550D/T2i raw video recording port official thread

[prabhath.mp]

Hi Friends,

I downloaded the latest nightly build for 550D and when I try to record in 960X540 16:9, I only get few seconds of record on my 30MBps SanDisk Pro 32GB card. Also, it indicates some bug (after it has stopped recording). I'm not sure if I have missed anything. I just replace ML and autoexec.bin to update to latest build.

Please, could anyone advise which RAW nightly build is more stable and has proven to record more frames?

Cheers,
Prabhath MP

[dafassi]

prabhath, maybe you recorded with 50/60 FPS?

nightly build means that it is possibe thats not bugfree ... this is the text in the download area for nightly builds.

Be sure your sd-card is formated an make a clean copy of the stabel version!
Then you should be able to record 960x540 16:9 @24fps

If not reply to this post and show all raw-record settings you did.

[Alex546]

I downloaded the latest nightly build for 550D and when I try to record in 960X540 16:9, I only get few seconds of record on my 30MBps SanDisk Pro 32GB card. Also, it indicates some bug (after it has stopped recording). I'm not sure if I have missed anything. I just replace ML and autoexec.bin to update to latest build.

dafassi  is right. If the fps limitation doesn't help, you could run a speed test for your sd card.
Have you done a low level format before you started recording? 
This can help sometimes to get more writing speed.






I have also done experiments with relativly low and diffuse sunlight.
I've tried to get the maximum grain out of the sensor.

[knoeterich]

Hey ho,

this is my first post and also first try in Raw filming with the canon 550d

Really much of a thanks to the magic lantern team!!!!

[prabhath.mp]

dafassi  is right. If the fps limitation doesn't help, you could run a speed test for your sd card.
Have you done a low level format before you started recording? 
This can help sometimes to get more writing speed.

Hi,

My knowledge in understanding write/read speed is limited, however, I tried the benchmark option on the firmware:-

I did a complete low-level format and tried to record RAW but still I only get few seconds of record on my 32 GB card whereas my 16 GB card can record for more seconds.

I assume that my 16GB card is better when compared to 32 GB card.

Cheers
Prabhath MP

[taapo]

Sorry if this is a newbie question, but just trying to get everything working my old 550d.

Problem is that when installing everything using eoscard, copying all files over from the "dual iso aug 25 2013 update" and setting the dial to "M" ... I can load ML perfectly, but I don't see the M tab (module) anywhere in the ML menu? -- Can anyone give me a tip on how this is not showing up? Thanks!

[glubber]

If memory serves me right, you have to enable "module loading" somewhere in that built before you cloud activate the modules.
Better try a recent nightly build for the 550D, the menu-layout is more straightahead.

http://builds.magiclantern.fm/jenkins/job/550D.109/113/artifact/platform/550D.109/magiclantern-v2.3.NEXT.2013Nov20.550D109.zip

[Alex546]

After you enabled modules, you are able to load each available module. It is recommended to load just the modules you need for your RAW record session.

Greetz, Alex546

[buffalohill]

Why not have a black and white raw option at higher resolution? Without the color info the 550D write speed might be good enough to get close to 1080p. Sure its only B/W, but for some of us that would be very cool indeed!!!

 

[NedB]

@buffalohill: the short answer is "because the camera does not have enough computing power to do any processing on the raw video before writing it out to the SD card. A conversion to monochromatic is not a matter of simply chopping off bits, and according to the devs at this point in time, even that (i.e., dropping the raw bit-depth from 14 bits to 12 or 10) is not possible." Cheers!

[a1ex]

Can you describe how to convert a raw file to monochromatic?

[dmilligan]

Can you describe how to convert a raw file to monochromatic?

Scrape the CFA off of the sensor like hardcore astrophotographers do.

Like this guy

[ItsMeLenny]

Scrape the CFA off of the sensor like hardcore astrophotographers do.

It will increase the resolution by 4x

[a1ex]

Have a comparison to prove it? (of course, against a good debayer algorithm, for example AMaZE)

[buffalohill]

Can you describe how to convert a raw file to monochromatic?

Sorry, I'm not that savvy at this level of data analysis. Is there no way to read just the data from the green pixels and discard red and blue directly off the sensor? This would already reduce the data by 50% on a Bayer grid (and either double the resolution or the crop factor). Yes, the resulting image would be orthochromatic, but that's not necessarily a bad thing (ask Anselm Adams!). Actually the orthochromaticity of the images could be a nice artistic motivator I think. If this works you could eventually even have a setting that lets the user choose which color channel to keep.

Like I say, I'm not familiar with exactly how the data is read from the sensor and how easy it would be to discard individual color channels but I do think that if there is any way of doing this, it could really distinguish these cameras as great for orthochromatic shooting.

[a1ex]

What I don't understand is how do you double the resolution by throwing away half of the pixels.

[dmilligan]

Is there no way to read just the data from the green pixels and discard red and blue directly off the sensor? This would already reduce the data by 50% on a Bayer grid (and either double the resolution or the crop factor).

No, it would actually reduce the resolution by some amount up to 50%, depending on how good a demosaicing algo you use. If you used a naive algo like 'super pixel', where each group of 4 pixels in the raw becomes one output pixel (maybe this is how you think raw is processed?), then yes you would increase resolution by 50%, but this is not how demosaicing is done.  With a good algo it's possible to recover almost all of the resolution by interpolating pixels based on their neighbor's values (e.g. a red pixel's green and blue values are interpolated from the green and blue pixels around it)

http://en.wikipedia.org/wiki/Demosaicing

It will increase the resolution by 4x

Have a comparison to prove it? (of course, against a good debayer algorithm, for example AMaZE)

It would increase resolution by up to 4x, depending on the subject matter and the quality of debayer algo.

If you were an astrophotographer taking a photo of a HII region that only emits light at one specific (red) wavelength, then yes it would increase resolution by 4x, b/c there would be absolutely no useful information in the green and blue pixels (This is why APers do this).

If you're taking a photo of a rather smooth subject without a lot of hard edges, in normal white light, then you'll be able to recover almost all the resolution with a good algo, and you'd see almost no improvement in resolution with monochrome.

[buffalohill]

No, it would actually reduce the resolution by some amount up to 50%, depending on how good a demosaicing algo you use.

Yes you're right, I was thinking in terms of super pixels. Nonetheless, the basic argument still stands I would think: Throw out the red and blue channel and you're left with half the amount of data (half the resolution, not double, as you pointed out), but due to the distribution of the green pixels on the bayer grid you should be able to interpolate the missing pixels very nicely with any number of straightforward algorithms. Obviously the interpolation would need to happen after shooting (i.e. on your PC) and not in camera. That way you could reduce the data stream substantially, therefore increase the imaged surface area and end up with a higher-res orthochromatic image.

Yes, half the pixels would be the result of interpolation, but with orthochromatic data and the checkerboard pixel distribution of the green channel I wouldn't be surprised if the actual result would be very beautiful and remember, within the same (or similar) data rate you could now double the resolution per above argument.

[a1ex]

If you were an astrophotographer taking a photo of a HII region that only emits light at one specific (red) wavelength, then yes it would increase resolution by 4x, b/c there would be absolutely no useful information in the green and blue pixels (This is why APers do this).

Correct.

If you're taking a photo of a rather smooth subject without a lot of hard edges, in normal white light, then you'll be able to recover almost all the resolution with a good algo, and you'd see almost no improvement in resolution with monochrome.

Afaik, most natural subjects (not astro) have a strong degree of inter-channel correlation, so a good debayer algorithm should recover almost all the resolution even with subjects containing fine detail, small text and similar things.

Of course, to check this, one needs a side-by-side comparison made with two cameras, one modified and one not. If you find a link or can do such comparison, with raw source images, I'm interested.

Yes, half the pixels would be the result of interpolation, but with orthochromatic data and the checkerboard pixel distribution of the green channel I wouldn't be surprised if the actual result would be very beautiful and remember, within the same (or similar) data rate you could now double the resolution per above argument.

Please prove it.

Of course, I hope you realize all this talk is just for fun (see the answer for NedB).

[Alex546]

This talk ain't just for fun.
It's the constant attempt to break the limits. :-)

[dmilligan]

Of course, to check this, one needs a side-by-side comparison made with two cameras, one modified and one not. If you find a link or can do such comparison, with raw source images, I'm interested.

I found some interesting things on the wiki page for the foveon sensor

However, the actual resolution produced by the Bayer sensor is more complicated than the count of its photosites, or its native file size, might suggest. The reason has to do with both the demosaicing and the separate anti-aliasing filter commonly used to reduce the occurrence or severity of color moiré patterns that the mosaic characteristic of the Bayer sensor produces. The effect of this filter is to blur the image output of the sensor, thus producing a lower resolution than the photosite count would seem to imply. This filter is largely unnecessary with the Foveon X3 sensor and is not used. The earliest camera with a Foveon X3 sensor, the Sigma SD9, showed visible luminance moiré patterns, but not color moiré.[24] Subsequent X3-equipped cameras have less aliasing because they include microlenses, which provide an effective anti-aliasing filter by averaging the optical signal over an area commensurate with the sample density, which is not possible in any color channel of a Bayer-type sensor. Aliasing from the Foveon X3 sensor is "far less bothersome because it's monochrome" according to Norman Koren.[25] Therefore, in theory, it is possible for a Foveon X3 sensor with the same number of photodiodes as a Bayer sensor and no separate anti-aliasing filter to attain a higher spatial resolution than that Bayer sensor. Independent tests indicate that the "10.2 MP" array of the Foveon X3 sensor (in the Sigma SD10) has a resolution similar to a 5 MP[26] or 6 MP[27] Bayer sensor, and at low ISO speed even similar to a 7.2 MP[28] Bayer sensor.

Here's a link to the relavent testing cited on the wiki page: http://www.dpreview.com/reviews/sigmasd10/page17.asp

[andy kh]

quick demonstration on work flow. u can sync audio easily. u can also cut in premiere pro cc. watch and see how easy it is

[lsg]

Hi guys, thanks for magiclantern
Not exist the possibility of connecting the camera to a laptop, tablet or other external drive and record the raw in them?
Do I have recorded the maximum resolution 24 frames with a sandisck 30\mg. would get more frames with a card most quickly?

[ib0oy]

little test with raw 

[dlrpgmsvc]

I was used to download the latest builds from MK11174 here : https://bitbucket.org/mk11174/magic-lantern/downloads/
But now it says it's a dead link ! Do we must now follow the normal nightly builds link for 550D ?
Many thanks in advance for the clarifications !  ;-)