View Full Version : I figured out the HVX200 CCD pixel count


Johann Adler
December 1st, 2005, 10:09 AM
Greetings everyone,

So, the HVX200's sensitivity has been stated to be F8 @ 2000 lux right? I know this isn't official, but that's pretty much what has been leaked out right now (supposedly from Panasonic themselves).

If this is correct, I did a little mathematical calculation and figured out the ballpark total pixel count of the HVX200.

It goes like this:

AJ-SDX900: 2/3" CCD with 520,000 pixels rated F13 @ 2000 lux --> 0.667/520000 = ~12.8x10^-7 pixel size (in inches)

AJ-HDC27H (Varicam): 2/3" CCD with 1,019,280 pixels rated F12 @ 2000 lux --> 0.667/1019280 = ~6.54x10^-7

Sony FW900 (CineAlta): 2/3" CCD with 2,200,000 pixels rated F10 @ 2000 lux --> 0.667/2200000 = ~3.03x10^-7

DVX-100: 1/3" CCD with 410,000 pixels rated F11 @ 2000 lux --> 0.333/410000 = ~8.13x10^-7

Ok, this information allows us to extrapolate a correlation between pixel size and sensitivity. So we divide the larger pixel size (higher sensitivity) by an unknown ratio (x) to equal the smaller pixel size (lower sensitivity)

I'll take two examples:

Panasonic Varicam (F12) and Sony CineAlta (F10)
6.54x10^-7 / x = 3.03x106-7
x = ~2.16

AJ-SDX900 (F13) and DVX-100 (F11)
12.8x10^-7 / x = 8.13x10^-7
x = 1.57

These two examples indicate that for every two stops decrease in pixel sensitivity, pixel size decreases by 1.57-2.16 times in size.

I compared the Varicam and CineAlta, because both are high end sensors with high sensitivity even compared to other sensors at equal pixel size. Sony sensors are known to be more sensitive than Panasonic sensors at the same pixel size. Likewise, the SDX900 and DVX100 comparison show us that within the Panasonic line, the pixel size need not be decreased as much for the equal two stop reduction in sensitivity. Taking these two examples in account, I conclude that on average, for every two stop reduction in sensitivity the pixel size decreases by half.

Applying this knowledge to the HVX200 reported sensitivity of F8 @ 2000 lux, we can estimate the total pixel count or gross resolution of the CCD.

Since the HVX200 fits the same nitch in HD as the DVX100 fits in SD in the Panasonic line (and thus the pixel sensitivity should be closest to each other) I will use the DVX100's pixel size as a starting point for calculating the HVX200's pixel size:

DVX100 pixel size = 8.13x10^-7 inch
HVX200 CCD size = 0.333 inch

0.333 / x # of pixels = (8.13x10^-7)/[(2)(1.5)]

Thus the HVX200's total CCD pixel count = 1,230,012

If we believe Panasonic's claims of a true 1080p CCD sensor (which I do), then we can simply divide this number by 1080 and get:

1139 X 1080 resolution of the sensor

So, taking into account the generality of the pixel size : pixel sensivity ratio, I conclude that the Panasonic HVX200 CCD pixel count or resolution is 1280 X 1080.

And guess what, I believe this number has been thrown around as a possibility based on "insider" reports or speculations. If this turns out to be true this camcorder can simply use horizontal pixel shift and generate a 1920 X 1080 resolution. Keep in mind that Sony's FX1/Z1 has 960 horizontal pixels; pixel shifted to generate 1440 non-square pixel resolution.

1440/960 = 1.5
1920/1280 = 1.5

Thus, it looks like pixel shift can be used to generate 1.5 times the resolution of the actual pixels. In my view, Panasonic is taking the purist route of actually pixel shifting to 1920 instead of 1440 (non-square).

In my opinion, this is a great step foreward for inexpensive HD camcorders. Throw in the fact that it has true progressive sensors and it will be great to see in action. Of course, F8 sensitivity isn't that great and noise will be an issue, but they have slightly improved that on the DVX100b and I'm sure they put in a cleaner gain in the HVX200 (just like Sony did with the FX1/Z1 to compensate for the reduced sensitivity/low light performance).

Happy Holidays to all,

Johann

Steven White
December 1st, 2005, 11:05 AM
You've presented a fine little argument here...

However:

The HVX200 doesn't need to produce a 1920x1080 image, it only ever records 1280x1080 (1440x1080 in PAL land?) to media via DVCPRO-HD. Whether Panasonic is pushing for the analog out to be true 1920x1080 4:2:2 or not remains to be seen, but in order to deliver most effectively to DVCPRO-HD, true 1280x1080 sensors seem unnecessary.

Furthermore, the release of the JVC HD100 with 3 1280x720P 1/3" CCDs has revealed there is a problem with temperature regulation on CCDs that small with this high a sampling rate. This lead to the split-screen effect on the HD100 which has received a lot of bad press.

It seems far more likely to me that the pixel count is closer to 960x1080P.

-Steve

Johann Adler
December 1st, 2005, 11:43 AM
960x1080 is definitely a strong possibility.

1139-960 = 179
1280-1139 = 141

So 1280 is statistically more likely, but not by much. Also, I used 2 as the ratio, but if I used something closer to the lower range 960 would be even more likely. Take into account that Sony's 1/3" sensor is 960x1080 and JVC's 1/3" sensor is 1280x720 and it does seem likely that 960x1080 would be the pixel count.

921,600 (HD100) versus 1,036,800 (FX1/Z1) shows that JVC's problem with split-screen effect may be due to another reason than simply too small pixels.

Steven White
December 1st, 2005, 11:50 AM
The split screen is due to the way the chip is sampled. JVC apparently had heating problems with the chips, so rather than sample the whole thing progressively at a very high rate, they sampled each half of the chip simultaneously at a somewhat lower rate. This created a matching problem with their amplifiers and noise levels which manifests as the split screen.

It's true that 960x1080 is still a higher count than 1280x720, so even that may be pushing it - though Panasonic/Matsushita/JVC has had an extra few months to work out the bugs.

-Steve

Keith Wakeham
December 1st, 2005, 11:50 AM
I'm sure you have a valid point but unfortunately I live in canada where things are in milimeters and centimeters and stuff. Time for me to start converting!

Daniel Rudd
December 2nd, 2005, 09:27 AM
Jonothan, I am completely unable to follow your formulas, but thanks for the work.

Were you unable to sleep or something?

:)

Lucia de Nieva
December 4th, 2005, 05:23 AM
Since all 2/3 inch CCD camcorders of Panasonic's DVCPROHD range have an effective pixel count of one million, which equals 1280 * 720 plus headroom for native 16:9, and thus apparently use "balanced" pixel shift for 1080p - multiply both values by 1.5 - we may securely assume that the HVX with 1/3 inch CCDs will have not more, if not even less pixels, especially with 16:9 and 720p confirmed so far. Anyway, we will be dealing with 1/4 inch CCD pixel density, and so I would be surprised if the DVX's excellent latitude and smear resistance are preserved. Let's hope the DSP helps a bit.