Tragic Lantern for EOS M

Started by coutts, April 17, 2013, 01:43:28 AM

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.


eosm1986

Hi All:

OK - I am thinking I may know why my camera is having such an affinity towards larger apertures.  I thought it might be trying to focus on dust on the lens, etc., - so I bought a cleaning kit to clean that, reseated the lens, no change.  I then flashed back to old firmware.  Still no change.  However, I see a field in the update firmware section that says my lens firmware is 1.8.0.  I'm wondering if when I flashed ML, it changed the lens firmware.  I guess I'm open to the idea that it *should* be using those larger apertures, but it's just such a big change from before when I flashed ML. 

So, I downloaded the current canon firmware - right from website - and flashed it.  No change.  Behavior still the same. 

So what is the most up to date version of lens firmware? 

Help!



Canon eos m

Quote from: gary2013 on September 06, 2013, 10:32:23 PM
Max, I understand everything after step 2.
Step 0- where? Manual mode in photo mode or video mode top dial?
Step 1- set Raw 2048, 2.39 how and where? If i do that in movie (top dial) in the movie menu raw sub menu, it always goes back to 1728 rez. ?? 12 fps how and where? Using FPS Overide?

Sorry for not understanding. I am a video guy and I never really do any of the photo mode stuff.
Gary

Quick questions:

I don't get more than 1728X on 5x mode. I do go higher with 10x mag though.

Any way of focus racking on EOS M.

I am on 1%'s latest hack version.
Canon 5D Mark III, Gopro Hero Blacks with 3D Casing, A Few Lenses, Adobe CC 2014, MacBook Pro, Windows 8 PC, Lots of Video Rig!

Started Nuke. Loved it but then the 15 day trial ran out. Back to After Effects and loving it :-)

maxotics

Towards A More Generalized PDR for the EOS-M?

I've been studying this Pink Dot issue.  1% will have to give his take, or someone else who knows more than I, but I believe the RAW files captured by the EOS-M, in video, are not the same RAW files that the camera takes for a photos.  There are just too many dots used for focus, edge detection, whatever Canon needs to read from the sensor to focus the lens.   I now suspect there is no real fix to the pink dot issue. 

I'm going to talk about a 16:9 aspect ratio.  The EOS-M sensor has 5,184 x 2,456 pixels.  Or, in an aspect ratio of 16:9, 5,184 x 2,903 pixels.  Because Canon knows they will down-sample the images to 1920 x 1080p, they have 3,264 pixels wide they can either discard or average down.  They can do the same for 1,823 pixels high. 

In order to illustrate this I did some graphics.  Here's a graphic showing the relative size of the sensor, to the size of the frame it will save as video in 16:9



Do you see any focus dots in that box?  No.  But they were there in the graphic before I shrunk it down.  Here is that file at the same relative size of the sensor.  This optical trick demonstrates how Canon can use all these pixels for focus without them ending up as pink dots in their videos.



And here again, are the pixels, used for focusing, at 100% size on my monitor



And here they are, blown up to 800%



As you can see, on a pixel level, there isn't just one pixel used for focusing.  There are clusters of pixels, a few pixels wide.

Again, after the camera down-samples all these pixels in video mode you don't notice them.

My guess is that in photo mode they shoot one of these pixels-used-to-focus frames, focus, then chuck it out and do another frame where each pixel acquires only image data.  1% your thoughts?

If I look at any frame that went through PDR effectively, I can still see many of these focus pixels, especially at any hard edges where colors make abrupt changes.  They blend out in any set of 50 pixels, say, that are generally the same color.  In short, these RAW movies are made using a large number of focus-debug sensor data from the camera.

Now, I'd rather have a high dynamic range image at 1280x720, with lots of noise/blurred color information, but I believe these essentially heavily masked video frames on the EOS-M, and other similar Canon cameras, will prevent them from reaching the same IQ as the Black Magic cameras.

From what I understand the black magic cameras, because they are reading from the WHOLE sensor, have more room to down-sample their image efficiently.  With the EOS-M, to get moire-less footage, you need to use a very SMALL PART of the sensor, 1280x720 pixels to be exact.  There are no extra pixels to throw out.  Each pixel that is used to focus, and there are many of them in that block,  are there, degrading the image.

If all this is true, the Pink Dot remover will never be able to completely remove the residue of focus pixels.  And if that's true, a simpler dots/spots average algorithm might be better.  (This whole post is the result of my PDR no longer working on my videos.  I have a feeling something changed in RAW output from the developer's new work.)

Another benefit of a different PDR is it could potentially post-process the moire issues away from the camera. Moire makes it imperative one use crop mode on the EOS-M, which means you can never get a very good wide angle shot.  I think 10mm ( or 16mm x 3) or 48mm effective is the lowest you can go.  If someone could "hide" pink dots and fix moire patterns in non-crop mode than the camera's 22mm pancake would make it quit nice.

Anyway, I look forward to the dev's take on this.  Maybe the way to go is higher dynamic range h.264 through HDR or the multi-shot ISO feature, which I've been trying out.   Has RAW reached it's limit on the EOS-M?




Rewind

QuoteAs you can see, on a pixel level, there isn't just one pixel used for focusing.  There are clusters of pixels, a few pixels wide.

What you see in your example is interpolated bayer pattern, that's why you think it's 'clusters'.
In order to see actual af dots locations, you have to use some soft which gives you apportunity to see actual bayer pattern, Photivo for example.
Then, you average a dozen of these shots to avoid noise, and finally you can get the actual af dots locations. A picture like this:
What PDR is doing, it applies a smart adaptive interpolation for the af dot pixels.
It does not working for crop modes anymore because of something changed in crop mode centering in latest builds, so we have to change dot data files.
Just for your information.

AnthonyEngelken

Quote from: Rewind on September 07, 2013, 07:38:01 AMWhat PDR is doing, it applies a smart adaptive interpolation for the af dot pixels.

If the dots are created by auto fucus, can they be avoided by turning auto focus off?

Rewind

Quote from: AnthonyEngelken on September 07, 2013, 10:52:28 AM
If the dots are created by auto fucus, can they be avoided by turning auto focus off?

Wish it was that simple ))

By the way, guys, while the devs of PDR are AFK, I found a way to help you with the crop mode.
I need you to shoot a bunch of frames (50-100) with the lens cap on, and send me a raw file (better upload it somewhere and give me the link)

Just recently a was able to modify the dot data files to succesfully remove all the dots from 720p mode for my 650D, now i can shoot 60 fps raw )

I figured out how the data in dot data files arranged, so now i can help you too.

dannydv

Quote from: BernardDigital on August 30, 2013, 09:51:19 PM
Is the intervalolmeter working for everyone else? For me it hangs and freezes up the camera after the countdown gets to zero to start the time-lapse...
Intervalometer is working fine for me. Stop can't be defined but using the MENU button will stop the intervalometer.
Canon EOS M with EF-M 22mm f/2.0 + Canon EOS 5D MII ML 2.3 with EF 50mm f/1.2L + EF 24-105mm f/4.0L IS + EF 14mm f/2.8L II + EF 100mm f/2.8 Macro

Malakai

Quote from: BernardDigital on August 30, 2013, 09:51:19 PM
Is the intervalolmeter working for everyone else? For me it hangs and freezes up the camera after the countdown gets to zero to start the time-lapse...

Try the power cycle trick before using the intervalometer. Power on the camera then power it off and back on within a second, if the SD light flashes red for a split second then you were too slow. start again.  If you did it, try the intervalometer. What lens are you using?
Hunting for that elusive EOS M shutterbug!!

maxotics

Quote from: Rewind on September 07, 2013, 11:00:48 AM
By the way, guys, while the devs of PDR are AFK, I found a way to help you with the crop mode.
I need you to shoot a bunch of frames (50-100) with the lens cap on, and send me a raw file (better upload it somewhere and give me the link)
I figured out how the data in dot data files arranged, so now i can help you too.

That's awesome Rewind!  Thanks for your help.  I've updated a zip of about 150 frames, shot as instructed.  It's 150 megs.  You can download it by going to

www.maxotics.com/photos

Then put this in for a key

EOSMCropPattern20130907.zip

You will then see a download link.

However, before you do that, I would download this file where I tried to get the dot pattern clear, it may be good enough and is small.  That file keyname is

EOSMCropPattern20130906.tiff

If that doesn't work, I hope the 150+ dngs above do.

Again, thanks for your help and explanation of the dots.   (though again, if you can point me to the explanation of how to read a dot pattern and change the config settings in PDR that would be greatly appreciated!)

Rewind

EOSMCropPattern20130907.zip — This key does not gives me a link.
Tif is not acceptable, because of interpolation. I need source dngs to see actual bayer pattern.

maxotics

put in the name without the .zip extension (SORRY!).  Thanks!  I'll work on getting bayer pattern also...

Here's one I created using a single DNG in Photoshop.  BTW, I find shooting, focusing, on some well lit white paper gives clearer dots than the lens cap. 

Try this photo key.  If it's not good enough, can you tell me why?  THANKS!

EOSMCropModeDNG20130907

Rewind

For those, who stuck with pink dots in EOS-M crop mode, this should do the trick:
PDR Rewind EOS-M Crop
Make sure you convert the RAW file, not dngs, just drag&drop them onto the PDR window, select EOS-M from dropdown menu, and chose Interpolate, then push Convert Files.
Then extract the dngs by raw2dng or whatever method you used to. Those dngs should be free from pink dots (at least for 1280x720 in 5x crop mode). Let me know if it works for you.


maxotics

BACK IN BUSINESS!  GREAT JOB REWIND! 

Rewind

Quote from: gary2013 on September 07, 2013, 09:07:27 PM
Thank you Rewind. Can someone modify PDR so it has an indication that it is working? A progress bar showing how long is still needed.
Gary
It has that indication actually, but... file-wise. If you feed a number of RAW files, progress bar will show the progress by each of the files is being processed.

maxotics

Was it difficult.  Was it just off by a pixel?  What do you think happened?  THANKS AGAIN!


maxotics

@Gary, it shows a progress bar if you process more than one RAW file at a time.  If you just do one, yes, there is no progress bar I can see. 


Rewind

Quote from: gary2013 on September 07, 2013, 09:14:06 PM
exactly how and where is this indication shown in the PDR window? I do not see anything except a pop up saying it is completed. I am using the PDR you posted here

I am  talking about doing a raw file only, not DNG files
Like this:

maxotics

Quote from: gary2013 on September 07, 2013, 09:16:31 PM
ok, cus I just sat here for a long time waiting for one raw file and it never did anything for some reason. must have been a bad file cus they usually work

That happened in the beginning to me.  Now I make sure I've clicked it!  Otherwise, simple enough to click again or, to be safe, close, re-open, and try again.  I have gotten warnings/stopping from corrupted files.  So I don't believe it will go on forever doing nothing.