Question O2 sensor calibration

[lermontov]

I wonder if/when CCR dive computers will track/store/test/alert to more information about sensor behavior, beyond just instantaneous ppO2 readings.

Beyond just flagging or voting out sensors based on ppO2 (or mV) alone, a modern dive comp could sense when a sensor is out of spec, responds slower than it did on the previous dives, or seems blocked--e.g. less mV change expected during depth change with no solenoid injection--or has an mV pattern that matches more complex known problems (with x% confidence), or track/profile/flag a more detailed history of individual sensor data over their whole lifetimes.

Many of us are already doing this ("keeping an eye on sensor #3...")

A few thousand detailed CCR dive logs with the relevant data could easily generate a 'baseline' profile for what normal sensor data should look like. Predicting outliers in realtime is increasingly trivial machine learning/classification

I *hope* this is included in whatever Shearwater is doing with the new default/mandatory clouding of any dive logs imported into its apps

can you already achieve this though? you can work out what the MV reading should be at 'X' depth at 1.4 and 1.6

I have it written im my wet notes so I can cross check at predetermined depths

 

[J-Vo]

can you already achieve this though? you can work out what the MV reading should be at 'X' depth at 1.4 and 1.6

I have it written im my wet notes so I can cross check at predetermined depths

Why go through the brain strain of looking at mv underwater? Your cells are calibrated, if PPO2 is outside of expected something is off. Dill flush to cell check.

As for the 1.6 check, why does it need to be @ 6m and with a full flush. You can check for current limiting at any point during the dive, just takes a couple puffs of O2, no O2 flush needed.

 

[inquis]

You can check for current limiting at any point during the dive, just takes a couple puffs of O2, no O2 flush needed.

Hitting 1.6 at the bottom doesn't necessarily mean it *is* 1.6 because of that potential nonlinearity. It's unlikely all cells would be similarly impaired and agree, but it's not impossible. Seeing them hit the value you *know* is correct eliminates all doubt.

 

[inquis]

can you already achieve this though? you can work out what the MV reading should be at 'X' depth at 1.4 and 1.6

I have it written im my wet notes so I can cross check at predetermined depths

The calibration records the conversion between mV and PO2. Why not use a consistent and intuitive number (i.e., the PO2)?

 

[LFMarm]

As for the 1.6 check, why does it need to be @ 6m and with a full flush. You can check for current limiting at any point during the dive, just takes a couple puffs of O2, no O2 flush needed.

Do you just add O2 and check that readings are moving or aim for a specific number?

 

[inquis]

You can check for current limiting at any point during the dive

I should walk back my earlier reply a bit. I do check for nonlinearity (of which I regard current limiting to be a severe form) periodically throughout the dive by looking at the *change* in PO2 in a descent. After a descent by 10m/33 fsw (from any depth), the PO2 should increase by the fraction of gas added when min loop volume is restored (and correspondingly less for smaller descent). I look for a 0.11 PO2 increase for 20 ft descent when adding 18/45.

I would expect to go from 1.2 to 1.5 for a 10 ft drop if I added O2, but the error uncertainties are dependent on both depth and fraction. Doing this starting at 30 ft would be far more tolerant of overshoot of min volume than at 230 ft.

 

[lermontov]

The calibration records the conversion between mV and PO2. Why not use a consistent and intuitive number (i.e., the PO2)?

yes of course but I was referencing justins desire to have

"A few thousand detailed CCR dive logs with the relevant data could easily generate a 'baseline' profile for what normal sensor data should look like.

so yes you can use you're PO2 knowing what it should read at x depth - and you can further cross check looking at your Mv if you have doubts about linearity

I usually look at my Mv reading after a dive on the downloaded data to see if any cells are flagging

see pic - not a great example but you can see the cells have some variance