Thread: p4k dynamic range evaluation

Page 1 of 2 12 LastLast
Results 1 to 10 of 20
  1. #1 p4k dynamic range evaluation 
    Senior Member Samuel H's Avatar
    Join Date
    May 2012
    Location
    Madrid, Spain
    Posts
    635
    edit: I was a bit worried that I might be pixel-peeping a bit too much, and wanted to see how everything would look once uploaded to youtube, so I packed a few of the tests and uploaded a short video. I'll edit the comments below to reflect what I learned with that as well.
    https://www.youtube.com/watch?v=6v_1DYwvwRM

    -------------------------

    So, I'm a DR nuthead. I believe DR is 90% of what makes nicer cameras look nicer. And I was excited to see if my new p4k would improve the DR over my old Sonys (with the RX100IV being the only one I keep anymore for video).

    So I did a quick test and... it went awful, with the RX100IV beating the p4k by about one stop. So I went ballistic and well, let's say I overdid this a little bit. I spent over a day doing not much else besides trying to measure dynamic range. Spoiler: it went well.

    My point of reference is the a6000 that I use for stills: if I can match the DR that I get on raw stills on that camera, I'll be a happy camper. Dxomark measures that at 13.1 stops, and since all my measures are always relative (camera A has X stops more than camera B), I'll post the numbers taking "a6000 has 13 stops of DR in RAW stills" as my point of reference.

    Before I start with the results: there are many ways of measuring DR. The traditional one, that I also used five years ago, just consists of counting patches on a chart or an exposure ramp: if the patch is visible, it counts, if you can't differenciate it from the background, it doesn't count. This kind of test is almost useless. What I did here was shooting several dark-ish objects (a textured shoe-box and a vintage bellows camera) and checking whether I could see the textures and details on them and whether noise was under control (i.e. if that part of the image is not so noisy that I wouldn't want to include it like that on a video). It is, therefore, a very subjective "usable DR" measure.

    Another important thing: I ran the tests on Resolve 15 for the video files, and Lightroom for the RAW stills. Software can matter a lot when shooting RAW...

    Now, the tests.

    First I ran my benchmarks witht he a6000 in RAW stills mode, and the RX100IV set at ISO 1600 with slog2 (flaat). The first one is my benchmark (what I aspire to get in video mode) and the second one is my baseline (what I currently have), which sits 2.5 stops below my target.

    Then I tested the p4k at ISO 400 RAW 4:1 and wow what a noisy mess it becomes when you push it: 3.5 stops less than my target, which means 9.5 stops of DR in the table below. I tried several things in order to improve this. First, using highlight recovery gets you around half a stop back. It kind of looks OK, as long as you don't have a side-by-side comparison with a truly unclipped shot. Let's be optimistic and keep that half-a-stop. Second, adding NR: as much as the shot would take, without ruining the textures in those dark objects I wanted to preserve in the shot. This worked beautifully, giving me another 2.5 stops of dynamic range that I could use. Then I brought out a Tiffen UltraContrast 1 filter, and this gave me the half-a-stop I was missing in order to match the a6000 stills. That's it: 13 stops. Happy me.

    I expected ISO 1600 to be noisier but well, I was very wrong. Without noise reduction or highlight recovery, the p4k has 1.5 stops more DR at ISO 1600 than at ISO 400, basically because the shadows are a lot cleaner out of the gate. Highlight reduction again gave me half a stop (being optimistic). Noise reduction didn't have that much room to help now: it "just" brought back one stop. In fact, if you're going to apply NR if needed, then ISO 400 and ISO 1600 have exaclty the same DR, it's just that ISO 400 will force you to add NR much more often.

    BTW, noise reduction works great here because the source footage is so nicely compressed. That 4:1 thing is lossy, but in a very mild form. Adding noise reduction to the RX100IV XAVC-S footage restored exactly zero stops of dynamic range: once textures start to go, the codec deletes any traces of them and NR simply can't bring them back.

    So that's it: 12.5 stops, just short of matching my pre-determined target, with the ability to get there by using my ultracontrast filter. And a whole two stops better than my previous camera. Also, one stop better than what you can get with the best Sony mirrorless cameras today, and on par or slightly above when compared with the XT3. The initial result was scary but it all ended very well. It does require more post work, though (and Resolve loves to complain about my GPU RAM as soon as I start adding NR to all the clips).

    Final note: this is exactly the performance I expected from the p4k, by looking at 12-bit RAW stills from the GH5s.



    a6000 ISO 100 RAW stills --- 13.0 stops (dxomark figure; I measure everything else against this)

    RX100IV slog2 ISO 1600 ----- 10.5 stops
    RX100IV slog2 ISO 1600 +NR - 10.5 stops

    p4k ISO 400 ---------------- 09.5 stops
    p4k ISO 400 +HR ------------ 10.0 stops
    p4k ISO 400 +NR ------------ 12.0 stops
    p4k ISO 400 +NR +HR -------- 12.5 stops

    p4k ISO 400 +TUC1 ---------- 11.0 stops
    p4k ISO 400 +TUC1 +NR +HR -- 13.0 stops

    p4k ISO-1600 --------------- 11.0 stops
    p4k ISO-1600 +HR ----------- 11.5 stops
    p4k ISO-1600 +NR ----------- 12.0 stops
    p4k ISO-1600 +NR +HR ------- 12.5 stops

    p4k ISO 1600 +TUC1 +NR +HR - 13.0 stops? (didn't test it )


    Now, for how it all looks when uploaded to Youtube. Watching from my coach all those textures (mainly the one on the black shoebox) are not visible anyway, so instead I'll just look at the area of the vintage camera on the right. Bare (no NR, no HR, no TUC filter) the p4k at ISO 400 is basically tied with my RX100IV. Applying HR+NR leaves them tied still (maybe with a small advantage for the tiny Sony). At ISO 1600 the p4k leaps ahead, getting an advantage of more than a stop both with and without HR+HR. And that's exactly how it performs ati ISO 400 but with the TUC filter. I didn't try the TUC filter at 1600, but I guess it will add an extra stop (or at least half a stop).

    So for youtube the p4k is still good in terms of DR, but not as good as it is when pixel-peeping on the original clips (the nicely-compressed-RAW advantage is awesome when looking at fine textures, not so much when looking at bigger shapes). It depends a bit on how much the TUC filter adds when shooting ISO 1600 (I didn't test it because I thought 400 would be better), but on YT the best possible p4k result ends up a bit over one stop ahead of the RX100-IV, which means it's tied to or slightly better than the best Sony mirrorless cameras today, and on par or slightly below the XT3.

    a6000 ISO 100 RAW stills --- 13.0 stops (dxomark figure; I measure everything else against this)

    RX100IV slog2 ISO 1600 ----- 10.33 stops
    RX100IV slog2 ISO 1600 +NR - 11.33 stops

    p4k ISO 400 ---------------- 10.5 stops
    p4k ISO 400 +NR +HR -------- 11.0 stops

    p4k ISO 400 +TUC1 ---------- 11.5 stops
    p4k ISO 400 +TUC1 +NR +HR -- 12.5 stops

    p4k ISO-1600 --------------- 11.5 stops
    p4k ISO-1600 +NR +HR ------- 12.5 stops

    p4k ISO 1600 +TUC1 +NR +HR - 13.0 stops? (didn't test it )



    Short video with some of the options I analyzed:
    https://www.youtube.com/watch?v=6v_1DYwvwRM
    Last edited by Samuel H; 11-05-2018 at 11:49 AM.
    Reply With Quote  
     

  2. #2  
    Senior Member
    Join Date
    Apr 2012
    Location
    Philippines
    Posts
    540
    Can you dxplain the iso1600 and why not 3200?
    Dustin Uy-Filmmaker
    Reply With Quote  
     

  3. #3  
    Senior Member
    Join Date
    Jun 2012
    Posts
    168
    You shouldn't really care about ISO on this camera especially when shooting RAW. Within the two ISO ranges, all the changes are digital so they're equivalent to a grade that pushing the grey point lower or higher. If you're exposing to the right, then it doesn't matter what ISO you're at. If you're exposing for middle grey, then you're gonna have less noise is if you expose for the lowest ISO within each range (100 and 1250) because you're exposing the sensor to more light.
    Reply With Quote  
     

  4. #4  
    Senior Member shijan's Avatar
    Join Date
    Apr 2012
    Location
    Odesa, UA
    Posts
    1,145
    Great! That's definitely more interesting measurements than Cinema5D. Do you have any plans compare it to BMMCC or Pocket?
    All my custom made accessories for BMMCC/BMMSC now available here https://lavky.com/radioproektor/
    Reply With Quote  
     

  5. #5  
    Senior Member
    Join Date
    Mar 2014
    Posts
    225
    Weird, that's the exact opposite of what these guys found. They say 1600 is basically unusable for its terrible dynamic range, that 400 and 3200 are vastly better. Thoughts?
    https://www.youtube.com/watch?v=ItdE_rsv014

    Jump to 8:55 in for the meat of their findings...
    Reply With Quote  
     

  6. #6  
    Senior Member
    Join Date
    Apr 2012
    Location
    Philippines
    Posts
    540
    Quote Originally Posted by Flabasha View Post
    Weird, that's the exact opposite of what these guys found. They say 1600 is basically unusable for its terrible dynamic range, that 400 and 3200 are vastly better. Thoughts?
    https://www.youtube.com/watch?v=ItdE_rsv014

    Jump to 8:55 in for the meat of their findings...
    Doesnt make sense why 1600 would be different from 3200.
    Dustin Uy-Filmmaker
    Reply With Quote  
     

  7. #7  
    Member
    Join Date
    Nov 2014
    Location
    www.gigaboots.com
    Posts
    78
    Because 3200 and 400 are the two base ISOs of the camera. 1600 is the most overstretched limit of the lower iso.
    My Gaming YouTube Channel: http://www.youtube.com/gigab00ts
    Reply With Quote  
     

  8. #8  
    Senior Member Samuel H's Avatar
    Join Date
    May 2012
    Location
    Madrid, Spain
    Posts
    635
    I didn't even test ISO 3200 because I'm doing ETTR and, as others have said, it's the same as ISO 1600. At least that's what the docummentation said. I guess someone should test it, and I may do it, but not in such a comprehensive way as what I did yesterday, just shooting something at boh ISO values (without changing f/ or shutter speed) and checking if there's a difference.

    And yes I plan to compare the p4k with the old pocket, but I have to borrow that one so no idea when it will happen.



    edit: I posted that before actually looking at the link. Now that I've watched it... it's a worthless test, I could have told you that it was going to look bad even before shooting. ISO 1600 and ISO 3200 are exactly the same, but if you change you expose for mid grey then you're going to change your aperture or shutter speed, and then at ISO 1600 your highlights will clip sooner but your shadows will hold up longer... even though ISO 1600 vs ISO 3200 is just a change in the metadata of the clip.

    If I think about a good way to present my results, I'll try to upload a video. Too many clips and crops and too little notes right now, I just organized the analysis for myself. Maybe next week...
    Last edited by Samuel H; 11-04-2018 at 05:29 AM.
    Reply With Quote  
     

  9. #9  
    Senior Member
    Join Date
    Apr 2012
    Location
    Philippines
    Posts
    540
    Ive done some tests and ive concluded shooting CDNG like the old bmcc cams, ISO doesnt matter. You can just adjust the ISO at resolve
    Prores on the other hand requires proper understanding if you want to fully maximize your DR. Prores actually compresses the highlights and shadow dependingn on your ISO. Its still best to shoot at the native imo
    Dustin Uy-Filmmaker
    Reply With Quote  
     

  10. #10  
    Senior Member Samuel H's Avatar
    Join Date
    May 2012
    Location
    Madrid, Spain
    Posts
    635
    Sure, all I said is only regarding RAW. I'm not shooting prores, ever. With BRAW just around the corner, I just don't see the point of shooting prores with the p4k, unless you're not shooting 4k.
    Reply With Quote  
     

Similar Threads

  1. BMMSC dynamic range
    By Greg Pickering in forum General Discussion
    Replies: 3
    Last Post: 09-13-2016, 09:42 AM
  2. Which has more dynamic range? 4k raw or 4.6K prores.
    By TheInternet in forum General Discussion
    Replies: 11
    Last Post: 05-24-2016, 03:37 PM
  3. GH4 Has More Dynamic Range Than You Think
    By Joshua Cadmium in forum Off Topic
    Replies: 3
    Last Post: 05-02-2014, 02:03 PM
  4. Dynamic Range from 5D Mark III
    By nickjbedford in forum Cinematography
    Replies: 8
    Last Post: 09-19-2013, 08:11 PM
  5. 4K Dynamic Range
    By mrbrycel in forum General Discussion
    Replies: 14
    Last Post: 04-26-2013, 10:19 AM
Bookmarks
Bookmarks
Posting Permissions
  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •