Olympus SensorTemperature

[Phil Harvey]

I'm posting this to document some tests by Eric Sibert to calibrate the SensorTemperature of his Olympus E-510 and E-620 cameras.  In his tests, he let the camera temperature stabilize for a "long time" before taking each picture.

This is a difficult measurement to make for 2 reasons:

1) It may take a really long time for the internal temperature to equalize with atmospheric.

2) The sensor heats up quickly when taking pictures.

For these reasons, the sensor temperature should rarely be lower than the ambient temperature (the exception is if the camera is moved from a cool to a warm environment).  If the sensor is warmer than ambient, the measured points in the plot will be pushed to the left as the raw SensorTemperature values decrease due to a higher actual temperature.

With this in mind, it seems that the Olympedia calibration (orange line) is not very accurate for the E-510 and E-620 at higher temperatures.  In fact, I can't see a strong case for using a fit with higher order than linear.  So I have drawn a linear approximation which to me looks reasonable (red line).



Points below the line are easy to explain as a sensor that is slightly warmer than ambient.  Points above the line may only be explained by measurement error or a camera that started off cooler than ambient, so I wouldn't expect any systematic shift of points above the line like we see for points below the line at low temperatures.

This is still a work in progress, and the new calibration is by no means finalized, but I'm leaning towards updating ExifTool with the new linear calibration.  There are differences of up to a few degrees between the measurements from the E-510 and E-620.  I'm not sure if these difference are significant.

Here are the raw values at each measured ambient temperature:

Code Select Expand

        Raw      Temp
        ----    -----
E-510:  534      22.7 
        550      20.6 
        552      20.8 
        558      19.3 
        564      19.1 
        567      17.8 
        576      18.6 
        582      17.2 
        599      13.8 
        631      10.7 
        642      12.4 
        652       9.6 
        692       5.2 
        714       3.3 
        895     -19.8 
        895     -19.2 
        900     -21.7 
       
E-620:  518      22.7 
        531      19.3 
        533      17.9 
        582      17.2 
        621      12.3 
        634       9.7 
        650       8.0 
        660       7.7 
        703       3.3 
        880     -20.6 
        880     -20.6 
        892     -24.4 
        892     -22.7

- Phil

[Phil Harvey]

I've been playing around with the fit, and have come up with what I think is a better linear approximation.  The fit is a regression of all circled points in the plot (ie. after discarding a number of points that I considered outliers).  Then I adjusted the fit constants to be nice round numbers, but they are the same as those given by the regression to 4 significant figures.  The new formula is:

TEMP = 84 - 3 * RAW / 26



- Phil

[Alan Clifford]

But there are no data between +2 and -20.

[Phil Harvey]

Hi Alan,

Yes.  Nothing between Eric's refrigerator and freezer temperatures.  Perhaps this winter he can add a few points in this region.  It would be hard to get a stable temperature in this range without the help of  mother nature.

- Phil