O2 sensor values on CCR dives

[Jef Driesen]

On 01-12-17 17:43, Davide DB wrote:
> On 1 December 2017 at 16:54, Jef Driesen <jef at libdivecomputer.org> wrote:
>> On 2017-12-01 15:57, Davide DB wrote:
>> That's already good to know. So that means that either the calibration
>> constant, our calibration formula, or even the decoding of the millivolt
>> values is wrong. But the problem here is how do we verify this?
>>
>> Look at the sample data in your xml export from shearwater desktop:
>>
>> <diveLogRecord>
>>      <currentTime>0</currentTime>
>>      <currentDepth>4</currentDepth>
>>      <firstStopDepth>0</firstStopDepth>
>>      <ttsMins>1</ttsMins>
>>      <averagePPO2>0.98</averagePPO2>
>>      <fractionO2>0.18</fractionO2>
>>      <fractionHe>0.45</fractionHe>
>>      <firstStopTime>0</firstStopTime>
>>      <currentNdl>99</currentNdl>
>>      <currentCircuitSetting>0</currentCircuitSetting>
>>      <waterTemp>20</waterTemp>
>>      <gasSwitchNeeded>false</gasSwitchNeeded>
>>      <externalPPO2>false</externalPPO2>
>>      <setPointType>0</setPointType>
>>      <circuitSwitchType>0</circuitSwitchType>
>>      <sensor1Millivolts>54</sensor1Millivolts>
>>      <sensor2Millivolts>55</sensor2Millivolts>
>>      <sensor3Millivolts>54</sensor3Millivolts>
>>      <batteryVoltage>1.52</batteryVoltage>
>>      <tank0pressurePSI>0</tank0pressurePSI>
>>      <tank1pressurePSI>0</tank1pressurePSI>
>>      <gasTime>0</gasTime>
>>      <sac>0</sac>
>> </diveLogRecord>
>>
>> There are millivolt values, but no corresponding ppO2 values. So I can
>> double check the millivolt values, but we have nothing to compare our ppO2
>> values against.
> 
> Hummm...
> If I enable the average pO2 display on Subsurface even that values are
> higher than the average calculated by Shearwater Desktop (SD).
> Of course if the input data have an offset, even the resulting average will
> be biased.
> You wrote:
> 
>> In the data there is only the millivolt, and libdivecomputer converts
> this into a ppO2 value based on the calibration values. This conversion is
> reverse engineered, so a mistake there is not impossible.
> 
> What I do not understand is how this conversion was made the very first
> time. I mean how did you reverse engineered those values? I mean did you
> have a reference data set (mV, calibration K and resulting po2) to compare
> against?

Anton implemented the conversion for the Petrel. It's a simple multiplication:

    ppO2 = millivolt * calibration / 100000.0;

So very straightforward to guess and then check whether the result is close to 
the average value (which is stored in the data as ppO2 value).

The Predator needed an extra scaling factor of 2.2. I found the scaling factor 
by doing the ppO2 conversion above, then average the three values and doing a 
linear regression against the average ppO2 from the data.

I think I still have that code lying around somewhere. I'll have a look again, 
but my weekend is fully booked, so that will probably be something for next week.

> SD optionally shows the individual mv values against the average pO2. Is
> this sufficient?

No, because both values are available in the data.

> Going Sci-Fi the hard way to solve this is having the computer recording a
> dive at a fixed depth and a fixed FO2 input. I have (nearly) all the pieces
> to setup a test like this except that the sensor cable doesn't pass through
> the pressure chamber :) I don't know if the Petrel can be forced to record
> a dive at ambient pressure.

Well, the easiest and most reliable solution would be if Shearwater would tell 
us how to do the conversion :-)

Jef