androidlad

21 minutes ago, TheBoogieKnight said:

II don't think this is true. You can't get more dynamic range from a 12 bit readout than 12 bits although you can get more *perceived* dynamic range if you start off with more bits and add dither. This is what happens in the audio industry (I worked/lectured in Music Tech for many years). But it actually reduces the true dynamic range, it's just that you can hear (or see in a camera) below the noise floor due to the dithering removing  quantisation distortion.

Cameras don't just read the sensor then dump the data straight to the cards, do they?

Noise floor can be lowered by noise reduction.

3 hours ago, HockeyFan12 said:

How do you get more than 12 stops of dynamic range from a 12 bit ADC readout? Isn't this impossible by definition? Wouldn't there by a hard ceiling of 12 stops from a 12 bit ADC? See the depreview link above.

My issues are that I don't know why Cinema5D uses SNR=2 as the noise floor or why they place the first wedge of the Xyla chart below clipping. SNR=2 seems arbitrary, and their best explanation of why they choose this is it's what gets them a 14 stop measurement from the Alexa, but Arri intentionally underrates its sensors and if you talk with their reps they'll tell you the Amira, Mini, etc. are 15+ stops. And the official number from Arri has always been 14+ anyway. That's why I see the comparisons as useful, but to me the actual measurement I'd derive from the S1H's result are the 13.8 they find at SNR=1 plus a bit more from them placing the first wedge below clipping. So about 14 stops. Likewise all their numbers seem low to me.

Numbers like "usable dynamic range" and Cinema5D's decision to sort of arbitrarily decide a cut off point for noise have been discussed in the past and never that productively. Canon took issue with their results on the C300 Mk II and released their own tests that correlate more closely with 14-15 stops if read traditionally rather than by Cinema5D's arbitrary (however useful for comparison) metrics. If you're interested in useable dynamic range in the field I find Geoff Boyle's over/under tests on CML more helpful anyway as you can see an entire scene and where different channels clip and there are some cameras that clip certain colors sooner or less attractively. Of course the Alexa does best. 

If the S1H results are heavily influenced by noise reduction I don't know. I didn't know the Ursa Mini had no NR at that, that's impressive if true. I haven't had issues with noise reduction on either camera, but I have read about the S1H having ghosting issues for other people so clearly others have. I'm not really concerned with that, though, as I'm not planning to buy either camera. I'm just trying to understand how you can get 14 stops of dynamic range from a 12 bit ADC.

The ADC bit-depth numbers you see (12, 14 or even 16) are quantisation accuracy, they themselves do not act as ceilings or floors.

It's the SNR measurement threshold that sets the floor, the last few stops are always buried in noise, and noise reduction is an extremely effective way to increase SNR, especially in video mode due to the temporal nature. That's how they got more than 12 stops of DR from 12bit ADC.

Current ALEXA sensors use 14bit ADC, but because the ADCs are off-chip, they themselves introduce a bit more noise than modern on-chip ones, so the camera's 16bit ISP does a bit of noise reduction (and other ARRI magic). For this reason, 14 stops is a good benchmark, so I wouldn't look at SNR=1 results, but they publish all the numbers and I respect each individual's own interpretation.

It's a purpose-built Panasonic sensor, it can output 8K 60P.

When you don't have power and thermal and form-factor constraints, it's not hard to build something like this. But it's entirely something else when you try to miniaturise it into a consumer product, this sensor assembly alone consumes a whopping 10W power.

42 minutes ago, thebrothersthre3 said:

Yeah its more like 12 actual stops of dynamic range. Numbers just depend on your testing methods though. Its one of the reasons I got the URSA mini 4.6k. The 12.6 stops in RAW measured by Cinema 5D are actually genuine as there is no noise reduction going on in RAW. 

BRAW is not really RAW in a conventional sense, it's partially debayered and processed. And there's absolutely noise reduction going on in BRAW, however BMD seem to have turned it down on 4.6K G2 vs G1 (12.1 stops in BRAW on G2 vs 12.6 on G1)

21 minutes ago, TheBoogieKnight said:

Everything I've ever read on the S1 says the converter is 14-bit. This would make sense as it outputs 14-bit RAW images and even Sony's technical documents state 14-bit so why is it only 12 bit? Just readout speed for video so they had to do that?

 

On an unrelated note, do people generally think an S1H is still worth buying with the R5 around the corner?

In stills mode, they use 14bit ADC. In video mode it's 12bit, because to achieve video frame rate (at least 24fps), ADC bit-depth has to be dropped to increase speed.

EOS R5 just passed global Bluetooth certification.

1 minute ago, Andrew Reid said:

RAW sensor data does not have a "format" other than being RAW sensor data. No LOG, no colour profile, no white balance.

It's output as RAW data but the Ninja V I think interprets RAW as V-LOG on the screen, so that you can use a LUT.

Well there are actually RAW formats, ARRI and Canon uses 12bit logarithmic RAW encoding.

In this case, ProRes RAW is designed to be linear.

https://***URL removed***/news/1623284310/canon-releases-100-firmware-update-for-professional-stop-motion-photography

It increase USB liveview output from 960x640 to 1920x1280.

10 minutes ago, Andrew Reid said:

Let's try and get a bit deeper into it... Not saying it isn't a bug, but we could do with knowing a few things... What lens was in use? What level of IBIS was turned on? Boost mode or just sensor shift? What focal length? Was it a manual focus adapted lens and was the focal length input into the menus? Or was it a native Fuji lens? If so, did it have OIS?

All the info is here:

My god he's HOT

35 minutes ago, Andrew Reid said:

https://www.raspberrypi.org/products/raspberry-pi-high-quality-camera/

Specs here imply it does output RAW?

• Sony IMX477R stacked, back-illuminated sensor, 12.3 megapixels, 7.9 mm sensor diagonal, 1.55 μm × 1.55 μm pixel size
• Ouput: RAW12/10/8, COMP8 
• Back focus: Adjustable (12.5 mm–22.4 mm)
• Lens standards: C-mount, CS-mount (C-CS adapter included)
• IR cut filter: Integrated
• Ribbon cable length: 200 mm
• Tripod mount: 1/4”-20

Yes obviously, those RAW formats are standard in MIPI CSI-2 interface protocol.

https://www.engadget.com/raspberry-pi-12-megapixel-c-mount-camera-084145607.html

It uses IMX477, a BSI & stacked 1/2.3" 12MP sensor.

It's capable of full 12MP readout at 60fps with 10bit ADC, and 40fps with 12bit ADC.

https://www.sony-semicon.co.jp/products/common/pdf/IMX477-AACK_Flyer.pdf

1 hour ago, hoodlum said:

An interesting comment from Sony on the 108mp sensors vs the latest 12mp sensor.

https://***URL removed***/news/8550800535/sony-provides-an-in-depth-look-at-the-sony-xperia-1-ii-camera-tech

"Sony says the conventional design of the sensor offers faster read-out speeds than the pixel-binning Quad-Bayer technology deployed in most current high-end phones. The entire sensor can be read out in 10ms versus 32ms for a 12MP image from a Quad-Bayer sensor."

Again, marketing BS.

Sony uses DBI stacked sensor with DRAM, Samsung's 108MP sensor is not stacked (but Samsung does have stacked sensors), of course it's slower.

https://videoai.topazlabs.com/

NTSC 720x480 with 0.9091 PAR with the padding removed is 704x480, then converted to square pixel results in 640x480.

Either keep it 640x480, or upscale with a dedicated upscaling tool to HD.

This is the proper way to do it.

5 minutes ago, OliKMIA said:

The Mavic 2 Zoom and Pro, and Mavic Air 1 are also very noisy in the shadows. The Mavic 2 Pro is 1 inch sensor. The previous generation of drone didn't show this type of noise. Apparently it has something to do with the Processor in these drones when DJI went the cheap route and skipped the ambarella processor. Which they corrected on the Mavic Mini but I haven't tried this drone.
All this noise wasn't there on the Phantom series. The Phantom 3  also had great 4k at 60mbps whereas the Mavic 1 was a disaster in 4k. The compression artifacts were so bad on the first mavic that I sold it after one week. All that suggest an image processing issue.

https://dronedj.com/2019/11/27/ambarella-processor-dji-mavic-mini-mavic-3-pro/

 

Those drones with 1" sensor have 12bit ADC, so the noise level will be noticeably better.

Mavic 2 Pro is a special case because the 4K is either heavily subsampled from 5.5K full width, or 1:1 sampled from the centre. Both of them amplify noise level.

Just now, sanveer said:

Any particular reason it is 10-bit. I remember a lot of smaller sensors are 12-bit. 

Mobile phone CMOS sensors are generally 10bit ADC only, for power consumption and speed.

Just now, sanveer said:

Perhaps their HDR photos compensate for this? 

The multi-frame stacked photos reduce some noise.

1 hour ago, User said:

And to export what about Frame Rate and Pixel Aspect Ratio? Somehow I'd like to ditch 29.97 for 23.976.

The original files are: Image Size: 720 x 480, Frame Rate: 29.97, Pixel Aspect Ratio: 0.9091

*Edit. I just read the Toutube/ Vimeo prefers square pixels.

From an academic point of view, what you are doing is essentially "digital curation" (well, a ghetto version).

You would want to preserve all of the "significant properties" of the digital objects and only "transform" the intended property, which in this case, is the resolution.

That means square pixel, HD with original 4:3 aspect ratio (960x720 or 1440x1080, depending on how good the upscaling algorithm is). The frame rate should also remain unchanged.

On current sub 1" CMOS sensors like the IMX586 used on MAVIC Air 2, the ADC operates at only 10bit, this is the reason why even at base ISO, in both photo and video mode, anything below middle grey is covered in noise.

2 hours ago, currensheldon said:

I'm guessing Canon is more interested in selling lenses than anything else. Of course, they want to sell cameras, but it would make sense that the margins are WAY higher on lenses than on cameras. EF becoming the standard for so long definitely helped and if the RF-Mount becomes the new mirrorless mount that everyone without their own mount uses (Red, Z-Cam, Black Magic, etc), then I'm sure Canon would be very happy with that. 

So Canon gets compressed raw (though only to about 5:1) and Red gets mounts and sensor help. Wonder what else is going on. 

RED may have helped Canon develop the DIGIC X Processor, and other yet to be announced ones

4 hours ago, Kisaha said:

I noticed that the new remote seems a lot more practical than the owful - older - design. That's a good thing I guess.

This is clearly their amateur line, while we are waiting obviously a new Mavic series, and probably keep the existing ones with reduced price. Probably the Mavic 2 zoom near 999, the Mavic 2 Pro for 1.199 and the new ones at 1.249 and 1.499. Something like that.

I am wondering what's the size difference between the two..

Also consider Autel EVO II series, they are rapidly catching up and are pushing frequent firmware updates.

Another thing is that with the switch to the new DJI Fly app, there's no longer the option to adjust the sharpening (and contrast, saturation). So in video mode you're stuck with the over-sharpened edge halo.

3 hours ago, Patrick B. said:

@cardinalallen good information to consider.  I overlooked the comment about the C300 seeing more into the shadows.  So even though the SNR1 rating was lower than the C500, it may be that you can actually eke out more in post.  Interesting point about the color channels in highlights.  I’ll be interested in seeing how it handles this.

Also, I noticed the Mavic Air 2 does some kind of HDR processing where different pixels are mapped to different exposures, similar to what DGO is doing.  Are other cameras doing this with Sony sensors?  That sounds like a useful feature for high resolution sensors that I didn’t consider before.  I’ve heard talk about quad Bayer sensors where blocks of pixels can become pseudo bigger pixels and record better low light at lower resolutions like we see in new camera phones.  But I never considered that high resolution sensors could actually record two exposures in lower resolution and essentially turn what is usually a negative sensor attribute for video into a positive.

The HDR video mode on Mavic Air 2 uses the DOL-HDR feature on the newer Sony sensors, it's very similar to dual gain but the two readout are not completely simultaneous (a few miliseconds apart), and requires 2 ADC for high and low gain, which halves the max frame rate and doubles the rolling shutter.