analyze data for correctness

[bewest]

@loudnate left us some great data to verify data decoding.

This must be a new pump or something, it came in blank, with lots of NoDelivery errors, and then it looks like everything changes. There are some novel records to be found here, and looks like we'll verify decoding for many.

---

Want to back this issue? Post a bounty on it! We accept bounties via Bountysource.

[bewest]

record 24

ChangeAlarmNotifyMode

4150 4/13/15 14:44:47 4/13/15 14:44:47 ChangeAlarmNotifyMode MODE=vibration, ACTION_REQUESTOR=pump 14913477664 53848595 1349 Paradigm 522
??

[loudnate]

I've got some analysis info and a PR decoding "Ian0b" as "AlarmSensor" coming your way tonight. Stay tuned.

On May 4, 2015, at 2:24 PM, Ben West notifications@github.com wrote:

@loudnate left us some great data to verify data decoding.

This must be a new pump or something, it came in blank, with lots of NoDelivery errors, and then it looks like everything changes. There are some novel records to be found here, and looks like we'll verify decoding for many.

—
Reply to this email directly or view it on GitHub.

[loudnate]

The 722 I picked up on Craigslist seems to have a bad CMOS backup battery. After swapping the AAA, each time it has entered "Error A21" mode with the system clock reset and the sensor data gone. I expect history disappears too; I will verify that soon.

On May 4, 2015, at 2:24 PM, Ben West notifications@github.com wrote:

@loudnate left us some great data to verify data decoding.

This must be a new pump or something, it came in blank, with lots of NoDelivery errors, and then it looks like everything changes. There are some novel records to be found here, and looks like we'll verify decoding for many.

—
Reply to this email directly or view it on GitHub.

[bewest]

Ah, try moving the clock back to some date before 2006?

[loudnate]

Interesting - before or after changing the battery?

[bewest]

@kenstack and others, it'd be cool to compare these two lists, methodically checking each one:

• https://github.com/bewest/decoding-carelink/blob/bewest/dev/analysis/578398/decoded_details.csv
• https://github.com/bewest/decoding-carelink/blob/bewest/dev/analysis/578398/details-no-sensor.csv

Here's a list of differences: https://github.com/bewest/decoding-carelink/blob/bewest/dev/analysis/578398/gap.diff

Most differences seem to be due to naming, just looking to match names and times...

[loudnate]

PR #85 is a list of manual comparisons for pump 736868, which I used when building PR #84.

A few other things that I noticed:

• UnabsorbedInsulin age value isn't handling 8-bit overflows (will need to find an additional byte somewhere)
• UnabsorbedInsulin curve is incorrect (not sure why yet)
• BolusWizardBolusEstimate carb_ratio and bg_target_high were incorrect (not sure why yet)
• BasalProfileStart includes the profile_index value in the second header byte. Not sure if that's important to anybody.

[bewest]

I now do something like this:

mm-pretty-csv <Carelink.csv> to get cleaner carelink csv.

To decode glucose pages, I use:
mm-decode-history-page.py --collate --data glucose --model 522 ReadGlucoseHistory-page-6.data | json -a date name op or similar

For history:
mm-decode-history-page.py --collate --model 522 ReadHistoryData-page-10.data | json -a timestsamp _type _head _date _body or similar.

The first two columns should be identical-ish/comparable in all the above cases, including carelink csv.
This is only useful for debugging/hacking decocare.

For people who just want the json exports, please consider using @openaps, which provides a nice app to drive this and other devices.