+ Reply to Thread
Page 4 of 4 FirstFirst 1 2 3 4
Results 31 to 34 of 34

Thread: CCR cells Checker 2.2.0 : Feed back

  1. #31
    RBW Member uwxplorer is on a distinguished road uwxplorer is on a distinguished road uwxplorer's Avatar
    Join Date
    Jan 2011
    Posts
    326
    rEvo mini III Hybrid rMS

    Re: CCR cells Checker 2.2.0 : Feed back

    Quote Originally Posted by Carlton  View Original Post
    Attachment 39271

    Herr, there is a screenshot of my cell 3 : not linear : bottom ppo2 is under the linear line (red one) and the upper ppo2 are upper. And the average voltage is 4mv mv under what they do.

    Vincent
    Your graph is showing exactly what I was warning against: cell lag (or inertia, or slow response time, or whatever you want to name it).
    The low pO2 values obviously correspond to the last few feet of ascent (BTW, you may want consider flushing your loop with O2 above 20 ft), which might very well last less than a minute, as shown by the handful of points.
    If you have two cells which are slightly slower than the first one, they could very well be used to measure the VOTED pO2 and your last cell will appear as if it read too high compared to them.
    Inversely, if the cell you are showing here is slower than the other two, it will be potentially voted out to compute the pO2, and it will appear as if it reads too low. This is the likely explanation of what you are seeing.

    As far as the high range of values, it probably shows the same phenomenon but compressed in a more limited range of pO2 values during the main portion of your dive, when your solenoid (or you) tries to maintain a constant pO2 despite depth changes or other events. This results in a "noisy" pattern, which if stretched out would look exactly the same as your lower data points.

    BTW, I am not quite sure why the curve seems to show ~50 individual points for a 56 min dive. Even taking into assuming that you got stuck for 1/2 hour on a spot trying to take a picture of a nudibranch, I would expect around 300 points...

    As I mentioned before, the mV (and pO2) profile recorded by the SW is a heavily downsampled version of the realtime display (and even further compared to the actual sampling rate of the electronics).
    I presume you have noticed during dives cells trying to catch up with others, which they usually do in a matter of seconds, but sometimes not quickly enough, and this results in a temporary "VOTING FAILED" message. This is all gone is the log, but you can sort of try to figure it out by comparing the mV records as a time trace. This kind of lag is usually very clearly visible (and it goes both ways depending on your profile).
    I'll try to post an example later.

  2. #32
    RBW Member Carlton is an unknown quantity at this point Carlton's Avatar
    Join Date
    Apr 2013
    Location
    Nantes, France
    Posts
    37
    rEvo 3 hccr

    Re: CCR cells Checker 2.2.0 : Feed back

    Quote Originally Posted by uwxplorer  View Original Post
    Your graph is showing exactly what I was warning against: cell lag (or inertia, or slow response time, or whatever you want to name it).
    The low pO2 values obviously correspond to the last few feet of ascent (BTW, you may want consider flushing your loop with O2 above 20 ft), which might very well last less than a minute, as shown by the handful of points.
    If you have two cells which are slightly slower than the first one, they could very well be used to measure the VOTED pO2 and your last cell will appear as if it read too high compared to them.
    Inversely, if the cell you are showing here is slower than the other two, it will be potentially voted out to compute the pO2, and it will appear as if it reads too low. This is the likely explanation of what you are seeing.

    As far as the high range of values, it probably shows the same phenomenon but compressed in a more limited range of pO2 values during the main portion of your dive, when your solenoid (or you) tries to maintain a constant pO2 despite depth changes or other events. This results in a "noisy" pattern, which if stretched out would look exactly the same as your lower data points.

    BTW, I am not quite sure why the curve seems to show ~50 individual points for a 56 min dive. Even taking into assuming that you got stuck for 1/2 hour on a spot trying to take a picture of a nudibranch, I would expect around 300 points...

    As I mentioned before, the mV (and pO2) profile recorded by the SW is a heavily downsampled version of the realtime display (and even further compared to the actual sampling rate of the electronics).
    I presume you have noticed during dives cells trying to catch up with others, which they usually do in a matter of seconds, but sometimes not quickly enough, and this results in a temporary "VOTING FAILED" message. This is all gone is the log, but you can sort of try to figure it out by comparing the mV records as a time trace. This kind of lag is usually very clearly visible (and it goes both ways depending on your profile).
    I'll try to post an example later.
    Thanks for your analysis :)
    I agree with you on the slow sampling and the importance of the lag on the analysis.

    If a cell is really slow, during the dive we see the yellow or red alert on the Shearwater... so for me, the software is not needed :)

    The goal of this soft is just to display the average voltage for a cell compare to the first dive done with it :)

    Not to say you what you have to do, but display data to think about. With a rebreather, i think it's a good thing :)

    If the 2 linear lines are very far : calibration is wrong, cell start to be old and not give right voltage or start to lag... keep it in mind and look up your next calibration value (my case with this cells), buy a new one... you probably have to change it.
    If the data samples go down in high ppO2 : cell start to be caped, change it.

    If i find calibration data, i can improve analysis of sample and compare the expected value and the recorded value, etc. A lot of works...

    The soft make average of voltage on each ppO2 recorded by the SW. I remove each ppO2 wich have less than 3 records (ie, recorded more than 30 sec or recorded 3 times).
    This is why you only see 50 points for a dive during 56 min (3000 samples analysis).

    I can build a new graph with candlestick and volume sample like on stock graph, to display if recorded voltage for the voted ppO2 are stable or not. I propose to display min, max and average voltage for a ppO2, what do you think about ?

    Vincent

  3. #33
    RBW Member uwxplorer is on a distinguished road uwxplorer is on a distinguished road uwxplorer's Avatar
    Join Date
    Jan 2011
    Posts
    326
    rEvo mini III Hybrid rMS

    Re: CCR cells Checker 2.2.0 : Feed back

    Quote Originally Posted by Carlton  View Original Post
    If the 2 linear lines are very far : calibration is wrong, cell start to be old and not give right voltage or start to lag...
    This is where we will have to agree to disagree.
    In my opinion, it is most likely that the calibration was not done right (loop not correctly flushed with O2, cell not fully stabilized before calibration, etc).
    The best way to check whether your cell is starting to "act up" is to write down the mV reading at pO2 = 1 and compare it from one calibration to the next. I write this down together with the ambient pressure (to correct for variations) and temperature (unfortunately I don't have a correction factor for that, but luckily I live in a place where changes aren't significant).
    In other words, I don't think you need a fancy software to figure out whether the calibration factor is changing or not.

    Quote Originally Posted by Carlton  View Original Post
    If the data samples go down in high ppO2 : cell start to be caped, change it.
    This is what I would like to see once, as it would indeed be very interesting to witness. In principle, if you flush your loop with O2 at 20 ft (6 meters) while deco-ing there for any significant amount of time, you should be able to accumulate enough data to compare the reading there versus the expected value extrapolated from your pO2 = 1 calibration factor. Of course there could be some variance (biased toward lower values) due to buoyancy issues, imitating the effect you would be looking for, so that is not as trivial as it sounds.

    Quote Originally Posted by Carlton  View Original Post
    If i find calibration data, i can improve analysis of sample and compare the expected value and the recorded value, etc.
    Again, use the calibration factor you get when calibrating. I wished SW was saving that in the log, but last time I checked, they were not.

    Quote Originally Posted by Carlton  View Original Post
    The soft make average of voltage on each ppO2 recorded by the SW. I remove each ppO2 wich have less than 3 records (ie, recorded more than 30 sec or recorded 3 times).
    This is why you only see 50 points for a dive during 56 min (3000 samples analysis).
    I see.

    Quote Originally Posted by Carlton  View Original Post
    I can build a new graph with candlestick and volume sample like on stock graph, to display if recorded voltage for the voted ppO2 are stable or not. I propose to display min, max and average voltage for a ppO2, what do you think about ?
    I think that representing the data points with an error bar would be more informative (or alternatively show all the data points). I don't think removing points which are visited only once is a very sound approach, as you can have both positive and negative bias depending on where you are in your profile. In other words, a point (pO2, mV) can be revisited several times, but it doesn't necessarily mean that it is more reliable... You could have two erroneous readings out of 3 in that sample!

    I personally make an histogram of the /mV ratios and perform a robust fit with a Gaussian to extract the relevant parameters (peak position and variance), but sometimes the dive is too chaotic for this to be satisfactory.

    Which is why it is important to have a look at multiple aspects of the mV time traces...

  4. #34
    RBW Member Carlton is an unknown quantity at this point Carlton's Avatar
    Join Date
    Apr 2013
    Location
    Nantes, France
    Posts
    37
    rEvo 3 hccr

    Re: CCR cells Checker 2.2.0 : Feed back

    Hi,

    Thanks for your feed back.
    I'm currently work on it.
    I will publish a new version in few days with a new file import module (selective file), an export/import to share data beetween device and a fix to keep single ppO2 value.

    I rename the soft un CCR Tools.

    I will implement your suggested request in few time : sharing portal, filter sample, error discover and a new calibration menu (until I don't, found data in memory dump).

    Vincent

+ Reply to Thread
Page 4 of 4 FirstFirst 1 2 3 4

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts