[ML] EFIC - CMOS temp

[Antony Newman]

Has anyone noticed what aspect of ML caused the temperature to rocket up?

+) Writing to CF
+) Heavy use of ARM
+) Tasks not sleeping for long enough
+) Tasks running at too high a priority
+) Using LCD rather than HDMI
+) Use of the' ML main Menu'

I'm trying to work out what (if action) I can take to 'slow' down whatever is causing the meltdown.

Thanks.
AJ

[Moira O'Brien LIPF]

On the 5D I have never had a problem with overheating so have never paid attention to the temp reading.

Moira

--
http://magiclantern.wikia.com/
 
To post to this group, send email to ml-d...@googlegroups.com
To unsubscribe from this group, send email to ml-devel+u...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/ml-devel?hl=en

[Antony Newman]

Thanks Moira,

I was testing camera for about an hour and got my first overheat.

If slowing down my (or the existing ML) task helps - I'll build in a suitable defence mechanism.

AJ

[inosak]

You'r referring to CMOS temp. from ML or overheating icon on lcd?

[Antony Newman]

Hi Inosak.

As the camera overheated with little warning - I will be introducing a warning system on the screen that gives you an idea that the temperature is going up (I've thought of a non-intrusive system - I'll add this by the end of the week).

If the temperature starts going into the Red - I may be able to dynamically scale back what ever it is that is cooking the camera.

Eg If the ML audio task that ways up every 16ms is contributing to the overheat - I could slow it down.  If this is the main cause - I could consider rewriting it (this is just an illustration - as I have not analysed the ML audio tasks 'carbon footprint).
 
It will take me a couple of days to analyse what are the contributing factors to camera-'cooking', and what I can do about it.
If people have a gut feel (maybe it only happens during recording?) - it'll save me time in my analysis.

Regards,
AJ

[Antony Newman]

Very quick first thoughts.

I think the temperature increase is caused by:

+) Rapid changes in the picture information
+) The rate at which memory is scanned (my task)
+) Rotation of the lens (not really sure about this, but something is being triggered by lens rotation that takes 100% of the CPU for a fraction of a second).
+) Writing to CF

I've implemented an active throttling of my task (forces frame rate down to around 5 Hz).

Here's the colour scheme:

         if (old_efic_temp <= 170)
            colour = 0 ;                    // BLACK       [0 - 150 - 170]
         else if (old_efic_temp <= 175)
            colour = 0x7F ;                 // DARK BLUE   [171-175]
         else if (old_efic_temp <= 181)
            colour = 0x9F ;                 // DARK YELLOW [176-181]
         else if (old_efic_temp <= 184)
            colour = 0xAF ;                 // DULL YELLOW [182-184]
         else if (old_efic_temp <= 187)
            colour = 0xCF ;                 // DULL ORANGE [185-187]                
         else if (old_efic_temp <= 189)
            colour = 0x6F ;                 // DULL RED    [188-189]
         else
            colour = 0x08 ;                 // BRIGHT RED  [190-192 - Bang]

First test with an 'overheating limit' set at 180 - Temperature hovered around 181.


AJ

[Moira O'Brien LIPF]

Excellent Anthony - looks good.

Moira

AJ

--

[Alex]

AJ,

> *Found the Battery 'Juice' indicator burried in a Property.
Which one?

[Antony Newman]

Alex,

I wrote a 'property exploder' and dump:

#define PROP_BATTERY_REPORT     0x8003001D     // this is the one I'm using
#define PROP_BATTERY_HISTORY    0x0204000F
#define PROP_BATTERY_CHECK       0x80030013

I compare what happened to the batteries when I attempted to melt them.

I found the 2nd integer in 'PROP_BATTERY_REPORT' out 76 (or so) in the report is the one that I need.

I page of how the rough number that corresponds to the segments of the Canon icon.

I could just copy the Canon Icon .. but think something else is called for.

AJ

[Antony Newman]

Alex,

Screendump for you .    

[Alex]

Nice. I hope the code for making those screens is not ASM :D

On Tue, Jan 18, 2011 at 10:21 PM, Antony Newman <antony...@gmail.com> wrote:
> Alex,
>
> Screendump for you .
>

[Antony Newman]

.. Not yet.