Rai & Markus' "Drake" HD camera

Steve Nordhauser · May 6th, 2005, 11:26 AM

Levan,
I'll address the first half since someone in cinematography is better suited to discuss log/linear.

The difference you are describing isn't CCD vs CMOS. It is single chip color vs 3 chip color. In a 3 chip camera, a beam splitter will separate out the R,G, and B parts of the incoming image and send them to individual monochrome sensors. As you say, with 8 bit A/D converters, this gives you 24 bit color although you can go to 10 or 12 bits per color per pixel and use a log encoding to bring you back down to 24 bits.

Single chip color almost always uses a Bayer color filter of 50% green, 25% blue and 25% red pixels. One 8-12 bit value is generated for each pixel. The other colors are interpolated. Rather than go into detail, try some of our website - tasty but less filling:

http://siliconimaging.com/ARTICLES/CMOS%20PRIMER.htm
http://siliconimaging.com/RGB%20Bayer.htm

The color from a single chip camera is not quite as good as a 3 chip but the cameras are much cheaper (a 3 chip is like 3 independent cameras) and recording is 1/3 the bandwidth. This does clobber your resolution when you look at blue or red objects however.

Michael Maier · May 6th, 2005, 01:28 PM

Quote:

Originally Posted by Levan Bakhia

What does Altasens camera record to? Is it hard drive or what? How can I then transfer the image to my Macintosh?

Well, as I understand it, there's no Altasens cameras. Altasens are just CMOS sensors. Not really a camera. It seems the new JVC 2/3" HDV camera will sue those. But I don't think any camera is suing them already. Also, if threre will be any industrial cameras using them, I think they will be expensive cameras, like the Dalsa. So not really a good optio for DIY I guess.

Levan Bakhia · May 6th, 2005, 03:45 PM

Great links Steve, I read it carefully. Now this definetely means that CCD is superior to CMOS, since in some way, 3CCD means that each pixel in the image has exact information of the shades of gray of Red, Green and Blue, while in CMOS, proportion (intensity of) Red Green and Blue are assumed base on Bayer filter, so it might have some (bot very little) mistakes itself.

This all is interesting, but when I look at images of CCD cameras, they look flat and unartistic. Could it be because they use different kind of lenses. I mean, as I know there is prism inside that splits the colors for 3CCDs, while in CMOS, the light travels direct and hits the sensor. Could it be the reason?

And one more question. In the artical there is info that visible wavelength is from 400-750nm. What about this sensors, do they sense full range? Or does this range also vary depending on the sensor?

Does anybody know what is the price for DALSA? Is their camera already available at all?

One more question to Michael. why would ALTASENS manufacture sensors, if nobody uses it? And why they call it camera, if it is just a sensor?

I also visited www.siliconimaging.com. There are two things I want to know. How good are the images from this cameras? Could I use those higher frame rate cameras for cinema projects rather than scientific observations? and I want to set up an array of up to 125 digital cameras (could be still), sycronised, connected to a computer and recording frames. To create special effect, like the one used in matrix?

I know I am going out of the subject of this thread, we are here to talk about and discuss DRAKE, but... :*)

Steve Nordhauser · May 6th, 2005, 06:16 PM

Levan,
I'll try again. One sensor or three. It has nothing to do with CCD or CMOS. I can make a 3 chip CMOS camera easier than a 3 chip CCD. There are many single chip CCD cameras.

Look at the data sheet for one of our cameras or at the Micron data sheets. You will see a set of response curves. There should be a monochrome response of the sensor over a range of wavelengths. There should also be the R,G,B respsonse that is a function of the Bayer color filter mask interacting with the monochrome sensor response. The response is not in any way linear. It is most sensitive in green.

As Michael said, Altasens is a sensor company, like Micron, Fill Factory and many others. Camera manufacturers (like Silicon Imaging and Drake - mentioning them since this is their thread) buy sensors like any other integrated circuit and build cameras using them.

People are using the Altasens sensor. We have a camera with one. They don't have their yields up high enough yet for high volume quantities which is keeping the camera manufacturers (us again) from announcing that we are ready to ship thousands of cameras. For now it is tens.

Yes, industrial cameras can be used as the head of an HDV camera. The cost is fairly high in single pieces because the cost of a single sale is high (someone has to pay for the time I spend on this board). OEMs pay much lower prices because it doesn't cost much more to support a sale of 50 cameras than a sale of 1. On the other hand, unless you are a student, the time spent designing a camera to own a single camera is just not worth it.

Rob Lohman · May 7th, 2005, 06:26 AM

Steve: thanks for expanding on my text with the "analog" part. I wanted to
describe this, but my knowledge about sensors (wells and noise) is too limited.
Therefore I just talked about the digital part (# of bits).

Levan: the number of bits has nothing to do with how many chips you have.
A 3 CCD or 3 CMOS camera will simply have more resolution (if the resolution
on each chip is the same and you just have more chips) to play with. Both
systems can output at 8, 10 or 12 bits *PER* pixel.

That value is always PER pixel. In the case of Bayer for each color there is
just one pixel (basically, you would interpolate this to three colors per pixel).
In a 3 CCD/CMOS setup there are 3 colors per pixel.

That is the reason that an RGB datastream will be 3x times as large as an
(unprocess/not-interpolated) Bayer datastream.

That is the reason we like to store the Bayer stream and convert this (again,
through interpolation) into the RGB stream on a computer after you have
recorded the footage.

Levan Bakhia · May 7th, 2005, 12:47 PM

Yes, Rob, I now understand that there can be both 3CCD and 3CMOS chips, and I guess this means that CCD can also use Bayer technology. But what this means is that, using one CMOS sensor is not defining each pixel color as it would with 3CMOS or 3CCD. As I understand, using bayer filter assigns final color to each pixel rather than 3 (RED GREEN and BLUE) gray levels of the color of this pixel. So by using bayer filter, you are not exact in color information. But, then on the other hand, all the 3 CCD camera images look more like video than even one CMOS image. I have never seen footage from 3CMOS sensors and I can not judge it. So what is the reason for that? Could it be that the prism that brakes light into RGB takes away DOF or anything else that make film different from video. All the CMOS sensor shot I have seen look more filmish, even those with not so good optics. Or maybe I am mistaken. But, I saw VIPER on NAB, and saw footage from this camera. I wouldn't be right to say that it was BAD, but as of my final decision I didn't like it. Anyways I couldn't afford it, but even if I could I would better buy the DARKE. Now, I am judging DARKE solely by it's sample from the web page of course, and I haven't seen it, and I haven't had a chance to play with it, and I definetely will as soon as it will be on sale, before I buy, but from this point of view I like it. I am no professional here, but I think that the good look of the images from DRAKE are all due to CMOS, maybe I am absolutely wrong and forgive me if I confuse other readers. So to finalize,what I tried to say here is that, I understand that technically 3CCD should be superior to 1 CMOS, but on the practice from a guy who is not a professional engineer or technician, 1CMOS works better. Maybe this is due to the prism.

I also saw P+S technics mounted to XL2, and it really looked nice, but all the compressions of mini DV format is not for me. I wonder if P+S technics would add some DOF to DRAKE.

That's it.

Aaron Shaw · May 7th, 2005, 07:01 PM

Personally, I think the reason CMOS looks so much more like film isn't because it's CMOS - rather it's the complete lack of compression and zero in camera sharpening. I suspect the results would be the same with a CCD.

Levan Bakhia · May 8th, 2005, 03:28 AM

Quote:

Originally Posted by Aaron Shaw

Personally, I think the reason CMOS looks so much more like film isn't because it's CMOS - rather it's the complete lack of compression and zero in camera sharpening. I suspect the results would be the same with a CCD.

But, VIPER, when recording in filmstream mode is also uncompressed and 4:4:4, and also if you have seen modificatoin of DVX100 by reelstream (www.reelstream.com) is also uncompressed, but it can not be compared to DRAKE at all. But then I really expect to see DVX100 modified by reelstream with mini35 and some good lenses. I expect to see samples of this setup.

Rai Orz · May 9th, 2005, 02:25 AM

CCD or CMOS ? Single sensor or 3sensors? Sensor size?
Viper is a geat camera,with high resolution 3x 2/3" CCDs + RGB prism. It is based on the same technology like other video cameras. ARRI is a movie camera manufactor and the D 20 use only a single 35mm CMOS with considerably less resolution. The result looks mutch more like film. The sensor is the main different and this example shows, a single CMOS can be mutch better as 3 x CCDs. But it need a lot of other things.

We had started this project, because we wonted a digital camera for film like pictures. Original it was designed only for one movie and only as a camerahead + long cable + PC Unit. We had top engineers for optics, mechanics, electronics and software. All with movie making know how. The result was so excellent that we develop this independent, little and lightweight Drake camera. Now the serial production has begun and we wait for parts.
Our engineers work on next projects, but in the future we will not discuss details on forums till it will be ready. We will not do the same inapt like Altasens or other companys who said: "...it will be available in n days..." and nothings comes. (BTW: Kinetta use also Altasens...). If we talking about future projects here, then it is always ready.

DRAKEs biggest advantage is, it is a modular camera. The main body contain hardware for recording mutch more than 1280x720p, 8Bit, also more than 1920x1080p, 12 bit. Thats why Drake is ready for future sensors and in the near future you can select (or upgrade) between different cameraheads with different sensors.

Levan Bakhia · May 10th, 2005, 12:58 AM

[QUOTE=Rai Orz]CCD or CMOS ? Single sensor or 3sensors? Sensor size?
Viper is a geat camera,with high resolution 3x 2/3" CCDs + RGB prism. It is based on the same technology like other video cameras. ARRI is a movie camera manufactor and the D 20 use only a single 35mm CMOS with considerably less resolution. The result looks mutch more like film. The sensor is the main different and this example shows, a single CMOS can be mutch better as 3 x CCDs. But it need a lot of other things.

Yes, that is exactly, what I thought, but could you in brief explain what is IT, that makes CMOS footage look more like film, and what about 3 CMOS sensor? Would it be better or do you think it might have a higher resolution but loose that special look of film?

Rai Orz · May 10th, 2005, 02:06 AM

CCD or CMOS?
CMOS dynamic range is higher than CCD. Its mutch more like film. Thats the main different.

one ore 3 sensors (CCD or CMOS)?
The different resolution is mentioned before. But practice and theory are often different. Colors and resolution are easy to handle with 3 chips. One chip nead a debayer software. But there are a lot of different debayers out. And not all are good. I like to compare debayer software with translation software. Each software translate words, but with bad software nobody understand sentences. That is one of the tricks.
The other big different is the prism itself. It need a longer optical distance between lenses and sensor, because the prism light path is longer. With one chip each film lenses (S16mm) work. With prism you need special lenses. And thats the third main different. A lens with longer distance (between lens and sensor) will never have those short DOF.

All together and you know why ARRI D20 (and Drake) go the one single CMOS way.

Levan Bakhia · May 10th, 2005, 02:56 AM

Thank you Roi!

I was almost sure, by logicaly thinking, that prism was the reason for loosing DOF, but I couldn't see the theoretical arguments. Now I understand.

Obin Olson · May 11th, 2005, 10:43 PM

Rai is Drake hardware? or is it software running on Linux/windows with a HIGHSPEED Frame Grabber card?

Keith Wakeham · May 12th, 2005, 09:34 AM

Rai, i tried emailing your new email, but it just bounced back, so i forwarded to your old one.

Email me again to confirm the address please.

Thanx
Keith

Rai Orz · May 13th, 2005, 10:03 AM

Quote:

Originally Posted by Obin Olson

Rai is Drake hardware? or is it software running on Linux/windows with a HIGHSPEED Frame Grabber card?

Drake is hardware and software. It is also ready for 1080p if you mean that, but that will not be all...

May 6th, 2005, 11:26 AM	#256
Steve Nordhauser Silicon Imaging, Inc. Join Date: May 2004 Location: Troy, NY USA Posts: 325	Levan, I'll address the first half since someone in cinematography is better suited to discuss log/linear. The difference you are describing isn't CCD vs CMOS. It is single chip color vs 3 chip color. In a 3 chip camera, a beam splitter will separate out the R,G, and B parts of the incoming image and send them to individual monochrome sensors. As you say, with 8 bit A/D converters, this gives you 24 bit color although you can go to 10 or 12 bits per color per pixel and use a log encoding to bring you back down to 24 bits. Single chip color almost always uses a Bayer color filter of 50% green, 25% blue and 25% red pixels. One 8-12 bit value is generated for each pixel. The other colors are interpolated. Rather than go into detail, try some of our website - tasty but less filling: http://siliconimaging.com/ARTICLES/CMOS%20PRIMER.htm http://siliconimaging.com/RGB%20Bayer.htm The color from a single chip camera is not quite as good as a 3 chip but the cameras are much cheaper (a 3 chip is like 3 independent cameras) and recording is 1/3 the bandwidth. This does clobber your resolution when you look at blue or red objects however. __________________ Silicon Imaging, Inc. We see the Light! http://www.siliconimaging.com

May 6th, 2005, 06:16 PM	#259
Steve Nordhauser Silicon Imaging, Inc. Join Date: May 2004 Location: Troy, NY USA Posts: 325	Levan, I'll try again. One sensor or three. It has nothing to do with CCD or CMOS. I can make a 3 chip CMOS camera easier than a 3 chip CCD. There are many single chip CCD cameras. Look at the data sheet for one of our cameras or at the Micron data sheets. You will see a set of response curves. There should be a monochrome response of the sensor over a range of wavelengths. There should also be the R,G,B respsonse that is a function of the Bayer color filter mask interacting with the monochrome sensor response. The response is not in any way linear. It is most sensitive in green. As Michael said, Altasens is a sensor company, like Micron, Fill Factory and many others. Camera manufacturers (like Silicon Imaging and Drake - mentioning them since this is their thread) buy sensors like any other integrated circuit and build cameras using them. People are using the Altasens sensor. We have a camera with one. They don't have their yields up high enough yet for high volume quantities which is keeping the camera manufacturers (us again) from announcing that we are ready to ship thousands of cameras. For now it is tens. Yes, industrial cameras can be used as the head of an HDV camera. The cost is fairly high in single pieces because the cost of a single sale is high (someone has to pay for the time I spend on this board). OEMs pay much lower prices because it doesn't cost much more to support a sale of 50 cameras than a sale of 1. On the other hand, unless you are a student, the time spent designing a camera to own a single camera is just not worth it. __________________ Silicon Imaging, Inc. We see the Light! http://www.siliconimaging.com

May 7th, 2005, 06:26 AM	#260
Rob Lohman RED Code Chef Join Date: Oct 2001 Location: Holland Posts: 12,514	Steve: thanks for expanding on my text with the "analog" part. I wanted to describe this, but my knowledge about sensors (wells and noise) is too limited. Therefore I just talked about the digital part (# of bits). Levan: the number of bits has nothing to do with how many chips you have. A 3 CCD or 3 CMOS camera will simply have more resolution (if the resolution on each chip is the same and you just have more chips) to play with. Both systems can output at 8, 10 or 12 bits PER pixel. That value is always PER pixel. In the case of Bayer for each color there is just one pixel (basically, you would interpolate this to three colors per pixel). In a 3 CCD/CMOS setup there are 3 colors per pixel. That is the reason that an RGB datastream will be 3x times as large as an (unprocess/not-interpolated) Bayer datastream. That is the reason we like to store the Bayer stream and convert this (again, through interpolation) into the RGB stream on a computer after you have recorded the footage. __________________ Rob Lohman, visuar@iname.com DV Info Wrangler & RED Code Chef Join the DV Challenge \| Lady X Search DVinfo.net for quick answers \| Buy from the best: DVinfo.net sponsors

May 6th, 2005, 03:45 PM	#258
Levan Bakhia Major Player Join Date: Apr 2005 Posts: 230	Great links Steve, I read it carefully. Now this definetely means that CCD is superior to CMOS, since in some way, 3CCD means that each pixel in the image has exact information of the shades of gray of Red, Green and Blue, while in CMOS, proportion (intensity of) Red Green and Blue are assumed base on Bayer filter, so it might have some (bot very little) mistakes itself. This all is interesting, but when I look at images of CCD cameras, they look flat and unartistic. Could it be because they use different kind of lenses. I mean, as I know there is prism inside that splits the colors for 3CCDs, while in CMOS, the light travels direct and hits the sensor. Could it be the reason? And one more question. In the artical there is info that visible wavelength is from 400-750nm. What about this sensors, do they sense full range? Or does this range also vary depending on the sensor? Does anybody know what is the price for DALSA? Is their camera already available at all? One more question to Michael. why would ALTASENS manufacture sensors, if nobody uses it? And why they call it camera, if it is just a sensor? I also visited www.siliconimaging.com. There are two things I want to know. How good are the images from this cameras? Could I use those higher frame rate cameras for cinema projects rather than scientific observations? and I want to set up an array of up to 125 digital cameras (could be still), sycronised, connected to a computer and recording frames. To create special effect, like the one used in matrix? I know I am going out of the subject of this thread, we are here to talk about and discuss DRAKE, but... :*)

May 7th, 2005, 12:47 PM	#261
Levan Bakhia Major Player Join Date: Apr 2005 Posts: 230	Yes, Rob, I now understand that there can be both 3CCD and 3CMOS chips, and I guess this means that CCD can also use Bayer technology. But what this means is that, using one CMOS sensor is not defining each pixel color as it would with 3CMOS or 3CCD. As I understand, using bayer filter assigns final color to each pixel rather than 3 (RED GREEN and BLUE) gray levels of the color of this pixel. So by using bayer filter, you are not exact in color information. But, then on the other hand, all the 3 CCD camera images look more like video than even one CMOS image. I have never seen footage from 3CMOS sensors and I can not judge it. So what is the reason for that? Could it be that the prism that brakes light into RGB takes away DOF or anything else that make film different from video. All the CMOS sensor shot I have seen look more filmish, even those with not so good optics. Or maybe I am mistaken. But, I saw VIPER on NAB, and saw footage from this camera. I wouldn't be right to say that it was BAD, but as of my final decision I didn't like it. Anyways I couldn't afford it, but even if I could I would better buy the DARKE. Now, I am judging DARKE solely by it's sample from the web page of course, and I haven't seen it, and I haven't had a chance to play with it, and I definetely will as soon as it will be on sale, before I buy, but from this point of view I like it. I am no professional here, but I think that the good look of the images from DRAKE are all due to CMOS, maybe I am absolutely wrong and forgive me if I confuse other readers. So to finalize,what I tried to say here is that, I understand that technically 3CCD should be superior to 1 CMOS, but on the practice from a guy who is not a professional engineer or technician, 1CMOS works better. Maybe this is due to the prism. I also saw P+S technics mounted to XL2, and it really looked nice, but all the compressions of mini DV format is not for me. I wonder if P+S technics would add some DOF to DRAKE. That's it.

May 7th, 2005, 07:01 PM	#262
Aaron Shaw Major Player Join Date: Jul 2004 Posts: 587	Personally, I think the reason CMOS looks so much more like film isn't because it's CMOS - rather it's the complete lack of compression and zero in camera sharpening. I suspect the results would be the same with a CCD.

May 9th, 2005, 02:25 AM	#264
Rai Orz Regular Crew Join Date: Jun 2004 Location: Germany Posts: 136	CCD or CMOS ? Single sensor or 3sensors? Sensor size? Viper is a geat camera,with high resolution 3x 2/3" CCDs + RGB prism. It is based on the same technology like other video cameras. ARRI is a movie camera manufactor and the D 20 use only a single 35mm CMOS with considerably less resolution. The result looks mutch more like film. The sensor is the main different and this example shows, a single CMOS can be mutch better as 3 x CCDs. But it need a lot of other things. We had started this project, because we wonted a digital camera for film like pictures. Original it was designed only for one movie and only as a camerahead + long cable + PC Unit. We had top engineers for optics, mechanics, electronics and software. All with movie making know how. The result was so excellent that we develop this independent, little and lightweight Drake camera. Now the serial production has begun and we wait for parts. Our engineers work on next projects, but in the future we will not discuss details on forums till it will be ready. We will not do the same inapt like Altasens or other companys who said: "...it will be available in n days..." and nothings comes. (BTW: Kinetta use also Altasens...). If we talking about future projects here, then it is always ready. DRAKEs biggest advantage is, it is a modular camera. The main body contain hardware for recording mutch more than 1280x720p, 8Bit, also more than 1920x1080p, 12 bit. Thats why Drake is ready for future sensors and in the near future you can select (or upgrade) between different cameraheads with different sensors.

May 10th, 2005, 12:58 AM	#265
Levan Bakhia Major Player Join Date: Apr 2005 Posts: 230	[QUOTE=Rai Orz]CCD or CMOS ? Single sensor or 3sensors? Sensor size? Viper is a geat camera,with high resolution 3x 2/3" CCDs + RGB prism. It is based on the same technology like other video cameras. ARRI is a movie camera manufactor and the D 20 use only a single 35mm CMOS with considerably less resolution. The result looks mutch more like film. The sensor is the main different and this example shows, a single CMOS can be mutch better as 3 x CCDs. But it need a lot of other things. Yes, that is exactly, what I thought, but could you in brief explain what is IT, that makes CMOS footage look more like film, and what about 3 CMOS sensor? Would it be better or do you think it might have a higher resolution but loose that special look of film?

May 10th, 2005, 02:06 AM	#266
Rai Orz Regular Crew Join Date: Jun 2004 Location: Germany Posts: 136	CCD or CMOS? CMOS dynamic range is higher than CCD. Its mutch more like film. Thats the main different. one ore 3 sensors (CCD or CMOS)? The different resolution is mentioned before. But practice and theory are often different. Colors and resolution are easy to handle with 3 chips. One chip nead a debayer software. But there are a lot of different debayers out. And not all are good. I like to compare debayer software with translation software. Each software translate words, but with bad software nobody understand sentences. That is one of the tricks. The other big different is the prism itself. It need a longer optical distance between lenses and sensor, because the prism light path is longer. With one chip each film lenses (S16mm) work. With prism you need special lenses. And thats the third main different. A lens with longer distance (between lens and sensor) will never have those short DOF. All together and you know why ARRI D20 (and Drake) go the one single CMOS way.

May 10th, 2005, 02:56 AM	#267
Levan Bakhia Major Player Join Date: Apr 2005 Posts: 230	Thank you Roi! I was almost sure, by logicaly thinking, that prism was the reason for loosing DOF, but I couldn't see the theoretical arguments. Now I understand.

May 11th, 2005, 10:43 PM	#268
Obin Olson Trustee Join Date: Jan 2003 Location: Wilmington NC Posts: 1,414	Rai is Drake hardware? or is it software running on Linux/windows with a HIGHSPEED Frame Grabber card?

May 12th, 2005, 09:34 AM	#269
Keith Wakeham Major Player Join Date: Jan 2005 Location: St. John's, NL, Canada Posts: 416	Rai, i tried emailing your new email, but it just bounced back, so i forwarded to your old one. Email me again to confirm the address please. Thanx Keith