.

The 720p Micron is doing well, but how is new (unreleased) multi mpixel version compare in picture performance? I notice that the more pixels sensors get the worst there picture performance compared to the lower res models.

I'm curiouse, what is the real resolution of your 1300 camera after filtering? Somebody mentioned a big real res drop to 1440 horizontal from bayer on 1920 chips. That is why I prefer 1080 if a single chip.

To all:

The problem, if the rumour be true, that Canon and Sony are going to use the same sensor out of a Sony Box camera. What ever we do it has to at least match this sensor, because in the end a portable solution (icluding the capture computer) is going to cost at least the same as the Sony HDV (though a single 720p could be cheaper).

Thanks

Wayne.

<<<-- Originally posted by Wayne
Now I have another idea, something like a lenticular array could be designed to split the individual direction the image primaries are projected to be read by three chips or three areas of one ;) A bit of single direction image compression would do the trick. This is based on an idea I have had a long time ago (as well as a projection idea): a lenticular array laid over a single sensor could take all the light per pixel area, mix it and split the primaries to 3 ajioning pixels (acting like little prisms). What you get is a completely acurate (less any abreviations) colour, each primary is sent 100% to it's own pixel (no major filter loses), all on a single chip, at very cheap price. The other advantages is that you get near 100% pixel area coverage, not 70% max, like in cmos, so you reduce the fly screen and bayer motion induced luma/chroma artifacts, and increase the used light (if you design it to miss the interpixel spacing. The other benefits of these screens is that they could be used as projection screens for the adaptors. I think I gave up on the idea after the foveon came out. So would this reduce the costs? When done right (with a couple of other adjustments) you could deliver all the transmittable light from a MF lesne right down to 1/2in chip. I have other ideas I am wanting to work on commercially aswell.

thanks

Wayne. -->>>


I remember where I saw something that gave me tis lenticular idea. I seem to remember in the information on Sony's HAD/hyperhad?? that they used a small microlense screen over the sensor to concentrate light on the pixels pads. Now if we could use something like this with a single chip (does Bayer allow this) filter we could get competely accuate colour (less some fidelity). Actaully if they could be an array of splitting prisms then we could get almost 3chip like findelity and light gathering power from a single chip (and the mass produced price of such an array would be a very small fraction of a proper prism), 90% of te benefits for 10% of the cost. What do you think?

I've been following this thread, the 4:4:4 thread and DVX100
thread with great interest. I really think the current industry
is moving in the wrong direction with HDV (MPEG2) and there
does not yet seem to be support for lots of things.

Most of you will probably know me as one of this boards
"moderators" (or wranglers as we call 'em). What not everybody
might know is that I'm also a professional programmer.

I mainly do pretty large intranet/extranet/website development
on the backend (business logic with connections between
systems and whatnot).

Some might even know me from a little device called the Iomega
Buz which was an analog capture card back in the days that
really didn't got off the ground as it could've been.

I've been dabbling with video (especially computer related)
for 10 years or so now and have programming experience with
a lot of languages including Assembly / C(++). I also have a
thorough understanding about low-level PC workings, protocols,
communications etc. etc.

Since Obin is desperetly looking for some programming assistance
I might be of some help. I do need to say that I only have real
experience with the PC and Windows / DOS / low-level and
not with things like Mac / Linux for example.

I'm willing to donate time to this project since it is of much interest
to myself and the community as a whole. The only problem is
that I do not have the money to but such a system myself just
yet.

What is the cost indication for the 1300 chip? Do you really need
that $500 box to connect the camera or is that just to make it
easy to get started?

<<<-- Originally posted by Rob Lohman :
I'm willing to donate time to this project since it is of much interest
to myself and the community as a whole. The only problem is
that I do not have the money to but such a system myself just
yet.
-->>>

Rob, I'm pretty much in the same boat. I don't have a lot of money, but I am interested in building one of these cameras in the near future. I am also a professional developer and have been thinking about developing software for this project.

I'm assuming we're talking about free/open-source sofware, so I've started to do a little digging to find GPL'd software that we can adapt for this purpose (The Gimp, CinePaint, transcode, Jahshaka, etc.).

I'm probably getting ahead of myself here, but here's a few of the first steps I was thinking ...

1 - Read raw files from SI-1300 camera
2 - Apply Bayer filtering
3 - Write files to OpenEXR format (preserving high bit depth)

The OpenEXR files can then be opened in CinePaint. Once we can do that, we'd have something working and could continue development to ...

1 - Support other file formats
2 - Perform color and gamma correction
3 - Support various lossless and lossy codecs and formats (QuickTime, AVI, Matroska, image sequence, etc.)

At some point, perhaps we should set up a SourceForge project for this...

Just FYI, I've added a page on my wiki to list possible software we can use:
http://obscuracam.com/wiki/wiki/run.php?iRequest=wiki/ViewPage&iPage=FOSS

One thing to watch - the Bayer filtering is a 3x data expansion with no new information added. That is why (aside from the real-time processing concerns) that most recording software records in RAW format. Also, not all Bayer algorithms are equal, so you are frequently better off doing it slowly as a post process.

I was also told (don't know for sure) that recording in AVI format is the only easy way to keep sound synced to video. That may not be a problem with editing software - I don't know.

Someone asked about lossless compression. Expect 1.5 to 2.3x compression for a non-noisy image for truely lossless. And never compress a raw color image - the Bayer will kill you. Point at a green background and you get a perfect checkerboard - completely uncompressable - most algorithms like RLE will expand the data (two bytes used to describe each byte - this color for this long).

If you want to go to YUV space on the fly, then you will have to do the Bayer on the fly also.

guys from what I can see sofar we need a VERY basic capture program that is easy to use unlike the camera link programs like streampix and norpix stuff...this needs to look and act as much as it can like a *normal* video capture app and we also need a way to view the live stream as it's captured on a TV or an LCD that can run by cable or wireless to the camera ..what would really be nice is if we could somehow take the cameralink signal and feed it to a harddisk AFTER we setup the camera settings via computer....

I agree with obin about the easy to use. I for one am much more an artist than a technician. The more "user friendly, one shot" type deal it is, the more condusive it will be on the set. And the more universal the finished captured footage is, the easier people can apply it to different platforms. Keep in mind most people using this stuff will be technical laymen in your eyes. This sounds awesome guys. Keep it up.

Software
 

Obin & Laurence -- I absolutely agree with both of you. Problem is, without a camera & card, I'm not sure I can develop that software.

As a first step, I thought it would be helpful to be able to successfully open & process a raw file. Then I might have enough confidence to buy the camera and continue development.

Obin, what comes with the capture card? For example, does it come with

1 - an SDK (Software Development Kit)
2 - documentation of the file format(s)
3 - sample files

Those would be helpful to get started.

Ask this of Steve Rob because I am not sure I *think* I heard somthing about a SDK. I will poke around and find out!! I may even work some type of deal like send you my camera for a couple weeks if you really need it for code writing...we can talk for sure about this as things progress . It sure would be great if you could find a way to cut the PC out of the loop once the camera was setup for capture...what about the device that Steve sells that takes cameralink and feeds it down a gigabit wire to a gigabit card on the pc? can this somehow be fed to a disk instead of pc? dunno just an idea....

<<<-- Originally posted by Obin Olson : It sure would be great if you could find a way to cut the PC out of the loop ... what about the device ...that takes cameralink and feeds it down a gigabit -->>>

I've had similar ideas, Obin, and I'm not exactly sure which way to go yet. I think it makes sense to go with CameraLink for now, since it supports a higher bandwidth than Gigabit Ethernet. We'd be set up for higher-resolution chips later on.

My current idea is to build the camera unit into a box with a Mini-ITX motherboard, CameraLink card and a RAID 0 2-hard-drive array. With appropriate software on-board, you'd be able to capture with a (pretty much:-) portable unit. Offline, you'd connect to the camera box via Ethernet & FTP and download the raw files; post-processing would take place on your PC later.

Either way, you need the software on your PC to take the raw files and process them, so that was where I was going to start.

I have some details here:
http://www.obscuracam.com/wiki/wiki/...esignApproach2

(I'd love to try out your camera, Obin, but I might end up buying my own before too long. We'll see!)

Steve -- does this capture card come with an SDK? Can you send me some of the documentation? Thanks!

Rob,
There is an SDK for the capture card. You can get Linux or Windoze support. The cost is $495 but there are no runtime fees so if someone developed a recording application, there would be no cost involved in sharing/selling it. It would be tied to the frame grabber somewhat. If anyone undertakes this and is offering the software free, I will give them aggressive pricing on the SDK and a discount on the camera.

There are two places software is needed - first, capture to disk in RAW format with display preview and basic camera controls (gain/digital exposure). The commercial options I know of are around $1500. CineForm may provide a lower cost option.

Second is the pre-post processing, if you follow that nasty phrase. Moving the RAW images through the basic pipeline - Bayer, YUV, any color balance, etc. to some generic format that can be used by the standard post processing tools.

I think Rob's computer would be perfect for the SI-1300 capture. Find a motherboard using the chipsets with integrated RAID - no bus traffic. Use cheap big drives.

<<<-- Originally posted by Steve Nordhauser :
Rob, There is an SDK for the capture card. You can get Linux or Windoze support. The cost is $495 but there are no runtime fees ... -->>>

That's exactly what I'm interested in!

Another question -- Do you have any idea if there are any similar boards in the Mini-PCI form factor?

<<<-- Find a motherboard using the chipsets with integrated RAID - no bus traffic. Use cheap big drives. -->>>

Exactly -- Mini ITX motherboards are affordable and many of them have integrated RAID and Gigabit Ethernet.

<<<-- capture to disk in RAW format with display preview and basic camera controls (gain/digital exposure) -->>>

Right, that software would be inside the box with a monitor output.

<<<-- Bayer, YUV, any color balance, etc. to some generic format -->>>

Yup, the conversion software would run on the user's PC or Mac after the raw images were downloaded from the camera box.

Why do we want to convert to YUV from RGB??

That seems like a big step down IMHO.

If possible, it would be great to keep the option for full-bit-depth 4:4:4 RGB files.

YUV
 

<<<-- it would be great to keep the option for full-bit-depth 4:4:4 RGB files -->>>

Absolutely, full 4:4:4 RGB (post-Bayer) would be available for output. (The OpenEXR format, for example, supports either RGB or YUV.)

In some cases you might need some processing that you can only do in the YUV domain. Some types of color correction, for example.

Also, some lossless (and near-lossless) codecs work only in YUV -- HuffYUV, for example, and Cinelerra's 10-bit uncompresed codec.

(edit - BTW, YUV isn't necessarily a step down -- it can still remain 4:4:4 and lossless. You don't lose any information other than a tiny bit because of the mathematical conversion involved.)

sounding VERY good. I did not know that ITX boards had pci or pci-x slots..ROB I jsut spoke with the cameralink card company and they will have a 64bit card in a few weeks forsale..that will allow a MUCH higher datarate = higher framerate and or higher bit depth all = good things ;) um what aobut this 10bit stuff? if I save as 10bit like the camera outputs how do you use that? I think that file codecs are 8bit 12bit and 16bit right? or?

I can't find any ITX boards with sata but lots of mini-atx with raid sata

http://www.directron.com/ideq200vb.html

<<<-- Originally posted by Obin Olson : I did not know that ITX boards had pci or pci-x slots -->>>

Yes, all the Mini-ITX boards I've seen have one PCI slot. (Mini-ITX is 170mm x 170mm)

<<<-- they will have a 64bit card in a few weeks forsale..that will allow a MUCH higher datarate = higher framerate and or higher bit depth all = good things ;) -->>>

Yes, definitely something to keep an eye on, but IIRC, we don't need anything beyond 32-bit for this particular camera.

<<<-- what aobut this 10bit stuff? if I save as 10bit like the camera outputs how do you use that? I think that file codecs are 8bit 12bit and 16bit right? or? -->>>

You're right, most codecs are 8-bit. (HuffYUV, for example, is 8-bit 4:2:2)

Most of the uncompressed codecs I've seen are 10-bit. For example, Apple's QuickTime has an uncompressed 4:4:4 10-bit codec. I'm not sure about AVI codecs -- I haven't seen anything other than 8-bit so far.

What file format will you be needing to use?

We need to keep the Rob's apart in this thread, heh.

Steve: I'm very interested (and the other Rob seems too) in such
an SDK and camera option. I'm all for open software on this since
there will be enough hurdles to take after we complete this.
Perhaps we could both work on the project together Rob?

A mini PC of some sort would be most ideal in my mind as well.
It should be more easy to work with than a chip which needs
programming and if we look for the right one has all the options
we need to hook it up to the rest of the world.

It's basically writing the camera's "firmware". Can you shoot me
an e-mail Steve with some more details?

well the question is if you save as a 16bit tiff then you waist the extra bits right? and make a file that is bigger then needed?

I will need 10bit or the max the camera can output so that in post with software like Combustion we can take all the bits and color grade the footage images are fine I don't have any need at all for avi or "video" files

<<<-- Originally posted by Obin Olson : I can't find any ITX boards with sata but lots of mini-atx with raid sata -->>>

Yeah, I'm having some trouble finding exactly the right board. I'm sure it's out there (or will be soon). This one is pretty close:

http://www.viaembedded.com/product/e...herboardId=261

... but has only 100 Mbps Ethernet. Painfully slow when downloading such huge files.

<<<-- Originally posted by Rob Lohman : We need to keep the Rob's apart in this thread, heh. -->>>

Too true! :-)

<<<-- Perhaps we could both work on the project together Rob? -->>>

Sure, I'd love to! Assuming, of course, that we can agree on the goals and detailed design. I suspect we're on the same page.

<<<--- A mini PC of some sort would be most ideal in my mind as well. ... It's basically writing the camera's "firmware". -->>>

Exactly. I looked into embedded chips and so forth, but that seemed a little intimidating to start out with. On the other hand, I'd like to write the software in a modular enough fashion that someone could adapt it to a fully custom hardware later on.

<<<-- Originally posted by Obin Olson : well the question is if you save as a 16bit tiff then you waist the extra bits right? and make a file that is bigger then needed? -->>>

That's true, assuming there is no compression (such as RLE). I'm not very familiar (yet) with what you can do with a 16-bit TIFF file. Assuming it can be compressed, the extra bits (containing lots of zeroes) will disappear.

<<<-- I will need 10bit or the max the camera can output -->>>

Absolutely. I think it's extremely important to find as many formats as possible to keep all the bits intact as far into the post-processing chain as we can. Ideally we won't "truncate" to 8 bits until we're doing the final encoding.

Can I assume that some of the software you use -- such as Combustion -- supports 16-bit TIFF files? I know CinePaint supports 32-bit TIFF, so I may take a look at its source code to see how it works.