Author Topic: Bilal's crop_rec_4k experiments for 100D / SL1  (Read 21059 times)

soundfan

  • New to the forum
  • *
  • Posts: 27
Re: Bilal's crop_rec_4k experiments for 100D / SL1
« Reply #125 on: March 02, 2023, 04:54:08 PM »
Tried this sd_uhs.mo. Same result.

Cold camera benchmark works on all three speeds, after some playing in the menues and some test recording speed drops to 20MB/s again.

160MHz:



192MHz:



240MHz (first attempt):



second attempt:




Seems not the be the sd_uhs.mo file...
EOS 100D... just beginning

theBilalFakhouri

  • Developer
  • Hero Member
  • *****
  • Posts: 1272
  • UHS-I
Re: Bilal's crop_rec_4k experiments for 100D / SL1
« Reply #126 on: March 02, 2023, 06:43:29 PM »
Thanks for the test.

The temperature isn't the only factor for 21 MB/s speed drops, some SD configurations perform better at high temps than others. See.
What you want is to find a better SD configuration than the one I found. I suggested to try some different GPIO values:

Tweaking GPIO registers in sd_uhs.c might help or might have no effect.

Current sd_uhs.c is using:

Code: [Select]
    MEM(0xC022C634) = 0x599;  <-- GPIO1

    MEM(0xC022C638) = 0x555;  <-- GPIO2
    MEM(0xC022C63C) = 0x555;
    MEM(0xC022C640) = 0x555;
    MEM(0xC022C644) = 0x555;
    MEM(0xC022C648) = 0x555;

Some info I wrote about them can be found here (under "New GPIO registers") , what you want for 100D:

for my 100D:
0x599 for GPIO1, 0x555 GPIO2 worked, and it seems the best for 240 MHz preset.

0x588 for GPIO1, 0x555 GPIO2 works too on 100D
0x599 for GPIO1, 0x566 GPIO2 works too on 100D
0x577 for GPIO1, 0x544 GPIO2 (700D values) don't work very well.

Without tweaking these registers, 240 MHz never been stable on 100D/EOS M, so they are required and they are a part of SD puzzle.

Try the other suggested combos, or try some new combos based on these info. if none of them worked, sorry it's hard to know exactly how to make it work. I might provide some tests later.



Okay, here are some tests, try these values (you can change them from SD Overclock submenu, every time after you change them restart the camera):

1- GPIO1: 0x588,   GPIO2: 0x555
2- GPIO1: 0x599,   GPIO2: 0x566
3- GPIO1: 0x577,   GPIO2: 0x544
4- GPIO1: 0x555,   GPIO2: 0x555
5- GPIO1: 0x566,   GPIO2: 0x555

This sd_uhs.mo, does any of above combos work better? try at 192 MHz first, if it give the same results, more likely there is no need to do the tests at 240 MHz.

soundfan

  • New to the forum
  • *
  • Posts: 27
Re: Bilal's crop_rec_4k experiments for 100D / SL1
« Reply #127 on: March 03, 2023, 05:44:13 PM »
Wow, that were a lot of proposals and a lot to test, too.

I did various combinations. All of your proposals and a few from my guts feeling.

Infact many work absolutely stable at 192MHz (I ended up using GPIO1: 0x5dd,   GPIO2: 0x599).

But unfortunately none of them keeps working at 240MHz. Even during recording with green data rate displayed camera stops out of the sudden. after that benchmark is at 20MB/s again.

Anyway, at least 192MHz is now stable with benchmark rates like these:



So far I'm pretty satisfied at the moment. Recording in UHD with 14bit works continuously.

Others like 3K, 1440p or Full-Res LV stop immediately with raw detect error.

UHD, 4K, 4.3K, 4.5K and 1920px1280p work perfectly.

I was actually hoping to get the Full-Res to work with whatever framerate is possible.

Do you know if there is a way to get there?


EOS 100D... just beginning

theBilalFakhouri

  • Developer
  • Hero Member
  • *****
  • Posts: 1272
  • UHS-I
Re: Bilal's crop_rec_4k experiments for 100D / SL1
« Reply #128 on: March 03, 2023, 06:10:52 PM »
Cool, thanks for running the tests!

Infact many work absolutely stable at 192MHz (I ended up using GPIO1: 0x5dd,   GPIO2: 0x599).

So some of these combos I shared eariler worked fine at 192 MHz on your 100D?

1- GPIO1: 0x588,   GPIO2: 0x555
2- GPIO1: 0x599,   GPIO2: 0x566
3- GPIO1: 0x577,   GPIO2: 0x544
4- GPIO1: 0x555,   GPIO2: 0x555
5- GPIO1: 0x566,   GPIO2: 0x555

I need to know which ones exactly worked fine at 192 MHz from above list on your 100D, so I can test them on my 100D at both 192 MHz and 240 MHz.
Although "GPIO1: 0x5dd,   GPIO2: 0x599" worked fine for you at 192 MHz, it doesn't really match Canon default patterns. I have made some tests like these in the past, some of them works yes.

I have one more left test to give for 240 MHz, might make it work on your 100D or might not, but firstly, please give what I asked for :)



Others like 3K, 1440p or Full-Res LV stop immediately with raw detect error.
..
I was actually hoping to get the Full-Res to work with whatever framerate is possible.
..
Do you know if there is a way to get there?

Sure, from first post:

-Quirks (not related to my build, but I have noticed it):

When selecting 3K and 1440p presets you need to turn off then on RAW video multiple times to let RAW video update RAW video resolution parameters.

Also:

Note: You will need to turn off and on RAW video multiple times after selecting 1440p or 3K presets, to let RAW video update its parameters!!

Same thing applies for Full-Res preset, this issue will be fixed in the next release from my side. However, commit is available online which fix the issue.

soundfan

  • New to the forum
  • *
  • Posts: 27
Re: Bilal's crop_rec_4k experiments for 100D / SL1
« Reply #129 on: March 03, 2023, 07:48:11 PM »
1. Downclocking read speed to 160 MHz instead of 192 MHz in 192 MHz / 240 MHz presets:
    Please try this sd_uhs.mo, try 192 MHz and 240 MHz, does write speed drop to 21 MB/s?

Well, the one from this post worked also stable on my 100D at 192MHz, but also crashed at 240MHz.

Same thing applies for Full-Res preset, this issue will be fixed in the next release from my side. However, commit is available online which fix the issue.


That is a little too much for me yet. Well, I tried many times to switch on and off the raw video. I found that jumping to live view and back (trash button) worked also well und made the modes update immediately. But not for the mentioned modes 3K, 1440p and Full-Res. Well, actually it worked in live view but didn't update the frame dimensions in RAW video.

Anyway, I will continue tying.
EOS 100D... just beginning

theBilalFakhouri

  • Developer
  • Hero Member
  • *****
  • Posts: 1272
  • UHS-I
Re: Bilal's crop_rec_4k experiments for 100D / SL1
« Reply #130 on: March 03, 2023, 08:09:28 PM »
Well, the one from this post worked also stable on my 100D at 192MHz, but also crashed at 240MHz.

That's make things totally different :) , in this case we need to redo GPIO tests without downclocking read speed to 160 MHz in 192 MHz preset:

Download this sd_uhs.mo: (has tweakable GPIO menu, read speed set to 192 MHz in 192 MHz and 240 MHz presets)
Just try the following values using 192 MHz overclock, does any of them perform better at 192 MHz (speed doesn't drop to 21 MB/s)?

1- GPIO1: 0x588,   GPIO2: 0x555
2- GPIO1: 0x599,   GPIO2: 0x566
3- GPIO1: 0x577,   GPIO2: 0x544
4- GPIO1: 0x555,   GPIO2: 0x555
5- GPIO1: 0x566,   GPIO2: 0x555

Perform a camera restart after you change GPIO1 and GPIO2, report back please.

soundfan

  • New to the forum
  • *
  • Posts: 27
Re: Bilal's crop_rec_4k experiments for 100D / SL1
« Reply #131 on: March 03, 2023, 09:03:48 PM »
Okay, let's see:

1- GPIO1: 0x588,   GPIO2: 0x555 drops to 20MB/s @ 192MHz
2- GPIO1: 0x599,   GPIO2: 0x566 drops to 20MB/s @ 192MHz
3- GPIO1: 0x577,   GPIO2: 0x544 drops to 20MB/s @ 192MHz
4- GPIO1: 0x555,   GPIO2: 0x555 drops to 20MB/s @ 192MHz
5- GPIO1: 0x566,   GPIO2: 0x555 drops to 20MB/s @ 192MHz


So 192MHz frequency for reading seems not to work on my 100D
EOS 100D... just beginning

theBilalFakhouri

  • Developer
  • Hero Member
  • *****
  • Posts: 1272
  • UHS-I
Re: Bilal's crop_rec_4k experiments for 100D / SL1
« Reply #132 on: March 03, 2023, 09:58:09 PM »
Okay, thanks, that was helpful, it seems GPIO values aren't the problem here.

Let's try three more tests for now before I give up, please make the test at 192 MHz preset:

1- I set 0xC0400624 and  0xC0400628 to 0x100 instead of 0x201 in 192 MHz, this sd_uhs.mo
2- I set 0xC040061C to 0x200 instead of 0x100                           in 192 MHz, this sd_uhs.mo
3- I set the changes from step 1 and step 2                                 in 192 MHz, this sd_uhs.mo

Does any of these sd_uhs work better at 192 MHz on your 100D?
(sorry, too many sd_uhs.mo copies)

soundfan

  • New to the forum
  • *
  • Posts: 27
Re: Bilal's crop_rec_4k experiments for 100D / SL1
« Reply #133 on: March 04, 2023, 08:47:03 AM »
Okay, thanks, that was helpful, it seems GPIO values aren't the problem here.

Let's try three more tests for now before I give up, please make the test at 192 MHz preset:

1- I set 0xC0400624 and  0xC0400628 to 0x100 instead of 0x201 in 192 MHz, this sd_uhs.mo
2- I set 0xC040061C to 0x200 instead of 0x100                           in 192 MHz, this sd_uhs.mo
3- I set the changes from step 1 and step 2                                 in 192 MHz, this sd_uhs.mo

Does any of these sd_uhs work better at 192 MHz on your 100D?
(sorry, too many sd_uhs.mo copies)

Okay, did the tests.

First only once each with cold camera:

Test 1:

Test 2:

Test 3:


There seemed to be a slight improvement with camera getting warmer so I did another 3 benchmark tests each and indeed a slight improvement was visible. So here are the results after the third test each:
Test 1:

Test 2:

Test 3:


In total I'd say the first versions shows the best results. Didn't do any recording though.
EOS 100D... just beginning

theBilalFakhouri

  • Developer
  • Hero Member
  • *****
  • Posts: 1272
  • UHS-I
Re: Bilal's crop_rec_4k experiments for 100D / SL1
« Reply #134 on: March 04, 2023, 08:52:41 AM »
Thanks for running the tests,

In total I'd say the first versions shows the best results.

maybe I wasn't clear in my previous reply, I am not looking for benchmarks numbers, I mean does 192 MHz drop to 21 MB/s after some time?
When using these sd_uhs versions:

1- I set 0xC0400624 and  0xC0400628 to 0x100 instead of 0x201 in 192 MHz, this sd_uhs.mo
2- I set 0xC040061C to 0x200 instead of 0x100                           in 192 MHz, this sd_uhs.mo
3- I set the changes from step 1 and step 2                                 in 192 MHz, this sd_uhs.mo

Didn't do any recording though.

Looking forward for recording tests! (to check if it drop to 21 MB/s or not).

soundfan

  • New to the forum
  • *
  • Posts: 27
Re: Bilal's crop_rec_4k experiments for 100D / SL1
« Reply #135 on: March 04, 2023, 10:59:37 AM »
Alright! I did some excessive testing now. Finally I found something reproduceable  :D

Version 1 drops to 21MB/s almost immediately, version 3 after a while. Even if recording timer is green, at some point it drops out of the sudden.

Version 2 seems to run stable until camera becomes too hot (beyond 10 minutes UHD 14bit @ 24°C indoor temp). I tested this several times, so it can be reproduced.

What I found during this testing is that camera temperature isn't updated on the live screen during recording. I started at 43°C and when I stopped after recording more than 10 minutes it jumped to 60°C. Is that a known issue (don't know if I should call it that)?
EOS 100D... just beginning

Danne

  • Developer
  • Hero Member
  • *****
  • Posts: 7702
Re: Bilal's crop_rec_4k experiments for 100D / SL1
« Reply #136 on: March 04, 2023, 12:02:54 PM »
Not temperature, correct registry setting needed. Seems version 2 is close.

theBilalFakhouri

  • Developer
  • Hero Member
  • *****
  • Posts: 1272
  • UHS-I
Re: Bilal's crop_rec_4k experiments for 100D / SL1
« Reply #137 on: March 05, 2023, 06:24:01 PM »
Okay, thanks for running tests again!

Version 2 seems to run stable until camera becomes too hot (beyond 10 minutes UHD 14bit @ 24°C indoor temp). I tested this several times, so it can be reproduced.

So does it drop to 21 MB/s when it's become not stable?
What does benchmarks show after SD becomes too hot (when it become not stable)?



What I found during this testing is that camera temperature isn't updated on the live screen during recording. I started at 43°C and when I stopped after recording more than 10 minutes it jumped to 60°C. Is that a known issue (don't know if I should call it that)?

It could be a bug (or maybe RAW video disables some Canon updates while recording), didn't pay much attention to these stuff, will make tests later.

soundfan

  • New to the forum
  • *
  • Posts: 27
Re: Bilal's crop_rec_4k experiments for 100D / SL1
« Reply #138 on: March 05, 2023, 06:30:45 PM »
Okay, thanks for running tests again!

So does it drop to 21 MB/s when it's become not stable?
What does benchmarks show after SD becomes too hot (when it become not stable)?

No, it runs stable at 192MHz. Only camera shut down due to overheating once. Second time it ran completely stable until SD was full  8)
Benchmark still at 79MB/s writing...


It could be a bug (or maybe RAW video disables some Canon updates while recording), didn't pay much attention to these stuff, will make tests later.

Not very important, but thanks anyway to have a look at it.
EOS 100D... just beginning

soundfan

  • New to the forum
  • *
  • Posts: 27
Re: Bilal's crop_rec_4k experiments for 100D / SL1
« Reply #139 on: March 11, 2023, 05:20:44 PM »
Hi Bilal,

it's me again. I'm actually pretty happy with the recording capabilities of my 100D at 192MHz now. Didn't have any instability yet.

Anyway, is there a chance we may find a way to get the 240MHz running, too?

Many thanks in advance!
EOS 100D... just beginning

theBilalFakhouri

  • Developer
  • Hero Member
  • *****
  • Posts: 1272
  • UHS-I
Re: Bilal's crop_rec_4k experiments for 100D / SL1
« Reply #140 on: March 13, 2023, 12:20:07 AM »
Hello soundfan,
Glad you are happy, thanks for the report :)

Unfortunately, more likely I won't be able to provide more tests this month. Maybe next month.
Anyway, I will let you know when I come up with new tests for 240 MHz.

iaburn

  • Member
  • ***
  • Posts: 192
Re: Bilal's crop_rec_4k experiments for 100D / SL1
« Reply #141 on: March 21, 2023, 11:39:18 PM »
Hi! I saw this thread and I couldn't resist. I got myself a 100D now that there is no "Zeek" channel in Youtube about it and it's still very cheap :D

On mine I noticed the vertical stripes on 1x3 modes, but solved with the "ADGT" version, and for now I only tried the Sandisk Pro 95MB/s and works well at 192Mhz (around 70MB/s).

theBilalFakhouri I'm still understanding how your build works and I would have a couple of questions:
Is it normal that, when I'm on the menu, at fixed intervals the live view flashes for an instant?
Is there any reliable way to set 3K mode? I was lucky just once, Now no matter how many times I activate/deactivate RAW, or go into Canon menu, I cannot get the RAW resolution to change.

And something extra about the camera itself. Is there a proper way to go from OFF to video mode and turn the camera off from video mode? I mean, do I have to wait a few seconds on the "ON" mode before going to Video or shutting it down?  ::)
Ah, and also, is it possible to use a dummy battery like in the EOS M? It has no hole and it forces you to close the battery door Found a suitable dummy battery  :D

I'm super impressed with this build. I know there is a lot of work behind, but I have the feeling that it came from nowhere when no one was expecting it, amazing!!

theBilalFakhouri

  • Developer
  • Hero Member
  • *****
  • Posts: 1272
  • UHS-I
Re: Bilal's crop_rec_4k experiments for 100D / SL1
« Reply #142 on: March 23, 2023, 01:30:21 PM »
Hi! I saw this thread and I couldn't resist. I got myself a 100D now that there is no "Zeek" channel in Youtube about it and it's still very cheap :D

Hello!
Cool, 100D is a good choice too.

On mine I noticed the vertical stripes on 1x3 modes, but solved with the "ADGT" version, and for now I only tried the Sandisk Pro 95MB/s and works well at 192Mhz (around 70MB/s).

Thanks for the report, I will delete other builds and keep ADTG version, seems to work fine on different 100D bodies.

Is it normal that, when I'm on the menu, at fixed intervals the live view flashes for an instant?

For now, yes, this problem does exist. I think it can be fixed, but didn't try yet. the fix could be similar to this one:

-Menus:
-fixed interference between ML menu and Canon overlays in LiveView, I took this fix from here, solves this issue.



Is there any reliable way to set 3K mode? I was lucky just once, Now no matter how many times I activate/deactivate RAW, or go into Canon menu, I cannot get the RAW resolution to change.

Here is what I am doing to make 3K preset work (I know, it's a dumb workaround :P):


And something extra about the camera itself. Is there a proper way to go from OFF to video mode and turn the camera off from video mode? I mean, do I have to wait a few seconds on the "ON" mode before going to Video or shutting it down?  ::)

I didn't get what you mean. You can set the power switch directly from OFF to Video mode and vice versa, what's the problem with this? that's how Canon made the camera work.

I'm super impressed with this build. I know there is a lot of work behind, but I have the feeling that it came from nowhere when no one was expecting it, amazing!!

Thanks, glad you liked it :)

I am actually discarding this branch (beside my 650D / 700D one). Will not work on them. There will be a new successor branch which is better in many ways.

iaburn

  • Member
  • ***
  • Posts: 192
Re: Bilal's crop_rec_4k experiments for 100D / SL1
« Reply #143 on: March 23, 2023, 01:58:06 PM »
Thanks a lot for the video, following these exact steps it works every time!  ;D

The question about the off/on/video switch was very dumb. You are right that Canon did it that way, so there should be no problem moving the switch between modes fast  ::)

This branch is already very usable, can't wait to see what's coming  :o

gabriielangel

  • Member
  • ***
  • Posts: 173
Re: Bilal's crop_rec_4k experiments for 100D / SL1
« Reply #144 on: March 23, 2023, 04:52:46 PM »
This is quite nice! How long can clips be recorded at 3K?

soundfan

  • New to the forum
  • *
  • Posts: 27
Re: Bilal's crop_rec_4k experiments for 100D / SL1
« Reply #145 on: March 23, 2023, 05:07:34 PM »
This is quite nice! How long can clips be recorded at 3K?

Until the card is full (if card is fast enough)  :D
EOS 100D... just beginning