SDI output, 4K and more...

[Alister Chapman]

The FS-700 has an 11 Million pixel sensor, more than the original RED 4K cameras and only 1 Million less than the so called 5K sensor of RED MX. I say so called because once upon a time 4K used to mean 4K resolution. Now it means 4K pixels across the sensor. 11 Million means about 4,700 pixels horizontally, so with Bayer your looking at 3.7K resolution at best.

We know it records internally using AVCHD, so internally it's 4:2:0, but the big question is what will come out of the SDI port during normal 1920x1080 operation.

Now comparing it to the C300 and F3 etc we must consider that the pixels will be a half the size of the F3 pixels and about 30% smaller than the C300's pixels. But as it's bayer the FS-700 won't have the double green sampling that the C300 uses to help reduce noise and improve sensitivity. On paper at least it would be reasonable to expect the FS-700 to be a little less sensitive than the C300/F3 and also a little noisier.

So the choice is far from clear, but I have to say I do like the idea of being able to shoot at high frame rates!

[Sareesh Sudhakaran]

From the PR:

The NEX-FS700U’s 3G HD-SDI and HDMI connectors can output full HD 50p and 60p, in addition to standard HD 60i, 24p, 25p or 30p frame rates with embedded time code and audio. With 3G HD-SDI, it can output native 23.98, 25, 29.97 progressive signals. Users can elect to output PsF over the 3G HD-SDI. Thanks to flexibility of the digital ports virtually any external recorder can be paired.
...Sony is planning a future firmware upgrade that will enable the NEX-FS700U to output 4K bitstream data over 3G HD-SDI when used with an optional Sony 4K recorder. The NEX-FS700U (body only) and NEX-FS700UK (with 18-200mm zoom lens) are planned to be available in June at a suggested list price less than $10,000.

AFAIK, a single HDSDI cannot handle more than 1080p 4:2:2 at 30fps, and a dual HDSDI/3G can at best do 2K (as Arri has done with their T-link technology). There is possibly no current standard (or method) to transmit uncompressed 4K 4:2:2 over 3GSDI, so I assume it has to be compressed at 4:2:0. HD, though, should be uncompressed at 4:2:2, otherwise 3G SDI is overkill. But this is just an assumption.

None of the current recorders can handle 4K over 3GSDI (very few handle T-link and ARRIRAW) in their present configurations, and I'm not sure why Sony would go to all the trouble to build a 4K recorder that can only handle 4:2:0 4K, when that is already recorded internally within the camera!

The camera plus recorder is already priced over the F3, and it's a great price for 4K + interchangeable lenses+uncompressed HD.

[Chris Medico]

I think everyone is not correctly considering what RAW is. RAW is RAW and it is only a single channel of luminance data. There is no colorspace to RAW. The data rate of 3G is more than capable of moving 4k RAW.

Within the packet arrangement for 3G-SDI there are 4 1080p streams. For normal video they are either encoded as RGBA or YUVA. Most people forget there is the "A" channel in there.

Now if you take a 4k image and divide it into 4 pieces you will completely and perfectly fill a 3G-SDI pipe. It won't be breaking the standard. You won't be following it either meaning (for RGB colorspace) that the R channel is not actually Red luminance info but will be "quadrant 1" luminance info. The bitdepth is 12bit. The same as what Sony has stated will be output from the FS700 will be. So with this method Sony can move 4k RAW using the 3G technology that is currently in place. It will be a bastardization but it will work.

Adding color/debayering the image will happen once it is in the computer system that is going to edit it.

[Monday Isa]

Thanks for that info Chris. I didn't know that is how it could possibly work. I wonder if the new recorder will also offer a couple of codecs to record to instead of only RAW for 4K.

[Chris Medico]

It may offer some compression but it won't likely be something you are used to seeing such as DNxHD or ProRes. Those codecs expect the data to be debayered with a color space applied. It will be very confusing for an existing codec to try and compress the native RAW stream.

If they did employ compression would expect it to be a flavor of the SR codec which does offer a compressed version of RAW (as used in the F65).

Strictly a guess on my part.

[Stuart OSteen]

@ Allister: Thanks for the good advice. Like you, I really like the C300's system of taking 2MB of two greens and 2MB each of blue and red, so that the capture doesn't require interpolation the way a Bayer sensor does. And, like you, I figure that the FS-700 will be less natively light sensitive, so that it will either be noisier at high ISO settings the C300 seems to deal with easily, or will use more aggressive, in-camera noise reduction, reducing resolution. But that's an assumption. The proof will be in the testing, as you imply. Can't wait to see some tests!

@ Sareesh: For my purposes, I'd need an external recorder whether I bought the F3 or the FS-100, so that's a wash for me. The reason I've been avoiding the F3 is because I know myself. If I bought an external recorder, I'd end up getting one that will take the 4:4:4, 10-bit S-Log, and that begins to run into real money ;-).

[Nate Weaver]

Originally Posted by Chris Medico

It may offer some compression but it won't likely be something you are used to seeing such as DNxHD or ProRes. Those codecs expect the data to be debayered with a color space applied. It will be very confusing for an existing codec to try and compress the native RAW stream.

If they did employ compression would expect it to be a flavor of the SR codec which does offer a compressed version of RAW (as used in the F65).

Strictly a guess on my part.

This all makes perfect sense, and I too agree that the recorder codec format will likely be HDCAM SR format at 220 (or a new, more compressed bitrate? 110ish?), into MXF, since Sony just put a lot of effort into getting that squared away for the F65.

BTW, F65 RAW is already supported in Davinci Resolve, and GPU accelerated. I've seen decent frame rates on my newish 17" Macbook, which gives credence to the reports of real-time debayering with F65 material on a MacPro with a proper GPU installed. More than you can say with Redcode, which has never supported real-time anything without a Redrocket card.

Additionally, I fully expect that 4K/RAW format coming out the HD-SDI spigot will be a Sony-only special sauce. They're paving new road here, only Arriflex has done something similar (Alexa T-link), and their solution is proprietary as well.

[David Heath]

Originally Posted by Alister Chapman

But as it's bayer the FS-700 won't have the double green sampling that the C300 uses to help reduce noise and improve sensitivity. On paper at least it would be reasonable to expect the FS-700 to be a little less sensitive than the C300/F3 and also a little noisier.

In 1080 mode, it's highly likely that the FS-700 and the C300 will read out in exactly the same way (2x2 double-green). In another thread I posted:

The statement I've read says: The FS700 has a newly designed 11.6M sensor, 8.3M which are used in the video mode. {EDIT - see Sony Expands NXCAM Line With New Full-HD Super Slow Motion Camcorder | Facebook }

Hmmm. 8.3M in video mode? Does that ring any bells? C300 anybody? (2x1920)x(2x1080)=8,294,400 - which seems pretty close to 8.3M for me! I'd assume direct read of 2x2 blocks like the C300.

And it has 11.6M total - hence you can expect some that don't get used in the 1080 mode to be available for feeding out for 4k.

What this means is straightforward processing for 1080 video (as the C300) or direct RAW output for 4k. (Hence little in camera processing.)

What I think interesting is that there was already speculation about the sensor in the FS100. It has been assumed for the last year that it was the same as the F3 - but if so you'd expect to see slight out of band coloured aliasing on zone plates as with the F3. And you don't - the only aliasing is monochrome, which implies a sensor with a much higher photosite count than the F3.

And now the FS700 comes along...... I would be prepared to put a pretty hefty bet that the FS700 and FS100 share the same chip, but the FS700 is engineered to make better use of it.

In that case, it's reasonable to expect that any 1080 output from the FS700 will be at least as good as from the FS100! I'd be interested to hear what Alister thinks of that conclusion? It also explains why the FS100 has such low power consumption relative to the F3. Like the C300, the direct read is less processor intensive than full de-Bayering. I had previously wondered if the FS100 was reading the F3 chip in the same way as the C300 and then upscaling, but the actual results are much better than that would predict.

Accepting that the FS100 has the same chip as the FS700 - and not the F3 - explains a lot of questions.

[Emmanuel Plakiotis]

"Now it means 4K pixels across the sensor. 11 Million means about 4,700 pixels horizontally, so with Bayer your looking at 3.7K resolution at best."

Alister my calculations are that for 16:9 ratio the wide side is 4412 pixels and not 4700, because
4412X2492=10994 total pixels and 4412/2492=1.77 (HD ratio)



"Accepting that the FS100 has the same chip as the FS700 - and not the F3 - explains a lot of questions."

Sony alway maintained that they share the same chip. On the other hand there were considerable differences between them to back up your claim.

Nevertheless I feel this is a new sensor altogether. They have leapfrogged in sensor technology this year to market a serious upgrade with last years technology.

[Sareesh Sudhakaran]

Originally Posted by Chris Medico

The data rate of 3G is more than capable of moving 4k RAW.

Within the packet arrangement for 3G-SDI there are 4 1080p streams. For normal video they are either encoded as RGBA or YUVA. Most people forget there is the "A" channel in there.

I agree with you. The confusion arises because there is a distinction between the SDI system and the SDI protocol. The SDI protocol is not specified to use RAW in any way. But the SDI system is basically a cable with two circuits on either end, and one can send a pig through it if one pleased, but it would no longer be called SDI.

Arri has used the same system to transmit ARRIRAW, but they have called it T-link. Sony might well do this and call it something new, but I doubt they will go to so much trouble over a $10K camera, especially since this will undermine the 4K F65. I can't imagine this camera outputting a 4K RAW signal, even as a firmware update, but we'll see.

[David Heath]

Originally Posted by Emmanuel Plakiotis

"Accepting that the FS100 has the same chip as the FS700 - and not the F3 - explains a lot of questions."

Sony alway maintained that they share the same chip. On the other hand there were considerable differences between them to back up your claim.

I'm trying to throw my mind back to the launch of the F3/FS100 and asking "do they have the same chip"? I seem to recall the answer wasn't direct, something like "they both have a s35 chip", but uncertainty beyond that. It seemed logical to assume that they must be the same.

It's quite plausible that many in Sony weren't in on the full story (if the assumption is true) and hence did what may be described as "lying by mistake". (Which I think is a wartime Churchill quote!?!) They assumed the F3 and FS100 were the same chip (both s35, both came out about the same time) and like all of us passed the message on in full belief it was true.

It's only when you start to examine zone plates that things don't add up. Make the assumption that the FS100/FS700 have the same higher pixel count chip than the F3 and the pieces slot into place much better.

[Jean-Philippe Archibald]

F3 and FS100 have a different DSP. That's it. Don't look for conspiration. They share the same sensor but the signal is processed differently.

[Jean-Philippe Archibald]

Originally Posted by Alister Chapman

The FS-700 has an 11 Million pixel sensor, more than the original RED 4K cameras and only 1 Million less than the so called 5K sensor of RED MX. I say so called because once upon a time 4K used to mean 4K resolution. Now it means 4K pixels across the sensor. 11 Million means about 4,700 pixels horizontally, so with Bayer your looking at 3.7K resolution at best.

Apparently, Andy shipside, from Abeline told thay the extra pixels are for a 4:3 still mode (or 3:2, I don't have the direct quote from Andy) So the sensor might not be 16:9.

[Matt Davis]

Whereas the FS100 had the same chip as the F3, sadly the FS700 doesn't have the same chip as the F65.

Can anyone else confirm that the HD-SDI output is 10 bit?

Edit: conflicting info from two trusted sources - 'quelle surprise' - Ned Soltz says his source says 8 bit... sigh, live in hope as the real info comes through.

[Chris Medico]

Originally Posted by Matt Davis

Edit: conflicting info from two trusted sources - 'quelle surprise' - Ned Soltz says his source says 8 bit... sigh, live in hope as the real info comes through.

The info from Ned agrees with the info I got directly from Sony. 8bit.