sensor size at DVinfo.net

Noah Yuan-Vogel · February 20th, 2007, 04:27 PM

I was thinking about this camera and its 1/2.7" 3MP sensor and I wondered what the actual recorded image dimensions are. I mean people like to say its better than a 1/3" sensor, but really, if its 3MP, and 1080p is 2MP, its only using 2/3 of that image area to record the image (or it's scaling, neither of which is good). Also, if the still image size is 1920x1440, it must be a 4:3 sensor, which are often smaller than their 16:9 equivalents. Some 1/3" sensors in widescreen are actually just widened compared to 1/3" 4:3 sensors, giving them larger surface area. 4:3 is not a good thing when you are shooting 16:9.

Rich Dykmans · February 20th, 2007, 05:24 PM

It starts with a full raster 1920 X 1080 image and then scales it down to meet the HDV standard. This is opposed to most of the 3 chip prosumer cams (V1, Z1, A1, H1, HVX) which start with smaller chips and scale up. That's one of the reasons the HV10 shows so much detail.

Noah Yuan-Vogel · February 21st, 2007, 12:02 AM

?
what im saying is im not sure its reasonable to say the hv20 has a larger chip. I'm not sure what you mean by starting with a smaller chip and scaling up, but I know they use similar size chips with larger pixels, which often means better latitude and a cleaner signal. also, i would imagine any kind of scaling is not a good thing, even if its from a higher res. I mean if they used fewer pixels (1440x1080 since thats the output res anyway), this might actually end up having ok low light performance and less noise. if it was a real 16:9 1/3" sensor with 1440x1080 anamorphic pixels, it could have pixels almost twice as large with that much more light gathering ability, right?

Noah Yuan-Vogel · February 21st, 2007, 12:07 AM

this is, of course, theoretical, and if it really does create a nice image with 24p and the lens isnt too poor, i would seriously consider building a nice housing for this camera to make it look nice. with a nice 35mm adapter and a hdd capture device, it might be a seriously nice camera system for the price.

Mathieu Kassovitz · February 21st, 2007, 05:48 AM

Je suis d'accord/Agreed. A true guerrilla killer.

Rich Dykmans · February 21st, 2007, 07:26 AM

I'm not sure exactly what you're not sure of?

The HV20 doesn't have a larger sensor then the HV10, it's exactly the same. This is the first sensor however (in the prosumer cost range) I believe that samples at true full HD resolution and that resolution is obvious to me when I compare video from my HV10 to my HVX200. More pixels are always better even though that's just one issue when comparing features these days.

The additional dynamic range of the larger photosites of the HVX200's 960X720 CCD's is pretty obvious to me as well (and the lens will always be an issue as one goes down the food chain.)

Thomas Smet · February 21st, 2007, 11:10 AM

over sampling always looks better. That is why film of photos transfered to video always look better. Of course there are other reasons as well but the fact that the source starts with more information is one of the key reasons.

Yes HDV only records 1440x1080 anamorphic pixels but in nature we do not view things in a anamorphic scale. turning 1920 into 1440 which gets turned back into 1920 when displayed will give a more natural look then 1440 displayed as 1920. It is very subtle but using a 1920x1080 chip is not a bad thing excpet for maybe low light. With that said however I am glad this camera is the way it is.

Instead of looking at the numbers why don't you take a look at some samples from the camera. The look of the imag should be all you need to judge how good it looks. To me the HV10/HV20 is the most photographic looking of any of the HDV cameras and that even includes the XDCAMHD series. It isn't perfect of course but it is a $1,000.00 camera after all. Just look at samples of what the camera can do and do not worry about the chip size.

Pieter Jongerius · February 21st, 2007, 01:11 PM

Hi All,

my 2cts..

first, I believe down-and-upscaling has about the same effect as upscaling alone, provided the step up is the same. But if I understand correctly, the upscaling in the prosumer cams mentioned is much more than the horizontal-only 1440>1920, so if that's correct the HV series have an advantage in terms of sharpness.

At the same time, the increase in chip resolution does improve color resolution because the downscaling is only a fraction of the primary color filter effect.
However, also true, with high resolutions in a given chip size, noise increases, indeed. I once read that the number of photons captured in a single CMOS cell during one exposure (you still with me? :) can be a mere couple of hundred. That's a noise floor that no-one can avoid :))

Finally the reason I think this thread was started: chip size. I believe its true that the chip size is somewhat misleading if it refers to the 4:3 diagonal size. So I did some calculations and the diagonal size of the 16:9 portion that gets used in HDV recordings is 1/2.94", or 0.340".

And utimately: "eye is king", I agree with Thomas.

Rich Dykmans · February 21st, 2007, 05:46 PM

One things for sure Canon is the king right now squeezing low noise images out of their CMOS sensors when it comes to DSLR's. One would think that technology will filter through to their CMOS video sensors as well, it's just a matter of time.

Robert Ducon · February 21st, 2007, 11:18 PM

Quote:

Originally Posted by Pieter Jongerius

Finally the reason I think this thread was started: chip size. I believe its true that the chip size is somewhat misleading if it refers to the 4:3 diagonal size. So I did some calculations and the diagonal size of the 16:9 portion that gets used in HDV recordings is 1/2.94", or 0.340".

And there I was hoping it was a 16:9 sensor. Darn. At least Canon gives us a biggie single. That new Fuji and it's 1/5th scares me. HV20 please!

Chris Hurd · February 22nd, 2007, 12:13 AM

The CMOS sensor is indeed 4:3. It has to be, in order to produce a three-megapixel still image, which utilizes all of the available real estate on the chip. Remember a still photo is 4:3 or close to it.

Video comes from a 16:9 target area matrix of pixels in the middle of the sensor. It's using the full width and almost the full height. It does not matter what shape the CMOS sensor is... whether it's 4:3, 1:1, oblong, circular, star-shaped, whatever. All that matters is the 16:9 target area matrix of pixels in the middle of the sensor, which gives you native 16:9 video.

Quote:

Originally Posted by Thomas Smet

Instead of looking at the numbers why don't you take a look at some samples from the camera.

Single best sentence in this entire thread!

The strange notion that "any kind of scaling is not a good thing," is of course entirely incorrect.

Mikko Lopponen · February 22nd, 2007, 04:59 AM

Well the HC1 has a 1920x1440 sensor and it doesn't look as good as the hv10. So maybe we should just look at the images and go from there?

Chris Hurd · February 22nd, 2007, 07:29 AM

Indeed, "how good the image looks" has a lot more to do with the camcorder's Digital Signal Processor than it does with the image sensor.

Robert Ducon · February 22nd, 2007, 09:48 PM

Chris (or anyone else),

Even with the large sensor, why is the low light of the XH-A1 so much better than the HV10? Lens size not admitting enough light? Same Digic DVII processor too. I love the XH-A1s low light.

Why is the low-light of the HV10 worse?

Noah Yuan-Vogel · February 23rd, 2007, 01:22 AM

definite possibilities:

1. lens f1.8 vs f1.6
2. pixel size depending on actual sensor dimensions, A1 could have pixels that are almost 2x wider and 1.5x taller than the hv10/20.
3. light loss difference between bayer filter vs. prism
4. who knows, poor IR filtration, poor quality cmos sensor, different processing, how much does divic dv ii handle? maybe the adc in the cmos is cheap.

any other ideas?

February 20th, 2007, 04:27 PM	#1
Noah Yuan-Vogel Major Player Join Date: Dec 2004 Location: New York City Posts: 613	sensor size I was thinking about this camera and its 1/2.7" 3MP sensor and I wondered what the actual recorded image dimensions are. I mean people like to say its better than a 1/3" sensor, but really, if its 3MP, and 1080p is 2MP, its only using 2/3 of that image area to record the image (or it's scaling, neither of which is good). Also, if the still image size is 1920x1440, it must be a 4:3 sensor, which are often smaller than their 16:9 equivalents. Some 1/3" sensors in widescreen are actually just widened compared to 1/3" 4:3 sensors, giving them larger surface area. 4:3 is not a good thing when you are shooting 16:9.

February 22nd, 2007, 07:29 AM	#13
Chris Hurd Obstreperous Rex Join Date: Jan 2001 Location: San Marcos, TX Posts: 27,368 Images: 513	Indeed, "how good the image looks" has a lot more to do with the camcorder's Digital Signal Processor than it does with the image sensor. __________________ CH Search DV Info Net \| 20 years of DVi \| ...Tuesday is Soylent Green Day!

February 22nd, 2007, 09:48 PM	#14
Robert Ducon Major Player Join Date: Nov 2006 Location: Victoria BC Posts: 400	Chris (or anyone else), Even with the large sensor, why is the low light of the XH-A1 so much better than the HV10? Lens size not admitting enough light? Same Digic DVII processor too. I love the XH-A1s low light. Why is the low-light of the HV10 worse? __________________ Mac + Canon HV20

February 20th, 2007, 05:24 PM	#2
Rich Dykmans Regular Crew Join Date: Dec 2006 Location: Discovery Bay, CA Posts: 138	It starts with a full raster 1920 X 1080 image and then scales it down to meet the HDV standard. This is opposed to most of the 3 chip prosumer cams (V1, Z1, A1, H1, HVX) which start with smaller chips and scale up. That's one of the reasons the HV10 shows so much detail.

February 21st, 2007, 12:02 AM	#3
Noah Yuan-Vogel Major Player Join Date: Dec 2004 Location: New York City Posts: 613	? what im saying is im not sure its reasonable to say the hv20 has a larger chip. I'm not sure what you mean by starting with a smaller chip and scaling up, but I know they use similar size chips with larger pixels, which often means better latitude and a cleaner signal. also, i would imagine any kind of scaling is not a good thing, even if its from a higher res. I mean if they used fewer pixels (1440x1080 since thats the output res anyway), this might actually end up having ok low light performance and less noise. if it was a real 16:9 1/3" sensor with 1440x1080 anamorphic pixels, it could have pixels almost twice as large with that much more light gathering ability, right?

February 21st, 2007, 12:07 AM	#4
Noah Yuan-Vogel Major Player Join Date: Dec 2004 Location: New York City Posts: 613	this is, of course, theoretical, and if it really does create a nice image with 24p and the lens isnt too poor, i would seriously consider building a nice housing for this camera to make it look nice. with a nice 35mm adapter and a hdd capture device, it might be a seriously nice camera system for the price.

February 21st, 2007, 05:48 AM	#5
Mathieu Kassovitz Major Player Join Date: Apr 2006 Location: paris Posts: 289	Je suis d'accord/Agreed. A true guerrilla killer.

February 21st, 2007, 07:26 AM	#6
Rich Dykmans Regular Crew Join Date: Dec 2006 Location: Discovery Bay, CA Posts: 138	I'm not sure exactly what you're not sure of? The HV20 doesn't have a larger sensor then the HV10, it's exactly the same. This is the first sensor however (in the prosumer cost range) I believe that samples at true full HD resolution and that resolution is obvious to me when I compare video from my HV10 to my HVX200. More pixels are always better even though that's just one issue when comparing features these days. The additional dynamic range of the larger photosites of the HVX200's 960X720 CCD's is pretty obvious to me as well (and the lens will always be an issue as one goes down the food chain.)

February 21st, 2007, 11:10 AM	#7
Thomas Smet Trustee Join Date: Mar 2004 Location: Milwaukee, WI Posts: 1,719	over sampling always looks better. That is why film of photos transfered to video always look better. Of course there are other reasons as well but the fact that the source starts with more information is one of the key reasons. Yes HDV only records 1440x1080 anamorphic pixels but in nature we do not view things in a anamorphic scale. turning 1920 into 1440 which gets turned back into 1920 when displayed will give a more natural look then 1440 displayed as 1920. It is very subtle but using a 1920x1080 chip is not a bad thing excpet for maybe low light. With that said however I am glad this camera is the way it is. Instead of looking at the numbers why don't you take a look at some samples from the camera. The look of the imag should be all you need to judge how good it looks. To me the HV10/HV20 is the most photographic looking of any of the HDV cameras and that even includes the XDCAMHD series. It isn't perfect of course but it is a $1,000.00 camera after all. Just look at samples of what the camera can do and do not worry about the chip size.

February 21st, 2007, 01:11 PM	#8
Pieter Jongerius Regular Crew Join Date: Jan 2007 Location: Netherlands Posts: 101	Hi All, my 2cts.. first, I believe down-and-upscaling has about the same effect as upscaling alone, provided the step up is the same. But if I understand correctly, the upscaling in the prosumer cams mentioned is much more than the horizontal-only 1440>1920, so if that's correct the HV series have an advantage in terms of sharpness. At the same time, the increase in chip resolution does improve color resolution because the downscaling is only a fraction of the primary color filter effect. However, also true, with high resolutions in a given chip size, noise increases, indeed. I once read that the number of photons captured in a single CMOS cell during one exposure (you still with me? :) can be a mere couple of hundred. That's a noise floor that no-one can avoid :)) Finally the reason I think this thread was started: chip size. I believe its true that the chip size is somewhat misleading if it refers to the 4:3 diagonal size. So I did some calculations and the diagonal size of the 16:9 portion that gets used in HDV recordings is 1/2.94", or 0.340". And utimately: "eye is king", I agree with Thomas.

February 21st, 2007, 05:46 PM	#9
Rich Dykmans Regular Crew Join Date: Dec 2006 Location: Discovery Bay, CA Posts: 138	One things for sure Canon is the king right now squeezing low noise images out of their CMOS sensors when it comes to DSLR's. One would think that technology will filter through to their CMOS video sensors as well, it's just a matter of time.

February 22nd, 2007, 04:59 AM	#12
Mikko Lopponen Major Player Join Date: Mar 2006 Location: Helsinki, Finland Posts: 393	Well the HC1 has a 1920x1440 sensor and it doesn't look as good as the hv10. So maybe we should just look at the images and go from there?

February 23rd, 2007, 01:22 AM	#15
Noah Yuan-Vogel Major Player Join Date: Dec 2004 Location: New York City Posts: 613	definite possibilities: 1. lens f1.8 vs f1.6 2. pixel size depending on actual sensor dimensions, A1 could have pixels that are almost 2x wider and 1.5x taller than the hv10/20. 3. light loss difference between bayer filter vs. prism 4. who knows, poor IR filtration, poor quality cmos sensor, different processing, how much does divic dv ii handle? maybe the adc in the cmos is cheap. any other ideas?