I dont get it... Effective pixels

Yik Kuen · October 20th, 2003, 07:17 PM

Heard that after DV compression, the final captured images of Mega 1CCD cams are very close if not exceeding the 3CCD siblings, unless we have higher bandwidth format such as the 50mbps DVCPro

Juan P. Pertierra · October 20th, 2003, 07:41 PM

That is completely possible. I do not know the specific math, but it is probable that with a large enough CCD, the mosaic approximation gets good enough that the human eye cannot tell the difference between the approximation and a 3CCD setup.

Juan

Jeff Donald · October 20th, 2003, 07:49 PM

There are many factors to consider. In a 3 chip camera each color has say 340,000 pixels (seems like a popular number). In your example of a four color Bayer matrix your million pixels are divided by 4, yielding only 250,000 per color. Signal to noise ratio is generally higher in 3 chip cameras also.

While single chip cameras do very well under many different shooting conditions, high contrast scenes generally suffer greatly. High contrast scenes shot with 3 chip cameras perform much better and capture greater detail, with lower noise.

Yik Kuen · October 20th, 2003, 10:57 PM

To make things complicated, Sony's new DCR-PC300/330 has 2 MP effective pixels for video and uses RGB filter instead. Under sufficient lights, I guess this cam can really produce images that are very close to 3CCDs.

So, do we still need 3CCD cams in the very near future when all these Mega cams become so popular?

Jeff Donald · October 21st, 2003, 04:12 AM

I think single chip cameras are the way of the future for professional use. It will take a couple of years, but eventually 3 chip cameras will fade away.

Peter Sieben · October 21st, 2003, 11:39 AM

Well guys, I just ordered my 3 ccd DVX100 to get a decent camcorder for short movies, and now you all say that a new consumer 1 ccd camcorder will also do the job. ;-(

Peter

Juan P. Pertierra · October 21st, 2003, 11:46 AM

Peter,

There isn't a single 1CCD camcorder out there that can put out as good images as the DVX100. Some people might argue the new JVC HD is better, but if you put HD aside, the DVX has better color, 24P and great manual control which the JVC does not have. The JVC is indeed the first of a kind, but it is extremely buggy, to the point where it is pretty much unusable proffessionally.

You made a great buy...and you will know this when you get it. There's nothing quite like the DVX. :)

I guess what was meant on here is that 'someday' 1CCD cams might be as good or better than 3CCD cams, but not today.

Juan

Peter Sieben · October 21st, 2003, 11:58 AM

Hi Juan,

That was already clear to me. I was just thinking of the fact that I had been thinking and counting a lot before doing such a big investment, and a couple of days later I read this forum thread.

I've already rented the DVX100 and know what a beauty it is.

Peter

Stephen van Vuuren · October 21st, 2003, 12:06 PM

Peter:

1 chip, 2 chip, 3 chip...

it's a red herring. The big news about the DVX100 CCD's is not that they are 3 chip, but they are progressive scan. That difference is much bigger than this chip count issue.

The fact that they are 24p progressive scan capable at under $20,000 is revolutionary.

Juan P. Pertierra · October 25th, 2003, 05:12 PM

I was playing with a Sony TRV19 (340k pixel 1 chipper) I just got for the experiment i'm doing, when I noticed something that might illustrate this whole effective pixel issue and one of the reasons why single chippers usually go towards larger pixel counts...

I had the camera on, aimed at the TV. Well, if you look at the red power LED on your TV through the camera, in a way such that the LED occupies approximately one pixel of the CCD(such that it looks very small), and then sloooowly move the camera up and down, or left and right, you will see the 'red' LED change colors.

The reason for this is that since the LED occupies approximately one pixel on the CCD(and the adjacent pixels are dark since it was a black TV), the ONLY data about the color of the LED is whichever color mosaic filter is in that particular CCD element. Thus, if you move the camera slowly up and down or left and right you should see a color sequence for the LED which corresponds to the mosaic configuration used in the camera. In my case it is Mg, Cy, Ye and G(i think).

Anyway, so this is a simple example of one of the problems of approximating colors with one chip. A simple solution, is to simply increase the effective pixel count, such that in the same situation as above, on say a TRV70, the LED would occupy at least 9 CCD elements, from which the actual LED color can be approximated regardless where on the chip it is. Of course, if you move farther away from the LED you can still get it to the point where it ocuppies one element and the problem is seen again, but the more the elements, the smaller the LED will be when the problem becomes apparent.

I hope this helps....
Juan

Agus Casse · October 26th, 2003, 02:35 AM

Nice experiment... thanks juan

October 20th, 2003, 07:49 PM	#18
Jeff Donald Warden Join Date: Mar 2002 Location: Clearwater, FL Posts: 8,287	There are many factors to consider. In a 3 chip camera each color has say 340,000 pixels (seems like a popular number). In your example of a four color Bayer matrix your million pixels are divided by 4, yielding only 250,000 per color. Signal to noise ratio is generally higher in 3 chip cameras also. While single chip cameras do very well under many different shooting conditions, high contrast scenes generally suffer greatly. High contrast scenes shot with 3 chip cameras perform much better and capture greater detail, with lower noise. __________________ Jeff Donald Carpe Diem Search DVinfo.net for quick answers \| Where to Buy? From the best in the business: DVinfo.net sponsors

October 21st, 2003, 04:12 AM	#20
Jeff Donald Warden Join Date: Mar 2002 Location: Clearwater, FL Posts: 8,287	I think single chip cameras are the way of the future for professional use. It will take a couple of years, but eventually 3 chip cameras will fade away. __________________ Jeff Donald Carpe Diem Search DVinfo.net for quick answers \| Where to Buy? From the best in the business: DVinfo.net sponsors

October 21st, 2003, 11:39 AM	#21
Peter Sieben Major Player Join Date: Mar 2003 Location: The Netherlands Posts: 345	Well guys, I just ordered my 3 ccd DVX100 to get a decent camcorder for short movies, and now you all say that a new consumer 1 ccd camcorder will also do the job. ;-( Peter __________________ Peter / Orphic Film www.orphicfilm.nl and www.myspace.com/orphicfilm

October 21st, 2003, 11:58 AM	#23
Peter Sieben Major Player Join Date: Mar 2003 Location: The Netherlands Posts: 345	Hi Juan, That was already clear to me. I was just thinking of the fact that I had been thinking and counting a lot before doing such a big investment, and a couple of days later I read this forum thread. I've already rented the DVX100 and know what a beauty it is. Peter __________________ Peter / Orphic Film www.orphicfilm.nl and www.myspace.com/orphicfilm

October 21st, 2003, 12:06 PM	#24
Stephen van Vuuren Space Hipster Join Date: Oct 2001 Location: Greensboro, NC Posts: 1,508	Peter: 1 chip, 2 chip, 3 chip... it's a red herring. The big news about the DVX100 CCD's is not that they are 3 chip, but they are progressive scan. That difference is much bigger than this chip count issue. The fact that they are 24p progressive scan capable at under $20,000 is revolutionary. __________________ stephen v2 www.insaturnsrings.com

October 20th, 2003, 07:17 PM	#16
Yik Kuen Regular Crew Join Date: Dec 2002 Location: Johor Bahru, Malaysia Posts: 135	Heard that after DV compression, the final captured images of Mega 1CCD cams are very close if not exceeding the 3CCD siblings, unless we have higher bandwidth format such as the 50mbps DVCPro

October 20th, 2003, 07:41 PM	#17
Juan P. Pertierra Major Player Join Date: Jul 2003 Posts: 479	That is completely possible. I do not know the specific math, but it is probable that with a large enough CCD, the mosaic approximation gets good enough that the human eye cannot tell the difference between the approximation and a 3CCD setup. Juan

October 20th, 2003, 10:57 PM	#19
Yik Kuen Regular Crew Join Date: Dec 2002 Location: Johor Bahru, Malaysia Posts: 135	To make things complicated, Sony's new DCR-PC300/330 has 2 MP effective pixels for video and uses RGB filter instead. Under sufficient lights, I guess this cam can really produce images that are very close to 3CCDs. So, do we still need 3CCD cams in the very near future when all these Mega cams become so popular?

October 21st, 2003, 11:46 AM	#22
Juan P. Pertierra Major Player Join Date: Jul 2003 Posts: 479	Peter, There isn't a single 1CCD camcorder out there that can put out as good images as the DVX100. Some people might argue the new JVC HD is better, but if you put HD aside, the DVX has better color, 24P and great manual control which the JVC does not have. The JVC is indeed the first of a kind, but it is extremely buggy, to the point where it is pretty much unusable proffessionally. You made a great buy...and you will know this when you get it. There's nothing quite like the DVX. :) I guess what was meant on here is that 'someday' 1CCD cams might be as good or better than 3CCD cams, but not today. Juan

October 25th, 2003, 05:12 PM	#25
Juan P. Pertierra Major Player Join Date: Jul 2003 Posts: 479	Case in point... I was playing with a Sony TRV19 (340k pixel 1 chipper) I just got for the experiment i'm doing, when I noticed something that might illustrate this whole effective pixel issue and one of the reasons why single chippers usually go towards larger pixel counts... I had the camera on, aimed at the TV. Well, if you look at the red power LED on your TV through the camera, in a way such that the LED occupies approximately one pixel of the CCD(such that it looks very small), and then sloooowly move the camera up and down, or left and right, you will see the 'red' LED change colors. The reason for this is that since the LED occupies approximately one pixel on the CCD(and the adjacent pixels are dark since it was a black TV), the ONLY data about the color of the LED is whichever color mosaic filter is in that particular CCD element. Thus, if you move the camera slowly up and down or left and right you should see a color sequence for the LED which corresponds to the mosaic configuration used in the camera. In my case it is Mg, Cy, Ye and G(i think). Anyway, so this is a simple example of one of the problems of approximating colors with one chip. A simple solution, is to simply increase the effective pixel count, such that in the same situation as above, on say a TRV70, the LED would occupy at least 9 CCD elements, from which the actual LED color can be approximated regardless where on the chip it is. Of course, if you move farther away from the LED you can still get it to the point where it ocuppies one element and the problem is seen again, but the more the elements, the smaller the LED will be when the problem becomes apparent. I hope this helps.... Juan

October 26th, 2003, 02:35 AM	#26
Agus Casse Major Player Join Date: Jul 2003 Location: Guatemala Posts: 322	Nice experiment... thanks juan