Canon 600D / T3i

[dfort]

@bpv5P - Yes, it is strange that we're directed to Pelican's website for old firmware versions but he's got the most complete collection. Canon takes down their old updates from the official support pages. I was on a mission to get all the ML supported cameras ported to the most recent Canon firmware updates but that seems to be a very low priority project. It isn't all that difficult and a good project when you get past compiling, which is very easy to do these days.

If you want to try to port a firmware update:

http://www.magiclantern.fm/forum/index.php?topic=19417.0

Easiest way to setup a Mac development environment:

http://www.magiclantern.fm/forum/index.php?topic=16012.0

Cygwin on Windows is also easy:

http://www.magiclantern.fm/forum/index.php?topic=15894.0

Though a more complete solution is to do it in the new Windows 10 Ubuntu environment:

http://www.magiclantern.fm/forum/index.php?topic=20214.0

[Walter Schulz]

Most of the time I simply ignore meta discussions and ad hominem. Let's make an exception here:
@bpv5P: I have to say that I'm not happy with your view on user pelican who is around us for more than 5 years. He not only offers a quite complete archive for firmware taken from Canon's web. He developed EOScard utility which for a very long time was the most convienient way to get ML running on cards >32 GB for Windows users. And it was standard program for getting ML v2.3 installed for Windows users. For free, without nagging ads.

Instead of going into sinister plots a regular may came up with (You're aware this kind of argument does apply to the devs, contributors, anyone offering a link, too?) your post would might get a warmer welcome (at least by me) sticking to the very risks involved in non-redundant structures and unsecure web hosting. I'm fine with pointing at pelican's householding duties. I'm fine with all ideas trying to get rid of single points of failure.

BTW: There are some prominent examples where big web based services failed, their user databases got public or - like recent - got unusable by changing terms of use.

[bpv5P]

@bpv5P - Yes, it is strange that we're directed to Pelican's website for old firmware versions but he's got the most complete collection.

That's the point, why not mirror the files?

@bpv5P: I have to say that I'm not happy with your view on user pelican who is around us for more than 5 years.

I don't have a bad view of "pelican" specifically, that's just my default mode.

You're aware this kind of argument does apply to the devs, contributors, anyone offering a link, too?

Yes, but the code is open source, the website don't runs too many nasty things as the pelican website and it's not distributing a compressed file with unknow binary. All ML builds has the code, changelog and (now) tests (on QEMU). We can compile and check for ourselves. There's a big difference between the two cases.

I'm fine with all ideas trying to get rid of single points of failure.

Decentralization without security/safety is for dummies.
You're trying to apply a good concept with the very wrong implementation.

BTW: There are some prominent examples where big web based services failed, their user databases got public or - like recent - got unusable by changing terms of use.

This is a admin issue. Database should always be encrypted. (edit: and a backup of this database)

[dfort]

why not mirror the files?

It is one thing if an individual does this and another thing when an open source project hosts these files which are closed source and contain Canon's copyrighted intellectual property. It is probably best not to go there.

I can easily demonstrate that the audio meters do not engage until you hit record or enter the ML menus and leave, playback a video, etc. I can repeat it each time I turn my camera off to on. After you do that you can see levels while non-recording which is what I want to do (want to nail settings before starting to record an interview for example).
...

The video you posted does indeed show the bug.

I recently spent some time straighten out a hack that used the address for StopASIFDMAADC in SoundDevShutDownIn on several cameras. It turned out a bit more complicated that I anticipated.

Maybe the bug you're exeriencing is related to this. The 600D doesn't have StopASIFDMAADC in stubs.S but I just looked at a disassembly and the camera does in fact have it. This is needed on for the audio meters to work properly on some cameras and maybe the 600D is one of them. On my Bitbucket downloads page you'll find a test build named StopASIFDMAADC.2017Oct28.600D102.zip. Please give it a try and report back. Of course other 600D users are also welcome to join in the test. Make sure you understand how to reproduce the bug that is clearly illustrated in the video.

https://www.dropbox.com/s/fxmjruhdixcaerp/C0006.MP4?dl=0

[dfort]

Seems I overlooked an important detail on rsmith02's video, the camera is recording H.264 and not mlv_rec/mlv_snd.

Ok--so there's something else that might fix this issue.

src/audio-common.c

Code: [Select]

    /* some models require the audio to be enabled using audio_configure() */
    #if defined(CONFIG_650D) || defined(CONFIG_700D) || defined(CONFIG_EOSM)
    int reconfig_audio = 0; // Needed to turn on Audio IC at boot, maybe neeed for 100D
    #else
    int reconfig_audio = 1;
    #endif

Let's see if the 600D needs this. Test build "turn_on_Audio_IC_at_boot.2017Oct29.600D102.zip" on my Bitbucket downloads page.

I'd still like to get feedback on the test build mentioned on my previous post. Start with the current nightly build, turn on only mlv_rec and mlv_snd. Check to see if the audio meters are working. Record a short clip, check the meters. Do this at least 4 times and see if the audio meters continue to function. On the 650D and 700D there was a problem where the audio meters stopped working after recording a clip. (Sort of like the opposite of the H.264 bug on the 600D.) If there is a problem with the current nightly build try the "StopASIFDMAADC.2017Oct28.600D102.zip" test build that I posted.

By the way, I found several of the missing ASIF stubs for the 600D. I created a dummy pull request if anyone wants to follow along:
https://bitbucket.org/daniel_fort/magic-lantern/pull-requests/10/600d-audio-meters-issue/diff

[rsmith02]

Hi dfort, sorry to not clarify but yes I was just using the standard H.264 camera features. The last time I tried ML raw was before there was sound support ; )

So I downloaded an installed the fix file and am happy to report audio meters now work when you go into video mode. I switched into the mode from photo modes and from power off and it works fine. I recorded 5 clips in a row (few seconds each) and the meters stayed live. If I notice anything odd during my course of normal use I will report back but I think this fixes it. Is there anything else in H.264 mode I can test? (not sure what the other stubs are for- happy to do tests with a mic, etc. if it is useful).

[rsmith02]

I attempted to help troubleshoot ML RAW but do not know enough about how it it supposed to work to be of much help.

So I used the last standard nightly build (not a dfort fixed file), loaded the raw and sound modules, enabled raw video in the menu, enabled sound. I set resolution to 960x540 16x9. I turned global draw on.

In live view there is a preview window in the center surrounded by a black box (because this is a crop mode?) While there are meters at the top they are not active while not recording.

I hit record, it says "busy ... please wait". A counter runs of seconds elapsed. The video cam icon is yellow. I'm recording at about 20mb/s. The audio meters are gone as are the histogram and everything else. If I try to change aperture, etc. it just says busy. After 2 minutes I give up, hit the record button to turn it off and now audio meters are registering. I realized 16x9 may not be continuous so I changed the aspect ratio, tried again and also only have meters after I exit recording. Aperture can't be changed using physical controls while recording though ISO can (after a delay). It says busy and is barely responsive.
Second test and again no meters after power on.

I swapped cards to the one with the "turn on" H.264 fix I reported on earlier. I did the same procedure to enable raw and MLV sound. Now there are meters before recording! Now when I record I have audio meters up top but also the full magic lantern display (histogram, settings at bottom). I can also change aperture while recording and the black mask stays active during recording. It no longer says "busy" and is responsive to starting and stopping movie recording now.

I have not tested the stop build as the turn on build seems to be what the camera needs?

[Walter Schulz]

Movie tab -> RAW video (MLV) -> [Q] -> Global Draw ON
Off by default ...
And there is a third option to surpress GD in Display tab -> Clear overlays. Off by default, so this one should be fine.

[rsmith02]

Thanks Walter. I think I must have accidentally had GD off the first time I tested the 10.4 build. With it on I have meters/full functionality in this build, too.

Re: sound, there are no meters in liveview on initial power on with MLV raw/MLV sound modules enabled. When you record a clip they wake up. When you stop recording the meters work in standby again (same behavior as H.264). I recorded 5 raw clips and upon return to standby mode each time the meters stayed live. They worked while recording.

I think the "turn on" fix fixes audio meters for both raw and H.264 with the 600D.

[dfort]

Thanks for the testing @rsmith02 - made a pull request and posted a new test build on my Bitbucket downloads page. Please make sure it still works and feel free to approve and/or comment on the pull request.

[rsmith02]

Hello,  tested the latest version and it worked fine with H.264 and RAW. I got meters before, during and after recordings with both.

Forgive my ignorance but as I didn't see how to comment on a pull request so I wasn't able to. A1ex is listed as the reviewer.
To date I've just been a ML user/forum commenter.

[Bob Hayes]

My Canon 600 is firmware version 1.0.3 but magic Lantern instructions for installation  says I need 1.0.2. Is this true  and do I need to go back to this earlier version? Also will I destroy the installation if at a later date I go to 1.0.3 or 1.0.4 if a later version is brought out? I need to know this for re-assurance before I proceed. Thanks.

[dfort]

No worries, the changes were merged and will soon be available on the nightly build. I'll take my test build down when I see a new one is available on the download page.

[rsmith02]

Excellent, that's good to hear.

I'll be away from my camera for a few weeks but would be happy to test other features for the 600D when I return.
Overall ML for the 600D is stable and mature so maybe there's no need, but it is a standing offer as ML has helped me so much.

[Its]

Can someone please give me a little advice on audio settings...
When using the internal mic, whats the wisdom on getting the best balance between Analogue Gain and Mic Boost for a given sound level?

UPDATE:
After a little trial and error here's the gain structure I've settled on.
Obviously avoid any 'Digital Gain'....
Analog Gain - around +6db
Mic Boost - around +20db
A decent enough noise floor without being undersensitive with the built in mic, works ok for a mid-level wireless pack system too (Boya WM6) cranked up to level 29 (of 32) on both the body pack transmitter and the receiver also.
Lav mic placed at mid chest position.
Wind Filter - DC = On   High Pass = On   HPF2 Cutoff Hz = 130  takes care of breathy bursts ok for me.

Headphone monitoring works fine, best to use a headphone amp such as something portable in the Fiio range.

[xaled]

Hi guys,

I can not get my 600d with san disk ultra 64GB 80MB/s exfat and latest experimental build to run continuously in 10 bit raw, 1280 @ 2:35 / 24fps mode with sound. Recording stops after 10-12 seconds.

Is there something special I should be aware of?

Thanks.

[bpv5P]

Hi guys,

I can not get my 600d with san disk ultra 64GB 80MB/s exfat and latest experimental build to run continuously in 10 bit raw, 1280 @ 2:35 / 24fps mode with sound. Recording stops after 10-12 seconds.

Is there something special I should be aware of?

Thanks.

The maximum speed of 600D is about 25MB/s. That's the bus interface limitation. You can test your card using the benchmark tool inside ML.
You can probably stretch some more seconds disabling Global Draw (using HacKeD preview on MLV menu) and disabling sound rec (use external recorder).

[xaled]

The maximum speed of 600D is about 25MB/s. That's the bus interface limitation. You can test your card using the benchmark tool inside ML.
You can probably stretch some more seconds disabling Global Draw (using HacKeD preview on MLV menu) and disabling sound rec (use external recorder).

thanks for the benchmark tool tip!

I was under impression from this posts:
https://www.magiclantern.fm/forum/index.php?topic=15360.msg178287#msg178287
https://www.magiclantern.fm/forum/index.php?topic=15360.msg176951#msg176951
that 600D can do more then I'm getting now.

So what is the highest continuous resolution you get from 600D with 10bits?

[rsmith02]

What resolution are you using exactly? Doesn't ML estimate the number of frames you'll get at different resolutions?

For me, I wasn't getting write speeds much over 20mb/s so do the math to see what resolution and bit depth will keep you under that.

[bpv5P]

So what is the highest continuous resolution you get from 600D with 10bits?

Sorry for the late reply...
On a SDXC Toshiba Exceria card, the speed is about 19.5 MB/s (according to the quick benchmark - global draw off).
So, the maximum resolution for continuos recording, in 16:9, is 960x540 10bpp.
On 2.50:1 aspect ratio, it's possible to record 1280x512 10bpp continuous.
I don't advice you to record above 640 ISO. Try to stay bellow 400.
I did a comercial project using MLV on 600D, it went not so well, although the I could finish the work and the cliend liked it.
If you want to know how MLV perform, compared to x264, on 600D I would say: try it yourself. Sometimes you'll get better results from a upscaled MLV, other times (especially at night) you will get very bad results.

If you want seriously experiment with MLV, try some camera with CompactFlash support (50D, for example).

[IDA_ML]

If you want seriously experiment with MLV, try some camera with CompactFlash support (50D, for example).

I strongly recommend the EOS 7D for serious MLV work.  This is a solidly built machine, very reliable and can be obtained at a very reasonable price used.  At 14-bit and 1728x972 resolution, MLV quality with sound is fantastic (continuous recording), also at 10-bit with sound in the 5x crop mode where resolutions of 2496x1200 are possible!  Due to the fast CF card write speed, (mine writes at 90 MB/s) and the excellent battery life this body is second best after the 5DMkIII for ML work.

[David_Hugh]

@ xaled

IDA_ML is right in saying that the 7D is perhaps the best option in terms of price/versatility, the fast cf write speed is really what sets this cam apart.

However, (at least without sound, not sure about sound recording) 1280 x 544 @ 2.35:1 is possible on a 600D, continous recording in 10bit. I shot this music video

completely on a 600D, at the resolution and ratio mentioned above.

Your problem is the card. The Sandisk Ultra does 80Mbit/s READ speed, but not WRITE. Both of my golden Sandisk Extreme can do 1280x544 continous. This, however, is the absolut max on a 600D at the moment.

Hope that helps,
Dave

[stevedraw]

Hey guys,

Have a slightly older build of ML on my 600D and was wondering if the continuous video shooting (ie start new recording when reaching gb limit/over heating) still has one second gap in between files?

I shot a gig with my 600D and 700D for over 30 mins and noticed the 700D has no gap between recordings, which I was amazed by! So just wondering if that has been looked at for 600D?

I say it every time but the work you guys do is phenomenal. Thanks again!

[ilias.2008]

Hi, i have firmware 1.0.3.
Is there a way to downgrade, so I can use Magic Lantern?

[Walter Schulz]

Use https://bitbucket.org/daniel_fort/magic-lantern/downloads/ until this problem is solved.