DoF values and other related items

barberofcivil

268

Re: DoF values and other related items

« Reply #40 on: 28 / November / 2009, 23:22:30 »

Advertisements

I've got some great news on finding a new function!
I got a bit of a hunch on camera focal lengths (after seeing FocalLengthConverter.c in the 720 reference) that the firmware should have the actual focal length in it, so I did a bit of searching and found it (in the SX10 at least).

Currently the focal length is estimated by linear approximation of a few hand input points in SX10/main.c:

Code: [Select]

static const struct {
	int zp, fl;
} fl_tbl[] = {
	{   0,   5000 },
	{  11,   6190 },
	{  41,  12610 },
	{  64,  25240 },
	{  86,  45470 },
	{ 105,  65880 },
	{ 128, 100000 },
};
#define NUM_FL (sizeof(fl_tbl)/sizeof(fl_tbl[0]))
#define CF_EFL 5600

const int zoom_points = 129;

int get_effective_focal_length(int zp) {
	return (CF_EFL*get_focal_length(zp))/1000;
}

int get_focal_length(int zp) {
	int i;

	if (zp<fl_tbl[0].zp)
		return fl_tbl[0].fl;
	else if (zp>fl_tbl[NUM_FL-1].zp)
		return fl_tbl[NUM_FL-1].fl;
	else 
		for (i=1; i<NUM_FL; ++i) {
			if (zp==fl_tbl[i-1].zp) 
				return fl_tbl[i-1].fl;
			else if (zp==fl_tbl[i].zp) 
				return fl_tbl[i].fl;
			else if (zp<fl_tbl[i].zp)
				return fl_tbl[i-1].fl+(zp-fl_tbl[i-1].zp)*(fl_tbl[i].fl-fl_tbl[i-1].fl)/(fl_tbl[i].zp-fl_tbl[i-1].zp);
		}
	return fl_tbl[NUM_FL-1].fl;
}

However, I found a function entry point in the firmware that returns the actual (at least internal set) focal length of the lens at FF941078:

Code: [Select]

FF941078:	E51F06CC 	LDR	R0, =0x90E4
FF94107C:	E92D4010 	STMFD	SP!, {R4, LR}
FF941080:	E5900004 	LDR	R0, [R0, #0x4]
FF941084:	E3500000 	CMP	R0, #0x0
FF941088:	059F1620 	LDREQ	R1, =0x14E
FF94108C:	059F05F0 	LDREQ	R0, =0xFF9409C8	**"ZoomLensController.c"
FF941090:	0BFB684D 	BLEQ	FF81B1CC
FF941094:	EBFFFFE6 	BL	FF941034
FF941098:	E8BD4010 	LDMFD	SP!, {R4, LR}
FF94109C:	EA046780 	B	FFA5AEA4
V
FFA5AEA4:	E3500080 	CMP	R0, #0x80
FFA5AEA8:	E92D4010 	STMFD	SP!, {R4, LR}
FFA5AEAC:	E1A04000 	MOV	R4, R0
FFA5AEB0:	828F0D06 	LDRHI	R0, =0xFFA5B038	*"ZoomLensConverter.c"
FFA5AEB4:	83A01068 	MOVHI	R1, #0x68
FFA5AEB8:	8BF700C3 	BLHI	FF81B1CC
FFA5AEBC:	E59F0170 	LDR	R0, =0xD2B4
FFA5AEC0:	E5900008 	LDR	R0, [R0, #0x8]
FFA5AEC4:	E0800104 	ADD	R0, R0, R4, LSL #0x2
FFA5AEC8:	E59000F0 	LDR	R0, [R0, #0xF0]
FFA5AECC:	E8BD8010 	LDMFD	SP!, {R4, PC}

Note that in the SX10 firmware, FocalLengthConverter.c is called ZoomLensConverter.c instead.

The next step will be to find the equivalent in the reference dumps for the signature genration (hopefully it is present and similar in the vxworks references). It will be nicer to work with in-camera calibrated focal lengths rather than having it estimated by zoom position (it isn't linear)

Logged

reyalp

14034

Re: DoF values and other related items

« Reply #41 on: 28 / November / 2009, 23:39:54 »

The existing values on most cameras are not estimated. They are taken from exif data at each zoom step. Only cameras with a lot of zoom steps (like sx10) interpolate between the exif values. I doubt the difference between the interpolation and real value is very significant.

That said, getting it directly from a canon function would be great. Would save effort in future ports, and have less room for error. Good stuff.

Logged

Don't forget what the H stands for.

barberofcivil

268

Re: DoF values and other related items

« Reply #42 on: 29 / November / 2009, 15:50:53 »

The exif value is not necessarily always accurate as it seems with the exif av and tv values from what I can tell (at least from comparing the apectual apex96 values used vs the exif data for these), however you are right that they should be very similar.

Here is the full comparison for the SX10 (I had to do every other value for posting size, but you get the picture). The bold values are the ones that are used for interpolation. It looks like they have been rounded to the nearest 10 vs the camera reported values.
The differences do vary up to 10%. Note that there is a large step from the zoom value 127 to 128 for some reason.

Quote

zoom firmware focal length old focal length %difference
0 5000 5000 0.0%
2 5247 5216 -0.6%
4 5441 5432 -0.2%
6 5641 5649 0.1%
8 5854 5865 0.2%
10 6077 6081 0.1%
11 6194 6190 -0.1%
12 6315 6404 1.4%
14 6567 6832 4.0%
16 6838 7260 6.2%
18 7126 7688 7.9%
20 7437 8116 9.1%
22 7773 8544 9.9%
24 8131 8972 10.3%
26 8521 9400 10.3%
28 8938 9828 10.0%
30 9392 10256 9.2%
32 9879 10684 8.1%
34 10408 11112 6.8%
36 10977 11540 5.1%
38 11595 11968 3.2%
40 12259 12396 1.1%
41 12612 12610 0.0%
42 12979 13159 1.4%
44 13751 14257 3.7%
46 14587 15355 5.3%
48 15481 16453 6.3%
50 16448 17552 6.7%
52 17476 18650 6.7%
54 18584 19748 6.3%
56 19766 20846 5.5%
58 21016 21945 4.4%
60 22350 23043 3.1%
62 23752 24141 1.6%
64 25239 25240 0.0%
66 26790 27079 1.1%
68 28423 28918 1.7%
70 30112 30757 2.1%
72 31876 32596 2.3%
74 33685 34435 2.2%
76 35559 36274 2.0%
78 37467 38113 1.7%
80 39427 39952 1.3%
82 41408 41791 0.9%
84 43432 43630 0.5%
86 45466 45470 0.0%
88 47533 47618 0.2%
90 49605 49766 0.3%
92 51706 51915 0.4%
94 53825 54063 0.4%
96 55948 56212 0.5%
98 58106 58360 0.4%
100 60276 60508 0.4%
102 62492 62657 0.3%
104 64733 64805 0.1%
105 65877 65880 0.0%
106 67035 67363 0.5%
108 69378 70330 1.4%
110 71803 73297 2.1%
112 74288 76264 2.7%
114 76876 79231 3.1%
116 79541 82198 3.3%
118 82324 85165 3.5%
120 85193 88132 3.4%
122 88180 91099 3.3%
124 91232 94066 3.1%
126 94359 97033 2.8%
128 100000 100000 0.0%

Logged

barberofcivil

268

Re: DoF values and other related items

« Reply #43 on: 30 / November / 2009, 08:19:13 »

Quote from: barberofcivil on 29 / November / 2009, 15:50:53

Note that there is a large step from the zoom value 127 to 128 for some reason.

I plotted the FW focal length vs the lens length and it makes sense now. There is a larger lens movement at the last step corresponding with this.

« Last Edit: 30 / November / 2009, 17:06:14 by barberofcivil »

Logged

barberofcivil

268

Re: DoF values and other related items

« Reply #44 on: 30 / November / 2009, 17:11:35 »

I have found the entrys for the 3 DryOS reference dumps, they all have the same structure so possibly only the first would be needed for signature building.

A720: FF603370
SD1100: FF92F558
SD980: FF947534

I'm still working on the vxworks. I think I may have found it on the A570, but I will have to test yet to verify on my own camera, then find it in the other reference dumps if I can.

One of the odd things about the function in DryOS is that it branches to a function that is generally equivalent to the GetCurrentZoomPoint function, but is not the one used by CHDK.

Logged

reyalp

14034

Re: DoF values and other related items

« Reply #45 on: 30 / November / 2009, 17:17:49 »

Quote from: barberofcivil on 30 / November / 2009, 17:11:35

A720: FF603370

Typo ?

What units does it return ?

Logged

Don't forget what the H stands for.

barberofcivil

268

Re: DoF values and other related items

« Reply #46 on: 30 / November / 2009, 20:50:16 »

Quote from: reyalp on 30 / November / 2009, 17:17:49

Quote from: barberofcivil on 30 / November / 2009, 17:11:35
A720: FF603370
Typo ?

What units does it return ?

Typo, should be FFD03370
(had a FFFFFFFF wraparound issue with the disassembly so I used FF500000 for the start address...)
Units is um (1000mm), actual focal length (not 35mm equivalent): 5000-100000.

Logged

reyalp

14034

Re: DoF values and other related items

« Reply #47 on: 30 / November / 2009, 20:58:08 »

Quote from: barberofcivil on 30 / November / 2009, 20:50:16

(had a FFFFFFFF wraparound issue with the disassembly so I used FF500000 for the start address...)

FWIW, you can trim the dump, or only load up to the point it doesn't wrap. There can't be valid ROM after FFFFFFFFF anyway

Logged

Don't forget what the H stands for.

barberofcivil

268

Re: DoF values and other related items

« Reply #48 on: 30 / November / 2009, 21:48:53 »

Quote from: reyalp on 30 / November / 2009, 20:58:08

Quote from: barberofcivil on 30 / November / 2009, 20:50:16
(had a FFFFFFFF wraparound issue with the disassembly so I used FF500000 for the start address...)
FWIW, you can trim the dump, or only load up to the point it doesn't wrap. There can't be valid ROM after FFFFFFFFF anyway

Perhaps, but I'm using the perl script from the wiki, so I just did it the easy way. I just have to remember to correct the address for it.

Logged

barberofcivil

268

Re: DoF values and other related items

« Reply #49 on: 30 / November / 2009, 23:10:05 »

OK, I successfully found and tested for VXWorks cameras I have:
A570 101A: FFE85A24
S2IS 100G: FFA86F24

Based on these, I searched the VXWorks reference dumps and came up with:
1: A610 100E: FFE521C0
2: A710 100A: FFE8E6BC
3: A570 100E: FFE859B0

the A610 & A570 are similar.
The S2 is similar to the A610 & A570 structures.

I assume the A570 100E reference is correct as it is pretty much identical to 101A which tested successfully.
I would assume the A610 is correct based on it's similarity with A570.
The A710 structure has the same general idea as these 2 and I found it using the same methodology, so it should be correct, but would need to be tested.

Logged