Luke Mason
-
Posts
306 -
Joined
-
Last visited
Content Type
Profiles
Forums
Articles
Posts posted by Luke Mason
-
-
36 minutes ago, jonpais said:
I continue to see this waxiness even at low ISOs in footage I’ve shot with the X-T2, just one more reason I stopped shooting Fuji a long time ago.
X-T2 does not offer a setting to turn off noise reduction in video (default at 0, greyed out), X-H1 and X-T3 allow you to turn it off completely using "-4",
-
23 hours ago, Andrew Reid said:
9ms sounds a bit optimistic and Cinema5Dracula have been wrong before on their rolling shutter tests.
Fuji themselves claim 17ms in the fastest sensor mode, 4K 60p.
The GH5S is similar, maybe a bit less around 14-15ms.
The Alexa is around 10-12ms.
I highly doubt 24p is 9ms when Fuji themselves claim 17ms for 60p!!
The 17ms quote cannot be found in any of the official Fuji press releases, DPReview mentioned "16ms" in a preview article but no source given.
Interestingly, Fuji's official X-T3 press release claims that the rolling shutter distortion "has been halved" compared to X-T2 and X-H1, which has a 21ms rolling shutter according Cinema5D lab test, X-T3's 9ms test result aligns nicely with this official statement and Cinema5D's consistent testing setup.
-
1 hour ago, jonpais said:
8 bit 4:2:2 can be better than 10 bit 4:2:0, but there are too many other variables to take into account, bit rate being just one, interframe or intraframe compression being another and so on.
The only reason is that none of the current ARM processors support HEVC Main 10 version 2 (which includes 10bit 422 up to 12bit 444)
-
23 minutes ago, amanieux said:
you don't even need a dedicated gpu, mainstream intel cpu laptop since generation 7 can decode h265 10 bit in hardware not sure if it can go up to 200mbit/s though (intel cpu generation 6 could not : https://www.cpu-monkey.com/en/compare_cpu-intel_core_i5_7200u-653-vs-intel_core_i7_6500u-588 https://www.anandtech.com/show/10959/intel-launches-7th-generation-kaby-lake-i7-7700k-i5-7600k-i3-7350k/6 )
It's limited to maximum 160Mbps
-
3 hours ago, JR Lipartito said:
Why only 4:2:0 when recording 10bit internally instead of 4:2:2? Will the blocky chroma subsampling shoot 10bit color in the foot ? I'm really interested to know which affects overall image quality more, the color depth or chroma subsampling.
Current ARM processors can only handle hardware accelerated HEVC Main 10 Version 1 up to 10bit 420, to get 10bit 422 or even higher (12bit and/or 444) you need version 2 which is not yet widely supported.
-
7 hours ago, ErikSwan said:
Do you have a profile you are using for correction in post, or are you just manually estimating the distortion parameters? I was also really struggling to edit the H.265 files in Premiere, but Resolve is much faster for me.
Do you have a source for this, or are you just guessing? The Ambarella chip in the Mavic 2 appears to be the H3 [1, 2] which should be more than capable of downsampling from the full sensor readout, seeing as the Ambarella H1 in the Phantom 4 Pro can do it, and the H3 can even capture and downsample up to 8K [3].
@Cliff Totten (who posted earlier in this thread), did some tests on the resolution of the sensor in the various modes, and it looks like Full FOV 4K does have slightly more resolution than the 2.7K mode [4], so it seems more likely to me that Full FOV 4K is using a pixel binning readout mode, rather than being upscaled from a 2.7K readout. Cliff's theory as to why the Full FOV mode is binned is due to sensor heating problems, which seems plausible to me.
Also, everyone is saying the sensor in the Mavic 2 is the IMX183, but is there actually any evidence for that?
Anyways, I think Cliff's test is a good argument for shooting in 2.7K instead of 4K if you want the full FOV. The resolution difference is barely discernible, so you might as well use that 100Mbps for better image quality/less compression rather than spending it on more pixels that aren't providing a real increase in resolution anyways. And hey, you also get 60p if you want it!
[1] https://mavicpilots.com/threads/possible-new-mavic-ii-ambarella-h-3-image-processor.45801/#post-538347
[2] https://tronicsfix.com/blogs/news/dji-mavic-2
[3] https://www.ambarella.com/products/flying-cameras/flying-cameras-products#H3
[4] https://youtu.be/EpETzfvbmD0?t=25m30s8K refers to hardware accelerated encoding, downsampling is almost always a computationally intensive CPU task.
If you look at the readout modes listed on the IMX183 page, 2.7K readout is pixel-binned, it's then upscaled to 4K in the FOV mode. I don't think it's the sensor that's prone to overheat, it's the processor since it's not capable of hardware accelerated 5K downsampling (Sony's BIONZ chip was purpose built for this).
-
6 minutes ago, HockeyFan12 said:
What model Alexa were you using that was 10-12ms? In my experience it seems like much less but I don't know the spec.
Alexa is under 5ms, otherwise the highest framerate 200fps wouldn't be possible.
-
1 minute ago, Andrew Reid said:
H.265 is hardware accelerated now unlike in the NX1 days. Resolve / FCPX will handle it WAY better than Adobe does with their dated software architecture.
What GPU do you have?
Either way, transcoding to ProRes is also hardware accelerated if you have EditReady to hand.
Ha, I've done my research, The bitrate Fuji implements at 200Mbps and 400Mbps exceeds the maximum for hardware acceleration (for now).
-
Right now the internal HEVC Main 10 Level 5.1 10bit files are extremely taxing on even the most high end hardware. Maybe Adobe CC 2019 with improved playback acceleration could make things a bit better, but for now if you really want 10bit, external recording is almost a must!
-
1 minute ago, jonpais said:
We’ve already seen that it works well; however, just because the green box is over the eye in the EVF is not always a guarantee that it nailed focus in the final clips. As far as the absence of video eye detect in Sony goes, while I hope they implement it in their new models, I’ve been shooting with and without face detect for months now and it still does a great job nonetheless.
With preproduction firmware, the performance looks on par if not better than Sony:
With Fuji's Kaizen philosophy it can only getter better from now.
-
9 minutes ago, Yurolov said:
I like how he keeps praising it and then in the end it completely fails and he just stops the video haha.
Sony can't do eye AF in video mode.
-
9 hours ago, Danyyyel said:
It might not be for the sensor, but if it make the scene less contrasty then it is technically making the sensor recording more from the shadows to the highlight. Same as using older less contrasty lens.
It does not reduce the dynamic range of the scene, it simply makes the black level "milky". You can take away light from a scene, for example using ND filter, but you cannot add light to the shadow with just a filter.
-
29 minutes ago, Snowbro said:
No exFAT? Can you format a card externally as exFAT & use it in the camera? That would suck to have 60 files
You can format card externally as exFAT and use it in camera, but the video files will still be split into 4GB chunks.
-
This tool does exactly what you want, Panasonic Vlog to LC709A:
-
-
6 hours ago, austinchimp said:
Looks great. Not sure I would risk flying over such a built up area myself.
I've been testing one at work and I find the full frame 4k mode to be quite soft compared to what I was expecting. Apart from that, the drone feels very pro and I'm very impressed with the battery life compared to the version 1.
Full FOV 4K is actually 2.7K upscaled, as the IMX183 sensor can either do full 5.5K readout or 2.7K readout, the Ambarella chip obviously cannot handle 5.5K supersampling.
See Table 3 Readout mode:
https://www.sony-semicon.co.jp/products_en/new_pro/may_2014/imx183_e.html
-
I wonder if DR400 in video mode is available on X-T3. On X-H1 it allows 12 stops DR using film simulation.
-
12 minutes ago, Andrew Reid said:
All YouTube footage has banding and compression, it's impossible to get a true picture, that determines what banding the camera's codec has or has not got.
I suggest wait until someone uploads the original files if Johnnie hasn't. Bit of an oversight if he didn't.
HDR videos on Youtube generally never show banding (10bit BT.2020)
-
3 hours ago, thebrothersthre3 said:
Not seeing why this camera would have any banding or moire issues beyond the normal 8-bit problems almost all cameras have. It only makes sense that 10 bit footage will hold up better then any 8 bit codec on the market. Plus no one has mentioned any issues with it including Cinema 5d which are usually harsh on things.
I am not a fanboy but this camera is just a big deal to me. 10 bit up to 120fps in HD and 60 in 4k, with an APSC sensor, and solid auto focus, at $1500... that's insane IMO.Another amazing thing is Cinema 5d measured it as having more then a stop more dynamic range then the GH5S or A7SII.
Cinema5D measured around 0.5 stop more than GH5S and A7S II, I think the new noise reduction algorithm and inter-frame noise reduction contributed to that.
-
5 minutes ago, Snowbro said:
Sadly, I have seen crazy banding in many xt3 videos. That killed it for me; now to see if Panasonic or Sony can release what we want...
They will all have crazy banding - after Youtube compression. These cameras are too clean at base ISO, there's no noise/grain to dither large gradients without much detail.
-
-
Bear in mind that:
1. Sony never use off-the-shelf sensors for their cameras, or in other worlds, their sensors in camera products are not for general sale.
2. This particular sensor is for industrial use
3. Sony's new APS-C flagship is not called A6700
4. You'll see a 26MP BSI sensor, sounds familiar right? ?
-
12 minutes ago, PannySVHS said:
Topic of this thread was misleading me to believe there was a rumor about a 10bit Sony. There has not been one yet.
The new Nikons on the other hand are 10bit full HDMI beasts for granted! And that with Nikon colors!
My headline for this thread: "Sony a7s3 with 8bit for sure and flip out screen, no word about 10bit"
10bit is confirmed. RX0 is Sony's first prosumer camera to output 10bit 422 via HDMI, there will be more to come.
-
Fuji X-T3 and X-T4 discussion
In: Cameras
Posted
Fuji never said 17ms, that came from nowhere, the official press release claims rolling shutter has been halved from X-T2/H1.
Also simple math, 60fps is approx 16ms per frame, so rolling shutter cannot exceed that.