Danne's crop_rec_4k, 5DIII

[Dmytro_ua]

Writing Speed test on a computer shows around 10 MB/s.
My main concern is about Proxy module. Does it work or nobody is using it?

[Walter Schulz]

Get a faster card.

[Dmytro_ua]

Thanks for an advice!
Although a regular Canon's video without RAW enabled, records continuously on this card.

[theBilalFakhouri]

Hi
Does Proxy work with any of the RAW presets? Even with a regular 1920p RAW it only records 5-8 seconds and then an Emergency stop appears.
SD card is selected as a preferred card. RAW goes strait to CF. The sound in RAW is corrupted when proxy is enabled.

Disable sound recording from RAW video submenu, it seems to cause problems with H.264 proxy.
I can record here 1920x1080 @ 30 FPS 14-bit lossless + H.264 proxy, recorded over 3 minutes then stopped manually.

I am using 1.2.3 firmware version, official crop_rec_4k build, SD card set to Standard card from Canon settings, RAW data goes to CF.

Note: 1.1.3 version probably has issue related to default LV RAW buffer, which might being used for both RAW data stream and H.264 video which causes early stop after few seconds.
In 1.2.3 version, a different RAW buffer is being used for RAW stream.

-Tests you can perform:
1- Make sure your SD card can record continuous H.264 video (without ML), if yes --> continue
2- Disable sound recording from RAW video submenu, record RAW video with H.264 proxy, use 1920x1080 @ 23.976 FPS at 14-bit lossless for RAW video, does it still stop early? If yes --> continue
3- Upgrade 5D3 firmware to 1.2.3, redo step 2, then feedback.

[Dmytro_ua]

theBilalFakhouri, thanks for the hint!!

Yes, it works without RAW sound.

Glad no need to buy another card 

[Leszek Wysocki]

For some reason the audio doesn't record at all or only part of the clip. Does anyone know what the problem may be?

Przykład na YT

[GullRaDriel]

Never had that problem before.
You may try to lower the video settings and or the audio output quality, try, and if it does not work use the way described by theBilalFakhouri a few posts ago

[12georgiadis]

Hi !
I checked all the hot topics and updates and the progress on ML is crazy !
One info was missing or not clear :
Can we get real time live view during raw rec in 1x3 5.7k anamorphic 1.78 AR and using 160mhz SD card overclock ? Or we really need 240mhz for that ? Does someone already made a test ? Can we imagine real-time also for 3.5k mode ?
Thanks

[Grognard]

5d mark iii:  No realtime preview exept with fullhd.
No link between Sd hack (160 mHz or 240 mhz)

Maybe, do you mean continuous recording ? It's continuous without sd hack for 5.7k anamorphic.  And it's continuous in 3.5k with 160Mhz hack (don't need 240MHz).

[12georgiadis]

5d mark iii:  No realtime preview exept with fullhd.
No link between Sd hack (160 mHz or 240 mhz)

Maybe, do you mean continuous recording ? It's continuous without sd hack for 5.7k anamorphic.  And it's continuous in 3.5k with 160Mhz hack (don't need 240MHz).

Thank you for the infos ! So we won’t get real time preview in the near future for 5dmkiii ? Or it is possible with external monitor or hdmi viewfinder ?
But we have real time preview with eosm and 70D while recording 1x3 anamorphic ? right ?

[Grognard]

Thank you for the infos ! So we won’t get real time preview in the near future for 5dmkiii ? Or it is possible with external monitor or hdmi viewfinder ?

No realtime preview with hdmi device.
Maybe We will get real time preview in the futur but nobody working on it.
Try this tip : put a Phone with a director's viewfinder application beside your camera to achieve an almost perfect realtime preview of what you record better than nothing...

But we have real time preview with eosm and 70D while recording 1x3 anamorphic ? right ?

Right on the Eos M and 700D (not 70d sorry) But only in 1440x1806 not with the new full anamorphic resolution (1736x 2178).

[12georgiadis]

No realtime preview with hdmi device.
Maybe We will get real time preview in the futur but nobody working on it.
Try this tip : put a Phone with a director's viewfinder application beside your camera to achieve an almost perfect realtime preview of what you record better than nothing...

Right on the Eos M and 700D (not 70d sorry) But only in 1440x1806 not with the new full anamorphic resolution (1736x 2178).

Thank you for the tips ! Will try smartphone viewfinder !

[GullRaDriel]

5DIII: To be precise there are some mode with realtime preview. Most anamorphic modes are giving a real time preview, even if distorted.

[Grognard]

5DIII: To be precise there are some mode with realtime preview. Most anamorphic modes are giving a real time preview, even if distorted.

Yes. Realtime previews.... but distorted (anamorphic) or cropped (3k, 3,5k).

[Wlad81]

From my experience, it's easy to get continuous recording.

card spanning, 14 bit lossless selected and also 10 bits enabled on top, in 3.5K 1:1 at the maximum resolution (for me 2:20 ratio and 3584x1630)

It works.

[has_abandonment_issu]

I don't know if this has already been reported, but I'm getting a massive difference in write speed on the april 07 3.123 build that bilal linked vs the may 15 3.123 build.

This is present in the 1080p 60fps preset. Clean install on both. Global draw off, card spanning off, small hacks and aewb+lvface on both builds.

I get around 60mbs without any hacks on for both builds, but 88mbs on the april build with hacks on and only 66mbs with the may build. Is the write speed indicator reading correctly? I haven't tested aggressively, but I've shot the same scene with both and there is a clear difference in write speed, even if maybe the indicator isn't accurate.

Is this happening for anyone else?

[rinski]

Full anamorphic 18 fps 240 MHz? 5d3

[dpjpandone]

Wrong thread, please delete

[names_are_hard]

I think if you do this that it won't work on older cams:

Code: [Select]

case 2:
            W = 640;

Very old cams I think don't have that high a resolution, 320x240 maybe?

[dpjpandone]

I just realized i posted this in the 5D3 thread,, I'm sorry guys this build is for eosm, please disregard

[vastunghia]

I don't know if this has already been reported, but I'm getting a massive difference in write speed on the april 07 3.123 build that bilal linked vs the may 15 3.123 build.

I think this is an important point that should deserve some attention. I myself have been reporting huge speeds with the 2022Apr07 build from Bilal (see here), which seemed to be unexpected.

I’ve always been on 2022Apr07 1.1.3 and I hoped I could find the time to test the 2022May15 1.1.3 build, but that didn’t happen

Now I would like to switch to 1.2.3 for testing of an external HDMI monitor, so I decided that I should give it a try to both 2022May15 and 2022Apr07 builds for 1.2.3.

Unfortunately, it seems that Bilal has pulled his 2022Apr07 build from here. Anyone knows where I can find it, or is kind enough to send it to me?

And: should the difference in write speed be confirmed by more users, would it be possible to single out the differences between these two builds? A kind request for Danne / Bilal of course 

Thank you

Sergio

[theBilalFakhouri]

Unfortunately, it seems that Bilal has pulled his 2022Apr07 build from here. Anyone knows where I can find it, or is kind enough to send it to me?

The 2022Apr07 builds are still in my Bitbucket download page, look for:

crop_rec_4k_mlv_snd_isogain_1x3_presets_2022Apr07.5D3113.zip
crop_rec_4k_mlv_snd_isogain_1x3_presets_2022Apr07.5D3123.zip


These builds are based on Danne's 5D3 source code, iirc last commit from Danne was db5dc42 when I created the builds, I have only implemented the new hacks in mlv_lite.c, I didn't change anything else.

More likely Danne implementation for new hacks has small bug which make them don't work in all presets, iirc I made some tests and it was from this line of code:

Code: [Select]

if (more_hacks && shamem_read(0xc0f383d4) != 0x4f0010) /* excludes mcm mode on eosm */
Removing:

Code: [Select]

&& shamem_read(0xc0f383d4) != 0x4f0010 might solve the problem.


Edit before posting the reply:

Ahhh, wait, there is a bug, in this commit, the function hack_liveview_more() is never being called, which make the new hacks don't work at all in 15 may build.

hack_liveview_more(); should be added in RAW recording initialization code, e.g. after these lines:

Code: [Select]

        hack_liveview(0);
        liveview_hacked = 1;

[vastunghia]

Thanks Bilal, I knew I could count on you!

No testing required then, a lot of time saved, thanks a lot.

Will stick to the 2022Apr07 build until bug is fixed in Danne’s one.

S

[Danne]

You mean adding:

Code: [Select]

        hack_liveview(0);
        liveview_hacked = 1;
        hack_liveview_more();
Did it now and uploaded new builds. Please test and report if fixing any errors.

[vastunghia]

Thanks Danne,

tested build Dec02 for 1.2.3, seems ok but I could achieve continuous rec. @max 133 MB/s (binary prefixes).

With Apr07 build for 1.1.3 I could reach 147 MB/s (see here).

I guess I should be testing also Dec02 build for 1.1.3? Just to rule out some weird 1.1.3 vs 1.2.3 difference. Let's see if I find the time...

Sergio