Messages

[Question:]

Old thread but after searching, seems most relevant. Might also be useful for others in the future searching for similar things. I read through the FAQ's and the first page of duplicates + done various searches so I hope this question doesn't annoy any one.

Question: Is it possible to make ML boot only when requested? In other words, make it so the camera doesn't load ML by default; ML has to be "turned on" before every use (via a firmware upgrade I think)?

Background: Just loaded ML onto my 7D after a long while without it. Somewhat dismayed to find the "wake time" has increased from near-instant to about 2 seconds. Missed some nice shots because of it. When the camera auto-powers off, it takes about 2 seconds to wake up after pressing a button, e.g. shutter. It wasn't like this when I used ML about a year ago, but a lot seems to have changed since then. I haven't found much information about this, but it appears to be a "known issue". It is mentioned here in the first reply.
If this wake time could be reduced back to normal, I would have no need for the above question. This seems unlikely though, so it might be easier for me to keep ML disabled by default and only boot it when I need it (I do wildlife for which I need fast wake time, then the odd bit of video work for which I need ML).

This may well fall into the [IMPOSSIBLE] category. Basically, it would be nice to leave my camera out for a several hours and have it record a video when it sees movement. That is currently possible using motion detect and using half shutter as movie start. However, it isn't very reliable (works about 25% of the time?) and also there's no way to stop the video.

The question is, would it be possible for the camera to stop the video if it doesn't see movement for a given amount of time. This would require detecting motion while recording and that might count as "Real-time video processing" which is impossible. Does the normal motion detect feature use Real-time video processing, but without any recording?

There's a couple of other aspects to it, but I'll just check whether it's possible first.

Thanks,
David.

+1 to kloned's suggestion of a "min,gold,max" thing. I'd been thinking of this myself a lot and didn't realise it was in a thread already

Would a min,gold,max solution even be possible? ML would need control over ISO, shutter and aperture individually. Does it have that? If it doesn't have individual control of them, would it be possible in program mode? ML could then change the ISO, and then use the program shift and exposure compensation to change Av and Tv.

The routine would be:

Check ISO,AV,TV
If any are outside their limits, changed them to their limit and compensate with the other two settings.

ML would need to decide how much to compensate and how to share the compensation between the other settings.

Presumable ML can't actually limit the values that the Canon firmware chooses from, but can only change them. In which case, it would need to change them just before taking the picture? That could be difficult?

In the end, I learned to shoot in manual mode to solve this problem, but then you risk not getting the exposure right in time, which often happens.

Just look at exposure indicator - if it changes too much, take a picture. Keep shutter pressed halfway.

Hadn't thought of that. How precise is the exposure readout? 1/3 stop wont get some things will it? Nice idea though, that would be low power!

Well, might be possible to engage mirror lock-up, flip up the mirror and get the sensor to monitor the image for motion/processing etc, without the whole live-view/screen thing on - it's the screen which sucks up a lot of power...

As a general feature, not just 7D specific, I thoroughly agree with this. But would it be easier to go at it the other way? Use normal motion detect WITH live view, but turn off the display, and anything else that wouldn't be needed, to help battery?

[@all:]

you do not "install ML", you just run it from a FIR.
in alpha versions you never install ML.

@all:
about the alpha 3: we are currently finding out a lot of interesting stuff.
as long we are still in this process, it makes no sense to release an alpha.
we are currently enjoying finding out how to use it for cool features.

then let us make those cool new features user-safe, and you will see a new 7D/5D3 alpha that will amaze you.

Liking the sound of this!  Any clues or is it  until it's released?

Btw, I've got some issues with the motion detect in the latest build. It's same as some I mentioned before. If the sensing area is "large" it always takes pictures. And if the sensing area (at other sizes) is moved to the top or bottom, it always take pictures. I'll get a video together when I have time and post it, but those are the same symptoms I had that I posted in a previous video.

http://www.youtube.com/watch?v=pCM_nI7Fc0Y&feature=youtu.be

Aspect ratio has messed up when I uploaded to youtube which is odd. Still shows the issue though. It seems there are zones at the top and bottom, if the detection area is moved into them, it triggers (the trigger number shoots up too). When the detection area is big, it's always in those zones. That's my diagnosis anyway. The big detection area will be really useful for me if/when it works.

Thanks,
David.

Ok, well I just keep you posted, but record the figure at the time of any crash. Can it not add that to the crash log?

I'll work on the video too. Cheers,
David.

Right now: "301K + 442K". But I've just upgraded, presumably you needed the figure from the time of the crashes?

I just upgraded my firmware now.

However, during the period I had those crashes, I had lots of memory most of the time (usually >3GB). However, I did shoot some time lapses every now and then which may have filled it up until I unloaded and deleted all the images. I can wait to see if it does it again, and then record how much memory is free on the SD card, then let you know?

Is it an issue with the Canon side of things than the ML side? Like I said, none of the crashes bothered or affected me. Just off/battryout/batteryin/on worked each time.

Btw, I've got some issues with the motion detect in the latest build. It's same as some I mentioned before. If the sensing area is "large" it always takes pictures. And if the sensing area (at other sizes) is moved to the top or bottom, it always take pictures. I'll get a video together when I have time and post it, but those are the same symptoms I had that I posted in a previous video.

Easier just to hoast them on my server than create an account I won't use. Link is here: http://www.spanglefish.com/davidjamesalban/documents/ml/logs.zip

I'll take it down ina month or so. Should be enough time.
Cheers.

Just the crash logs (4) or the other "logs" as well (presumably for HDRs?). If the other .log files are for HDRs, I've had no trouble with HDR.

How do I attach them? When I click "Attachments and other options" I just get 3 checkboxes. If it's just the crash logs, I can probably just copy and post them.

I'm about to update to the latest build. Got some crash logs and some other logs (7 in total). They're from the 20th Jan build?
Do you guys want them or are they too old to be of any use? The crashes happened straight after boot. I took out the battery and started up again and all was well. Never a perminant problem.

Thanks for ML!

David

Can you (or anyone else) record a video showing bug #2? I need to see the exact timing of events.

Been busy. Just tried to re-produce this and film it, and typically, it worked fine. (This is intervalometer with bulb ramping in JPEG). For now, I suppose we'll just assume I did something wrong before, and that it works fine. I don't think I did do anything wrong, but who knows. I'll keeo a look out for it, hopefully I'll have more opportunities to use it in the coming weeks.

Good stuff as always,
David.

1. Could it be set to work out if one image is being displayed and then change it or warn if necessary? It could test whether there is a black grid is present in the displayed image, or it could simply zoom out many times then zoom in twice, which would guarantee it's displaying only one image. It would only need to do that during the first test. Subsequent tests would be the same. It's probably not the most urgent or important thing, but it would be nice for a final "perfect" version.

2. No, auto HDR seems to be unaffected. I just tried and it worked with JPG large.

Thanks

Think I've found a couple of bugs.

1. When shooting with auto-HDR. If you previously viewed images in grid view (zoomed out twice), the auto HDR routine seems to try to meter the whole 9 images, which leads it to take far too many (possibly indefinitely). To recreate: cold start, take a few pictures if the card is empty, press play to view them, zoom out twice (if not already) so you view 9 at once, then set up an auto HDR image, it will take pictures pretty much indefinitely. I need to turn off to get out of it. If you press play and zoom in twice, so as to view images one by one, then try and auto HDR image, it works.

2. I'm not sure about this one, but it seems to be repeatable. If you use bulb ramping with image quality set to JPG LARGE (smooth), it doesn't seem to change the exposure. It just takes photo after photo, even if they are completely blown out, or black. However, if I switch to shooting in just RAW, it works fine. If this isn't a bug, but is just how it's meant to be, it's very limiting, because it reduces the number of possible images in a single time laps by a factor of four. On an 8GB card, that means only about 30 seconds at 30fps.

Still loving ML and hope these things can be easily ironed out.

Thanks to the ML team

Given the bulb ramping a try and it seems to be working. Pretty dull grey weather here though, so can't test whether it is "ramping" correctly. It did the calibration though and started taking the intervaled shots, so looks promising.

The motion detection still has the image mis-aligned but that wasn't meant to be fixed yet was it? Will it planned to be fixed in the future or is it not worth it given the motion detection technically works.

Thanks, can't wait to use these things in anger
David.

Faster response than most things I've paid for, brilliant  Look forward to giving it a try and will report back

Cheers

Hi,
Sorry the video took so long - lack of compatible cameras/memory cards/lenses - ended up shooting using an old screw on 100mm f2.8 + 1.4x converter + extended + adapter. Bizzar setup :p

Anyway, video is here: http://www.youtube.com/watch?v=LFuZxTgyaDc
Again, the camera generated a log file which is bellow. Hope this is useful.
Thanks,
David.

Code Select Expand

ML ASSERT:
ABS(bramp_last_exposure_rounding_error_evx1000) < 500
at ../../src/shoot.c:6087 (hdr_shutter_release), task shoot_task
lv:0 mode:3


Magic Lantern version : v2.3.NEXT.2013Jan04.1100D105
Mercurial changeset   : 1f026f967e02 (unified) tip
Built on 2013-01-04 03:04:10 by [email protected].
Free Memory  : 346K + 322K


1. Hard to tell what a cycle is so I'll just play it by ear.

2. I used my ipod on full brightness in a dark room, moving it into view. It changes from ~4% to ~50% when the screen is under the spot. Tried it both horizontally and vertically. I'll include it in the video.

4. Don't see any "error" message on scren, just things like "checking display gain", "increasing ISO" etc. I wasn't in manual mode, though it did change to manual mode. Just trying it in manual mode and it seems to be having the same problem. Video probably needed.

Thanks for such a quick reply 

1. I'll do that but it'll take a little bit of time. How long do you want the video? Like, I said, it did whatever it was doing for over an hour. I'm sure it was doing the same things in some kind of pattern or cycle, but I wouldn't know if that pattern or cylce was a minute long, or 10.

2. Spot metering seems to be working fine even with motion detection on. I'll include a demo of that in the video.

3. Got it, shoult have check, thanks + sorry.

4. The log files were just there when I looked on the SD card. Opened them with notepad. When I was shooting: set up scene, took a few test shots, went to "shoot menu", everything off, set timelapse ramping to "Smooth ramping"/"OFF"/"OFF", turned timelapse ramping on, set intervalometer to "16s"/"0s"/"Disabled", turned intervalometer on, pressed the delete button to come out of ML and it started it's checks.

Cheers,
David.

Hi, couple of things:

First:
I tried to do a bulb-ramping timelapse today. When I set it going, it started doing the calibration stuff - It took a few pictures then tried to draw the graph thing, but soon it got stuck in a cycle of testing the display gain, changing the ISO, and the shutter speed, checking luma and stuff. It did this for ages. I've only had bulb ramping work form me once, on a previous build. However, when it did work that one time, it took ages and did something similar to this cycle of checks. So I decided to trust it and left it to do it's checks then hopefully start shooting the times lapse. However, over an hour later, it was still doing the checks and was nearly out of battery. This is the fourth time out of four tries that this has happened, though this is the only time I've left it for so long. The scene was well lit - I was shooting the sky, (ISO100, f5.6 1/200 was well exposed, though I tried other settings). The scene wasn't moving at all, it was on a tripod too. I've attached the log files from the two tries I did today. I stopped the first one after about 5mins but I let the second go for over an hour. Strangly though, both are only 1kb.

Second: Did you see my video on youtube showing the problem with the motion detection? I'm not asking for any results, just checking you saw it. I can provide more info on request, but if I'm the only one with this problem, it may not be worth it.

Third: I've notice something when viewing images. When I press the zoom button, it goes to full magnification, and when I press it again, it becomes unzoomed completely. Is there any way to turn this off? I find the standard zooming controls easier to use.

Thanks for the continued efforts. I might have a go at compiling myself, to at least see if I can understand it, though I'm not sure I understand the instructions 

Happy snapping!

Edit: I can't seem to add attachments. There is not option for it. I've copied the log files bellow:

Code Select Expand

ML ASSERT:
ABS(bramp_last_exposure_rounding_error_evx1000) < 500
at ../../src/shoot.c:6087 (hdr_shutter_release), task shoot_task
lv:0 mode:2


Magic Lantern version : v2.3.NEXT.2013Jan04.1100D105
Mercurial changeset   : 1f026f967e02 (unified) tip
Built on 2013-01-04 03:04:10 by [email protected].
Free Memory  : 345K + 311K


Code Select Expand

ML ASSERT:
ABS(bramp_last_exposure_rounding_error_evx1000) < 500
at ../../src/shoot.c:6087 (hdr_shutter_release), task shoot_task
lv:0 mode:2


Magic Lantern version : v2.3.NEXT.2013Jan04.1100D105
Mercurial changeset   : 1f026f967e02 (unified) tip
Built on 2013-01-04 03:04:10 by [email protected].
Free Memory  : 346K + 334K


I update the firmware to the lates Canon one. Then updated the ML firmware onto it, still got the same problem. Here's the video of what's going on. While shooting it, I discovered a related problem where the motion detection is triggered when the sensing square is moved to the top of bottom of the screen.
http://www.youtube.com/watch?v=1e9Z6LRX_dg&feature=youtube_gdata

hope this helps.

Hi,
Do we have a solution for the motion detection issue? Am I the only one that has the problem?

Cheers,
David.

i can also put that code into a plugin to make sure its available to those who need it and hidden to those who dont.

Sounds like a good idea. Could also include a few (3?) profiles that are considered to be good - maybe that have different good points and bad points. Average user then has a choice with those, advanced user can get the plugin and make their own profiles taylored to their needs.