Luke Mason

It's carbon fibre reinforced polycarbonate, it's about 30% actual carbon fibre.

The sensor assembly IMX294CJK has IR cut filter built-in.

It was shot on a rainy overcast day with a variable ND filter smeared in rain drops - most of the shots were fogged up.

Another issue with IBIS is that if you want it off, the sensor is locked in place by magnetism, some camera movements could cause the sensor to slightly vibrate.

exFAT format for files larger than 4GB will be added via future firmware update. HLG is broadcast HDR, there's no requirement for bitdepth or colour subsampling, can be 420 or 422, even 8bit. Gimbal these days are quite lower than $1000. Implementation of H.265 is purely for hardware-accelerated 10bit.

The soon-to-be-announced Sony APS-C flagship offers a bit better features than X-T3 (in some way) ?

it's cropped in 50p/60p.

H.265 is chosen because the ARM processors have 10bit HEVC hardware-accelerated encoding. At high bitrate, encoding efficiency/quality between H.264 and H.265 is identical.

It's due to both the body and the lens, newer lenses with quad linear motors are better for smooth AF, for example the f/2.0 primes and f/2.8 zooms. X-H1 also added customisable AF speed setting. You could try manual focus.

Why? Canon Log is available internally as 8bit 420 in all resolutions and frame rate. Externally it's also available in 8bit 422 and 10bit 422 with the additional option of BT.2020 colour space. 10bit Canon Log option is only available externally.

They have, I worked closely with ML team for a while. For example, a firmware dump can be extracted from the camera's AF assist LED (blinking binary code to a receiver). The whole reason the job hasn't been done is because they choose not to.

EOS R: 6720/3840=1.75 5D4: 6720/4096=1.64 Just to let you know, there are many ways ML can obtain a firmware dump without any public update.

We have review units in our hands.

X-T3 has a crop.

Internal is 8bit 420 only, HDMI outputs 8/10bit 422, C-log with BT709 or BT2020.

HDMI outputs HD/FHD/4K in 8bit 422, UHD 4K only at 10bit 4:2:2, Canon Log available both internally and externally, with BT709 and BT2020 gamut options.

What's your take on this? Alexa seems to have a distinct bias towards cyan.

I used QCTools to check 10bit quantisation, I can confirm it's true 10bit colour.

A6500 uses the same sensor as A6300, it's not BSI.

4K will be a 1:1 crop of the sensor, approx. 1.6x

Please read carefully, the HDR mode is achieved with a single exposure, through interleaved readout (alternating lines between high gain/low gain). Exposure bracketing is a separate feature. Many Sony sensors have this feauture and recently introduced a second gen algorithm called Quad Bayer HDR.

As I've said earlier, the higher dynamic range quote is for HDR stills mode, the Ambarella SoCs are capable of interleaved exposure, essentially it's like Magic Lantern's "dual ISO" feature. On Mavic Air, HDR stills offer a DR of 13.2EV.

That crop mode is quoted in the IMX183 sensor spec: https://www.sony-semicon.co.jp/products_en/new_pro/may_2014/imx183_e.html 1/1.4" equivalent. H265 100Mbps is not equivalent to H264 200Mbps, the 50% bitrate reduction using H265 is only observed in very low bitrate, highly compressed siturations (eg. streaming). In higher bitrate H265 is the same as H264 in terms of encoding efficiency.

It has nothing to do with heat, most of them are capable of continuous 12bit/14bit RAW. 10bit video wouldn't be a problem. But it's the limitation of the processing architecture. Mirrorless/DSLRs have either RAW output or JPEG output, a 12/14bit and 8bit engine. It takes extra resources to implement an additional 10bit processing path. DJI uses off-the-shelf Ambarella SoCs so 10bit is readily available. Fuji XT3 is confirmed to have 10bit also, I guess it's easy for them to implement as Fuji uses ARM processors for their cameras.