Messages

This should help you understand better how it works and how to expose the pictures:

http://acoutts.com/a1ex/split_MG_8792.dng

What did you drink?

Yes, but it's very scary, you have to go in Canon menu...

if you set a short interval, it should do that.

By default, ETTR slows down the shutter as much as it can, so adding an artificial limit doesn't make sense (it will only prevent it from exposing properly).

around hCopyLumaLockKey: 00010001 00000000 00030006 00020006

code around this uses EDMAC 9, 0 and connection 6 for both read and write. Therefore, blocks 2 and 3 can be read/write connections or viceversa?

LV_ResUnLockTripleRamClearPass seems to confirm that block 2 is for write connections and 3 is for read:
00030006 00030007 00030003 00020006 00020007 00020004 and others
connect calls: r/w for connections 6 and 7, r for 3 and w for 4.

What command did you use for exiftool?

You can limit the slowest shutter speed, but not the fastest one (for obvious reasons).

hjulenissen: you seem to know your stuff, but from your question, I bet you did not read my paper and you did not download any DNG samples.

For now it's just bursting silent pics without proper sync. Most of them are good.

From now on, it's your job to get it fully working.

This means I'm confident you can do it

Updated this chart with theoretical values.

If you look at DxO dynamic range numbers, you will notice there's no point in going above ISO 1600 on 7D. I've left the higher ISOs so you can experiment, but that doesn't mean you should use the extreme settings.

The 5 EV difference means the highest analog ISO is 3200, which confirms this: http://www.guillermoluijk.com/article/isos5dmkii/index.htm

BTW, got something cool for you: https://bitbucket.org/hudson/magic-lantern/commits/c2a01891156e

From now on, it's your job to get it fully working.

Fixed the 11-second waiting instead of 10 (was caused by post deflicker running with a very high priority - so the internal clock was no longer updating while post deflicker was running). This bug also tricked me into thinking that post deflicker only needs 0.12 seconds to run, when it was requiring around 1 second.

Could not reproduce the irregular cadence though. Here, it was taking a picture every 11 seconds, but it was regular.

About camera starting up again... that's very likely a sign of memory corruption. That's going to be hard to troubleshoot, but if you can find a way to reproduce it, will help a lot.

vsync should fire exactly once per frame.

Here's the correct way to fix pink frames: http://www.magiclantern.fm/forum/index.php?topic=6740.0 (under research)

The problem is present on all cameras, just the probability of visible side effects is different (ranging from say under 0.1% on 5D3 at common resolutions, a little higher at 3K, and goes all the way to 100% on 60D in crop mode, for example). Enabling overlays is just changing the timing (I bet the problem is a race condition), and here it only happens to increase the probability (but this doesn't mean it's the cause or the solution).

Using the 640x480 mode also changes the timing (the probability of Canon firmware requiring EDMAC access at the same time as ML is much lower, so it's a lot less likely to get pink frames in that mode).

The test from don't click me is relevant for pink frames - when that one will work without side effects in LiveView, the pink frames will be gone.

Come on, anyone with some very basic C knowledge can do it. Even some basic programming knowledge in whatever other language and some google search skills should be enough.

But yeah, it's easier to ask for someone else to do it...

1) normal Canon behavior (ever used a Canon camera without ML?)

2) try working in a short directory (e.g. c:\test)

3) I expose to the right for ISO 100; if the scene needs dual ISO, it would look very dark in LiveView at ISO 100, but the histogram will touch the right side.

I think the cr2hdr needs a little bit of tweaking to leave alone the mids, and work with only  highlights and shadows.

Did you read my paper?

That's why you should try to cover your main subject with both ISOs, and keep the extra range for highlight/shadow recovery (e.g. detail in clouds).

With the crop tool in Paintbrush.

I think I've got raw histogram and zebras working in photo mode, sort of. Zebras say almost the entire picture is overexposed, most probably a wrong offset. Quick and dirty patch:

Code Select Expand


diff -r af0907c14c7d platform/7D.203/Makefile.setup.default
--- a/platform/7D.203/Makefile.setup.default Thu Aug 01 14:44:07 2013 +0300
+++ b/platform/7D.203/Makefile.setup.default Thu Aug 01 15:32:27 2013 +0300
@@ -12,5 +12,6 @@
video_hacks.o \
misc.o \
ptpbuf.o \
+ raw.o \
afma.o

diff -r af0907c14c7d platform/7D.203/internals.h
--- a/platform/7D.203/internals.h Thu Aug 01 14:44:07 2013 +0300
+++ b/platform/7D.203/internals.h Thu Aug 01 15:32:27 2013 +0300
@@ -102,3 +102,5 @@

/** We can record movies in regular photo modes - M, P, Tv, Av... */
#define CONFIG_NO_DEDICATED_MOVIE_MODE
+
+#define CONFIG_RAW_PHOTO
diff -r af0907c14c7d src/raw.c
--- a/src/raw.c Thu Aug 01 14:44:07 2013 +0300
+++ b/src/raw.c Thu Aug 01 15:32:27 2013 +0300
@@ -58,6 +58,8 @@
#define RAW_LV_EDMAC 0xC0F26208
#endif

+#define RAW_LV_EDMAC 0xC0F26208
+
/**
  * Photo-mode raw buffer address
  * On old cameras, it can be intercepted from SDSf3 state object, right after sdsMem1toRAWcompress.
@@ -67,7 +69,7 @@
  * and http://a1ex.bitbucket.org/ML/states/ for state diagrams.
  */

-#if defined(CONFIG_5D2) || defined(CONFIG_50D) || defined(CONFIG_500D) || defined(CONFIG_600D) || (defined(CONFIG_DIGIC_V) && !defined(CONFIG_FULLFRAME))
+#if defined(CONFIG_5D2) || defined(CONFIG_50D) || defined(CONFIG_500D) || defined(CONFIG_600D) || (defined(CONFIG_DIGIC_V) && !defined(CONFIG_FULLFRAME)) || defined(CONFIG_7D)
#define RAW_PHOTO_EDMAC 0xc0f04A08
#endif

@@ -140,7 +142,7 @@
  * It will also work with the values from some other camera, but colors may be a little off.
  **/

-#ifdef CONFIG_5D2
+#ifdef CONFIG_7D
     //~ { "Canon EOS 5D Mark II", 0, 0x3cf0,
     //~ { 4716,603,-830,-7798,15474,2480,-1496,1937,6651 } },
     #define CAM_COLORMATRIX1                       \
@@ -242,7 +244,7 @@
  * This is only used in photo LiveView, where we can't compute it
  */

-#ifdef CONFIG_5D3
+#ifdef CONFIG_7D
static int dynamic_ranges[] = {1097, 1087, 1069, 1041, 994, 923, 830, 748, 648, 552, 464};
#endif

@@ -531,6 +533,13 @@
         skip_left = 72;
         skip_top = 52;
         #endif
+       
+        #ifdef CONFIG_7D
+        width = 5360;
+        height = 3516;
+        skip_left = 160;
+        skip_top = 52;
+        #endif

         dbg_printf("Photo raw buffer: %x (%dx%d)\n", raw_info.buffer, width, height);
         dbg_printf("Skip left:%d right:%d top:%d bottom:%d\n", skip_left, skip_right, skip_top, skip_bottom);
@@ -559,6 +568,7 @@
         prev_zoom = zoom;
     }

+    #ifdef CONFIG_RAW_LIVEVIEW
     /* in zoom mode: yuv position changed? force a refresh */
     if (lv_dispsize > 1)
     {
@@ -574,6 +584,7 @@
         prev_delta_x = delta_x;
         prev_delta_y = delta_y;
     }
+    #endif
     
     if (dirty)
     {
@@ -700,6 +711,8 @@
     int preview_skip_top = skip_top;
     int preview_width = raw_info.jpeg.width;
     int preview_height = raw_info.jpeg.height;
+   
+    #ifdef CONFIG_RAW_LIVEVIEW
     if (lv_dispsize > 1)
     {
         int delta_x, delta_y;
@@ -742,6 +755,7 @@
             preview_height = vram_hd.height / zoom_corr;
         }
     }
+    #endif
     
     raw_set_preview_rect(preview_skip_left, preview_skip_top, preview_width, preview_height);

diff -r af0907c14c7d src/state-object.c
--- a/src/state-object.c Thu Aug 01 14:44:07 2013 +0300
+++ b/src/state-object.c Thu Aug 01 15:32:27 2013 +0300
@@ -17,6 +17,7 @@

#ifdef CONFIG_7D
#define LV_STATE (*(struct state_object **)0x4458)
+#define SDS_FRONT3_STATE (*(struct state_object **)0x32DC)
#endif

#ifdef CONFIG_7D_MASTER
@@ -358,7 +359,7 @@
     int ans = StateTransition(self, x, input, z, t);
     int new_state = self->current_state;

-    #if defined(CONFIG_5D2) || defined(CONFIG_550D)
+    #if defined(CONFIG_5D2) || defined(CONFIG_550D) || defined(CONFIG_7D)
     // SDSf3:(0)  --  3 sdsMem1toRAWcompress-->(1)
     // SDSf3:(1)  --  3 sdsMem1toJpegDevelop-->(1)
     if (old_state == 0 && input == 3 && new_state == 1)
diff -r af0907c14c7d src/zebra.c
--- a/src/zebra.c Thu Aug 01 14:44:07 2013 +0300
+++ b/src/zebra.c Thu Aug 01 15:32:27 2013 +0300
@@ -827,6 +827,7 @@

int can_use_raw_overlays_photo()
{
+    return 1;
     // MRAW/SRAW are causing trouble, figure out why
     // RAW and RAW+JPEG are OK
     if ((pic_quality & 0xFE00FF) == (PICQ_RAW & 0xFE00FF))


I'll be away for a few hours; if you can clean it up (e.g. by copying dynamic range values from DxO, color matrix from dcraw, and tweaking the offsets), it will be very helpful.

Yes, but not on Canon cameras.

Would be very nice if you can upload some DNG files (or CR2 if you still notice aliasing artifacts).

I've tried it right now:

5D2, auto ETTR, M mode, shutter speed 2 seconds, raw zebras, raw histogram, post deflicker on or off, around 50 test shots.

Timing was perfectly consistent. The intervalometer countdown was starting from 5 seconds, so there was plenty of idle time.

Can you record a video of the camera screen that shows the issue?

Isn't it easier to choose the number of pics from ML menu?

Yes, camera- and mode-specific quirks.