Llaasseerr

I've actually been okay with the the approach of releasing firmware after the fact, but they need to keep at it. I'll try again with the email contact.

Hm, that's disheartening. I just bought a Powrig DC adapter for the fp that I want to use with two batteries that has an "up to" 7.4V output. I don't know if that's any better than the dummy battery you bought. Usually I would use a d-tap to dummy battery, but "reasons". I'm waiting for that to turn up, so fingers crossed it doesn't fry the camera otherwise I'll have to bin it because I imagine the repair is not cheap and it's out of warranty now.

Forgot to add that Apple also released LUTs. For a PHONE.

Just wanted to add that it's interesting that Sigma confirmed what I have suspected for a while. But it's like Arri putting the Alexa out there without having Arri Wide Gamut defined - not helpful! They can't say that this is a professional tool for dp's and colorists and be making amateurish moves like this. I particularly don't like the fact they are chastising some users saying that there is a "responsibility" when shooting raw to know what you're doing, when they are just not providing the right info. Look at what Apple has done with the iphone 15. They have knocked it out of the park. They defined an Apple Log spec, released a white paper and there is an ACES IDT already in Resolve. And internally you can shoot 4444 so it is effectively RAW anyway. If Sigma just added a few more features to the fp via firmware, this camera would be a beast.

Thanks for doing this work. It's definitely a problem with OFF mode that Sigma have not supplied some kind of input transform for their original sensor gamut. It's possible they don't know how to define the gamut, or they are just lazy. For the user, the information should be in the DNG metadata so that the image is correctly interpreted in Resolve. Another alternative might be if they provided a LUT from the native sensor space to something like Rec2020. Some people are creating their own input transforms LUTs with tools like dcamprof, but that can be a complex process. It's an interesting observation you made that using DNG Converter actually means that the gamut behaviour changes so that the other color modes start to behave like OFF. I had a look in DNG Profile Editor after running through DNG Converter, and that seems to be the case. From some of my early tests with other libraw readers besides Resolve, the DNG conversion does seem to actually reconfigure metadata to convert to other color spaces more accurately. The matrix tags are definitely being changed, confirmed by exiftool. So far, I just tested with OFF files, but the implication is that once ran through DNG Converter that it doesn't matter what the color type in-camera was. I do remember contacting Sigma through their web page wanting to give them feedback on the fp as an industry professional and I never heard back. Can you tell me how you actually got them to respond? Or if you have an email address you can pass on either here or as a DM?

I think the FX3 sensor is superior to the FX30 regardless of it being full frame vs s35. It has excellent sub-10ms rolling shutter and more dynamic range, plus an amazing high base ISO - if those things are important to you. Due to the fact the sensor is “only” 4k and is full frame, each photosite (pixel) is quite large - about the size of an Alexa - and is excellent at light gathering.

Thanks for posting this! Do you know what the difference is between the IMX410AQL in the a7III and the IMX410DQL in the FX9? I’m actually interested from the perspective of the Sigma fp which I assume uses the AQL.

OK I don't get it because that's what I was saying I would plan to do, but you said that wouldn't work. You said "There is not a single camera on earth that I would want to trust to keep accurate timecode with a timecode box on it." Maybe you meant _without_?

I know you're a Sony guy (mainly the older cams) and a sound guy, so do you think the FX3 is just never going to work well enough for timecode? I do not yet have a lot of experience with timecode on set - my realm has been mainly post. I had assumed that keeping the TC unit plugged in was the only way to guarantee a more low-end camera with true TC-in to stay in sync vs. syncing it at the start of the day and unplugging it, and then the camera drifting a few frames by EOD. Are you saying a camera needs a more expensive solution like genlock? Overall, do you think the TC solution for an FX3 or even an FX6 (which has proper BNC TC in) is a waste of time? For either cameras I would want to use a Deity or TentacleSync unit, but I would have to use the proprietary cable for the FX3. There were some people reporting issue with the Sony cable, which is why I mentioned it. But I don't know the specifics. I figured that the Deity solution with their device plus cable specifically for the FX3 would be pretty tight.

Continuing this OT on 4K mainly being a Netflix and consumer TV-driven phenomenon vs 2K for theatrical with a lot of VFX, this article gives a pretty accurate breakdown considering it's a non-industry publication: https://www.vulture.com/2022/07/4k-tv-cgi-review.html Also it reiterates the point I made that indie films are easier to finish at 4K than bug budget ones. Back on topic: still waiting for the Ninja V to get a firmware update to support 24P DCI out of the FX3 now it has the firmware update. It's unclear Atomos will support it at all, but here's hoping. It would have to be a 4.2k raw 24p, I guess with a few lines of pixels cropped off the top and bottom to conform to the DCI aspect ratio.

Definitely interesting findings. I really liked the look of this camera for about a week (lol) as a kind of lower stakes FX3, but after the latest FX3 firmware updates, I'm back in on that. I'm super interested to see how the upcoming 120p plays out, also given these observations about the thermal behaviour. I wish it wasn't 119.88, but I don't think there's true 120p available in the FX3 or FX6 either, sadly.

So this is super interesting. The imdb tech specs have been updated and the Alexa 65 is not mentioned any more. It now just mentions the FX3 and states "Negative Format Sony RAW(4.2K)", which is the raw output via the Ninja V. Obviously I don't know who is in a position to edit these tech specs, but it makes a lot more sense if they did shoot it raw. By "Sony RAW" I assume they mean ProRes Raw since there's no X-OCN on these cameras. Also if they shot true 24p, then I'm guessing they had access to some unreleased firmware from Atomos which I hope becomes available soon.

Yeah, that's exactly it, but you'll see that in the big budget shows as well! Eventually it will all be 4k across the board, but right now it's mostly only lower budget projects or the extreme high end meaning like an Imax-shot sequence in a Nolan film. If $100-150m projects are shot on an Alexa 65 and are very vfx heavy, it's still normal to see them being finished at lower than 4k despite the res of the camera original. Besides Netflix and their 4k push (and good to point out it's associated with flogging tv's), I am a bit out of the loop with what all the streamers are doing for their spec requirements, but it would make sense for any footage for the Volume being shot as hi-res as possible. Also stuff like VR, that's a good place for 8k+. Yes the 4k hype has pushed Arri forward with the Alexa 35, which is an astonishing camera. And watching a 4k remaster of an old movie scanned off the neg is amazing. Ah yeah there you go, I shouldn't have said it was a rumor. The dual gain thing I mentioned WAS a rumor though, that sadly hasn't panned out: https://www.sonyalpharumors.com/fx6-features-19ev-wdr-via-v2-x-firmware-and-wdr-license/ Was hoping to see that in the FX6 and in my fever dreams, the FX3. I did a bit of online searching, and there were a bunch of papers published my Sony engineers in Japan that were using this WDR approach, but it hasn't made it into any mainstream camera releases. I'm always interested to see dual gain approaches in other sensors like Canon and the GH6. What I noticed is Canon have focused more on cleaning up the shadows, while Arri focused on the highlights. I think this is to do with the power requirements of using a DGO approach to capture highlights since the Alexa has a much higher power draw, but also probably a culture difference where Arri's film background made them more focused on capturing filmic highlights. I can imagine that the engineers at Canon are more interested in a lower noise floor. Another issue with Canon's DGO is slower rolling shutter compared to when it's turned off, so again an area to improve on.

It's one of the reasons I like Sony's stance on the FX3/FX6/a7sIII etc sensor being "only" 4k. It has about the same pixel pitch as the ALEV III sensor. I hope they do another version of these cameras that unleashes the sensor's rumored dual gain ability, if it really is 8k being binned to 4k. I would really like to see cameras in the affordable price range focusing on higher dynamic range and reduced rolling shutter or having a global shutter. We still haven't quite seen that breakthrough where we are getting a true 13 stops to match the original Alexa in a camera in the sub-10k market.

Oh, absolutely. I just want people to know that super hi res footage isn't a requirement for vfx. It's definitely good pixels over more pixels (Steve Yedlin's mantra, I guess). 4k+ is definitely becoming more popular with vfx shots, but I continue to see work in the region of 2k-3.5k on the most cg-heavy projects because otherwise the work is just too expensive unless it's for an IMAX sequence that can budget for it. Netflix is an outlier here.

Right, but that's not to do with the vfx. It's to do with the dp's choice for the format and the camera. The extra resolving power gives him more dynamic range in the shadows due to less noise, for example, compared to a single ALEV III sensor. Also it's to do with the joys of shooting a digital version of a 65mm film back and using 65mm lenses. I just came off a very high profile vfx heavy film shot on the 65 and the pulls were 3.5k. So that extra res is typically not making it to the vfx vendors by the time they start. But the extra IQ is baked in on the downres. And the vendors would have reduced them down further before they start work because the delivery required for DI is below 3.5k. So they will shrink resolution wherever they can in order to get the job done on time.

I'm not sure how the whole idea of "higher resolution is good/required for vfx" started, but it's a myth. Practically everything I've worked on has been Alexa plates at 2-3.5k. Particularly for a cg-heavy film like this, it's common to do it at 2k or a bit above, just because it gets way too slow and expensive. I think ironically 4k happens more at the low budget end, and then of course there's Netflix originals (lol). It makes a lot more sense if this was primarily shot on Alexa 65, but I do think the FX3/FX6 sensor is very good. It has the same highlight range as the Venice/Venice 2, which is to say it doesn't hold a candle to an Alexa 65 but for the money it's very good. Where the Venice has it beat is that it's cleaner in the shadows, but that's not accounting for the ISO 12800 option with the FX3/6 if you don't mind half a stop less in the highlights.

That makes a lot more sense.

On the issue of timecode, it seems like people want it to stay synced if you unplug the generator. I mean it should and that would be great, but if you need to keep it connected I can live with that. Those units are small enough.

Any bets on how much of this was shot with the FX3? I also heard the FX6 was used with Atlas Orion anamorphics. Looking forward to confirming details though. Got a feeling all the latest firmware updates were requested by Greig some time ago.

Just watched this. I feel like some of his criticism is unwarranted, but he does make some real points that need to be addressed. 1. Timecode glitches. I'm kind of hoping that if you get the Deity unit and their cable (instead of the Sony cable) that it works properly. 2. DCI 4K being cropped FOV compared to UHD. To me this is a nothing burger since they are both derived from the full width of the 4.2k sensor. 3. 24P is DCI only. This is just how Sony roll. It's the same on the FX6. Technically it's correct, since UHD is 23.98. 4. Ana desqueeze. Agreed the lack of other aspect ratios is weird, but it's not a dealbreaker. 5. No open gate. It would be nice to have if data rates permitted, but open gate recording on full frame does not relate to existing 35mm film ana lenses because they are meant to record a film back that is about the height of the DCI 4k sensor area. If anything, it would be nice if there was a recording mode that cut off the sides so we could record 1.2:1. 6. No Prores RAW for new features. Right now, I'm expecting Atomos to release an update for 24p. I don't understand why he would say that the ana desqueeze would not work though because it's just a preview on the camera itself. As for DCI, raw already records at the full 4.2k width, so maybe there will be the option to record the DCI aspect at 4.2k. So besides the unforgiivable lack of shutter angle (wtf Sony) I'm going to let Sony off the hook for now as long as Deity's dedicated FX3 timecode solution works, and as long as Atomos releases a firmware update supporting 24p.

Open gate for the fp is not needed for 35mm anamorphic as is usually shot - just crop the sides. The current height of a 16:9 recording on the fp and similar affordable full frame cameras matches the height of an Alexa 35 camera recording in open gate. You do lose some resolution from cropping the sides, but it's fine and it also improves rolling shutter speed.

The same obviously applies for the FX6, a camera I quite like. Sony are just ratcheting up the pro functionality with each model. I suppose it's possible that the superior cooling of the FX3 would reduce image noise a little compared to the A7sIII. It would need a comparison between ProRes Raw recorded to a Ninja V on both cams to circumvent the in-camera noise reduction. Also you get more control over highlight latitude and to some extent, the appearance of a lifted noise floor, when shooting CineEI rather than a baked ISO image that depending on the scene brightness, might clip bright highlights. Of course, CineEI is really just an in-camera preview for shooting underexposed footage at base ISO. So you can replicate shooting with it on an a7sIII just by shooting at base ISO with an external monitor with a preview LUT that increases the exposure. I was interested in buying an FX3 when it was rumored, but I consider it only viable now with this latest firmware update since I stan true 24p. Definitely interested, but will see how NAB shakes out. I'm kind of also hoping that we finally see some completely new cameras now that supply chains and post-pandemic conditions are improving.

If you're just playing the footage back then it's hard to see the loss in chroma information in a typical compressed 4:2:2 file, except maybe on very saturated primary colours where you see the steps on edges. But if you're doing heavy post manipulation then it's much more obvious that it's lacking. If you change the colour space to YUV (YCbCr) and look at the chroma channels (Y being luminance) on something like internal XAVC 4:2:2 Sony footage, there is practically nothing there. It's so blocky and compressed and undetailed. 4:2:0 is obviously a lot worse. That's the difference. But for most people with a camera that costs a few thousand dollars, they aren't going to be bothered by that.

I have looked at 8-bit log footage recorded from a Sony a7s to an Atomos recorder in UHD 4k ProRes, managed through an ACES workflow and then output to HD and viewed on a full size cinema screen and it looked truly stellar. I was amazed at how much that camera could do. But it doesn't stop me from preferring to shoot raw where possible to get that extra chroma information, ability to do exposure changes and apply more aggressive film emulation all without breaking the image as quickly. The "thick neg" comment applies. Since mid-range cameras have more or less bypassed internal 4:4:4 log recording options, possibly due to file size, raw recording is a decent "workaround". It is after all 1/3 the size of an equivalent RGB image, with generally less aggressive compression applied that something like XAVC. Recording log 4:4:4 ProRes is still an amazing option though that is basically indistinguishable from raw, but it seems to be an endangered species.