Author Topic: Danne's crop_rec_4k experiments for EOS M  (Read 239980 times)

cedricp

  • New to the forum
  • *
  • Posts: 28
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #3200 on: September 14, 2022, 10:59:39 AM »
No luck for me  :( SanDisk ExtremePro 170MBs 128Gb "made in China" card here... I hope we can find a fix for those
@192Mhz -> 60.5MB/s
@240Mhz -> 20MB/s

Walter Schulz

  • Contributor
  • Hero Member
  • *****
  • Posts: 8487
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #3201 on: September 14, 2022, 11:28:50 AM »
Highly unlikely "Made in China" makes a difference. Mine are chinese, too and do work with 240 MHz.
Do you own a cardreader compatible with Sandisk's 170 MByte/s read mode?

wyup

  • New to the forum
  • *
  • Posts: 20
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #3202 on: September 14, 2022, 11:49:42 AM »
No luck for me  :( SanDisk ExtremePro 170MBs 128Gb "made in China" card here... I hope we can find a fix for those
@192Mhz -> 60.5MB/s
@240Mhz -> 20MB/s
Exact same results by me with Sandisk ExtremePro 200MBs.
EOS M, 15-45 IS STM, 64GB Sandisk Extreme Pro UHS-I 200 MB/S

Walter Schulz

  • Contributor
  • Hero Member
  • *****
  • Posts: 8487
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #3203 on: September 14, 2022, 12:00:43 PM »
R200 variety comes in 2 flavours:
R200/W140 for 256 GB and up
R200/W90 for 64 and 128 GB

Which one do you own?


cedricp

  • New to the forum
  • *
  • Posts: 28
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #3204 on: September 14, 2022, 12:03:07 PM »
I tested with :
32Gb 90MB/s -> 50GB/s @ 192 & 240MHz
64Gb V30 U3 microSD (no speed indicated) -> 40GB/s @ 192 & 240MHz
No real gain here...
Frustrating..  ::)

Walter Schulz

  • Contributor
  • Hero Member
  • *****
  • Posts: 8487
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #3205 on: September 14, 2022, 12:12:58 PM »
Please give detailed information about the card you are actually using. Your descriptions are confusing.

If you kept your blister packages: They contain Model Name like "SDSDXXD-256G-GN4IN" and AFAIK this is the only way to tell different card generations apart.

cedricp

  • New to the forum
  • *
  • Posts: 28
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #3206 on: September 14, 2022, 12:17:19 PM »
I have not kept the blister, unfortunately....

Danne

  • Developer
  • Hero Member
  • *****
  • Posts: 7392
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #3207 on: September 14, 2022, 12:19:31 PM »
You have a 128gb card. That should work with 240Mhz.

Walter Schulz

  • Contributor
  • Hero Member
  • *****
  • Posts: 8487
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #3208 on: September 14, 2022, 12:27:57 PM »
And your Extreme cards won't. And nobody claimed they should.

cedricp

  • New to the forum
  • *
  • Posts: 28
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #3209 on: September 14, 2022, 12:40:27 PM »
I know, just a test, weird thing though, the speed does not fall to 20MBs with the non pro card

ink

  • New to the forum
  • *
  • Posts: 7
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #3210 on: September 14, 2022, 01:22:26 PM »
Hi guys,

Using Magic Lantern for about a year or so. Always used it with the Sandisk Extreme Pro 128Gb SD cards (on the picture above). @Danne, Is it possible to use the latest build with the Micro SD card like this?


gabriielangel

  • Freshman
  • **
  • Posts: 86
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #3211 on: September 14, 2022, 01:48:43 PM »
Please give detailed information about the card you are actually using. Your descriptions are confusing.

If you kept your blister packages: They contain Model Name like "SDSDXXD-256G-GN4IN" and AFAIK this is the only way to tell different card generations apart.

I still have my original boxes. Both of my cards are SDSDXXY-128G-GN4IN. The only difference I could find (Visually) is the China/Malaysia difference (First doesn't work, second does, and very well).
I tried using chipgenius to see if it could expose something, but the software only works with usb sticks.
Is there a known way to expose the chip types or revisions of the cards?

BTW, the Sandisk extreme 256GB Micro SD now works @192, where it previously worked @160. 192MHz gives 2.5K 1:1 12 bit continuous and  14bit 2+ minutes on EOS M, which is still a good improvement.

Walter Schulz

  • Contributor
  • Hero Member
  • *****
  • Posts: 8487
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #3212 on: September 14, 2022, 02:17:55 PM »
Using Magic Lantern for about a year or so. Always used it with the Sandisk Extreme Pro 128Gb SD cards (on the picture above). @Danne, Is it possible to use the latest build with the Micro SD card like this?

Not Danne but I use MicroSD a lot. All tests done for Bilal happened with microSD exclusively.

ink

  • New to the forum
  • *
  • Posts: 7
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #3213 on: September 14, 2022, 02:23:31 PM »
Not Danne but I use MicroSD a lot. All tests done for Bilal happened with microSD exclusively.
So are they working fine or are there any issues with these microSDs?

Danne

  • Developer
  • Hero Member
  • *****
  • Posts: 7392
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #3214 on: September 14, 2022, 02:28:17 PM »
BTW, the Sandisk extreme 256GB Micro SD now works @192, where it previously worked @160. 192MHz gives 2.5K 1:1 12 bit continuous and  14bit 2+ minutes on EOS M, which is still a good improvement.
Good to see that there's improvement also in other ways. Bilal's code includes a few different approaches so I guess more stuff could be refined.

Walter Schulz

  • Contributor
  • Hero Member
  • *****
  • Posts: 8487
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #3215 on: September 14, 2022, 02:44:44 PM »
So are they working fine or are there any issues with these microSDs?

If you ask for any issue: Yes, if you insert an adapter without microSD card cam will stall at startup without any error message. Plain Canon firmware throws an error.

ink

  • New to the forum
  • *
  • Posts: 7
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #3216 on: September 14, 2022, 03:43:53 PM »
If you ask for any issue: Yes, if you insert an adapter without microSD card cam will stall at startup without any error message. Plain Canon firmware throws an error.
:D

I bet I can cope with that!

TrincaEspinhas

  • New to the forum
  • *
  • Posts: 5
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #3217 on: September 14, 2022, 06:13:07 PM »
For some reason with theBilalFakhouri's sd_uhs.mo I can get W 77MB/s & R 86MB/s at 192MHz but with the sd_uhs.mo included in the latest release I can't get past 20MB/s in both W&R at 192MHz.

Danne

  • Developer
  • Hero Member
  • *****
  • Posts: 7392
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #3218 on: September 14, 2022, 06:15:45 PM »
For some reason with theBilalFakhouri's sd_uhs.mo I can get W 77MB/s & R 86MB/s at 192MHz but with the sd_uhs.mo included in the latest release I can't get past 20MB/s in both W&R at 192MHz.
Interesting. Could people having issues with my version of sd_uhs.mo replace and test it with the original one here from bilal:
https://www.magiclantern.fm/forum/index.php?topic=26634.msg240128#msg240128

Danne

  • Developer
  • Hero Member
  • *****
  • Posts: 7392
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #3219 on: September 14, 2022, 06:30:40 PM »
For some reason with theBilalFakhouri's sd_uhs.mo I can get W 77MB/s & R 86MB/s at 192MHz but with the sd_uhs.mo included in the latest release I can't get past 20MB/s in both W&R at 192MHz.
@TrincaEspinhas.
Could you test latest again? I added a missing code bit in sd_uhs code. Might work better now for 192Mhz.

TrincaEspinhas

  • New to the forum
  • *
  • Posts: 5
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #3220 on: September 14, 2022, 06:39:21 PM »
Yeah, the latest sd_uhs build did the trick. It's working again. Thanks!

Danne

  • Developer
  • Hero Member
  • *****
  • Posts: 7392
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #3221 on: September 14, 2022, 06:48:05 PM »
Thanks for test and report 👍.

gabriielangel

  • Freshman
  • **
  • Posts: 86
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #3222 on: September 15, 2022, 06:23:40 AM »
Some Hi-res results with the latest build:

crop_rec_4k_mlv_snd_raw_only_2022Sep14.EOSM202.zip
Sandisk Extreme Pro 170 MB/s 128GB SD @240 MHz
only  lvface hack enabled (as opposed to  lvface+aewb)
Shutter fine-tuning set to +0.87ms when using the 2.5k preset (Value changes depending on the preset chosen. It gives roughly 1/48s shutter speed)
Histogram on while recording.

Summary:

100GB worth of clips recorded over the course of 1 hour
Only 1 clip was corrupted.

3k (3000x1256, 2.39:1 AR)

10bits @23.976fps : 1 minute +
12bits @20fps : 1 Minute +

2.8K (2880x1170 2.39:1 AR)

14bits @23.976 : 1 Minute + (But often stops before 1 minute)
12bits @23.976 : 1 Minute +
10bits @23.976 : Continuous (Green / Orange Light)

2.5k (2520x1054 2.39:1 AR)

14 bits @23.976: Continuous (Green Light and tolerates some overexposure)

The latest mods added the possibility to record 3k 10 bits @23.976fps, 2.8k 14 bits under controlled conditions; and made the 2.5k mode more robust at 14 bits.
So 2.5k is the recommended mode for moving subjects and situations where the variations in brightness cannot be controlled.


I limited the recordings to about 1 minute for the test, but it is possible to go to 3-4 minutes with proper exposure and
beyond 10 minutes by underexposing the image (But the camera will get above 50 degrees a lot faster).

Follow the method below, but once you are done exposing using the false colors,
drop the exposure so that the histogram reads 2.0 instead of 1.0

Details:

In order to get longer recording times, it is very important to properly expose the image.
I usually use the function Danne added a few builds ago to toggle the False Colors on/off with the SET button.
I also enable the histogram (and I leave it on while recording).

The false colors should be mostly Grey shades and Green for the Main Subject, with some bright Orange on the bright parts;
and only hints of yellow here and there for the highlights BUT NEVER ANY DARK ORANGE or RED!!!
Histogram should be between 1.2 and 1.0
Here is a video of a few clips I recorded using this method: https://bit.ly/3eKo5e9

I use a variable ND filter  on top of a circular polarizer. The polarizer helps taming the highlights so that you can push exposure a little more.
Even when I record 2.8K or 3K, I scale down to 2.5K for editing and export. This gives a finer-grained noise and give you a crisp picture to feed to your favourite 4K upscaler.

I posted pictures of the MLV App screen and one of the Files on disk.
Those who want to verify the actual data rates can have a look at those.
You can also see how the recordings went. For example, because it was a cloudy day, as soon as the sun became brighter while recording, the recording stopped.

If you see a clip of 1 minute or more, it means it was going well (Orange Light) and I stopped it myself.
A clip of less than 1 minute means it stopped by itself.

It is preferable to use a dummy battery and power bank. That way, I only stopped recording for a  minute after recording clips for 43 minutes. The temperature was 54 degrees C.
I tried on my second camera which has an LP-E12 battery and it gets to 52 degrees after recording only a few clips.

Thanks to Bilal and Danne, a lot of us will have even more fun :)

Screenshots:

Files-mlv-a" border="0

Files-mlv-b" border="0

Files-finder" border="0


Recording longer clips:

Long-Recordings-Trial" border="0








Danne

  • Developer
  • Hero Member
  • *****
  • Posts: 7392
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #3223 on: September 15, 2022, 06:56:47 AM »
Thanks for sharing tests.
What if you disable histogram? Do you enable both hacks as well from bilal? lvface + aewb?

roterabe

  • New to the forum
  • *
  • Posts: 4
Re: Danne's crop_rec_4k experiments for EOS M
« Reply #3224 on: September 15, 2022, 09:15:54 AM »
New to the forum guys, but just a bit of quick info from me.

The new changes do in fact stabilize 192MHz quite a bit. Running a Samsung Evo Plus 128GB MicroSD at 192MHz now results in around 52-54MB/s versus the 48-52MB/s I used to get at 160MHz and 192 previously resulted in a sudden drop of around 20MB/s.
I have yet to test video recording though. I'll probably do a test with a SanDisk Extreme Pro next month, but I'll probably take the forum's word for it.