Are we talking about the number of pixels on the sensor or the width of the sensor in pixels as used by RED and the Alexa?

That is interesting. I can see the thinking but I am going to wager it will end up being lower. A RED rig now costs about $35 grand once you put all the bits in place to make it a usable package. I can't see Sony putting out the F3 package with the recorder for any more than that. You also have products like the Cinedeck recorder (for 12ish grand) that when paired with a similar camera will yield the kind of material the Sony recorder will put out. Sony has to be mindful of this and this is with existing product.

We are in the early days of shooters and production companies finding out about this product so once the pricing hits then people start to do the comparative analysis and then other options that weren't even on the table now become viable, if Sony's pricing is off. The other big question is other unannounced product. We are still 5 months away from NAB. Announcements generally start coming out in January, then a bit more in March and then a lot of surprises in April.

I was initially talking about the total number in megapixels, but I do now note David says "3.5k" which would presumably refer to 3,500 horizontally. Sorry, misread it the first time.

Hence for 1080 video, total number of recorded pixels is 1920x1080 - or just over 2 megapixels. Because of the Bayer patterning, you need more to get anywhere near that luminance resolution, and 2x is taken as a good approximation, so 4 megapixel total is normally seen as the minimum necessary.

3.5k corresponds to a bit less than 7 megapixel, and that amount is necessary if you're hoping for true 4:4:4 from a Bayer sensor.

I've just been looking at the Alexa spec and I now see it's not 3.5k horizontally - it's actually 2880x1620 ( see Camera Specs | ARRI Digital ).

Doing the maths, that's 4,665,600, or about 4.5 megapixels. That ties in much better with other things I'd heard - that around 4 megapixels hits a sweet spot for 1080 output of resolution v sensitivity. It's not possible you heard "3.5 megapixel" instead of 3.5K, is it?

That would tie in far better with what's been heard about photosite sizes?

The Alexa is actually 3392 x 2200, but that includes the look around, the recorded area is as you say. It's called 3.5K by ARRI themselves.

Perhaps the quote I read was meant the Sony is analogous to the Alexa 3.5K, but is 2.8K, and really meant it uses the same math, but with no look around area on the sensor.

I went to a Sony event at Visual Impact in London yesterday. Got hands on with the F3.

One of the Sony reps admitted that the F3 can't produce a true 4:4:4 image due to the Bayer filtered sensor, although he said recording 4:4:4 should still yield an advantage as that is what is produced by the DSP.

In short I though the images from the F3 were great but the ergonomics shocking. Sony promised me that the NXCAM S35 camera would share the F3's sensor and have 10bit 4:2:2 output so I'm waiting to see what happens with that at NAB. In the meantime, I'm loving my PMW350. The images I got from it today are incredible. I also got to see a few other goodies like the new HDCAM-SR solid state recorder. I've put up info, pictures and video on my blog: Mike Marriage - Blog

You can't get a true 4:4:4 signal directly from any bayer pattern sensor, that includes the Alexa and RED. It's simply because there are twice as many green photosites as there are red and blue. Two green, one red and one blue per pixel.

Once it's been through the de-bayer it becomes a 4:4:4 interpretation of the original bayer pattern. How accurate it might be depends on the algorithms used.

Hopefully the next Sony rep will agree with yours (mike) on the sensor of the NX cam. :)

Not quite sure I fully agree with that, David, at least not without qualifications. But I suppose you do say "directly"....... :-)

4:4:4 implies full resolution for R,G and B, to the specs of the final output signal. Hence, if we're talking about 1080, that implies AT LEAST 2 million photosites for each of R,G, and B (or more). In the case of a Bayer, that effectively means an 8 megapixel chip (4 million green, and 2 million each red and blue) which (theoretically) should be capable of true 4:4:4 1080 after processing.

It's luminance resolution should also be inherently much better than 1080 - which will obviously be lost in giving the 1080 output. And that's the case with Red in 1080 mode.

It seems as though the F3 is about 4 megapixel (similar to the Alexa), but that shouldn't necessarily be seen as a bad thing. It will mean the photosites are larger than a similar sized 8 megapixel chip, so whilst it won't give "true" 4:4:4, it will give substantially better S/N and/or sensitivity. I suspect that for the vast majority of the time, that's the best compromise.

Mike - I also went to one of those showings, and fully agree with you about ergonomics, at least if you're talking handheld. Why didn't they at least base form factor on an EX3 rather than an EX1? It was promoted as a B camera to such as an F35, so might be expected to be used in difficult to get to and running around situations. That said, what's the alternative? The AF101 seems OK, if not up to the F3 standard in quite a few ways - but it's even WORSE for ergonomics!

ergonomics ...
 

I have the EX3 and find the ergonomics reasonable but not perfect. I wonder why the JVC ergonomics for their 700 have not been adapted for these small cameras? That seems a very good camera style for light and small cameras!

Hey if we get to vote, I'd vote EX3 any day. I LOVE the compromise between full shoulder and handheld. I'm 6'4" and find the EX3 to be nearly ideal for me.

F3 ...
 

I would add that the viewfinder on the EX3 is excellent and wonder why Sony didn't incorporate this in their design or design the camera around it. The one flaw that's apparent on the F3 is that the viewfinder on the back is probably going to be useless!

Go argue with Sony. I was just explaining what the rep meant to Mike.


It's has 8294400 recorded photosites if the Sony figures quoted are accurate. 23.6mm wide sensor, 12 micron pixels, bayer sensor. 23.6mm/12um is 1966. A few rounding errors there obviously, so each photosite is @ 6um give or take.

David, the rep I spoke to said that I could quote him that the NXCAM would definitely have the same sensor as the F3. I'm waiting to see what that is like. From the concept model it looks to be more modular and so I could possibly build it to suit my own needs. The £5K price difference would buy a Ki Pro Mini, additional VF and all the bolt on rails and supports I would need with a little left over for the "PL glass fund."

interesting look at the camera