Wanted: temperatures from your cams

[a1ex]

5D3, camera in fridge for 2-3 hours. EXIF says 8 degrees C, debug menu says 3. Will check again after some more hours.

A few hours later: EXIF says 4 degrees, debug menu says 0.

[escho]

Thanks, Alex.

I have to do some more measurements, which takes a lot of time to acclimate the camera to the ambient temperatures without internal heating. But the trend, I can see at the moment, doesn´t look very good. When I´ve got the values of the higher temeratures, I can say more... 

Edgar

[RenatoPhoto]

Just to confirm that 7D is linear and accurate display.. I got my camera warmed up in the stove and got red thermometer alarm from 61 to 67 C.

Here is the list of temps I gathered.  First temp is read from camera and second using exif tool

21   21
23   23
27   27
30   28
28   28
35   35
37   37
42   42
47   47
59   59
67   67
61   61
59   59
55   55
58   60
54   54
47   47
54   54
56   54
47   50

[escho]

Great, Renato

So we can be sure, that for 7D this function is true:

Code: [Select]

#define EFIC_CELSIUS ((int)efic_temp - 128)
For the red alarm. Do you mean the ML-alarm or the Canon-alarm. in MagicLantern the warnings are harcoded to 50 and 60°C:

Code: [Select]

static LVINFO_UPDATE_FUNC(temp_update)
{
    LVINFO_BUFFER(8);
   
    int t = EFIC_CELSIUS;
    snprintf(buffer, sizeof(buffer), "%d"SYM_DEGREE"C", t);
    if (t >= 60)
    {
        item->color_bg = COLOR_RED;
    }
    else if (t >= 50)
    {
        item->color_bg = COLOR_ORANGE;
    }
}
I remember, someone in the forum told, that 7D switched herself off at a certain temperatur. Do you remember this temperature, maybe?

Edgar

[RenatoPhoto]

Hi Edgar,

This was the Canon high temp icon appearing on the LV screen.  I got it up to 68C and the camera was still recording video, never got to shutdown temp.  Probably 70 C.

[escho]

Testing the temperatures of my 600D

Resulting plot


600D_result von seescho auf Flickr

Description

1. External heating. Warming up the camerabody very slow during 5 hours.
1.1 The blue points show the ambient temperatures, measured with an external thermometer.
1.2 The red points show the corresponding camera-temperatures, extracted from the exifs.

2. Internal heating, capturing a mov-video, video-duration 45 minutes
2.1 The green points show the results

3. Function for converting ML-temp into real-temp
3.1 The violett graph show th converting-function, as it is used by ML at the moment


Analysis

4. External heating
4.1 The slope of the graphs in 1.1 and 1.2 are quite the same. It is "efic_temp - 128", like it is used now (little differece of 4°C in y-axis)

5. Internal heating
5.1 The resulting function differs to the default one very strong. On higher temperatures, there are differences of more than 20°C.

It seems to be sure, that there are different sensors used for exif-temperature and the temperatures , shown by ML (efic_temp). Heating the camera in normal use (capturing a video) results in wrong temperatures. The "efic-sensor" meter very high temps, the "exif-sensor" follow these temperatures much more slow. The more cold the camerabody is, starting the test, the more difference is between the default function and the resulting function.

Conclusion

1. The temperatures, shown by ML on the LV-screen are not useable.
2. The ML-warnings are hardcoded to 50°C btw. 60°C. Looking at the temperatures on the LV-screen and the ML-warnings, you think, something is wrong with your 600D. You think, your camera is much too hot.

Other cameras

1. 5D3 seems to react like the 600D. Highly wrong values, which cannot be corrected, because they depend on the starting temperature on your recording.
2. 7D fits to the default function. No problem with this camera
3. 650D looks a bit like 7D, but I´m not sure about this.

What do I think about this

1. Showing temperatures on LV-screen should be disabled on LV-screen
1.1 Seeing these high temperatures makes me nervous
1.2 I see no useful information in showing the temperatures on the LV-screen, too much info in this screen
1.3 If somebody wants to know temperatures, he can look in the debug-menue

2. The ML-warnings should be disabled
2.1 No useful info, for the best a double info, because Canon warns too.

3. Debug menue
3.1 A hint should be shown, that these temperatures may be wrong ("measured somewhere in the cam" is not enough)

What do you thing about this stuff?

Edgar

[Walter Schulz]

3. 650D looks a bit like 7D, but I´m not sure about this.

Don't get me wrong but do you have data to back this opinion? Other than mine, of course ...
I've written before: I have strong reservations about my own datasets because of the very same problem you run into with 600D.

Ciao, Walter

[escho]

I have no other data than yours, Walter. That´s why I wrote: I´m not sure about this...
There are so many 650D-users, I think, but nobody seems to be interested in helping.

Edgar

[RenatoPhoto]

Personally I would ignore the external heating test.
Take the green data (internal heating) and draw the line to be used for ML temp display. 
Calibrate ML data to this line.

If the camera gets hot, it will show on the LV, if it is too hot, it will shut down.  I dont think there is any camera damage reported due to temperature.

[escho]

Personally I would ignore the external heating test.
Take the green data (internal heating) and draw the line to be used for ML temp display. 
Calibrate ML data to this line.

If the camera gets hot, it will show on the LV, if it is too hot, it will shut down.  I dont think there is any camera damage reported due to temperature.

That sounds reasonable.

I think, I´ll create a pull-request for the formulas of the 4 known cameras next week. The shown temperatures will be not correct, but more realistic than now.

I hope, some other guys do this test for other cameras, too.

Thanks for helping

Edgar

[dmilligan]

Ambient Temp: 21.1 C

1100D:
Turn on temp: 25C

Code: [Select]

36 32
39 34
41 38
46 40
46 42
46 43
51 45
51 46
51 47

60D:
Turn on temp: 25C

Code: [Select]

28 24
27 27
34 35
37 37
42 39
42 41
42 42
44 42
44 44
49 45
49 45
49 47
53 47

I think these are close enough not to matter, really would need a IR thermometer to check it.

[escho]

Thanks for testing

Plots of 60D and 1100D added
http://www.magiclantern.fm/forum/index.php?topic=9673.msg93158#msg93158

Edgar

[escho]

And updated all other plots too

http://www.magiclantern.fm/forum/index.php?topic=9673.msg93158#msg93158

Edgar

[escho]

On my harddisk, I changed ML-code to make my 600D working with the new temperature-function. Looks fine on my cam, much more realistic temperatures.

I think, for testing-purposes I need the efic_temp in the debug menue. Will do a pull-request for this later.

Second I saw in debug menue: Fahrenheit is shown as x.x. This precision cannot be provided by efic_temp. So, a integer temperature-value for Fahrenheit should be enough. This reduced precison makes the convertion from Celsius to Fahrenheit a bit simpler. I will do a pull request for this too.

Edgar

[RenatoPhoto]

Great work!  This will ease some of the concerns about high temps.

[escho]

These are the convertion-functions, I will use, generated from the test-data, known up to now:

Code: [Select]

5D3: EFIC_CELSIUS = efic_temp * 63 / 100 -72
60D: EFIC_CELSIUS = efic_temp - 130
600D: EFIC_CELSIUS = efic_temp * 63 / 100 -72
650D: EFIC_CELSIUS = efic_temp - 125
1100D: EFIC_CELSIUS = efic_temp - 133
others: EFIC_CELSIUS = efic_temp - 128
Edgar

[a1ex]

Great analysis.

I'd say we can just consider two camera groups:
- 5D3/600D (efic_temp * 63 / 100 - 72)
- 60D/650D/1100D (efic_temp - 128).

The difference between the offsets is small and it's probably the sensor not being very accurate (that is, one 650D might efic_temp - 125 and another one maight fit efic_temp - 130).

For the other cameras, maybe we should display just the raw temperature (this might convince the users to post here).

For the implementation, I suggest to delete EFIC_CELSIUS from propvalues.h and move it to platform/*/consts.h. Then, in the menu (or wherever the temperature is used), we can say:

#ifdef EFIC_CELSIUS, print it
#else, print the raw value and maybe point the users to this thread in the help text.

[escho]

For the implementation, I suggest to delete EFIC_CELSIUS from propvalues.h and move it to platform/*/consts.h.

That´s  what I wanted to do btw. what I did for my 600D on my own compile for testing.

Yes, I see these two groups too. 5D3 is similar to 600D. the others looks like what is suggested by exiftool (x-128).

Showing only raw-temperatures for the non tested cameras is a great idea. I will try to implement this.  But with the helptext I will need some help for myself, I guess. Give me some days to play with the code.

Edgar

[escho]

I think, I,ve got it.  It is running fine on my 600D (for testing with both cases, celsius and raw). I found, how to handle the help text, too.  I will push it to my repo this weekend.

Edgar

[ayshih]

Here's temperature data for the 50D (nightly build magiclantern-v2.3.NEXT.2014Jan09.50D109.zip)

Code: [Select]

21 25
27 27
31 31
36 36
40 40
44 44
47 47
49 49
50 50
52 52
52 54
54 52

It looks like it belongs to the 60D/650D/1100D group.

[escho]

Thanks for testing 50D. Yes, you are right. It belongs to the "efic-128"-group.

Updated plots: http://www.magiclantern.fm/forum/index.php?topic=9673.msg93158#msg93158

Edgar

[escho]

I just saw:
7D and 7D-master is in the maintained and in the unmaintained src-tree. Which one can I use for the temperature-functions?

Edgar

[Audionut]

http://www.magiclantern.fm/forum/index.php?topic=9848.0

[escho]

created pull-request. I hope, I didn´t forget too much.

Edgar

[escho]

From now on, your test results are used to show accurate temperatures in ML.

But only cameras, tested by you, show celsius or fahrenheit. All other cameras only show raw temperature-values. And the temperature display in LV-screen is disabled for these untested cameras.

I just updated the introduction on the top of this thread: http://www.magiclantern.fm/forum/index.php?topic=9673.msg93005#msg93005

Edgar