Jump to content


Photo

Is Bigger Better? CMOS Pixel Size That Is...

  • Please log in to reply
11 replies to this topic

#1 mmalik

mmalik

    DSLR camera modifications

  • *****
  • Vendors
  • Posts: 5480
  • Joined: 13 Jan 2012
  • Loc: USA

Posted 04 March 2013 - 08:42 PM

Let’s leave the pixel density (megapixels) aside for a moment and look at the individual pixel sizes for Canon CMOS... sensors. Still vs. video CMOS imaging seem to have been going in somewhat different directions in the recent past. Is there a trend here or may be NOT? How do these trends compare to CCD... imaging? What’s best for astro-imaging in the long run, CCD or CMOS; video or still; small or large pixels? Is Canon on to something?


Note: um=micrometer (micron)

CMOS Still Imaging:
Master (Full Frame) [Trended MIX/UP]
1Ds Mark II: 7.2 um (2004)
1Ds Mark III: 6.4 um (2007)
1D X: 6.95 um (2012)

Master (APS-H) [Trended DOWN]
1D Mark II N: 8.2 um (2005)
1D Mark III: 7.2 (2007)
1D Mark IV: 5.7 um (2009)

Advanced (Full Fame) [Trended DOWN]
5D Mark II: 6.4 um (2008)
5D Mark III: 6.25 um (2012)

Midrange (APS-C) [Trended DOWN]
20Da: 6.4 um (2004)
60Da: 4.3 um (2012)

CMOS Video Imaging: [Trending UP]
Older CMOS Sensor for video [C300]: 6.4 um (2012)
New CMOS Sensor for video: 19 um (2013)

CCD Video Imaging:
Mallincam Xtreme Color (SONY ICX418AKL / ICX428AKL): ~9 um
Astrovid StellaCam3 Mono (SONY ICX418ALL): ~9 um


According to Canon regarding their NEW 35 mm full-frame CMOS sensor for video capture..., “The newly developed CMOS sensor features pixels measuring 19 microns square in size, which is more than 7.5-times the surface area of the pixels on the CMOS sensor incorporated in Canon's top-of-the-line EOS-1D X and other digital SLR cameras.


Here is the math:
1D X: 6.95 um=6.95^2=48.30 um square
New CMOS Sensor: 19 um=19^2=361 um square

7.5 Factor:
48.30x7.5=~361 um square

Posted Image

#2 Hikari

Hikari

    Viking 1

  • *****
  • Posts: 940
  • Joined: 05 Jan 2011
  • Loc: Maine, USA

Posted 04 March 2013 - 10:32 PM

The problem with a simple baseline comparison like pixel size is it assumes all pixels are created equal. They are not. What is important is the actual performance. If a smaller pixel performs better, it is better. CMOS has made great advances in controlling noise in the shadows. Also, there is some indication that the Bayer filtration is weaker giving more signal--color fidelity seems to be lower in newer sensors because of that. But that is hard to measure.

#3 piaras

piaras

    Viking 1

  • -----
  • Posts: 827
  • Joined: 26 Jan 2009
  • Loc: Niagara Region

Posted 04 March 2013 - 10:39 PM

I do not know what Canon is thinking but I would like a chip like the one in my 1Ds. Update it to the latest effiency and keep the 8.8 pixel size. Apparently the larger pixel size is better for low light, yeah we knew that already, so maybe this is an offshoot of high speed photography? It seems that the new chip would be swamped if used during daylight at regular settings.
Pierre

#4 Falcon-

Falcon-

    Aurora

  • *****
  • Posts: 4935
  • Joined: 11 Sep 2009
  • Loc: Gambier Island, BC, Canada

Posted 04 March 2013 - 10:51 PM

*nod* as Hikari said, not all pixels are created equal.

From a pure S/N ratio perspective if you have two sensors build with the *same process* and in identical cameras (same read/processing electronics) then the larger pixels will have a higher S/N.

Thankfully in the real world of DSLRs those sensors with smaller pixels are using newer technology then those with larger pixels so there is no real loss in S/N performance.

So when looking at current vs older DSLRs that answer is fairly clear - do not worry about pixel size except when considering spatial resolution.


With this new *video* sensor though - that is interesting. Naturally the spacial resolution of that chip will be relatively low (it is at most 1920x1080 as 19u x 1920 = 36.48mm and a "full frame 35mm" sensor is 36mm wide). That is 2.07megapixels BTW. So if the trade-off of low resolution was acceptable and you took that sensor and, say, a 1Dx and put them on scopes that produce the same per-pixel spatial resolution then yes the crop of the 1Dx's center 2.07mp would be beat easily by the new sensor. (There are a lot of hypotheticals there though as we do not *know* that the video-focused sensor could be set up to do long exposures in the same way.)

(Edit: oops, put the wrong value for the "megapixel" measure of the new sensor - fixed now!)

#5 Falcon-

Falcon-

    Aurora

  • *****
  • Posts: 4935
  • Joined: 11 Sep 2009
  • Loc: Gambier Island, BC, Canada

Posted 04 March 2013 - 11:10 PM

Just because I was curios I looked at what it would take to match the spacial resolution (arc-seconds per pixel) of this chip vs a standard DSLR.

Here is what I worked out:

Stick this new sensor with it's 1920x1080 resolution and 19u pixels on a 8" SCT at f/10 (2032mm focal length) and you get 1.93 arc-seconds/pixel. Not bad, good match for seeing most nights.

To get the same spacial resolution out of a 1000D (3888 x 2592, 5.7u pixels) you would instead need to use a 609mm focal length scope.

If both scopes are f/10 and if you crop the 1000D output and ignore everything except the center 1920x1080 pixels then the new sensor would have beat the 1000D hands down. This ignores the fact that the new sensor will have a field of view only 1.015 deg wide while the 1000D has a FOV 2.089 deg wide.

While we are speculating wildly what would be REALLY curious to see is if the 610mm f/4 AT6IN and 1000D would beat the new sensor on a C8 at f/10. :thinking:

#6 mmalik

mmalik

    DSLR camera modifications

  • *****
  • Vendors
  • Posts: 5480
  • Joined: 13 Jan 2012
  • Loc: USA

Posted 04 March 2013 - 11:13 PM

Mallincam Xtreme astro video uses CCD sensor; does anyone know the specs/brand of the CCD sensor in that cam? How does it compare at least at the pixel level with Canon's new chip? Is Canon going to provide this chip to other vendors (like Kodak does) or will it be Canon video cam exclusive? Is this in anyway a SONY CCD astro video (Mallincam/StellaCam) killer? Thx

#7 mmalik

mmalik

    DSLR camera modifications

  • *****
  • Vendors
  • Posts: 5480
  • Joined: 13 Jan 2012
  • Loc: USA

Posted 04 March 2013 - 11:32 PM

Stick this new sensor with it's 1920x1080 resolution and 19u pixels on a 8" SCT at f/10 (2032mm focal length) and you get 1.93 arc-seconds/pixel.


Can you elaborate on the math Falcon; details of calc. will be appreciated. Thx

#8 Falcon-

Falcon-

    Aurora

  • *****
  • Posts: 4935
  • Joined: 11 Sep 2009
  • Loc: Gambier Island, BC, Canada

Posted 05 March 2013 - 12:47 AM

Stick this new sensor with it's 1920x1080 resolution and 19u pixels on a 8" SCT at f/10 (2032mm focal length) and you get 1.93 arc-seconds/pixel.


Can you elaborate on the math Falcon; details of calc. will be appreciated. Thx


For most existing cameras you can use let this web site here do the calculations for you. :smirk:

However if you want to do the math yourself....

First you need the arc-seconds per mm image scale of your scope. For that I went to this site here and entered in the 8" diameter and 2032mm focal length of a C8. It will calculate all SORTS of info, most of it totally unimportant for imaging. However way down at the bottom in the "Esoterica" section is the "image scale at prime focus". For a C8 the image scale at prime focus is 1.692 arc minutes per mm. That is 101.52 arc-seconds per mm.

We are told the pixels are 19u in size. That is 0.019mm per pixel. So we multiply the arc-sec/mm by the size in mm of the pixels (101.52*0.019) and we end up with 1.92888 arc-seconds per pixel

How bout the FOV? Normally I would just look up the sensor dimensions but this is not known currently. We know it is a video chip and we know it is "35mm". Video is *PROBABLY* 1080p. 1080p video is 1920x1080 pixels in size. 1920 pixels multiplied by 19u is a sensor 36480u wide (36.48mm wide). Wikipedia told me that a "normal" 35mm sensor is 36mm wide, so that seems likely to be correct. So, 36.48mm wide multiplied by the image scale (101.52 arc-sec per mm) is 3703.4496 arc-seconds wide. This is a Field of View of 1.028736 Degrees. (Do the same math for the vertical dimension and we find the total FOV is 1.028736 deg by 0.578664 deg)

#9 Falcon-

Falcon-

    Aurora

  • *****
  • Posts: 4935
  • Joined: 11 Sep 2009
  • Loc: Gambier Island, BC, Canada

Posted 05 March 2013 - 12:48 AM

Oh, and for the megapixel value that is easy. 1920 pixels wide multiplied by 1080 pixels tall = 2073600 pixels total (2.07 megapixels)

#10 LeCarl

LeCarl

    Vostok 1

  • -----
  • Posts: 183
  • Joined: 16 Mar 2011

Posted 05 March 2013 - 11:47 AM

Just because I was curios I looked at what it would take to match the spacial resolution (arc-seconds per pixel) of this chip vs a standard DSLR.

Here is what I worked out:

Stick this new sensor with it's 1920x1080 resolution and 19u pixels on a 8" SCT at f/10 (2032mm focal length) and you get 1.93 arc-seconds/pixel. Not bad, good match for seeing most nights.

To get the same spacial resolution out of a 1000D (3888 x 2592, 5.7u pixels) you would instead need to use a 609mm focal length scope.

If both scopes are f/10 and if you crop the 1000D output and ignore everything except the center 1920x1080 pixels then the new sensor would have beat the 1000D hands down. This ignores the fact that the new sensor will have a field of view only 1.015 deg wide while the 1000D has a FOV 2.089 deg wide.

While we are speculating wildly what would be REALLY curious to see is if the 610mm f/4 AT6IN and 1000D would beat the new sensor on a C8 at f/10. :thinking:


Hey Falco, exactly what I thinking (and use)
This is why I use an aps-c with an focal reducer .56x on my at6rc
I guess it's more/equaly sensitive than an 35mm on regular native F9

and less $$$

4 time luminosity vs 4 time the sensitivity ?

#11 mmalik

mmalik

    DSLR camera modifications

  • *****
  • Vendors
  • Posts: 5480
  • Joined: 13 Jan 2012
  • Loc: USA

Posted 06 March 2013 - 03:17 AM

Not sure how CMOS pixel size compares to CCD unit cell size in terms of CMOS video and CCD video respectively. I have added Mallincam CCD video (SONY sensor) specs in the first post. Thx

#12 mmalik

mmalik

    DSLR camera modifications

  • *****
  • Vendors
  • Posts: 5480
  • Joined: 13 Jan 2012
  • Loc: USA

Posted 06 March 2013 - 03:24 AM

Someone rightly so pointed out in this... thread, "Canon would not have made the Da models (20 Da and 60 Da) were it not for a quirk that one of their chief execs is also an amateur astrophoto hobbyist. We are lucky they went that far. But it was luck... not economics." That seems true here as well.

Nonetheless, it is a 'small' step forward from Canon to help photo astronomy, whether it was a by-product or real goal. Nikon as always remains MIA. Thx






Cloudy Nights LLC
Cloudy Nights Sponsor: Astronomics