Canon 1100D / T3

[snakesanti]

thank you very mach nanomad...and there's no way to free ram?
Y turned off, took out batt...no way. May i need to deselect some options to free memory?

[ans86]

Hi, I'm new in this forum.
Thanks for the creation of ML!
I'm here to ask and propose an useful thing in my opinion.
Can you enable the "red eyes light" for use as AF light?
As all of you know, Canon use prelamp of the flash for AF with few light and indeed Nikon has a light that help the AF without use the flash.
What do you think?
I hope to have explained good.. 
Thank!

[nanomad]

I had a look into it and couldn't find anything useful. As most things, it is probably controlled by another chip

[xanax78]

hi, nanomad,
im ml user with 5d mark II and it was great,
iv just purchased a 1100D and modded him for astronomy picture, i just need the intervalometers and bulb mode active, does it work with the last version 1100D RC3 ?
I dont need all function for now with this modèl
it work properly ?
THX for respond

[nanomad]

Sure. 99,9% of ML features work on the 1100D just fine

[xanax78]

cool thank you

[ubbut]

Hi, first of all thanks for ML and thanks for finally porting it to the 1100D (I just found out;-)

I know ML for the 1100D is still in dev, but just for the record I wanted to report, that some image effects in movie mode don't seem to work (nothing happens), these were:

cartoon (weak)
oil painting
sharpness

can someone confirm this?

My cam also freezed the first time I tried cartoon (weak) but I can't reproduce this...

[nanomad]

Confirmed, the same happens on the 600D (which is the big brother of the 1100D)


https://bitbucket.org/hudson/magic-lantern/issue/1487/new-image-effect-not-working-on-600d


Thanks for the report

[a1ex]

They should probably be disabled for menu. None of these effects works on 5D3 (either some things changed, or I'm changing them from the wrong spot - just before Canon code overwrites them).

[ubbut]

Hi,

in FPS override the frame rate won't go above 25.054 (30.041) since the timer minimum values are restricted to the original values in fps-engio.c for 1100D:

Code Select Expand


#undef FPS_TIMER_A_MIN
#define FPS_TIMER_A_MIN fps_timer_a_orig


I guess there is a reason for that so I am not going to change anything here, but
I thought I might disable the frame rates higher than 25 (30) in the menu as a first contribution to ML if that's desirable...

[nanomad]

Yeah, I totally forgot about that. The reason is that the camera uses weird timing values (or at least I couldn't compute them properly....)

Feel free to post a pull request for it on bitbucket so that we can credit you

[ans86]

I had a look into it and couldn't find anything useful. As most things, it is probably controlled by another chip

Another chip? And so can't you do nothing for that? 

[nanomad]

Not yet (or in the near future)

[ubbut]

Hi, good news I think...   :-)

I played with timer values A and B, first thing I did was changing the minimum values to the minimum original values used for 30p, so you could select framerates up to 30fps no matter if you select 25p or 30p (in 25p mode fps was restricted to 25)

minimum values are:

25p : A= 1000 B=1280
30p:  A= 960   B=1112

I than tried to find the minimum working value for timer A, which seems to be 872, lower values first result in vertical purple stripes and even lower ones in a completely black screen.
This changed the maximum possible fps to 33 :-)

Minimum value for timer B seems to be 1040, below that you get a nasty ERR70 ;-)

This further improved the maximum fps to 35 :-)


I noticed that the algorithm tends to choose a lower timerA value then the original minimum even for framerates below 30. Is this a problem, or does it help to reduce yellow?


I would upload the code to the repository but I have to admit that I don't know how to do so...(I tried the pull request button, but it says "access denied")

[ans86]

Not yet (or in the near future)

Ok thanks!
Let's hope for the future!

[nanomad]

25p : A= 1000 B=1280
30p:  A= 960   B=1112

Thanks, I really appreciate it. To upload code to the repository, you'll need to fork it and then submit a pull request (you need a bitbucket accout of course)

[a1ex]

I noticed that the algorithm tends to choose a lower timerA value then the original minimum even for framerates below 30.

The algorithm depends on the optimization setting:

- low light (for low frame rates): try to increase only timer B
- exact FPS: try to get as close as possible to the exact value; if there are more solutions, the one with smallest timer A is chosen (that results in lowest jello effect)
- high FPS: try to decrease timer B first (no side effect on shutter speed)
- high jello: try to increase timer A first

At extremes, in all cases the algorithm should hit the hard limits for both timers (at least in theory).

On new cameras (including 1100D), timer B can be changed either with or without the side effect of altering the shutter speed (you can choose whether you want the side effect, or not). So, low light mode includes the side effect, to get slower shutter speeds, but the other modes will cancel the side effect.

Technical: timer B can be altered via engio (with side effect) or via table patching (without side effect, but requires a video mode change to take effect).

[ubbut]

Something else:

I created a new small font, which should look much better on the low-res screen of the 1100D. It is basically what a1ex suggested some time ago, a small font of 6x3 scaled to 12x6. After downscaling in the camera by nearest neighbor this could look just fine.

To create the font I changed the font-small.in and ran mkfont to get the corresponding font-small.c. Worked, but can't see any changes on the cam.
Do I have to create a new FONTS.DAT?

thanks

[a1ex]

Yes, and it's not quite easy. I'd suggest another approach:

1) add this to 1100D Makefile:

CFLAGS += -DCONFIG_STATIC_FONTS=1

2) replace font-dyn.o with the 3 static font files in Makefile.

Advantage: this will load the fonts in the malloc memory pool, which will free some RAM from the AllocateMemory one (where the 1100D has some trouble). The 1100D has 512K reserved for autoexec.bin, and binary size after doing this will be around 450K.

Disadvantage: shadowed fonts will not work, but it shouldn't be very hard to fix. For this, you will need to port some init steps from font-dyn.c (maybe split it into another file called font-shadow.c and use it in all cameras).

[ubbut]

Hi, just updated the pull request:

lowered minimum timer A,B values to achive higher maximum framerates of up to 35fps and restricted menu selection to 35fps. In zoom fps is restricted to 32.

Created new small fonts for 1100D and changed to static fonts. To make the shadow fonts work I introduced a new struct 'sfont' which is always dynamically generated (not depending on CONFIG_STATIC_FONT)

Maybe there is a way around this, but I could not come up with one.
depending on CONFIG_STATIC_FONTS font is:

Code Select Expand

struct font
{
        unsigned        height;
        unsigned        width;
#ifdef CONFIG_STATIC_FONTS
        unsigned        bitmap[];
#else
        unsigned*       bitmap;
#endif
};

however it needs to be

Code Select Expand

unsigned*       bitmap;
for the shadow fonts.
So I introduced a new struct sfont:

Code Select Expand

struct sfont
{
        unsigned        height;
        unsigned        width;
        unsigned*       bitmap;
};

How could this be avoided?

[nanomad]

The new font is simply stunning, thanks for everything

[a1ex]

Screenshots please

Applying the same trick to medium and large fonts might improve the readability, and might be helpful on 5Dc/40D too.

[ubbut]

Yes I would like to create a new medium font, too. I am currently looking for a font which looks similar to the used helvecia but which renders nicely at a hight of 7px. (because that would mean the same hight than the current med font after upscaling)

Should also be GPL licensed, I believe. 

So far I couldn't find anything, but I will be more than happy to create the font-med.in file, If anyone points me to a suitable font :-)


Oh btw. I noticed, that there is no '?' character in font-med.in, font-large.in or font-small.in even though it seems to be created by generate-font...
maybe one has to run generate font once more ?

[nanomad]

Well, the "start at odd pixels" is not actually a hack, it's a work-around for the LCD screen scaler, which uses line skipping.

Btw, as a suggestion, we should avoid 1 px high horizontal lines (e.g. the on in B) since they get cut away

[ubbut]

Yes, the idea is to create a font where every "pixel" is a 2x2 pixel square.
That's what I did for the new small font (it's almost not used, but you can see it on the audio meter).

I'm currently working on a similar medium font...