Messages

Again, please suggest better math. I've spent a lot fine-tuning it and I don't feel like doing it again.

Yes, if battery discharge would happen while the camera is off, you won't be able to power it back on without taking the battery out first.

It can happen if ML leaves DryOS in some inconsistent state (memory corruption or stuff like that).

Active and moving? that's not quite good.

The only easy solution that I see right now is to try globaldraw on and all other things disabled.

Or... hack the CONFIG_KILL_FLICKER code so Canon GUI can be disabled from some other tasks too. Not sure what is the best way to implement this; maybe a reference counting like LV RAW mode from raw.c.

Doesn't make sense; when that happens, the camera will not power on without taking the battery out first.

The algorithm is just adapting itself to the contrast of the scene. Feel free to suggest a better formula.

Yes. Does Canon histogram stop updating while shooting raw video?

Canon histogram is not relevant for exposing raw btw. Use ML raw histogram for that.

On 5D3, I've ran a sensor cleaning (not sure what exactly, I've tried all the options from that Canon menu) and pam! all of them were gone.

Right, the digital ISO from image finetuning is effectively exposure compensation applied "in post" (after the raw data has been captured). But since it's applied before H.264 compression, it can help a lot.

I've pointed this out one year ago I think, but it got buried in the google groups

A little scratchpad (feel free to polish it, fix mistakes and turn it into a nice article): http://magiclantern.wikia.com/wiki/ISO . That one represents my opinion from one year ago, so take it with a grain of salt.

Those lv_af_raw dots are edge detection info. That's why I didn't recommend using it.

What image problems are you having without it?

The benchmark code wasn't touched since then...

90MB on write speed? that puts it in the same league as 5D3...

Is ML properly disabling Canon GUI while recording raw? (to check, enable Canon histogram, it should freeze or disappear while recording)

In the footer of a raw file, or in the exif info of a DNG from raw2dng. The value is not constant (it's somewhere around 1024, 2048, sometimes in-between).

WARNING: may not record faster than my 28May13 build.  so if you are shooting a music video this weekend, you may not want to "upgrade".

Huh?!

Well, there are no Canon engineers here that could answer you (and I doubt they have to justify why their 600D code is a little bigger than the 550D one...)

As said elsewhere: Changing picture quality in the canon menu from RAW to JPEG (L/M/S) gives me a bigger shoot_malloc buffer. I get 100MB extra on my 60D:

^^^ that is the exact setting.

Good and bad.

Good: you can get the same video quality as the other APS-C canons;
Bad: we need to figure a way to work around the H.264 (very low bitrate, or some other way to get into that video mode without actually recording).

For now, I don't think it makes sense to post performance data for the EOS-M.

Why do I have 100GB free space on the hard disk? My neighbour has 200!!!

Updated, can you confirm the other fields? I don't know the maximum resolutions, read them from menu.

Because 1728x672 covers the 3:2 area, not 16:9.

If you can't get into menu, take some silent pics while recording and upload a DNG.

Good point, because ISO (and shutter on certain cameras) can be changed every single frame, without delays.

2. collecting the *same* amount of photons but at different iso levels

Here it's all about the analog amplifier between the sensor and the ADC. If that amplifier is less noisy than digital gain (e.g. multiplying underexposed data by 2), you can start thinking to increase the ISO.

For digic 4, the analog amplification from ISO 1600 to 3200 is just as noisy as the digital one, so there's no point in going above ISO 1600.

Sure, by raising ISO you lose some dynamic range. You increase ISO by 1 stop => you lose 1 stop of DR from ADC saturation and you gain less than 1 stop of DR because of cleaner shadow noise. So, if you can increase shutter or aperture, do that first (because you don't lose DR by doing that, you just shift it).

But if you have more DR than your scene needs, you can throw away some of it, and increase ISO to get better results.

I recommend reading this: http://www.guillermoluijk.com/article/isos5dmkii/index.htm

Some major changes to EDMAC things (5 minutes ago), should fix these issues. Some EDMAC writes were not properly aligned at addresses multiple of 4096 bytes.

This should fix the issue (at least for me): https://bitbucket.org/hudson/magic-lantern/commits/150a9c5810834e60da6311282d61aabc267ad9a7

Needs some testing at all resolutions though.

Another tip: now that we can tell the location of the raw crop frame in the full image, we can minimize those delays (e.g. wait until Canon code changed the window, not a fixed delay). If changing the window takes 200ms (rough estimation based on how long the image freezes when you move the zoom box around), a 3x4 pic would take 2.5 seconds. Not that bad.

There's a small delay when changing the crop window (at least via properties). Maybe in the future we'll figure it out.

On 5D3/6D, the crop window has over 3500 pixels, and panning makes a bit more sense there.

Here you go: http://www.magiclantern.fm/forum/index.php?topic=6215.0