Is the EOS-M *THE* Digital Super-8 Camera?

[byuri734]

Great. Any application of this for sound recording as well? My understanding is that sound does not work with lossless compressed, right? Perhaps with higher speeds one can record uncompressed with sound.

[Alpicat]

12 hours ago, byuri734 said:

Great. Any application of this for sound recording as well? My understanding is that sound does not work with lossless compressed, right? Perhaps with higher speeds one can record uncompressed with sound.

I think someone posted a build with the mlv_lite module able to record sound a while back even before the hack was available. It's in the magic lantern forum somewhere. I'm sure it should be possible.

[tupp]

On 5/12/2018 at 11:19 AM, Alpicat said:

I've uploaded another video with the SD card hack:

Great footage (and music)!  Thanks!  

 

On 5/12/2018 at 11:19 AM, Alpicat said:

It takes about 3 or 4 minutes to do the two tests before the sd overclock is ready to use, and then the camera runs as normal.

Interesting.  The tests are required every time one starts the camera?

 

On 5/12/2018 at 11:19 AM, Alpicat said:

You should be able to shoot 1920x1080 in 5x zoom mode without overclocking in 10 bit lossless - you won't get much record time but it's feasible. I can already get nearly continuous shooting in 1800x1030 12 bit lossless movie crop mode.

1800x1030 works for me.  What's the longest run time you've achieved with that mode?

 

On 5/12/2018 at 11:19 AM, Alpicat said:

Also, unsure if I mentioned that Viltrox are planning to do a 0.71x focal reducer for EF-M mount - they told me it should be out in the summer. I don't have a bmpcc speedbooster any longer so this should be a cheaper alternative, with the advantage of electronic contacts for EF lenses.

I've heard some good things about the Viltrox focal reducers.  I have the RJ focal reducer for the EOSM with a Nikkor mount, and it was used in a couple of the shots in this test (along with the 18mm-55mm kit lens):

 

 

 

 

 

 

All of this footage employs the All-I frames video hack in a Magic Lantern build from 2015.  Frame rates were 23.98 fps and 59.94fps (yielding 1280x720 which was scaled up to 1920x1080).  The Flaat 10 picture style was used in all shots with the first shot ungraded, but I gave a touch more snap to the contrast of the rest of the clips.

 

The ISO was set to 800, but noise was prominent in some footage.  So, I'll probably dial the ISO back down to 100 or 200 when possible.

 

The bit rate was boosted to 1.5x, but I got a hiccup (dropped frame or doubled frames) four seconds into one of the shots.  So, regardless of the cause, I will probably back off of that setting and bring the bit rate down to 1.3x in subsequent shoots.

 

A screw-in ND and a polarizer were stacked, so. occasionally, there was slight vignetting.  However, there was some weirdness with the vignetting moving/changing in the middle of shots, which could be attributed to OIS, except that same behavior seemed to appear in the shots with the manual Nikkor 20mm with the RJ focal reducer.   Have to take a closer look at this apparent phenomenon.  Anyone have ideas on the cause?

 

This test was rushed, and I forgot to change the shutter speed from 1/60th to 1/120th in the first 60fps (slo-mo) shot, so that clip looks slightly smeary.

 

I see a little moire/aliasing and a moment of banding, but I think that this imaging quality could work for a lot of situations.

[Alpicat]

@tupp thanks for posting that video - All-I frames video looks like a great option when recording with the full sensor!

As for 1800x1030 resolution (without the sd card hack), with 12 bit lossless you can record for a long time if there's not too much contrast or highlights in a frame. 10 bit lossless is continuous

There's some slow progress being made with the sd card hack, it's now possible to use it without having to run tests (which took around 3 minutes every time the camera was turned on), so it's quick to start up now. However that build only works with certain fast cards, and it's still not safe to use. If you try that build with an unsupported card, I don't know what would happen to it.

Below is my latest test, this time with people! I'm using a Sandisk extreme pro 95mb/s 64gb card I'm still getting a max of 55mb/s write speed. At 2520x1080 resolution in 12 bit lossless, recording runs for quite a while unless there's a lot of highlights and contrast in a frame. 10 bit lossless is usually completely continuous. I shot this when it was 26 degrees celcius outside, but the camera worked - it didn't go above 42 degrees C.

Also I've now bought a Canon efs 10-18mm lens. Waiting to get a viltrox 0.71x speedbooster when it's released. That speedbooster combined with this lens at 2.5k resolution in crop mode will result in a field of view of 24mm (35mm full frame equivalent). That's wide enough for me and will make crop mode totally usable.... and that lens has IS which works nicely.

 

 

[tupp]

On 6/2/2018 at 10:18 AM, Alpicat said:

As for 1800x1030 resolution (without the sd card hack), with 12 bit lossless you can record for a long time if there's not too much contrast or highlights in a frame. 10 bit lossless is continuous

That's good to know.  Isn't there some bit depth mode of 9-11 bits?  If so, I wonder if that would give a better, continuous image than 10 bit.

 

 

On 6/2/2018 at 10:18 AM, Alpicat said:

There's some slow progress being made with the sd card hack, it's now possible to use it without having to run tests (which took around 3 minutes every time the camera was turned on), so it's quick to start up now. However that build only works with certain fast cards, and it's still not safe to use. If you try that build with an unsupported card, I don't know what would happen to it.

I actually read about this in the ML thread.  I assume that a Sandisk Extreme Pro works without running the tests.

 

By the way, I was at the Cinegear trade show two days ago, and both Sandisk and Kingston had a booth.  I told both of them that ML folks are starting to overclock cards, significantly increasing the write speeds.  I suggested that they should consider designing cards meant for overclocking.  Neither company had heard of ML nor card overclocking, but they seemed interested.

 

 

On 6/2/2018 at 10:18 AM, Alpicat said:

Below is my latest test, this time with people!

It looks great!  Thanks for all of your tests!  Seeing people in the frame definitely helps.  The dynamic range looks wide, but it seems like there might have been open sky (and/or a white building) behind the camera.

 

 

On 6/2/2018 at 10:18 AM, Alpicat said:

I'm using a Sandisk extreme pro 95mb/s 64gb card

So, you format this card in the camera with an exfat filesystem?  Do you install ML with this 64GB, exfat card to give the card the boot flag, or do you use some other program to do so?

 

Thanks!

[Alpicat]

@tupp the lossless bitrate menu settings seem to keep changing with each build I install - sometimes I see 10 bit lossless in the menu, and other times 11 bit. I don't know why but in any case I've never been able to see a difference. 

Danne's newest test builds which bypass the overclock tests are meant for 95mb/s Sandisk Extreme Pro cards which is what most people seem to be using. It's possible it works on other brand cards but I don't know.

Thanks for talking to the card manufacturers, sounds like a great idea! What would be nice is if they could increase the write speeds for UHS-I SD cards (of course there's UHS-II, but that's no good on an EOS camera) - I'll write to Sandisk. I wonder if it's the card that's being overclocked or the camera's internal card read/write interface? 

Yes in that video there was just the open sky in the background which was blown out. But using Juan Melara's cdng grading technique the highlight rolloff looks more pleasant (see here: http://juanmelara.com.au/blog/a-better-way-to-grade-ursa-mini-cinemadngs ) 

My card already came formatted to exfat since it's 64gb size (I think the 32gb ones are fat32). When I put ML in this card I just copied pasted the files on there after formatting the card in camera I think, didn't do anything special. You can format the card to fat32 but I haven't seen any improvements in max write speeds when doing so. 

 

 

[Alpicat]

Here's some more footage shooting 2.5k raw video:

 

[tupp]

@Alpicat

Thanks for the videos and for the link to Juan Melara's page!

[Alpicat]

The 700D, 100D and EOS M are now able to shoot around 2520x1320 in crop mode (previously 1080 was the max vertical resolution). 3k is also possible although liveview doesn't really work at that resolution.

Record times are very short as it's only possible to shoot in 14 bit lossless for now. Here's my first successful test with the EOS M recording at 2520x1320 and 2336x1314:

 

[tupp]

@Alpicat

Thanks for all the updates and videos!

 

At some point, I hope that someone will create a simple tutorial demonstrating how to configure the EOSM with ML to enable these capabilities.

[PannySVHS]

Hey guys, another massive contender for THE digital super 8 camera would be... the Panasonic DVX100 Andromeda! 720p 444 10bit, 3 ccd magic!

See this footage and info from this thread. Would love, if this could be revived:

 

[Alpicat]

1 hour ago, tupp said:

@Alpicat

Thanks for all the updates and videos!

At some point, I hope that someone will create a simple tutorial demonstrating how to configure the EOSM with ML to enable these capabilities.

Thanks! Hope I can do a tutorial someday. Right now the SD overclock hack is simple enough to use, at least compared to the next phase of development that is being carried out to increase the resolution beyond 2520x1080, which is an extremely complex process.

We're hoping it will be possible to overclock the SD card interface even further at some point. With the current overclock hack the write interface runs at 160 mhz (up from 100 mhz natively used by Canon). If the developers find a way of getting the frequency up to 208 mhz (the max UHS-I cards can achieve), then the camera's write speed could increase up to 80mb/s apparently (as opposed to 55mb/s with the current hack). That would make 2520x1320 recording feasible, possibly even 3k recording. 

In fact I just saw that a new build has been posted enabling 4k recording for the EOS M and 100D. Will test tomorrow but doubt it'll be possible to record longer than a few frames! 

Here's another really quick test recording 2520x1320 and 2336x1314:

 

@PannySVHS that looks very interesting and would be great if someone could look into reviving that hack. The colours look amazing, must be thanks to that CCD sensor! Unfortunately I have no coding knowledge so wouldn't know how to help. 

Magic Lantern development on the EOS M has progressed quickly since this thread started and now in crop mode it's become a 16mm camera (thanks to an increase in resolution capabilities) - so the title of this thread is slightly out of date. Maybe I should start a new thread! Of course it's still possible to shoot super 8 at 720p without using any of the experimental builds.

[PannySVHS]

@Alpicat, three of them ccd sensors even, yeah! global shutter included! Unfortunately only a few enthusiasts on this forum are interested in my thread above , which I find astonishing. I love this eos m super 8 thread though! Still, a bit disappointed. Well, no biggie. cheers :)

[Alpicat]

12 hours ago, PannySVHS said:

@Alpicat, three of them ccd sensors even, yeah! global shutter included! Unfortunately only a few enthusiasts on this forum are interested in my thread above , which I find astonishing. I love this eos m super 8 thread though! Still, a bit disappointed. Well, no biggie. cheers

It would be amazing to get the dvx100 hack working. It looks like they were connecting a laptop to the camera and running it from there? I just wouldn't have a clue how to do it!

Here's 4k filming with the EOS M at 9fps (4096x2552), it's only possible to shoot 5 frames at a time though. At this resolution, the crop factor becomes 2x (micro four thirds). I don't think it'll ever be feasible to get this to work at 24fps. Even with the extended SD overclock hack writing at 80mb/s, it would only be possible to record a second or two, the data rate is just too high. Don't know if it would be possible to hack this camera with an external recorder and increase the write speed that way, but that's just too much work!

 

[PannySVHS]

@Alpicat, yeah, it was a hardware hack, making it possible to record to a computer, such as a mac mini. If one looks for DVX100 Andromeda, it tells the whole story about a few happy campers having purchased the enhancement and filming with a 720p+ resolution 3CCD raw camera with global shutter, massive color, nice resolution image, in the beloved form factor of the legendary DVX100! It would be the perfect Digital Bolex substitute. Love this thread! Super fun read, with beautiful footage!

[tupp]

On 6/21/2018 at 12:50 PM, Alpicat said:

It looks like they were connecting a laptop to the camera and running it from there? I just wouldn't have a clue how to do it!

I think that they tapped into the signal coming out of the A-to-D converter.  Doing so is a little tricky as one must:

1. find the part of the circuit to tap;
2. bring that raw digital signal up to line level/voltage for a recorder;
3. work with extremely miniaturized components.

 

Over the last year, I have considered trying such a hardware hack with my EOSM, but the skills required for steps #1 and #3 are beyond me.  Here are two videos showing the EOSM innards:

 

 

 

 

 

 

 

 

 

 

Don't know if it would be feasible to start a business making such hardware mods on EOSMs, as I can only guess at the demand/quantities involved.

 

However, such a hardware hack could work just as easily on the EOSM-5, EOSM-6 and EOSM-50 as it would on the original EOSM, which could transform these more recent models into raw, mirrorless monsters.

[Alpicat]

@tupp very interesting videos, thanks for sharing! If it were simply possible to swap out the UHS-I SD card interface inside the EOS M for a UHS-II one, that would make higher resolution recording feasible. But I imagine the amount of coding required to make it work might make it very difficult. 

In any case I doubt you'd be able to sell and make any money from a hardware hack like this - unless you get permission from Canon!

I spoke to a Canon rep at the media production show in London recently where they were showing off their cinema EOS cameras, and he was aware of the EOS M 2.5k hack and seemed totally positive about it even saying he'd like to try it himself. I showed him the hack in operation. He also said he hoped that magic lantern would be available for the M50 one day! 

[PannySVHS]

On 6/24/2018 at 11:56 AM, Alpicat said:

@tupp very interesting videos, thanks for sI showed him the hack in operation. He also said he hoped that magic lantern would be available for the M50 one day! 

That would be really something! One of my all time favorite threads since I´ve been posting and regularily reading here! Thanks guys and Alpicat, keep it going! Though the DVX super8 global shutter raw monster would have been really awesome. Would love a ccd raw camera in that formfactor and stabilized zoom lens!

[tupp]

The latest development with ML and the EOSM is that one can shoot raw with the full height of the sensor -- and it looks fairly clean (minimal moire/aliasing).  The resolution is 1736x1120, and I think that the pixels are binned 3x3.

 

Here's @Alpicat's test at 14-bit lossless raw:

 

Here are the instructions on how to setup the EOSM in this new crop/resolution.

[Alpicat]

Thanks @tupp. I guess this update makes the title of this thread slightly out of date now!

Here's the build I used: https://bitbucket.org/Dannephoto/magic-lantern/downloads/magiclantern-Nightly.2018Jul03.EOSM202_mv1080p.zip   With this build you're stuck with shooting full sensor only (3x3 pixel binning), you can't switch to 5x zoom mode. They may have partially fixed that in the newest build but unsure if you can then access the full sensor height, I'll need to test. See here: https://www.magiclantern.fm/forum/index.php?topic=19300.msg204564#msg204564  

I shot that footage at iso 100 and iso 400, f3.5 and above. I later noticed that if I export to prores on MLVapp I get less noise in the shadows that exporting to cdng (which can look really excessive if you push up the shadows). However I don't know if I tested this fairly and will have to experiment further. The above video is cdng.

Right now the main issue is liveview preview is streched vertically 1.6x (preview looks like you have an anamorphic lens attached) and you can only see the top half of the image. It's still usable though and half-pressing the shutter gives you a low fps / pixelated preview of the whole frame.

The moire/aliaising can get fairly bad depending on what you're filming, and can become very obvious if cropping in on an image. I guess it's now the same amount as you get with all the other aps-c Canon DSLRs, or it's similar to the bmpcc. Exporting to prores rather than cdng on mlvapp can sometimes reduce the amount of moire - I need to test this further though.