Oxygen Sensor Fundamentals

[LFMarm]

Interesting presentation on redundancy and voting logic and how it should affect replacement and monitoring protocols.

​

 

[mr_v]

We discussed the video in a different thread. While the video contains good ideas, one could debate the assumptions made there, e.g., what is the exact degree of statistical dependence between the sensors, and the definition of "failure," and "failure's" implications for trained divers.

Please do not get me wrong - I am all for solid state sensors for multiple reasons, but that video requires substantial analysis, thought, and debate. Also, there is a bit of irony there--Poseidon CCRs are known for locking out divers due to somewhat subjected pre-check failures.

 

[LFMarm]

I am not advocating the voting logic implemented by Poseidon. I found the discussion on statistical redundancy interesting:

This is what I am taking away for myself is:

1. Avoid sensors from same manufacturing batch
2. Avoid installing sensors at the same time (same aging profile)

All others are impossible or impractical to adopt with my setup.

 

[mr_v]

@LFMarm, also, consider that manufacturers are aware of statistical dependencies and may minimize by rotating part suppliers or batches. In many industries that is a standard process that reduces operations risk, so two sensors bought from the same manufacturer at the same time may have ingredients from different batches.

 

[craig haymaker]

@LFMarm, also, consider that manufacturers are aware of statistical dependencies and may minimize by rotating part suppliers or batches. In many industries that is a standard process that reduces operations risk, so two sensors bought from the same manufacturer at the same time may have ingredients from different batches.

We discussed the video in a different thread. While the video contains good ideas, one could debate the assumptions made there, e.g., what is the exact degree of statistical dependence between the sensors, and the definition of "failure," and "failure's" implications for trained divers.

Please do not get me wrong - I am all for solid state sensors for multiple reasons, but that video requires substantial analysis, thought, and debate. Also, there is a bit of irony there--Poseidon CCRs are known for locking out divers due to somewhat subjected pre-check failures.

I’ve been diving the Se7en since 2017, and have never been “locked out”.

@LFMarm, also, consider that manufacturers are aware of statistical dependencies and may minimize by rotating part suppliers or batches. In many industries that is a standard process that reduces operations risk, so two sensors bought from the same manufacturer at the same time may have ingredients from different batches.

 

[piranhaman00]

Very interesting video, thanks for posting back in 2020 @tbone1004
@stuartv asked similar questions but I am still curious if someone could help.

Using the video as example:

If I have a sensor with 10.0mV in air I expect 47.6 mV in O2 but during calibration I get a corrected value of 43.0 mV. The Petrel 3 is now calibrated to say that on this sensor 43.0mV is 1.0.

The video states that this is a linear deviation and at 1.0 I am actually at 1.11. How does this make sense? If I calibrate the Petrel to say that 43.0mV is 1.0 how does the linear deviation affect the actual po2 in the loop?

 

[tbone1004]

Very interesting video, thanks for posting back in 2020 @tbone1004
@stuartv asked similar questions but I am still curious if someone could help.

Using the video as example:

If I have a sensor with 10.0mV in air I expect 47.6 mV in O2 but during calibration I get a corrected value of 43.0 mV. The Petrel 3 is now calibrated to say that on this sensor 43.0mV is 1.0.

The video states that this is a linear deviation and at 1.0 I am actually at 1.11. How does this make sense? If I calibrate the Petrel to say that 43.0mV is 1.0 how does the linear deviation affect the actual po2 in the loop?

The deviation is relative to the calibration point. So in this case you'd say that the cell is reading 90% linear. When it's at 1.0 it will read 1.0, but when it reads 1.3 then you're actually at 1.44
That's pretty severe deviation and if that is the case then you want to keep the ppO2 very near where you calibrated, though at 90% you're having other issues and need to replace the cell ASAP. If you must dive on it then you can, just keep the setpoint as close to 1.0 as possible and do not try to run anything above about 1.2 until you can do a dil flush to verify. I.e. you get to 100ft on EAN32 and know that it should be reading 1.28 but it's only reading 1.12 so you now know that 1.12 on the screen is really 1.28 so you can run there. Your deco computer won't know though which is a different problem.

 

[piranhaman00]

The deviation is relative to the calibration point. So in this case you'd say that the cell is reading 90% linear. When it's at 1.0 it will read 1.0, but when it reads 1.3 then you're actually at 1.44
That's pretty severe deviation and if that is the case then you want to keep the ppO2 very near where you calibrated, though at 90% you're having other issues and need to replace the cell ASAP. If you must dive on it then you can, just keep the setpoint as close to 1.0 as possible and do not try to run anything above about 1.2 until you can do a dil flush to verify. I.e. you get to 100ft on EAN32 and know that it should be reading 1.28 but it's only reading 1.12 so you now know that 1.12 on the screen is really 1.28 so you can run there. Your deco computer won't know though which is a different problem.

Thanks this helps.

Again this was hypothetical situation taken from video, not real.

 

[rjack321]

The deviation is relative to the calibration point. So in this case you'd say that the cell is reading 90% linear. When it's at 1.0 it will read 1.0, but when it reads 1.3 then you're actually at 1.44

Except that a cell this far off expected is highly likely to be non-linear. In all likelihood be incapable of reading even 1.3 at all.

 

[tbone1004]

Except that a cell this far off expected is highly likely to be non-linear. In all likelihood be incapable of reading even 1.3 at all.

that's what the next sentence was for saying that you probably have other issues and need to replace.... Though I have seen them that are really non-linear but will still read 1.6. This is actually where the old ISC Meg with the Apecs controls or the Divesoft Liberty are nice. The ISC did 2-point with air and O2, and the Liberty can actually do 3-point calibration with the cal kit for the head so that handles the linear deviation output issues. Thankfully this all goes out the window in a few years when we are all using solid state cells and don't have to worry about this anymore.