S95 Porting Thread

Andras

Thanks Waldo for the great work!

Is there any chance that the built in HDR scence can be altered in any way by CHDK? (saturation for example).

Thanks again,
Andras

waldo

philmoz

EDGE_HMARGIN is used to reduce the overlay area height and set the y_min and y_max values for creating the overlay. Changing this value causes the edge overlay to not include a strip at the top and bottom on both the SX30 and G12.
If it's not what is causing the behaviour on the S95 then I don't know what is.

I've found that after changing values in camera.h it's best to do clean build so that all the code gets recompiled.

Thanks, Phil. That must have been what I did. When I force edgeoverlay.c to be rebuilt, it now draws to the edge of the screen. Personally, I like it better when the edge overlay does not draw over the Canon OSD, but someone complained that it did not look correct that way.

BTW, does anybody know why the standard CHDK makefile always forces a rebuild of gui.c ?

scroaticle

Hey, all -

I haven't read through the entire thread, so forgive me if this has been addressed.

Can the S95 be hacked so that it allows auto-focus during video recording? This is my only hang-up with the S90/S95. I like to record do-it-yourself/demonstration videos every now and again and the fact that you cannot get up close to an object if you began recording while at a distance is a major disappointment.

waldo

Microfunguy
Incidentally, try setting PARAM_FILE_COUNTER to 0x3A and see if raw saving still crashes.

Yes, in fact, that prevents the crashes and allows it to create a raw file. The raw file appears to be just garbage data that does not seem to be an image. I have also noticed that the corresponding jpeg image gets corrupted at the top of the image. The top of the jpeg image is filled with colored noise in an arc, presumably due to the in-camera barrel distortion correction. I'm hoping this is not due to CHDK running out of memory and colliding with the raw buffer.

Microfunguy

Turn-off any CHDK options that reserve memory.

I will try it on SDM 'tomorrow', it has more memory available.

David

Microfunguy

Well, tomorrow is here and this is what the DNG looks like.

Does your version look similar ?

waldo

Microfunguy
Well, tomorrow is here and this is what the DNG looks like.

Does your version look similar ?

Yes.

Trying changing the raw buffer address in hook_raw_image_addr() to 0x46000000.

Microfunguy

OK, I will have to ask the tester to confirm this.

Is that for 100H ?

Maybe two buffers are used (as on the SX30), look at this :-

http://tools.assembla.com/chdk/browser/trunk/platform/ixus980_sd990/sub/101b/lib.c

waldo

Microfunguy
OK, I will have to ask the tester to confirm this.

Is that for 100H ?

Maybe two buffers are used (as on the SX30), look at this :-

http://tools.assembla.com/chdk/browser/trunk/platform/ixus980_sd990/sub/101b/lib.c

That is for 100E, but may also work for 100H.

Ask your tester to also take several shots and look at the JPG files as well. They may see some corruption of those images. It appears as though the raw processing is affecting the jpeg chromiance buffer.

Microfunguy

The raw looks just the same with the new address.
The JPG is fine but this version of SDM creates the DNG on the PC and also has more memory available.

For IXUS980 101B :-

char *hook_raw_image_addr()
 	{
            return (char*) (*(int*)(0x5520 + 0x1😎? 0x42CFB780 : 0x40EBAFA0);
        }
 // similar to a560, searched for first raw address in a table with others, get variable from functions that use table

Need to find the equivalent of (0x5520 + 0x1😎

Maybe Phil can advise on this based on his G12 experience ?

Microfunguy

@Waldo

Have there been any changes to the original code that you posted for edge-overlay ?

philmoz

Sounds like two issues, picking the right raw buffer and putting the call to capt_seq_hook_raw_here in the right place.
Comparing S95 firmware 1.00H to G12 1.00C I would suggest the following.

In lib.c:

char *hook_raw_image_addr()
{
	if (*((int*)0x2ECC) != 0)
		return (char*) 0x46000000;
	else
		return (char*) 0x41269150;
}

In capt_seq.c try adding the call to capt_seq_hook_raw_here at the location below:

"loc_FFAF9A20: \n"
"		ADD	R1, SP,	#0x14 \n"
"		MOV	R0, R5 \n"
"		BL	sub_FFAF97F4 \n"

"		BL      capt_seq_hook_raw_here\n"      		// added

"		B	loc_FFAF9A5C \n"

The lib.c changes should work for 1.00e firmware; but the address of the code for capt_seq.c will probably be different.

Hope this helps,
Phil.

Microfunguy

Thanks Phil, you beat me to it.

I was just looking at this thread :-
https://chdk.setepontos.com/d/5722/212

How do you find the three addresses that you give above ?

I have sent the new version to the tester.

Guess we will have to wait until tomorrow morning.

David

waldo

Microfunguy
@Waldo

Have there been any changes to the original code that you posted for edge-overlay ?

No, the existing code works fine.

waldo

philmoz

Sounds like two issues, picking the right raw buffer and putting the call to capt_seq_hook_raw_here in the right place.
Comparing S95 firmware 1.00H to G12 1.00C I would suggest the following.

In lib.c:
[code]
char *hook_raw_image_addr()
{
if (*((int*)0x2ECC) != 0)
return (char*) 0x46000000;
else
return (char*) 0x41269150;
}
[/code]

In capt_seq.c try adding the call to capt_seq_hook_raw_here at the location below:
[code]
"loc_FFAF9A20: \n"
" ADD R1, SP, #0x14 \n"
" MOV R0, R5 \n"
" BL sub_FFAF97F4 \n"

" BL capt_seq_hook_raw_here\n" // added

" B loc_FFAF9A5C \n"
[/code]

The lib.c changes should work for 1.00e firmware; but the address of the code for capt_seq.c will probably be different.

Hope this helps,
Phil.

Thanks, That works for the 100e. I left the call to capt_seq_hook_raw_here where it was - it works fine as is.

philmoz

waldo
I left the call to capt_seq_hook_raw_here where it was - it works fine as is.

If it's at the location I mentioned in an earlier post (later on in the same function) then it will work; but there is a problem if you use 'RAW Develop' with i-Contrast turned on in the camera settings. With this combination you will get ghosting of highlights from the second shot onto the 'RAW Develop' image.

Moving the call to the earlier location in the function fixes this problem.

Regards,
Phil.

asm1989

The color of the Raw files is ok for this port?

In the ixus1000 using the RGGB Pattern the color is wrong, looks like the green chanel is off

about the location of the addreses, the place to look its the sema in the latest dryos

https://chdk.setepontos.com/d/5722/212

Microfunguy

Hmmm .. my first attempt did not work.

Folder '200CANON' is not created even though I returned '200' as folder number.

Maybe just a simple coding error.

The Canon firmware creates folders like 200_2911 (for the 29th of November).

@Waldo

What folder number are you using for CHDK raw's ?

emutier

scroaticle
Can the S95 be hacked so that it allows auto-focus during video recording?

very very unlikely. (no. and if ever - to slow to call it continous)

@waldo
maybe you should this to the build info on page 1 to keep guys from asking: "not working yet or ever: continous focus while recording movies" 😉

@microfungyy
this streak photography thing sounds amazing!!! want! PLEEEEEESE

waldo

philmoz

waldo
I left the call to capt_seq_hook_raw_here where it was - it works fine as is.

If it's at the location I mentioned in an earlier post (later on in the same function) then it will work; but there is a problem if you use 'RAW Develop' with i-Contrast turned on in the camera settings. With this combination you will get ghosting of highlights from the second shot onto the 'RAW Develop' image.

Moving the call to the earlier location in the function fixes this problem.

Regards,
Phil.

That location for capt_seq_hook_raw_here does not work. The function is not called after taking a picture.

« Previous Page Next Page »