EVA1: New firmware

[Erik Wittbusch]

Just FYI, here's the official press release:

http://shop.panasonic.com/about-us-...02212018-EVA1firmware2.html?tab=press&start=0

Unfortunately, there's nowhere one word how the 5,7k RAW will be recorded unless.
We'll have to wait, but nevertheless, I'm looking forward to that firmware.

I also think that the LCD cloning feature is great if you don't like the included LCD or mirror.
They should also allow this on one of the SDI outs.

 

[filmguy123]

Aha! That Newsshooter link is great.

RAW DATA OUTPUT
5.7K- SUPER-35 FRAME 1-30fps
4K- FOUR-THIRDS FRAME 1-60fps
2K- FOUR-THIRDS FRAME 1-240fps

So, it's bascically the 240fps sensor crop mode that we get right now @ 240fps for any RAW above 30fps. No benefit of oversampling on any of those modes. But, unlike the internal 240fps, we of course get the robust RAW codec which should clear up the majority of the quality issues.

When shooting off of a sensor crop mode, we lose oversampling, but we also lose pixels and hence photosites, right? Does shooting in a sensor crop mode reduce sensor sensitivity/low light performance and dynamic range?

One MAJOR problem/confusion to me is why in the world the GH5 can do *better* ALL-I internal codecs for 2K/FHD at 24p and other modes at 200mbps ALL-I and the EVA1 has it capped at 100mbps. So the $2000 GH5 has better internal 2K/FHD codecs than the $7000 EVA1? I'm all for ALL-I finally being here but am flabbergasted as to why any ALL-I codec for FHD is capped so low. 100mbps for an ALL-I codec is actually disconcertingly low.

Finally, Mitch, are there other small improvements to the EVA1 via the firmware that have not been listed on these sites? From my post months back, you had mentioned the following were being worked on and wanted to follow up to see which of these, if any, may be making it into Firmware 2.0?

Suggestions:

- LongGOP 422 150, or at least 100, should be available in 1080p HD (this is crazy).
- Base ISO 800/2500 toggle should be bindable to a user button
- User button 9 on the grip should have a "press and hold" modifier option that could change the function of the iris dial to either shutter, vfr, or ISO
- Syncho Scan shutter option for flickering lights and monitors, without having to "corrupt" my presets to add a ton, and drill way into menus to adjust
- Zebra information on LCD should show, an option for both a permanent overlay of percentage + zebra setting (1, 2) and/or on button tap as you rotate through "zebra 1 80%, zebra 2 100%, zebra off".
- HOME screen should be able to tap on Base ISO to change there

- 1920x1080 is limited to LongGOP 422 50m? This is bogus to me and borders on infurariating. It's a worse internal codec than GH5, and it's the mode where VFR modes are used and the extra bitrate would be very needed. It's a step backwards from a $2000 camera.

- No way to have both VFR and shutter controllable on camera body wheel? There is a user switch for the rolling dial, and it can be set to VFR (great) or shutter (great) but the problem is these are both highly desirable features to have quickly accessible. And I have to choose between one or the other?

- No Zebra % display on the LCD. I can set Zebra1, Zebra2. But no reminder as to what each Zebra is or which one I am on, either immediately upon switch, or permanently on LCD? Also, when I bind the zebra on/off to a user button, it does not appear to be toggling between zeb1/zeb2/off? Only zeb on/off? (I could be mistaken on the later part?).

- Unable to bind a user button to BASE ISO swap between 800/2500. Seriously? So, I can set to run on Base800 and then user the wheel to jump up and down in ISO here. And then I can pop back into the menu to change to Base 2500, and use the wheel. Or I can make the swap locked between Base 800-Base 2500 only. Fine. But what I need most is a quick user button to bind to base 800, or base 2500, and then the wheel to toggle settings in between these. But I can't set any user button to change BASE? Not only that, when I jump to the Varicam HOME screen and tap the Base 800 it takes me to another screen to cycle ISO within the same base, and I can't even change there? I have to call up the menu and drill in?

- Increments to shutter and VFR have to be manually added to presets to scroll through. I don't mind this usually, but there should be an option to "freescroll". Here is the reason. Take shutter. So, I bind my user switch to shutter, and now I can use the wheel and rotate through shutters... 144, 172, 180, 247, etc. Cool. Anything I want to add, is a laborious process in the menu to add custom values (174, 175, 176, 177, 178, 179). But, I do like that I can customize this as often you don't need all of that. However there needs to be a free flowing synchro scan option because often with flickering lights and slow motion settings, or various electronics, you need a very quick way to just cycle through to get rid of flicker, and as of now if I am on set I have to go into the menu and start fiddling and adding presets. And once I do that, now my "cleaner" settings of "144, 172, 180, 247" are ruined and I have messy menu. So I like keeping the current system (main presets with option to add) plus a free-flowing syncho scan option so if I am using the wheel on the camera body I can get any shutter setting, immediately. Same would be nice for VFR as an option, though shutter is critical, VFR is a nicety. With shutter it's very difficult to just lock a common setting if you are using various FPS settings under various lighting settings.

- Everytime I changes sensors mode and come back to S35/4k in system settings, it defaults back to 420 100. This is annoying as during quick swaps to different modes (ie HD for VFR, back to 4K) it needs to remember my last setting, not make me reset my codec everytime. Or at a minimum, default to the 422 codec because I don't know many EVA1 users who will ever prefer defaulting to 8bit 420...

 

[Paul F]

One MAJOR problem/confusion to me is why in the world the GH5 can do *better* ALL-I internal codecs for 2K/FHD at 24p and other modes at 200mbps ALL-I and the EVA1 has it capped at 100mbps. So the $2000 GH5 has better internal 2K/FHD codecs than the $7000 EVA1? I'm all for ALL-I finally being here but am flabbergasted as to why any ALL-I codec for FHD is capped so low. 100mbps for an ALL-I codec is actually disconcertingly low.

That's part of the fun of having competing divisions. No one can say what the inter-divisional rivalries/co-operation may be within the company. There's nothing worse than forced co-operation between divisions. It slows down product releases and makes for mediocre products. So if they are competing, and one is doing better than the other, it's actually a good thing.

 

[filmguy123]

So if they are competing, and one is doing better than the other, it's actually a good thing.

That's one way to look at it! In that case, COMPETE Eva1 team!

 

[Mitch Gross]

The sensor modes on the RAW output are actually different from the ones used when in video mode.

5.7K RAW 1-30fps
The full 5.7K Super-35 sensor is utilized, so this one is the same size as the Super-35 sensor modes in video.

4K 1-60fps
A center crop of 4096x2160. This is NOT the same size as the MIX 2.8K mode in video, as that frame is derived from the 5.7K Super-35 frame. However, if you are in video mode and use the Digital Zoom feature, that goes to the same 4K native image area as this frame. At 4K RAW, you will have the same resolving capabilities as other 4K cameras. I don't have the numbers handy, but this is a four-thirds sensor area size that's a little bit bigger than the sensor on the GH5.

2K RAW 1-240fps
This mode uses the same sensor area as the 4K mode. This is NOT the same size as the Mix 2.2K mode in video, as that is derived from 4.5K resolution.

How does this work? It is similar to the "Mix" part of what we do in the different video modes. Take a cluster of photo sites (top row red/green, bottom row green/blue). Group four of those clusters together (two on top, two on the bottom). Average the representative photo sites from each of the four clusters to create a single "virtual cluster" of red/green/green/blue. Thus without debarring we create a 4K to 2K "mixing" conversion without reducing the field of view and the smaller data throughput allows use to go up to 240fps. This is similar to what we do in the video modes where we take the 5.7K sensor and mix it to 2.8K, then deBayer and down convert that to 2K/HD, or we crop the sensor to 4.5K and mix it to 2.25K, then deBayer and down convert to 2K/HD. Very efficient and works surprisingly well.


Other notes:

LCD cloning to SDI has been available in the camera since Day 1. The new firmware will add the function to HDMI.

For the most part the other features for this firmware have been noted but there will be more detail upon release.

 

[Barry_Green]

To add to what Mitch is saying, and address filmguy123's question in a different way:

No, there is no oversampling and scaling being done. This is raw. The new firmware is giving you the ability to record the raw data directly off the sensor, in all its Bayer-pattern raw-ness. It's not possible to scale raw and have it still be raw, so no, the 4K raw will not have as wide a field of view as the 5.7k raw does.

The 2K raw is pretty clever in that it's still the field of view of the 4K raw, but it's still raw too, as the pixels are mixed on a color basis. It's not de-Bayered or de-mosaic'd, the output is still a raw Bayer pattern.

If you want oversampled raw, that's what the 5.7k raw is for. You can keep it at 5.7k resolution, or you can scale it down to 4k (or 2k or whatever).

 

[filmguy123]

Got it. This is helpful. Let me see if I got it straight, and ask one more question then:

||||||||||||||||||||||||||||||||||||||||||||||

VIDEO Modes

On the video 2.8K and 2.2K mix modes, 2.8K is actually derived and oversampled from 5.7k, but processed at 50% resolution to save bandwidth and allow for the higher FPS. Hence no sensor crop. 2.2k is derived from a mild sensor crop (4.5k) and thus is derived and oversampled from this 4.5k and processed at 50% resolution to a 2.2k image, in order to make headway for the higher FPS.

RAW Modes

In RAW, only up to 30fps uses the full sensor. Everything above that uses a 4K sensor crop as a starting place. This crop is actually more aggressive than either mix mode (2.8 from no crop 5.7, or 2.2 from no crop 4.5). Which means in any RAW over 30p, more pixels are being "thrown away" than in any other video mode. 4K RAW 31p-60p outputs full 4K resolution from the sensor crop; 2K RAW 61p-240p derives and processes from the 4K center crop but processes out at 2K to save bandwidth.

||||||||||||||||||||||||||||||||||||||||||||||

While on the downside RAW after 30p is utilizing less overall pixels from the sensor than any other mode (including 2.2k MIX), the benefits of a robust RAW codec have many pros over any internal codec (which, internally, tend to stretch the EVA1 image very thin).

Tracking Correctly?

****So, My One Final Question:****

When taking a center crop of any amount, in this case 4K, it is my understanding that many pixels and photosites are left unused. Does this mean that Dynamic Range and sensitivity should be expected to decrease proportionate to the amount of photosites being left "off" in this 4K 4/3 center crop mode? Perhaps about 1.7k worth of photosites? Or, do sensors not work that way - is this STRICTLY a detail/resolution loss?

 

[filmguy123]

For the most part the other features for this firmware have been noted but there will be more detail upon release.

Thanks Mitch. Do you anticipate the team being eager to address any of the following issues?

* Higher mbps bitrates for FHD/2K, in both LongGOP and ALL-I (to rival the GH5?)
* Base ISO 800/2500 toggle bindable to a user button
* Zebra Information on LCD display, (percentages, 1/2 display, etc.) - currently all guesswork :/
* UB 9 (on the grip) being given "press and hold" modifier (to allow the grip's iris dial to control either shutter, vfr, or ISO)
* Syncho Scan shutter option for flickering lights and monitors, without having to "corrupt" favorite presets to add a ton, or drill way down into the menus to adjust

The release didn't say, only that the home button had more functionality now, but I'll assume the firmware knocked off this obvious one:

* HOME screen should be able to tap on Base ISO to change there

Thank you!! I know I've bugged you a bunch on these, but they are key usability issues for myself and seemingly others based on the response, and they seem (likely deceptively ) easy to implement

 

[Barry_Green]

VIDEO Modes

On the video 2.8K and 2.2K mix modes, 2.8K is actually derived and oversampled from 5.7k, but processed at 50% resolution to save bandwidth and allow for the higher FPS.

No, not really. Sort of. The data from the pixels is used, but it is not oversampled. It's mixed. So the net effect, effectively, is that it becomes a Super35-sized 2.8K sensor with 2.8K resolution. It is not oversampled from 5.7K. Think of it as that the pixels are now 4x bigger. But because there are fewer of them to process, you can get higher frame rates from the sensor.

If it was reading the full 5.7k and processing the full 5.7k, then ... there'd be no savings and no way to thus increase the frame rate. When processing all 18 million pixels, there's only so fast you can read that data. By using the pixel mix process, it drops the # of pixels processed down to about 4.7 million pixels, a much easier to process load, and therefore the frame rates can increase.

Hence no sensor crop.

Yes, there's no sensor crop. But there is a resolution crop. Instead of being 5.7k x 3k, it becomes effectively 2.8k x 1.5k. Instead of 18 million pixels of resolution per frame, it's 4.7 million per frame. The thing to understand is: it's not scaling. It's not digitally scaling 5.7k down to 2.8k. It's sampling the pixels as combined pixels instead, so it's literally only reading off 2.8k worth of individual pixel resolution.

Now, that 2.8k is then used as oversampling to create a 2K or 1920x1080 final deliverable, so you're still getting the benefit of oversampling, but not as much as you'd get from the 5.7k sensor source.

2.2k is derived from a mild sensor crop (4.5k) and thus is derived and oversampled from this 4.5k and processed at 50% resolution to a 2.2k image, in order to make headway for the higher FPS.

Well, the same restrictions apply. Yes there's a mild sensor crop, but you're not getting oversampling within that. Instead, the pixels are 4x bigger, so you're getting literally 2.2k worth of individually-sampled pixels on each row. Because the sensor is reading less data (now it's only handling about 2.8 million pixels per frame), it can process at much faster rates.

RAW Modes
In RAW, only up to 30fps uses the full sensor. Everything above that uses a 4K sensor crop as a starting place.

This is correct.

This crop is actually more aggressive than either mix mode (2.8 from no crop 5.7, or 2.2 from no crop 4.5).

It is a very little crop when compared to the 2.2k crop.

Which means in any RAW over 30p, more pixels are being "thrown away" than in any other video mode.

Not remotely true. 4K raw at 60 fps is retaining almost 10 million pixels per frame. The 4K video modes use more, yes, but the crop & mix modes use far less pixels. 4K raw is 10 million pixels per frame; 2K or 1080p video is working from about 4.7 million pixels per frame, which is oversampled, and gets scaled down to about 2 million pixels per frame.

Keep in mind -- 4K raw at 60 fps is delivering every bit as much resolution as just about any other 4K camera on the market. If the sensor is 4k, then it's the same resolution and same pixels per frame as the EVA1's 4K, whether that's at 1fps or 60 fps. You're getting the full benefit of 4K raw, all the way to 60 fps. It's just that you get even more benefit when you keep the frame rate between 1-30 fps, because in those scenarios the EVA1 is a nearly-6K camera.

2K RAW 61p-240p derives and processes from the 4K center crop but processes out at 2K to save bandwidth.

Not sure what you mean, so I'll go back to the start. 2K raw is an actual raw Bayer pattern signal, with 2048 big pixels per line. It uses the same surface of the sensor as the 4K raw. It mixes each group of four pixels together upon scanning to create one large superpixel. The net result is a true 2K resolution, true raw signal, in true Bayer format.

****So, My One Final Question:****

When taking a center crop of any amount, in this case 4K, it is my understanding that many pixels and photosites are left unused. Does this mean that Dynamic Range and sensitivity should be expected to decrease proportionate to the amount of photosites being left "off" in this 4K 4/3 center crop mode? Perhaps about 1.7k worth of photosites? Or, do sensors not work that way - is this STRICTLY a detail/resolution loss?

I should defer to Mitch to answer that. I'll take a bit of a stab at it though.

When you mix pixels, you're gaining usable DR because you're retaining the light gathering ability of the mixed pixels, while averaging the data together which results in an overall reduction in the visibility of noise. So -- while I haven't played with it, I would expect that the 2K raw would give you a cleaner image, that would survive more gain, than the 4K raw image would. Just like scaling 4K down to 2K results in a reduction in visible noise, the pixel mix process gives a similar reduction in noise.

Regarding the 5.7K vs. 4K situation: you're going to see a small reduction in sharpness, because you're not having the benefit of oversampling. And you're going to probably notice a hair more noise in the raw image, because it's having to fill your display device starting from a smaller source. It should be akin to the difference between shooting 5K on a Red, vs shooting 4K on a Red. Red's been doing it for years and years. This is similar technology -- a slight crop in on the sensor results in fewer pixels being read, which results in faster readout rates. It won't be identical quality, but it won't be hugely different either.

You can preview it for yourself: try the Digital Zoom feature; that's pretty much what you're going to get from the 4K raw. Plug the camera straight into a monitor, and set the camera in 5.7k mode, and get an image, and then toggle the Digital Zoom on and off. Ideally you'd also be using a zoom lens so you can swap from, say, 25mm with D.Zoom on, to 35mm with D.Zoom off, to get a similar field of view. If you notice any difference in DR or noise in that experiment, then you'll have (a pretty good approximation of) your answer as to how the 5.7k raw will compare to the 4k raw.

 

[Mitch Gross]

****So, My One Final Question:****

When taking a center crop of any amount, in this case 4K, it is my understanding that many pixels and photosites are left unused. Does this mean that Dynamic Range and sensitivity should be expected to decrease proportionate to the amount of photosites being left "off" in this 4K 4/3 center crop mode? Perhaps about 1.7k worth of photosites? Or, do sensors not work that way - is this STRICTLY a detail/resolution loss?

Barry answered most of your points so I'll just take on this one.

The Dynamic Range of a sensor functions in large part down to an individual photosite. How many stops of Dynamic Range can it handle? Or more realistically, two nearby photosites - how much difference in sensitivity can one be compared to the other with neither falling into the noise floor or saturating to clip? If we are simply talking about the 5.7K RAW mode v the 4K RAW mode, meaning comparing 1:1 actual photosites, then there is no difference. Dynamic Range and sensitivity remain as constants on the sensor level.

However...

That's rarely what people actually mean when they ask the question. Because they don't think about the actual sensor and its resolution, they think about a finished output format no matter how it is derived. So in the case of 5.7K RAW scaled in post to derive a 4K video image compared to a 4K RAW turned into 4K video, the higher resolution sensor data means that the noise floor can be hidden and lowered some in the post process, effectively giving greater functional DR. The same is true on the other end, meaning that when using a 4K RAW sensor area to create a virtual 2K RAW signal via averaging ('mixing'), any errant noise is averaged out and the resulting image shows a lower noise floor. Think of it this way: one red photosite gets hit by a bit of static or gamma wave or whatever and goes black for a frame as a standard bit of "noise." At full resolution that photo site is black and it will take the borrowed info of blue and green from the sites around it to create an RGB pixel that looks a bit off. Classic noise. But in a Mixed mode, that one photo site only has 25% of the information averaged into the final value for the virtual photo site. Let's say we were talking 10-bit values and the photo sites were 512, 536, 500, and 0 (the black one). Average those and you get 387. Yes the red value takes a bit of a hit but it's sure better than a black hole of zero.

 

[filmguy123]

Blown away by the resourcefulness, knowledge, and quick yet thorough and detailed responses. Thank you! Makes very much sense!

 

[StMad]

Agreed...it's impressive and appreciated.

In my next camera I must have raw, but I also must have a quality monitor...now I'll be on the lookout for a light monitor that can record the Eva's output.

 

[filmguy123]

Ok Barry & Mitch, I've got a couple follow up questions if that is ok!

1. EVA1 4K crop RAW vs competitor native 4K sensor RAW: How much does the EVA1 rely on its oversampling to compete with s35 native 4k sensors in terms of noise and perceived DR? My understanding from a few months back was that one of the real feats of the EVA1 was ballpark equivalent noise/DR of 4K s35 competitors like the FS7, but, with a big bump in resolution to 5.7k - impressive! But it also seems that one of the keys to achieving this is in fact what was mentioned: higher resolution sensor data to compress down that noise floor in the post process via oversampling. Is it likely that the 4K sensor crop RAW modes which do not provide extra resolution for this process may have a hard time competing in terms of noise floor and perceived DR with a native 4K sensor w/large initial photosites outputting RAW?

2. 4K 60p video output to 10-bit ProRes 422 on Atomos VS 4K 60p RAW output to RAW on Atomos.... the 10-bit ProRes would still give a lot of color grading flexibility in a very robust codec, but with no sensor crop, and a full 5.7k oversampled image. When might one prefer RAW over this mode, and why? Or vice versa? To me the big benefit in 4K raw > 30p is VFR and being able to output 36/48 fps in an HQ, non 8-bit coded. But, for 50p/60p, you could get this WITH 5.7k oversampling if not using RAW. Thoughts on best practices / use cases where RAW wins VS prores, and where prores wins?

3. What method of uncompressed RAW does the EVA1 deliver in, and what are the different methods and formats on the market? What is the colorspace of RAW (4:4:4?), mbps (400+?) and the other benefits? I know I can find some of this online, and I HAVE, but if i'm TBH Barry & Mitch do a better job explaining this stuff in general than pretty much any other place I've been able to find so I'd rather hear your take on it. I actually have never shot RAW or even been interested in it prior to the EVA1, but I got very interested in it for a number of reasons on the EVA1 so a bit of a RAW primer from you two would be pretty killer.

Thank you for all the feedback and explanation! Much appreciated.

 

[Mitch Gross]

Ok Barry & Mitch, I've got a couple follow up questions if that is ok!

1. EVA1 4K crop RAW vs competitor native 4K sensor RAW: How much does the EVA1 rely on its oversampling to compete with s35 native 4k sensors in terms of noise and perceived DR? My understanding from a few months back was that one of the real feats of the EVA1 was ballpark equivalent noise/DR of 4K s35 competitors like the FS7, but, with a big bump in resolution to 5.7k - impressive! But it also seems that one of the keys to achieving this is in fact what was mentioned: higher resolution sensor data to compress down that noise floor in the post process via oversampling. Is it likely that the 4K sensor crop RAW modes which do not provide extra resolution for this process may have a hard time competing in terms of noise floor and perceived DR with a native 4K sensor w/large initial photosites outputting RAW?

2. 4K 60p video output to 10-bit ProRes 422 on Atomos VS 4K 60p RAW output to RAW on Atomos.... the 10-bit ProRes would still give a lot of color grading flexibility in a very robust codec, but with no sensor crop, and a full 5.7k oversampled image. When might one prefer RAW over this mode, and why? Or vice versa? To me the big benefit in 4K raw > 30p is VFR and being able to output 36/48 fps in an HQ, non 8-bit coded. But, for 50p/60p, you could get this WITH 5.7k oversampling if not using RAW. Thoughts on best practices / use cases where RAW wins VS prores, and where prores wins?

3. What method of uncompressed RAW does the EVA1 deliver in, and what are the different methods and formats on the market? What is the colorspace of RAW (4:4:4?), mbps (400+?) and the other benefits? I know I can find some of this online, and I HAVE, but if i'm TBH Barry & Mitch do a better job explaining this stuff in general than pretty much any other place I've been able to find so I'd rather hear your take on it. I actually have never shot RAW or even been interested in it prior to the EVA1, but I got very interested in it for a number of reasons on the EVA1 so a bit of a RAW primer from you two would be pretty killer.

Thank you for all the feedback and explanation! Much appreciated.

1. Quality of the 4K Cropped Sensor
The quality of the 4K native sensor area is still very high. It doesn't really hide that much noise and therefore expand the usable dynamic range. That was described above as more of a theoretical advantage, but in fact the difference is quite small. The performance of the sensor and camera processing is the real performance. If you want to see how well the camera does, just shoot in S-35 5.7K and select the Digital Zoom, which will then only look at the 4K native sensor area. Other than zooming in you'll see that the image quality is generally retained as far as noise and DR go. So yeah, I think it will compete quite well.

2. 4K60 video v. 4K60 RAW preferences
Honestly, this will get into a Coke v Pepsi discussion. Best is to wait until the firmware is available and test them out for yourself to determine your own opinion. Depending on how you shoot it might not make a big difference or it might make a big difference, and that difference could go either way depending on what is important to you.

3. Uncompressed RAW format
The EVA1 will output an uncompressed RAW format in Log structure. Now what will be recorded by third party partners will be something to be announced once the firmware is released. There are several possibilities out there and not all is being shared at this time.

 

[Barry_Green]

3. What method of uncompressed RAW does the EVA1 deliver in, and what are the different methods and formats on the market?

Mitch has answered that as much as Panasonic will answer at this time. There may be more details in the future.

As for what the different methods are... well, pretty much everyone is using a Bayer-pattern sensor nowadays, so the only real difference is going to be in overall resolution and in bit depth. The EVA1 has outstanding resolution (nigh unto 6K) but the bit depth hasn't been released yet AFAIK.

And then there's compression -- some output uncompressed raw, some output compressed raw. Panasonic is outputting uncompressed raw. How will that be recorded? That's up to the recorders -- I don't know if the recorders will require compressing the signal to fit on the drives. I imagine there will be uncompressed recorders (like something from Codex, I would assume) but I don't know if uncompressed 6 gigabits per second raw can be stored on an SSD reliably. I'm sure the answer's out there, I just don't know what it is.

What is the colorspace of RAW (4:4:4?)

There is no such thing. Such a concept does not exist. Raw video is raw, it is unprocessed. Color sampling (4:2:0, 4:2:2, 4:4:4 etc) is something that happens when you process raw and convert it into something else. The EVA1's raw will have the identical same color characteristics as every other camera's raw signal, since all of them are using Bayer-pattern sensors. There is no such thing as 4:4:4 raw (unless you had three sensors, either stacked or using a beamsplitting prism). All single-sensor sensors will have the exact same color resolution as all other single-sensor cameras that use the Bayer pattern.

There was at least one other option, the RGB Striped sensor, that Sony used in the F35. It could deliver a sort of 4:4:4 but it did so by limiting its overall resolution to HD-only. You can accomplish the same thing with a 4K Bayer-pattern sensor (getting a net 1080p 4:4:4 after processing).

mbps (400+?)

Well, as far as the output, it could be as much a 5,940,000 megabits per second. In uncompressed raw the bitrates are determined by how many pixels there are, multiplied by how many frames per second there are, multiplied by the bit depth (i.e., 10-bit raw, 12-bit raw, 14-bit raw, etc). So without knowing the bit depth, we can't back-figure out what the bitrate would be, so there's no way to know what the bitrate would be. Now, the one thing we do know is that it will be six gigabits per second or less, because the raw signal will be transported on the 6G SDI cable, and 6G SDI carries a max bitrate of 5.94 gigabits per second IIRC.

But that's how it's output. How will it be recorded? If it's recorded uncompressed, then yeah, you get all that. If the recorder has to compress it, then ... we don't know what the final compressed recorded bitrate will be.

and the other benefits?

Well, the benefits to raw are that you retain everything the sensor captured. But raw sensor data needs processing -- a lot of it. Video cameras generally do all that internally. If you're skipping having the camera do it for you, and taking the raw image, then you're going to have to do all that processing yourself, in post. So the best way to say it is you'll get out of raw what you put into it. If you work your footage in post thoroughly, you can extract the best from a raw image. If you need quick turnaround, it would be kind of insane to shoot raw. Raw is not inherently "better", it's just a different path to get to the final goal, which is finished displayable footage.

But there are two ways that raw could be actually "better", and that's in overall image resolution (starting with that 5.7k sensor), and potentially in bit depth. If the raw bit depth is 10 bits, then that's not an advantage, but if the raw bit depth ends up being 12 bits (or more), then there'd be more information in the source signal for you to work with.

 

[wilecoyote]

For a good explanation on whether RAW is really better than a conventional codec, and when it makes sense is use it, this is a really useful post:

https://prolost.com/blog/rawvslog

 

[Erik Wittbusch]

That article does not fit to the EVA1 potential RAW files.

As Mitch stated above, the EVA1 RAW will be outputted in a log structure to overcome the shortcomings of RAW in lower (<16bit) bitrates.
A 12bit RAW in 5,7k will then give you much more information than a 10bit compressed log format.

If you are able to actualy make use of this - nobody knows today.

Why not stop speculating and wait for the actual firmware?

 

[wilecoyote]

That article does not fit to the EVA1 potential RAW files.

As Mitch stated above, the EVA1 RAW will be outputted in a log structure to overcome the shortcomings of RAW in lower (<16bit) bitrates.
A 12bit RAW in 5,7k will then give you much more information than a 10bit compressed log format.

Yeah. I should have clarified that the article is more of a general explainer about RAW vs. non-RAW, specially about some of the myths surrounding the former. It's not specific to the EVA1 or any other camera.

And as many people pointed out in its comments, you can also have log-encoded RAW files.

 

[Mitch Gross]

Log-encoded RAW files are an important note. A mathematically Log encode of RAW data is far more efficient at capturing useful data compared to Linear math encoding. A Log 10-bit RAW holds more useful data than a Linear 12-bit RAW.

Linear RAW means that the top stop of exposure holds half the available data bits. The next stop down holds half that again, and so on. So the first two stops eat up 75% of the data, leaving just 25% for all the rest. If a modern RAW camera has at least 12 stops of Dynamic Range you can easily see how quickly even the midtones will be data-starved, and the shadows barely get any info at all. Log math is far, far more efficient.

 

[filmguy123]

Hi Mitch,

I'm confused, an email from Panasonic was sent out stating RAW OUTPUT: "2K | Four-Thirds frame | 1-120fps". I think the 240p figure has been confirmed from Cinema5D, yourself, and a YouTube video, but I just wanted to confirm this was in fact a typo in the email?