HVX chip size???

[Toke]

I’m not a dsp programmer, but in addition to gamma correction, they make all kind of low pass filtering, color correction, white balance, edge enhancing, skin tones, knee, toe, elbow & other body parts and lot of more.
If the end product is 8bit and changes to picture is made in a intelligent way (not too recursively),
I don’t see any reason for more than 12bit in dsp.

And for the resolution, I hope it isn’t more than 1280×720, that’s enough resolution and more would be always away from sensivity.

[Jan Crittenden livingston]

When it comes down to an image, it starts with the lens, and then the chip set. The chip set is an analog device. The signal coming off of them is an analog signal. The next thing that happens is turning that signal into a digital one, and that is doen by sampling. How often it is sampled is probably as important as the size of the chip set. From a higher sampling frequency we can expect a more accurate repesentation of the image in front of the camera. Then the sample is quantized into the DSP. The DSP is responsible for the gradations of color, tones and edges. The greater the Bit Depth the better the ability to control the picture and present it to the recording section of the camera.

The size of the pixel can actually work against the picture if it is too small, it can add noise to the picture and fight the latitude and low light performance. Too large and you don’t have resolution. I have also seen discussion on these boards that pixel shift technology is a bad thing. Well, we call it spatial offset and maybe Sony does too, but without it, all of the cameras would have stipes going down the pictures. You can prove to yourself that your camera does it by capping the camera, and boosting the gain to 18dB, no stripes, your camera is utilizing some sort of spatial offset. What it does is to cover the non-photosensitive areas of the chip set and give resolution back, by offsetting the red and the blue chip by half a pixel. Then depending on the manufacturer the green signal is either simply gated or put through a delay to wait for the the red and blue signals. It then moves onto that sampling frequency after another process that defines the shading of the camera. White shading and black shading is how the camera determines what white and black are in all parts of the screen. In the Panasonic cameras the shading is done in the analog domain but monitored in the digital domain, any fluctuation and it is corrected digitally. This way we can match cameras with identical setups.

Another piece of this puzzle is how the chips are held to the prizm block. When we patented our solution some 15 years ago, we discovered that if you attach the chips to the prizm with the same material that the prizm is made of that as the camera gets warm, the chips don’t shift at a different rate than the prizm itself. In digital cameras, it does get warm. If this stuff was easy, anybody could do it; it really seems like rocket science to me when you consider all of the pitfalls that the signal could hit and be ruined. And all of it contributes to the image quality of the camera. Pixel size is not everything, it is portion of the overall picture. High pixel density without some of these other considerations could result in signals that look inferior.

Anyhow, I hope that adds to the discussion.

Best,

Jan

Jan Crittenden Livingston
Product Manager, DVCPRO, DVCPRO50, AG-DVX100
Panasonic Broadcast & TV Systems

[Toke]

AFAIK you don’t _have_ to have pixel shift to avoid striping.
You can take a sample from center of a pixel and represent it as it would be from the whole area of pixel.
Then you of course need very precise timing in quantization to sample the centers of a pixels, not the sides.
But that’s why video cameras usually have constant sampling rate.
Am I wrong?

[Graeme Nattress]

Excellent post!

Does the camera have a physical anti-alias filter, and if so, does it go infront of the prism block, or infront of each CCD? Also, how the anti-aliassing filters “strength” work with the pixel offset??

The only mathematical paper I’ve read on anything close to pixel offset was showing how that the CCDs need to alias so that you can use the pixel offset to create an image of higher resolution that’s not aliassed, and the figure they were using was if the undersampling factor is n, then you need n^2 shifted CCDs to recreate that. In our case, we have 2 shifted CCDs, which would give a 1.414 undersampling factor. The paper was using vibration in a single CCD over time as the shift to recreate a higher resolution black and white infra-red image, so I don’t know how applicable their findings are to our 3CCD colour situation.

If anyone has any more information on this, then I’d really like to learn more.

Graeme

– http://www.nattress.com – Film Effects for FCP

[James L. arthurs]

[Graeme Nattress] “1280×720 is the current “best guess”, but as no details have been revealed, it’s just a guess. All we were told at NAB is that Panasonic are optimizing for dynamic range, resolution and low noise, and that they’d give us the best on all three (or words to that effect.) “

On my first past at the Panasonic booth last Monday, I was told that the sensor was the “same as the Varicam… 960 by 720”. I asked about pixel shift and he said, “yes, both directions”. This was around 11:00 A.M.

I’m not taking this as a given, as different folks had different answers at different times… I’m sure they just didn’t know… but his was said in the most authoriative tone… 🙂

Regards,

Jim Arthurs

[Graeme Nattress]

But the Varicam has 1280×720 CCDs, so somehow I don’t think the information you got was totally correct.

Graeme

– http://www.nattress.com – Film Effects for FCP

[James L. arthurs]

That’s why I put it in quotes… 🙂 As I said, different stories at different times of the day from different folks.

Whatever it winds up with, as long as it allows for a reasonable amount of rez for 1080, I’ll be happy.

BTW, it was nice meeting you… at the gate to go into the DVX users event on Monday at 6:00… you were at the head of the line…

Regards,

Jim Arthurs

[Graeme Nattress]

Thanks. Yes, that was a good event.

Indeed yes, as long as the end result at 1080p looks good, then everyone will be happy. As I’m quoted as saying, I want high definition, not just high resolution!

Graeme

– http://www.nattress.com – Film Effects for FCP

[Bob Cetti]

Thanks Jan for your excellent description of the workings of the CCD and DSP, etc.

Bob Cetti
Audio Video Services
cetti@aol.com