Show Posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.

Messages - gabriielangel

Pages: [1] 2 3 ... 8
MOD intervention: User is giving incorrect information. I just tested my Samsung PRO Plus 256 GB (2021 edition, MB-SD256K) with Mood Crop 2023Aug19 build and recorded for 5:03 with manual stop. Recording size is 25,569.4 MByte resulting in 84.3 MByte/s. Benchmark is showing > 90 MByte/s in PLAY-PH mode (cam in photo mode).

=============Original post below========================================

This Samsung card is not compatible with overclocking on eos m,don't buy it!

MOD: Without the Recording Resolution, Bitrate, and a description or a screenshot of the scene you shot, your statement can be misleading.

For example:

a Sandisk Extreme 256GB 170 MB/s will show 94.1 and 95.1 MB/s write speeds in the benchmark;
a Sandisk Extreme Pro 256GB 200 MB/s will show 94.8 and 95 MB/s.

but at 2.8k/14 bits, only the Sandisk Extreme Pro will be able to record long takes in a reliable fashion in difficult / well exposed  scenes.
In order to do so at -1 EV (1.0 showing on the histogram) , a card needs to be able to handle 85 MB/s continuously and tolerate 93 MB/s peaks (And those peaks can last up to 4-5 seconds on the Extreme Pro).

The benchmark doesn't give you the whole picture. In order for any  speed figures to be really meaningful for actual use, comparison tests between cards shall be made by recording the same scene with identical parameters.

Recording a dynamic scene at -4 EV / 10 Bits will work with a lot of cards, but the noise level and subpar resulting image will seriously misrepresent the actual capabilities of the cam...

Performance hardly makes a difference. They all do > 90 MByte/s in benchmark and Lexar 1667x tops at over 100. Results are within 10 percent range.
10 percent in bandwidth translates to less than 5 percent difference in horizontal resolution and that is nothing to write home about.

This is exact. But a 4% increase in bandwidth will give you about 1 EV of extra exposure latitude at 2.5k/14 bits. So although marginal, the gain can help mitigate camera hiccup related issues.

I tried updating our MLVApp AVFoundation export for Apple SDK, for macOS newer than 10.13. @Danne and other mac users, who are able to compile: could you please test that? If someone could test if the old options still work on older macOS, I also would be interested in test results. Thx!

The update brings ProRes Proxy, LT and HQ via Apple engine.

If someone can provide a compiled Macos 10.13+ version, I can test (when I compile myself, I get errors and the resulting version is too slow (No MP support).
AVFoundation is the option to choose to get an output file with the proper gamma and accurate color translation. So your update should be very interesting!

Shutter fine-tuning is on and shutter set to 1/48th @23.976FPS

Try setting your shutter to 1/42 (or the closest value, depending on the resolution you are using). That way, you don't need to use shutter fine tuning, and there is no appreciable difference. The image remains sharp down to 1/30 (unless you are filming fast cars, etc.). This allows you to use all the hacks and record 3K 12 bit for at least 30 seconds on a freshly formatted card. The gain in clarity / fov trumps the motion blur difference (if any) in my opinion.

By the way, to get the exact same motion blur I am used to at 1/48 on a Sony cam, I need to set my eos m to 1/44 - 1/46.

Just tested Crop mood's 1x1 full resolution 14 bit lossless fps = 3 mode on 650d. Basically, this mode is not working. The script windows pops up for no reason. Also, at 3 fps AR 2.4 the calculated data flow is supposed to be around 40-50 MB/s according to ML, yet after only 1 frame recording stops. This behavior is very similar to the old version before Crop Mood.
I am redeploying my 650d to a dedicated time lapse camera, so this mode is critical to me. I am sure many other MLers like the full res 1x1 mode too. Once trying some full res 1x1 footage, I guarantee more will be addicted to it.
Thanks a lot to Bil's great work!

Have you tried turning sound recording off before recording? The way Bilal stated it in the first post suggests that no sound will be recorded when using that mode, but for it to work, you actually need to turn sound recording off.

I tested HDR video today and it works surprisingly well!
@bilal, the way you made it alternate between both ISO's slowly is genius. It makes it possible to adjust the exposure taking both ISO's into account without having to fiddle with a toggle switch.
It is just superior to dual ISO, as there are no additional aliasing artifacts in overexposed areas, no flickering, and where there is a lot of motion, it just looks like a slow-shutter / slo-mo effect. Very pleasing to the eye.

Is there a way to see the "Merged" image in mlv app? It would be nice if it behaved like dual ISO does, with metadata and all.
To make sure I'd be able to fully recover highlights, I exported a PNG 16 sequence, but right out of the box, the blending is pretty good!

1080p mode is more reliable to me in the Danne build than Crop Mood build.

What is good with Danne's build is that it is set up to work as good as possible by default. Bilal's build requires a little more work before you can use it.
For 1080p mode to work, you need to get into x10 zoom mode and press the onscreen down arrow once or twice to offset the preview. The black screen will disappear.
Also, Kill Global Draw must be set to ON.

@bilal, it would be a good idea to ask the forum admin to add a second post right after the first one on your thread, where you could just add version history / new releases and found fixes. As a lot of things must be set manually and it varies depending on which mode you are using, this can get confusing pretty fast for someone just starting using it... :)
For example, some very important info is only available on pages 5 and 8 (Custiomize Buttons, Preview Toggle,etc.)

Tested the Autofocus on

-10bit is now almost on par with 14bit when autofocus while recording (Small Hacks must be off), quite nice!
It works within the limits of the camera (Some lenses work better than others depending on brightness, and it works with every mode where it is possible to record with small hacks off.)

But when you autofocus while recording, it goes into framing mode. This makes it hard to know when focus has been nailed.
I think it would be more useful if it went momentarily into zoom mode if feasible, so we would be able to see focus being nailed in realtime.
This is how it is working on Danne's build.
EDIT: Preview Toggle must be set to OFF. It defaults to Half-Shutter. When set to Off, there is no zoom, but the display remains in realtime preview if you focus while recording.
So, for Autofocus to be usable while recording, Small Hacks must be set to Off, and Preview Toggle (Located above Small Hacks) should also be set to OFF

-in 1x3 modes, it is possible to change the aspect ratio in the crop mood menu. I don't know if it is intended to be this way, as it is not possible to do so in 1:1 modes (Where you have to change it in the Raw Video Menu)

-When in 2.5k 1:1, the preview shows vertical lines on screen when using lower bitrates. The lines become more visible as you go down bitrates (No lines at 14bits--->Very visible lines at 10bits). Those lines are not recorded in the file.

Could it be posible to add an option to change the prefix of the mlv filenames so each camera has distinctive file names? When using different cameras if they are set to the same date sometimes the file names are the same and opening them in MLVapp can cause a big mess.

This is a good suggestion. I often use 3 cams (So that I don't have to switch lenses) and also when using footage from last year, I need to use a batch-rename program to make it more manageable. There is a module to have custom image names, but no such luck for MLV files. Even just One character at the beginning of the filename would do.

And another problem: If I toggle between 2.8K and "Full-Res LV", I get crashes. After restart, Full-Res LV doesn't let me record (stops very quickly with invalid MLV).
When trying to use Full-Res LV 1:1, I get a similar problem to what @masc reported, but in my case, I cannot record with Full-Res LV at all, even if I load it first or if I come from the 2.5k preset.
it gets into record mode for less than 1 second then stops. The MLV is empty.
Build is

    My humble contribution to the discussion:

    • Then, for each of the 4 videos, I averaged the same number of frames to get rid of random noise in the same way.
    • Then I graded to taste the 4 resulting single frames in MLV App.
    Does step 2 make any sense? I think it does, but of course would be glad to hear from you.

    It does to see clearly any differences in gain, color and contrast.
    But the amount of noise and its quality are an important component. When using luts, any instability in the noise will be magnified and usually manifests itself as blotchy uneven color dots dancing all around.
    So, if one plans to use luts or lift shadows significantly, 12 and 14bits should be preferred.
    If the scene is well lit and there are few shaded areas, overexposing slightly (Without Clipping) and lowering in post makes the latest 11bit lossless quite clean and useful to get 5.2k 1x3 continuous, where 10bit lossless should be avoided as much as possible.

    On a side note, Bilal's build is quite tolerant when there are a few clipped highlights (Black when using false colors), whereas the previous builds would stop recording almost immediately.

    For those who didn't know, it is possible to record Uncompressed (As opposed to Lossless) with some presets.
    The advantage is that the Data Rate being constant (No Compression), Scene Complexity and Brightness have little effect on recording times.

    As this build is less prone to yield compression errors with very bright scenes, and there is now an 11bit lossless option;
    10bit Uncompressed will most likely be advantageous only when the scene is very bright and busy. (Lots of tree branches, bushes, fur, pine trees, busy landscapes etc.)

    Possibility is as follows:
    2520x1048,  10bit Uncompressed, Orange (Tested over 1 minute of recording)
    1472x1846,  10bit Uncompressed, Orange
    1736x720, up to 14bit Uncompressed, Green (!)

    In order to test, I recorded the same scene at 3 degrees of brightness (F2, F4, F11).
    As you can see in the screenshots, varying brightness with Lossless changes the Data Rate, but the latter remains constant when using Uncompressed.

    Recording 10bit ettr and lowering exposure in post will yield a cleaner image than 14bit recorded severely underexposed.
    Also, 10bit Uncompressed has cleaner shadows than 10bit Lossless (No red tint)

    You can see the data rates at various resolutions in the screenshots.
    This could be particularly interesting for those using 1080p 3x3.

    Tip to get longer recording times: Low-Level Format your card every time it is full (As opposed to deleting the files). There is an option to keep the ML files now (Not sure how this happens, but it works). It is easier to write on a freshly formatted SD card than on one with files deleted.
    There is a good primer on that topic here:

    Test-f2" border="0
    Test-F4" border="0
    Test-F11" border="0

    In order to test it and validate it on your end:

    For 10bit Uncompressed, set the bitrate in the RAW video Menu. Choose 10-bit (The bitrate becomes grayed out in the crop mood menu and has no effect. Even if it shows a different bitrate on the main screen when recording, the recorded file will be 10bit Uncompressed. You can choose 12-bit or 14-bit if using 1080p 3x3)

    For Lossless, You put back 14-bit lossless in the RAW video menu, then you set the desired bitrate in the Crop mood menu.

    Tested with:
    240Mhz Overclock
    Sandisk Extreme Pro 170MB/s 128GB

    This build is the most robust so far.

    Also wonder if custom modes would be good in this build as we have no fast acces to presets otherwise. This way one could create 5 or more custom presets with all ratios, bitdepths etc already set and remembered.

    I have a suggestion here: I have noticed that it is possible to press&hold the SET button to have an extra function, in the same way we press&hold the trash can icon to go into ML Menu.
    This button would be a good candidate for a preset menu, because the tap screen is prone to false triggers, and once you tap it (In Danne's build), it will change presets even if you exit without selecting.

    One more argument is that, for example, my Sandisk Extreme Pro card can handle 2.5k/14bits, but the Sandisk Extreme card is good at 2.5k/12bits.
    Presets makes it possible to switch modes in a more tailor made fashion.

    If you have time could you check af possibilities while recording in x10zoom and while in regular preview and check behaviour compared to my build? I simply don't have the time atm to dig deeper.

    About Autofocus capabilities: (Some of this has been mentioned in other posts in the past few days, but will repeat nonetheless to facilitate the comparison)

    With Danne’s build, hacks lvface and AF on.
    With Bilal’s build, hacks off. (To be able to use it while recording)
    Both with a Sandisk Extreme 128G, 170 MB/s.

    Tested with 2.39:1 Ratio

    AF is bitrate dependant (On both Bilal’s and Danne’s builds). I did not notice a huge difference between 14 and 12 bits, but 10bits misses focus noticeably more often than the other 2.
    The difference in AF accuracy between 10 and 11bits on Bilal’s build is not significant enough for me to quantify it.

    On Bilal’s build, 4.4k 1x3 will keep recording while autofocusing, up to 12bits (Because hacks are off).
    4.8k and 5.2k just can’t record more than a few seconds with hacks off, so autofocus will not be very useful there.

    On Bilal’s Build, It’s not possible to use x10 mode while recording (Recording stops as soon as you hit half-shutter), but when idle, it never misses focus.

    AF at X5 mode is not as robust as x10 ATM, but as Danne’s build seems to be using x5 mode while recording, it can still work pretty well.

    So, on Danne’s build:

    -2.5k 1:1 Centered 12/14 bits: Focus is near perfect before and while recording and everything is pretty much centered;
    -2.8K 1:1 12 bits: Can focus while recording, but misses a lot. When enabling the Canon GUI to see the focus box, putting it a little to the right and 1 step up makes it miss less often, but it’s still not perfect;
    -3K 1:1: Will work slightly better than 2.8k, but you need to lower frame rate to 20fps to run at 12 bits, so…
    -5k ana frtp 12/14 bits: Focus is very good before and while recording, but you sometimes have to aim slightly lower to nail focus;
    -5k ana flv: Can only focus before recording. Trying to do so while recording will lead to corrupted frames.
    -1080p: Will Focus almost as well as 2.5k 1:1 Centered, but screen turns pink while using AF. The recording is always  corruption-free nonetheless.

    On Bilal’s build:

    -AF works evenly across all modes and resolutions;
    -The preview always remains clean while autofocusing, even while recording;
    -X10 mode AF while idle never misses with the good lenses. Has the same accuracy as Photo mode;
    -X5 mode AF while recording is not as robust as Danne’s 2.5k, but it is precise enough. Will nail focus 8-9/10, but will sometimes get lost and completely lose focus for a while;
    -AF will sometimes stop working at x5, so you just need to go into x10 mode, focus, then it will start working again.

    The things to keep in mind are:

    When moving the AF box in x10, it remains where you put it. So if moving the focus box to the side to get  a certain framing/focus subject,
    In order to be useful in x10 while recording (If this becomes possible), the AF box would need to be automatically moved to the center of the frame as soon as you hit record. Otherwise AF would become very hard while recording.
    This is not a problem in the current implementation, because in x5 mode the focus box is always at the center.

    The focus box should be at the center of the sensor area being scanned. It is not a big problem anymore, as most presets in both builds are now very close to the center, the exception being 5k ana frtp on Danne’s build.

    Some lenses work better with AF than others; the best lenses I have tried are the ef-m 22mm f2.0 and EF 35mm F2.0 IS USM.

    Also, when in x5 mode, the zoom looks larger on Danne’s build than on Bilal’s.

    If I forgot something, let me know.

    If you have time could you check af possibilities while recording in x10zoom and while in regular preview and check behaviour compared to my build? I simply don't have the time atm to dig deeper.

    Yes, You can check back tomorrow morning and I will have something for you.

    @bilal, your build works flawlessly at x10 zoom, but only when idle (Pre-recording). This is why I said earlier that it just works, because it works exactly as you advertised it.

    Dannes build works differently. In 2.5k  you can focus flawlessly before and while recording, but in other modes, steps have to be taken and it isn't 100%. The problem I found is that the focus box was sometimes too big compared to what you were focusing on, and it wasn't fully centered. So you had to re-enable the Canon GUI to see the arrows, put the focus box where you want it, etc.

    I will give you all the details I have tomorrow morning also.

    Turn off crop mood and raw video from Movie tab, can you take pictures?
    Yes, it works normally. Turning off Crop mood and leaving Raw video on also works (All the modules mentioned in your video are loaded)

    Some of overlays from "Overlay" tab take an effect and it's being applied on image review, that's not a bug. Turn off all overlays, re-test then report back.

    I am unable to reproduce this issue now. I will let you know if I figure out what happened. (This was regarding the garbled black frame around the picture preview)

    Alternative method would be to put module load flag in build zip file in SETTINGS folder, e.g. put "mlv_lite.en" inside SETTINGS folder.
    This way the module will be loaded, and when it's needed to unload the module, you can do that from ML menu.

    I were under the impression that pre-setting the flag which put the green light on next to the required modules (Whatever happens after we turn those on ourselves) was less involved.
    I agree that all the modules should be in the list, ready to be enabled/disabled. In addition to what you pointed out, It makes it easier for a user to try something he/she read on the forum, without having to track down and add a module in the folder after the initial install.

    Let's give the guy a break😂. I spent so much time on my version. Porting should be done not in a rush. And why the hurry 😎.

    Ahaha! This is so true. But I think everyone is more than happy with what has been done so far, there's no doubt about it.
    When someone looks at what Bilal and yourself have done in the past two years alone, there's a reason for everyone to be overly enthusiastic :)

    What's the case with Danne build? same issue?
    This is an EOS M issue only, other supported models don't have it. Is it the same as shutter bug known issue?
    With Danne's build, as long as you shut down the camera before going to the photo modes, everything works fine.
    This is not the same as the shutter bug issue. In my case the shutter bug issue happens only with the ef-m 15-45mm.
    When this happens, you just unscrew the lens to break the connection, re-screw it and the shutter works again.

    With your build doing this won't fix it, you really need to start with Set button pushed.

    Does Photo mode work if modules are not loaded? Why if they are loaded but turned off?
    With no modules loaded: You can focus and take pictures, but something is broken. You take the picture, and when the picture appears briefly, it is surrounded by a noisy frame before going back to preview.

    I don't know how to turn off modules. I can make the module dim and not load at the next restart but don't know how to turn it off

    I might not able to help with solving this issue, mainly because I don't have EOS M. If this issue is also presented in official ML builds, that mean it's not exclusive here (I didn't break photo mode on EOS M, it was already broken).
    Danne's build never exhibited this issue for as long as I can remember (About 2 years), so I guess it should be fixable.

    You could exit by pressing half-shutter button right? or by using arrows and press on "Exit" on MLV play GUI?
    On 700D when MLV playback is running, MENU button doesn't do anything too.
    Yes this works, my bad!

    What did you excpet it to do? was pressing MENU button in Danne build make the camera exit from playback to LiveView?
    With Danne's build, pressing Menu would exit. As clicking Exit on the UI or half-shutter works (and as Danne said, you can just touch the screen), I wouldn't change the behaviour. The less unnecessary things or redundancies, the less likely problems are to occur :)

    Quote from: Danne on Today at 07:37:20 AM

        I would probably automate install and exclude most of the modules..

    Not a fan of this step for many good reasons. But yes, you can do that.

    I think that at least pre-enabling all the necessary modules (those you enable in the video) would be better. Users who know what they are doing will be able to go in there and disable/enable what they need.
    Otherwise, we will have to watch your video each time we install your build. (Proof being that most of the problems reported on facebook are because one or some of the modules weren't enabled)
    Danne's approach of having a usable cam as soon as you turn it on is a good one. Most users are more "Artists" than "Engineers" :)

    Regarding framing mode when using half-shutter, We can already use the * button (Left Button) to see framing.
    So it would be more useful to have the cam go to x10 mode automatically and then focus, when using half-shutter.
    Would help a lot when using a gimbal, and the cam misses focus very often without x10 mode anyways.
    (Right now, when custom setting half-shutter to x10 , the Framing preview seems to be taking precedence, so we cannot see the x10 preview)

    I probably would put in tap display too  :P.
    Overall very stable build. Very good work.

    If putting tap display, the arrows on screen  near the focus box (Those appear when in x10) need to be available without interfering, as those are needed for proper autofocus.

    @danne, could you please confirm that you get inaccurate focus even when in x10 mode? (INFO, Loupe icon, focus)

    If so, I will need to test Lens/Cam combinations with my other cams to avoid surprises...

    The following applies to the eos m,
    Any suggestions for more choices?

    -It would be a good idea to add false colors to SET and/or INFO. When recording Bright colors (Like yellow or red tulips, or neon signs) even if the histogram shows 1.0, some colors will clip.
    The false colors catches all the peaks, so as long as you don't see any red, you are good to go. This allows you to stay as close to 1.0 as possible without underexposing too much)

    -The x10 options you added will make it easy to pre-focus in one step, nice! (As opposed to having to press INFO, tap the loupe, focus, tap the loupe again and then press info to return to preview and then hit record)

    -The autofocus just works! (in idle state) I tried with 3 different lenses (EF and ef-m) and the focus is always spot on, as long as you use the  Zoom x10  and you use the arrows on screen to bring the focus box right where you want it.

    -Recording times are comparable to Danne's build, but now those are a lot more consistent.
    The 11bit option is a nice addition, as it allows to get decent recording times with the Highest 1x3 option (5.2k) with nicer noise in the shadows than 10bits.

    There is one thing requiring a workaround:

    It is not possible to take pictures in photo mode. Workaround is to press and hold the set button while powering on the camera. The camera will then work as expected in Canon mode.
    Steps to reproduce the issue:

    Choose one of the two photo modes
    Power on the camera
    Half-shutter to focus, take a picture
    Try taking a second picture, it won't work. You need to restart the camera. (You may be able to take a picture if you go to x10 mode and snap from there)

    Also, when pressing play to review clips, it is not possible to go back to recording mode by pressing MENU, you have to restart the camera or do a battery pull.

    For those who miss Danne's tap screen presets:
    As every resolution needs a different bitrate to get optimal recording times, switching resolutions can be tedious. You may want to use 5.2k/11bit or 4.8k/12bit to get a wider field of view and then switch to 2.5k/14bit  or 2.8k/12bit on the next shot to get a closeup without the binning artifacts (thin vertical lines, diagonals, etc.)

    You could take the easy way and always use 12bits and use 4.8k instead of 5.2k;

    OR (this was my favorite method as it kept you from tapping the screen accidentally and changing presets when you didn't feel like it)

    By using the Config preset option, you can have a total of 4 different configs at startup, 3 by holding either INFO, MENU or PLAY while turning on the camera, and one by just pressing on/off.

    How to do it:

    Go in ML menu (Hold Trash can)
    Go in Prefs
    Go in Config options (first choice)
    Go in Config preset, set it to Startup key

    Now, turn off the camera.

    Turn on the camera while holding either INFO, MENU or PLAY;
    The first time, You will have to re-enable the modules and do the whole setup Bilal shows in his walkthrough video:
    So each restart will have to be done with the same startup key combination until you are done enabling everything;
    Set the proper Crop Mode (or rather Crop mood), Preset, bitrate and button customizations;
    Turn off the camera.

    Repeat for every Startup key combination.

    You will now be able to switch presets on the fly without having to remember all the details.

    And finally, for those who forgot:

    Pressing the Right button (+/-  button) once will allow you to use the wheel to set the aperture. Right button again to go back to shutter speed (It will go back to shutter speed selection by itself after a while)

    On Facebook and Youtube it looks like Magic Lantern is still a thing, many people talks about it, but it's sad to see how the actual interest is lower than it seems  :(

    People mostly go to the fb group to ask questions and share their videos, so it will take a few days before everyone gets to see the announcement.
    The other thing is that fb and YT compression makes the 1080p mode look "Acceptable" (And some of the members have so much talent that the images actually look quite good!)
    Some time will be needed for people to realize how significant Bilal's latest work is (and what he means with the post with Drake in it, as some users never used those modes before  :) )

    This is quite nice! How long can clips be recorded at 3K?

    Is this something specific on my fork or an issue in MLVApp in general? Could you please point out an example illustrating the issue?

    Here are a few examples (2.5k and 5k frtp, with and without dual ISO 100/800):

    In order to trigger artifacts in a clip easily, you must:

    Move the camera vertically slowly;
    Have areas where clipped and non-clipped detail-rich elements are next to one another (Like bare tree branches towards the sky);
    Have horizontal elements moving through High and Low ISO scanlines;
    Clip must be at least a few seconds long for you to see movement happening.

    The exposure flickering is more visible in large, relatively flat areas (few details), where blending occurs.

    If you use a tripod and avoid what is stated above, artifacts are sometimes imperceptible.

    Yes, Windows cmd line is limited. But only 260 frames on Mac? Currently I process GoPro RAW timelapse footage in MLV using DNG import and I also had 300+ frames... I've seen no error so far.

    Sorry, the Official Mac 1.14 version can transcode and import 420+ frames (Don't know the exact upper limit)
    Official 1.14 PC Version: 305 frames Max (Gives an error at 306)
    Iaburn's openmp version: Cannot transcode and import

    Your fork makes it less obvious, but this is an issue in general, as it also does it with the official build.
    I will try to make a few examples in a controlled / repeatable manner over the weekend (I cleaned up my previous tests).

    @iaburn, I tried the Windows version you posted and it really makes a difference on some clips.
    So far, it works best when the whole scene is relatively well exposed without clipping and only the shadows needs lifting.
    When there are a lot of highlights and mid exposures in a given area, we get the horizontal lines and resolution artifacts (Especially when not using a tripod).

    I also saw that the "Transcode and import" option will not allow to import 10 second DNG sequences. On the official Mac version, it is possible to import around 260 frames before it gives an error.
    This function is quite handy to match footage coming from motioncam, when the film emulation filter is used in MLV app.

    Is this a Windows limitation?

    And what will be the point of using dual iso, if you expose so the highlights are not clipping with iso 800?

    The shadows of the iso 800 will be cleaner than the shadows of 100 pushed +3 stops up.. you will probably get more noisy image by using dual iso in this scenario, because lines with 100 iso will be very underexposed

    You're right.  This particular scene has both extremes.

    But once you identify how far you can push it, you'll be able to compromise between the amount of aliasing you are willing to tolerate in the overexposed region and the amount of noise in the shadows.

    As long as the portion you overexpose in the highlights is in the high white region, aliasing will be a lot less obvious. But the minute you begin overexposing a portion of the image where there are a lot of details, problems begin. (Until the Devs find a clever way to McGyver it  :) )

    I posted a few examples here:

    Pages: [1] 2 3 ... 8