On 14-04-17 22:04, Anton Lundin wrote:
On 14 April, 2017 - Jef Driesen wrote:
On 2017-04-13 17:04, Jef Driesen wrote:
I tried a different approach yesterday. Instead of adding 1024 to
the calibration value, I simply used the stored value as is, and
calculated the average ppO2 over all three sensors. Then I plotted
all those values against the average ppO2 reported by the device.
And guess what, there is a nice linear relationship between the two!
Doing a linear regression on the data gives a scaling factor of 2.2.
Interesting find.
For the example above this gives:
Sensor 0: 0.642 = 33 mV * 885 / 100000.0 * 2.2
Sensor 1: 0.630 = 29 mV * 989 / 100000.0 * 2.2
Sensor 2: 0.674 = 26 mv * 1179 / 100000.0 * 2.2
As you can see, the average ppO2 (0.648) is now very close to the
average ppO2 reported by the device (0.65). This seems to be true
for all samples. The largest difference is now 0.018.
The next question is of course what's the source of this factor 2.2?
Could it be that the calibration value is stored in some odd format, and
thats where the / 100000.0 * 2.2 part comes from?
My first thought was that it represents some kind of relative value. If you look
at one of my previous emails, you'll notice that the default value for the
calibration (when not yet calibrated) is 2100 for the Petrel. And the
calibration values for the Predator are all near the value 1000. Relative that's
a factor of 2.1. That's not close enough to 2.2 to produce correct ppO2 values,
but it's the only explanation I have that makes some sense to me.
BTW, the fact that the calibration values are near 1000 is also why adding 1024
worked for most samples. If the value is a bit higher than 1000, then the result
is pretty close to multiplying with 2.2.
And is correct for all Predators?
How did it line up on the petrels?
I only checked for the predators. For the petrel, the results are already
correct without the 2.2 scaling factor. So applying it there too, will produce
wrong results. So this is most likely something specific to the predators only.
Jef
_______________________________________________
subsurface mailing list
subsurface@subsurface-divelog.org
http://lists.subsurface-divelog.org/cgi-bin/mailman/listinfo/subsurface
