Jump to content

Luke Mason

Members
  • Posts

    306
  • Joined

  • Last visited

Posts posted by Luke Mason

  1. 31 minutes ago, Cliff Totten said:

    Damn,...are you pretty sure about this? The Ambarella H1 image processor was a massive improvement over that Ambarella A9 junk that the GoPro3 and Mavic Pro had. However, Ambarella surpassed the H1 with the H2 and H22 and now H3 which can do 4k 120fps and even 8k 30p.

    I'm hoping DJI would have chosen a much newer Ambaralla chip than the H1. I'm waiting for the first Mavic-II tear down video to get on YouTube.

    CT

    Mavic 2 Pro uses Ambarella H2 for the 10bit HEVC Main 10 Level 5 Profile.

    Sensor readout mode is mostly dependent on the sensor though.

  2. 1 hour ago, Cliff Totten said:

    I bought a Mavic 2-Zoom AND Mavic 2-Pro 30 seconds after it hit their web site..

    I also own a Phantom 4 Pro+ V2.0 today. My goal was to sell my Phantom 4 Pro+ V2.0 and keep the Mavic II Pro as my main 1 inch aircraft. (Keep the Zoom too as backup) However, this issue with the Mavic II Pro's "HQ" mode and "FOV" mode has me concerned.

    Questions:

    Is the Phantom 4 Pro V2.0 doing a full 16x9 pixel readout and scaling to 4k? (it looks pretty damn good today)

    Is the Mavic Pro-II now pixel binning or line skipping it it's "full" 16x9 readout? (NOT the same scan as the Phantom 4 Pro V2.0)

    Is the Mavic Pro-II cropping and using a 1:1 readout and calling THIS mode "HQ"?? WTF?

    IS the Phantom 4 Pro V2.0 actually using a higher quality full pixel read out than the Mavic-II Pro??

    This is one topic that really does concern me. I might not sell my P4P-2.0 if it's sensor scan is better than the Mavic-II Pro!

    DAMNIT!

    P4P uses the same downsampling algorithm (pixel-binning) as Mavic 2 Pro in full FOV mode up to 4K 30p.

    P4P 4K 48/50/60p use pixel-binning + line-skipping.

    Mavic 2 Pro 4K HQ mode uses 1:1 readout for 10bit HLG/D-Log.

  3. 30 minutes ago, Cliff Totten said:

    I ordered both Mavics (Zoom and Pro) but I can tell you one thing; 

    This DJI claim of 13 stops for the 1/2.3 sensor and 14 stops for the 1inch-type....is a GROSS marketing lie. That is the biggest whopper that DJI had ever told. LOL!!

    Not even the A7S-II, FS5, FS7 or GH5-S sensors get "14 stops". (The FS7 gets close)

    CT

    That quote is based on HDR mode for stills photo.

  4. 13 minutes ago, Cliff Totten said:

    There is a good chance this is a line skipping or pixel binning process. It's hard to say for sure though. We know that Sony has created a high tech "full pixel readout" for their fantastic scaling engines but DJI does not buy Sony SOC's. DJI buys Sony sensors and uses Ambarella image processors on Phantom and Mavics. 

    The original Mavic Pro used a dinosaur Ambarella A9 processor that has the absolute worst temporal noise reduction algorithm. It SEVERELY scrubbed the life out of everything below middle grey it's temporal analysis artifacts caused a "pulse" every 15 frames. Shadows are 90% washed, scrubbed and flattened SEVERELY! (a lot of people blamed 60mbp/s h.264 when it was really that disgusting noise reduction circuit to blame)  Total junk chip that dates back to the GoPro 3. The Phantom 4 Pro uses the Ambarella H1 and it's pretty good. The Mavic Air used the Ambarella H1 and doesn't suffer that same noise reduction Hell that the Mavic Pro does and gives you 100mbp/s too.

    Today we have the Ambarella H2, H22 and H3 chips which are far more powerful than the H1 now. It's going to be interesting to see what video processor that DJI used for these new Mavics. The Ambarella H3 can do 4k at 120fps and 8K at 30p!

    It's very possible that DJI is making these motherboard designs usable for the new Phantom 5 too. This allows them to buy H3 or H2 chips, do all the programming work together and only activate features the need for each model and leave others hidden. This saves lots of programming man hours.

    To see this HORRIBLE Ambarella A9 vs Ambarella H1 noise reduction lab analysis comparison, check out my video here. I think it will answer a lot of people's questions about DJI Mavic image quality problems. It's NOT 60mbps vs 100mbp/s like we have all though for years.

    https://youtu.be/5WDc9PKCDm0?t=1m30s

    CT

    p.s....DJI is using H.265 HEVC but they are only using "main" profile and NOT "high" profile. This means it is only using a fraction of the HEVC tool sets and calculations that are available to the full HEVC/H.265 specification. So,...the "50% gains" over H.264/AVC are reduced even further. This is a "light" or "partial" HEVC profile designed to keep calculations per second lower and easier.

     

    I believe Mavic 2 Pro still uses Ambarella H1 chip, which limits H.265 to main profile. Also the Sony sensor IMX183CQJ doesn't support full sensor readout for video.

  5. 1 hour ago, Robert Collins said:

    Are you sure 4k FOV is 'pixel binned' as opposed to 'downrezzed'? Not questioning whether you are right or wrong - I simply dont know. It doesnt make a lot of sense in video terms to produce a 1 inch 20mp sensor and then 'pixel bin' half the pixels for video. But this maybe a choice that DJI have made... (I mean now you are  effectively telling me that there is no full sensor read out for video at all - you have a choice between 'pixel binning' or 'cropping'. Not the end of the world but we are getting to the stage that we dont really have a true 1" sensor for video in the first place.)

    Pixel binning is one of the methods of "downres", for example RX100M5/6 use supersampling to downres 5.7K to 4K, it's full sensor readout which is the best quality but computationally demanding. Other downres methods include line-skipping.

    With all of the processing going on in a small drone, supersampling 5.7K doesn't seem practical due to power/thermal constraints. Ideally you'd want a 1" sensor with less pixel to strike a balance between photo and video, but the Sony 1" 20MP one is the only model DJI has access to.

  6. 3 hours ago, KnightsFan said:

    Compression also hurts digital noise. Noise from uncompressed Raw files looks great; it's only after compression that it becomes video-y.

    Only Alexa's digital noise comes the closest to film grain, the others have blotchy, chromatic unpleasant noise patterns, even uncompressed

  7. 2 hours ago, Robert Collins said:

    I saw a video where someone pointed out something interesting about one of 'ositalv''s photos. This one.

     

    Clipboard Image (26).jpg

    Look at the number of connector pins to the lens cpu - there are 12 - that is a lot. Off hand the Canon mount has 8, ef-m 9, Sony E 10 and M43 11.

    The video's theory was that the extra pins were for electronic zoom (like you can do via the app with M43.) Sounds plausible but if we look at the X7 camera and lenses, we can also see the possibility of primes with a built in ND that can be controlled by the app and a leaf shutter (to reduce rolling shutter.)

    That's debunked as fake, just a custom P4P

  8. 46 minutes ago, cantsin said:

    I was referring to vertical resolution (720p vs. 1080p) in the posting that you quoted - so actually, we're saying the same thing.

    The 13 stops specification for the Pocket 4K refers to RAW and log capture. But dynamic range specifications in f-stops are rather unscientific anyway since they leave a lot of room for interpretation. (For example: If the lowest stop in the shadows is 99% noise and 1% signal, do you still count it or not as one stop dynamic range?) 

    Nevertheless, these values tell something when they are given by the same manufacturer and differ between two camera models. Since BM specifies the Pocket 4K with one stop less DR than the old Pocket, it is safe to assume that it also will have one stop less DR in reality. [A rather logical consequence of smaller sensor pixels with lower full-well capacity.]

    Both the old pocket and pocket 4K are marketed as 13-stops of DR.

    Pocket 4K has a larger sensor with much more advanced architecture from Sony (IMX294CJK). I'd speculate that the production version could offer slightly better DR than the old pocket.

  9. 3 hours ago, Andrew Reid said:

    It was a bit odd and a shame that Nikon cancelled the DL cameras before they even had a chance. I would have bought one. A good range of lenses, better ergonomics and colour than Sony, plus better specs than the Panasonic 1" cameras.

    If they are going to dump the fast aperture zoom and make it a consumer super zoom, at least call it RX200. Besides we are on version 6! RX100 VI looks confusingly similar to RX100 IV. Roman numerals get a bit silly after 5.

    If you are interested in the RX100 6 for the very fast AF and clean 120fps + 240fps (albeit buffered) consider what more you could get for the same price. Panasonic G9 - nice 180fps with a much larger sensor, much larger EVF, better handling body and bigger battery, not to mention 4K 60fps on an interchangeable lens camera, which can take a 24-200mm equivalent as well as fast primes, so that is a better solution for sure, and much cheaper than a GH5 with nicer ergonomics, snappier AF, more responsive handling and better EVF.

    Unless you absolutely must have a pocketable 24-200mm.

    I don't know why Panasonic has not already brought out an LX200 and put it in the hands of every high profile YouTube channel on the planet. Is it begging to sell like hot cakes.

    What are they scared of?!

    We know the compact market is in an incredible era of decline but the Instagram / YouTube camera market is the answer to pumping up sales of the higher-end compacts.

    I don't even consider the consumer compact and the LX100 as the same thing... A high spec compact is really a mirrorless camera with a fixed lens. There will always be a space for a small high spec camera.

    The 1" sensor in RX100 VI performs better than G9/GH5 due to more advanced architecture - BSI and stacking (Exmor Reverse Stacked)

  10. 3 hours ago, Eric Calabros said:

    This is the result of having direct communication with your customers (Japanese should take note), yes its bulky, but in this case bulk is not necessarily a negative thing for me.

    With this developments in stabilizer industry I bet camera makers will ditch IBIS eventually. Its costly complicated mechanical thing that also exacerbates the heat issues. 

    IBIS helps with micro vibrations when using gimbal.

  11. 38 minutes ago, CyclingBen said:

    It's really impressive.  I've found Eterna works better on the X-H1 and X-T2 for indoor and lowlight shots and the Fuji WDR works better for brightly lit and outdoor scenes. 

    If only these camera's had better battery life, that's about all I can find to complain about any more, they actually make shooting fun!

    What I've found is Fuji WDR retains more highlight detail than Eterna.

  12. 1 hour ago, Django said:

    internal FLOG + Eterna LUT is a game changer. Dynamic range gets a major boost and the final look is so much more filmic then before.. it's like a whole new camera!

    HFR options is also pretty cool. overall at under a grand second hand with the grip & around half the cost of new XH1, i feel XT2 is among best value cameras around thanks to this update.

    Have you tried the Fuji WDR LUT? IMO it looks better than Eterna.

  13. Essentially, HLG gamma is just BT.709 with aggressive highlight roll-off. It's not advisable to do any intensive grading due to this highly non-linear distribution of bits.

    If you have used BT.2020 gamut (often combined with HLG gamma as a default). Just map the gamut to BT.709 with some minor adjustment. The dynamic range will remain unchanged after export. After all, HLG is designed to work on both SD TV and HDR TV.

  14. 13 hours ago, Axel said:

    I noticed that. Without being an expert, could this have to do with what in Sonys custom PP setting is influenced by "knee" (the point where the gamma curve starts to "fade out") and "slope" (the steepness of he curve from there)? Something that's baked in. As it is now in ProRes. They could address it with a new "PP". But see below.

    I've never - anywhere - seen compressed video that has so little noise even in completely underexposed parts. The NAB pre-production model that the brave thief used to record the 4 minute clip on was set to  ISO 800 ("variable framerate"?):

    File size                                : 11.2 GiB

    Duration                                 : 4 min 9 s

    Overall bit rate                         : 384 Mb/s

    com.blackmagic-design.camera.uuid        : fdee50bc-5896-477b-be85-b1a07a8382c5

    com.blackmagic-design.camera.projectFPS  : 24

    com.apple.proapps.shootingRate           : 50

    com.blackmagic-design.camera.cameraType  : Blackmagic URSA Pocket

    com.blackmagic-design.camera.shutterAngl : 190°

    com.blackmagic-design.camera.iso         : 800

    com.blackmagic-design.camera.whiteBalanc : 5450

    com.blackmagic-design.camera.whiteBalanc : -20

    com.apple.proapps.customgamma            : com.blackmagic-design.camera.filmlog

    com.blackmagic-design.camera.look.LUTNam : Blackmagic 4K Film to Video.cube

    com.blackmagic-design.camera.guides.aspe : 2.35:1

    com.blackmagic-design.camera.guides.safe : 45

    com.blackmagic-design.camera.firmware    : 5.5

    (...)

    com.blackmagic-design.camera.colorScienc : Blackmagic URSA Pocket, Color Science Gen 4

    (...)

    Format profile                           : 422 HQ

    Codec ID                                 : apch

    Duration                                 : 4 min 9 s

    Bit rate mode                            : Constant

    Bit rate                                 : 369 Mb/s

    Width                                    : 3 840 pixels

    Height                                   : 2 160 pixels

    Display aspect ratio                     : 16:9

    Frame rate mode                          : Variable

    Frame rate                               : 12.507 FPS

    Minimum frame rate                       : 3.429 FPS

    Maximum frame rate                       : 24.000 FPS

    Color space                              : YUV

    Chroma subsampling                       : 4:2:2

    Scan type                                : Progressive

    Bits/(Pixel*Frame)                       : 3.554

    Stream size                              : 10.7 GiB (96%)

    Writing library                          : bmd0

    Language                                 : English

    Encoded date                             : UTC 2018-04-12 01:43:44

    Tagged date                              : UTC 2018-04-12 01:43:44

    Color primaries                          : BT.709

    Transfer characteristics                 : BT.709

    Matrix coefficients                      : BT.709

    matrix_coefficients_Original             : BT.709

     I think I vaguely remember that the upper native ISO was 3200 (where clean shadows and some harshly blown out highlights would be expected, if the A7s' native ISO of 3200 is anything to go by). What was the lower native ISO? 800?

    Please also note that on the official Blackmagic site, they list 4k 60p as ...

    An error?

    You are in an entirely separate universe. The, er, native virtues of BM cameras have nothing to do with intelligent AF or things like that. That's why all BM cameras, no matter how cumbersome they are in terms of usage (ever lifted the classic URSA?), remain classics.

    If you are tuned into the BM user mindset, you witness parallel efforts of Son, Pan, Can with serenity. So they found another smart gimmick to free you from the burden of having  to do something manually? Damn amazing indeed!

    Based on current info, the dual ISO on P4K is native at 400 and 3200.

×
×
  • Create New...