Were you trained on Oxygen sensors with this level of detail?

[Capt Jim Wyatt]

Everyone can see the poll results now.

 

[_Slacker_]

Bobby,

From your paper you use an example of a cell reading 10mV in air, 43mV in O2, and 65mV in O2 at 20ft. This hypothetical cell is showing 10% less the expected mV from .21 to 1.0 but 15% less mV from .21 to 1.6.

.....If you are 10% off at 1.0 and diving 1.4 as a set point you could easily be at or above 1.6 simply due to lack of linearity.....

You did not list the mV @ 1.4 for this cell so I can't give the exact answer because we don't know how the cell changes between our datapoints from 10% to 15% deviation. However if I was diving 1.6 setpoint on the computer the actual ppO2 in my breathing gas would be 1.7. The computer is looking for 68.8 mV for 1.6 but we tested that the cell only outputs 65mV at 1.6.

A deviation of 0.1 and not 0.2 or greater as you suggest in the quote above. Diving at 1.4 would be off even less than 0.1

This is why we want the calibration routine to calibrate against 100% O2. Your examples are baselining against calibrating the cell in air. However the computer is using the 43mV as the baseline which even in this example of a cell being almost 15% non linear at the high end isn't quite as drastic a problem as stated. Also, if I dive a setpoint of 1.0, then my breathing gas is exactly 1.0 and not 10% off.

 

[Bobby]

Slacker,
43mv is the actual 02 output which gives a 10% linear deviation. "If" the cell follows the same linear deviation then it will continue to deviate past 1.0. If the PO2 is bumped up to 1.6, for decompression, then the actual set point would be 1.84 when reading 1.6. Linear deviation is not normally a straight line from linear it is normally a curve. I was attempting to keep it basic and simple so that it would be easier to follow. I also need to spend some time putting in graphs that properly visualize how linear deviation normally works. I believe that working from 0.21 up and showing a normal curve from that point is the easiest for most to understand however there could be a better way to present it, which I'm very open to suggestions.

 

[_Slacker_]

Slacker,
43mv is the actual 02 output which gives a 10% linear deviation. "If" the cell follows the same linear deviation then it will continue to deviate past 1.0. If the PO2 is bumped up to 1.6, for decompression, then the actual set point would be 1.84 when reading 1.6. Linear deviation is not normally a straight line from linear it is normally a curve. I was attempting to keep it basic and simple so that it would be easier to follow. I also need to spend some time putting in graphs that properly visualize how linear deviation normally works. I believe that working from 0.21 up and showing a normal curve from that point is the easiest for most to understand however there could be a better way to present it, which I'm very open to suggestions.

Bobby,
I understand that linear deviation is not a straight line which is why I used the values you picked in my example. Baselining on 0.21 may be easy to present but it is not the way our computers work. I calibrate against O2 not air. If I have calibrated my cell in O2 at 43mV my DC shows 1.0 at 43 mV. It doesn't care that it was 10% non-linear before or after that. I have changed the scale to reflect the 43mV reading. So your example that showed 10% deviation from 0.21 to 1.0 and 15% deviation from .21 to 1.6. So using air as your reference you say the setpoint is 1.84 when reading 1.6, but the computer would actually be showing 1.67 because it was calibrated @43mV and now reading 65mV.

Wanting 1.6 and getting 1.84 could be a big deal, but getting 1.67 instead is less of an issue.

 

[Bobby]

Slacker,
I'm not saying that there is 15% deviation from air, I'm saying that there is 15% deviation from O2 @ 1.0. 1.6X1.15=1.84. You would be seeing 1.6 however actually be diving 1.84. This is with a correlated linear curve which easily might not be the case, reality is that you could be over 2.0 with no way of knowing. This is what we call limiting however it is really linear drift that we can't see because of the human limit on PPO2 tolerance.

I apologize if I'm not stating this clearly.

 

[_Slacker_]

Slacker,
I'm not saying that there is 15% deviation from air, I'm saying that there is 15% deviation from O2 @ 1.0. 1.6X1.15=1.84. You would be seeing 1.6 however actually be diving 1.84. This is with a correlated linear curve which easily might not be the case, reality is that you could be over 2.0 with no way of knowing. This is what we call limiting however it is really linear drift that we can't see because of the human limit on PPO2 tolerance.

I apologize if I'm not stating this clearly.

I'm referring specifically to the example in your pdf. (cell reads 10mV in air, 43mV at 1.0, and 65mV at 1.6) where you correctly state 15% linear drift (@1.6 reference) from air. You then state with a setpoint of 1.4 you could easily be above 1.6. I'm saying the math there is incorrect and your will not be above 1.6.

 

[tarponchik]

Bobby,
I understand that linear deviation is not a straight line which is why I used the values you picked in my example. Baselining on 0.21 may be easy to present but it is not the way our computers work. I calibrate against O2 not air. If I have calibrated my cell in O2 at 43mV my DC shows 1.0 at 43 mV. It doesn't care that it was 10% non-linear before or after that. I have changed the scale to reflect the 43mV reading. So your example that showed 10% deviation from 0.21 to 1.0 and 15% deviation from .21 to 1.6. So using air as your reference you say the setpoint is 1.84 when reading 1.6, but the computer would actually be showing 1.67 because it was calibrated @43mV and now reading 65mV.

Wanting 1.6 and getting 1.84 could be a big deal, but getting 1.67 instead is less of an issue.

Using O2 as calibration standard will give you higher accuracy than using air but only when your sensor is new. However, since the sensor's linearity is lost starting with the higher O2 values, using O2 as a standard will give you higher error than using air if your sensor is aged. And this will always work one way, giving you values lower than the real numbers. Also, using O2 for calibrations will age your sensor sooner.

 

[tarponchik]

Dave,
Please let me know what isn't correct in what I wrote, either through a post or PM.

Thanks,

Bobby

Bobby, from the paper, it appears that you are not aware what is causing sensor aging and how exactly the linearity is lost. If you knew, you would not have said that "Voting logic simply shows, or votes out, a cell that has reached enough deviation from the other two cells. It does not determine which cells are correct." Because if you have 3 analyzers calibrated to 21% with the same air tank side by side, and then you check the same Nitrox tank using these, the highest reading will be the most accurate. Your cell with a 10% deviation at 100% will most likely have only a 1% deviation at 32-36%.

And, Bobby, with all due respect, who taught you to write papers? Where is your CONCLUSION(S)?

Is there a button to vote "I know this better than Bobby"?

 

[tbone1004]

I'm referring specifically to the example in your pdf. (cell reads 10mV in air, 43mV at 1.0, and 65mV at 1.6) where you correctly state 15% linear drift (@1.6 reference) from air. You then state with a setpoint of 1.4 you could easily be above 1.6. I'm saying the math there is incorrect and your will not be above 1.6.

You and @Bobby are both correct, the problem is with the specific numbers chose. Again, please remember the paper is a draft that was done years ago and hasn't been formally reviewed, so the actual numbers given may be a touch off, and in this case, they are by a bit. It is meant to explain the concepts of how cell drift and linearity can cause large fluctuations in your shown ppO2 vs. your actual ppO2 and that you have to use the grey matter between your ears to use the cell readings themselves instead of the computers interpretation of those cells to know what you're actually breathing.

Napkin math says this
air to surface O2=90% linear
surface O2 to O2 at 1.6=94% linear
air to O2 at 1.6=85% linear

You as a diver need to determine which one is correct and why.
If we assume that at the 1.6 check we did get a full and complete loop flush, then you have two points of reference that are we have bounded our setpoint range so we can assume that the 94% is linear in which case a setpoint of 1.4 will actually be around 1.5, happy days.
If we assume that the linearity at the surface is right, then our setpoint of 1.4 will have us a lot closer to 1.6
If we assume that the lineary from air to the 1.6 check is right, then our setpoint of 1.4 will be around 1.65 actual

You as a diver need to determine which curve is the least wrong and then hopefully are smart enough to have the expect mV's for your chosen setpoints written on your wetnotes to make sure that you have your setpoints on the controller chosen correctly.
It may also be in your best interest to have a standalone computer running your deco calculations and CNS calculations if you believe in that. If you want a setpoint of 1.4, but you need your handset to say 1.3, then your computer will run those and you'll just have some more deco than you bargained for. If you have cells that are more than 100% linear, then you'll be setting a setpoint of say 1.5, and actually be breathing 1.4 and that may not end so well for you...

 

[Capt Jim Wyatt]

Again, please remember the paper is a draft that was done years ago and hasn't been formally reviewed

Thank you for that reminder. The paper perfectly illustrates the concepts we are trying to discuss here.