What is the X70 sensor's horizontal resolution?

[Ricky Sharp]

For me, if money were no object, I'd pick up a Canon C100 Mark II. One of the main reasons being I actually do not want 4K and am pleased that you truly get 1920 x 1080 pixels (4 photo sites per pixel). Thus, no debayering.

The first gen of the Canon has a horizontal resolution of 1000. Though I found that on B&H's site and not in Canon's own specs. Strangely, the Mark II doesn't reveal a value. I'm thus assuming both are actually at a horizontal resolution of 1080.

Now then, I've been unable to make sense of the 1-inch sensor in the X70. At pro.sony.com, the overview says it's a 20 MP sensor. But the detailed specs reveal it as 14.2 MP effective.

Can I assume this single sensor uses a bayer layout much as what the C100 does? But 14.2 is much greater than ~8 MP of the Canon. But, it doesn't have enough photosites to do the same trick, but at 4K (well, UHD). It would need to be a 7680 x 4320 photosite sensor (4 sites per pixel).

Does anyone know any details on how many photosites are used for each pixel in an HD signal? Is there still interpolation going on?

And ultimately, what is the horizontal resolution of this sensor in HD mode?

Thanks!

[Tom Grushka]

Given (assumption): All 14.2M "effective pixels" are read out and used to create the HD image (if this isn't the case we don't have enough data for an exact calculation, but it should be close anyway);

Given (assumption): The 14.2M effective pixel count applies when digital stabilization is OFF (there may be fewer effective pixels used when it is on due to the "borders" needed for stabilization -- someone might be able to calculate the border based on relative scale of the image with stabilization on vs. off);

Given (assumption): "Full Pixel Readout" or "Direct Pixel Readout" as Sony calls it: All 14.2M effective pixels are used without line or pixel skipping; that is, they are all interpolated and downscaled to a 16:9 HD image (this is suggested by Sony's marketing on the AX100 and CX900, which both have the same sensor; the lack of moire and aliasing in the final image; and also the lack of noise in the optical zoom range but the presence of noise in full Clear Image Zoom -- where CIZ "zooms in" by reducing the effective pixel area on the sensor and reduces the ratio of physical X70 pixels to final HD pixels).

Calculate number of horizontal and vertical lines on the sensor:
h = SQRT(14.2 x 10^6 px) * 1.3333 = ~5020 lines
v = SQRT(14.2 x 10^6 px) * 0.75 = ~2830 lines
(inexact due to only 3 significant digits)

(Check: 5020 lines * 2830 lines = ~14.2M px)

Pixel count of a "standard" HD sensor with exactly 1920x1080 px would be:
1920 lines * 1080 lines = 2,073,600 px = 2.07M px


:. The horizontal resolution of the X70 sensor is:
5024 lines / 1920 lines = 2.6 times the horizontal resolution of HD;

And, there would be nearly 14.2M px / 2.07M px = about 6.8 X70 sensor pixels used for each final HD pixel after interpolation and downscaling.

So, it is not an exact multiple of 4 ... which might be ideal due to the Bayer filter, but given how fast the hardware and software is becoming, it probably does the Bayer interpolation before the pixel interpolation, so does that really matter anymore?


Now, what about photosite size?

Given (assumption): Gapless photosites.

Given (assumption): The 1-inch sensor itself has a 16:9 aspect ratio, and the 1" diameter encompasses all 20M pixels (photosites) exactly.

Given (assumption): 1.0-type sensor dimensions:
13.2 x 8.8 mm (Sony RX100 - Compare camera sensor sizes: full frame 35mm, APS-C, 4/3, 1", 1/1.7", 1/2.5? | PhotoSeek.com | World travel photos, advice, best digital cameras reviewed)

h = SQRT(20M) * 1.3333 = 5962 photosites horizontally

13.2mm / 5962 lines = 0.00221mm --> ^2 = 4.90 x 10^–6 mm^2 per photosite
5962 lines / 13.2mm = ~452 lines per mm or 204,000 photosites per mm^2

[Tom Grushka]

... but oops I screwed up a bit on the second part because it appears that the 1" X70 sensor does not have a 16:9 ratio but rather a 3:2 ratio, so my numbers for photosite size are wrong. :(

[Paul Anderegg]

Wait....3:2? The X70 doesn't have a 3:2 mode, isn't that a DSLR still camera aspect ratio? Why don't they just glue some pixels onto the back of the battery and add those to the pixel count, just because they exist on the unit. :)

Paul

[Ricky Sharp]

Thank you, Tom!

From what I've been reading, the downscaled HD images look great. Still, I would imagine they'd look even better if the scaling involved an integral scale factor vs a fractional one.

And, yes, the sensor aspect ratio is 3:2 (13.2mm x 8.8mm from Sony's spec sheet).

16:9 would be 13.2mm x 7.425mm. That's almost 16% of the sensor going unused.

But the cam offers a 4:3 aspect optional when recording in DV. So here, I'd assume the full height of the sensor is used and that its sides aren't (~ 11.73mm x 8.8mm; about 11% going unused).

I'm not familiar with Sony's photography line, but my guess is this same sensor is used in some of their still cameras. That does make some sense as manufacturers love to use the same part if they can.

[Noa Put]

I left my calculator at home because otherwise I would have given you a more detailed breakdown of Tom's simplified calculation but seriously, what do all these numbers tell you? Will they help in any way to make a better decision when buying a videocamera?

What is it you exactly expect from a camera and for what budget, also what type of productions are you planning to use the camera on?

If ultimate resolution on a budget is that important for you then go for a 4K camera, even if you don't "need" 4K as outputformat, 4K downscaled to 1080p in post looks as detailed as it gets and you can already have that for below 1K these days.

[Ricky Sharp]

Originally Posted by Noa Put

I left my calculator at home because otherwise I would have given you a more detailed breakdown of Tom's simplified calculation but seriously, what do all these numbers tell you? Will they help in any way to make a better decision when buying a videocamera?

What is it you exactly expect from a camera and for what budget, also what type of productions are you planning to use the camera on?

If ultimate resolution on a budget is that important for you then go for a 4K camera, even if you don't "need" 4K as outputformat, 4K downscaled to 1080p in post looks as detailed as it gets and you can already have that for below 1K these days.

I don't need (or want) ultimate pixel resolution. I want the best possible HD footage in the USD 2,000 to USD 3,000 range and at a reasonable bitrate. 10-bit, 4:2:2 HD at 50 Mbps is thus really ideal.

As to my question, I just want to have a better understanding of the technical aspects. Pure numbers are definitely not solely what drives my decisions. But they are helpful.

If it were up to me, I'd require all camera manufacturers to provide the same set of numbers in their spec sheet. e.g. along with horizontal resolution, it would really be nice if all spec sheets showed bit depth and chroma subsampling.