These are chat archives for nightscout/intend-to-bolus

20th
Jan 2018
Scott Leibrand
@scottleibrand
Jan 20 2018 00:07
@live4sw in the future it's much easier to read long lines if you paste them with single backticks, like this:
Jan 19 17:42:38 [rig name] pump-loop.log: Last carbs 302 minutes ago; remainingCATime: 10.6 hours; 82% carbs absorbed
Jan 19 17:44:27 [rig name] pump-loop.log: {"carbs":50,"nsCarbs":50,"bwCarbs":0,"mealCOB":8,"currentDeviation":-1.68,"maxDeviation":1.17,"minDeviation":-1.66,"slopeFromMaxDeviation":-0.321,"slopeFromMinDeviation":0,"allDeviations":[-2,-2,-1,1,1,-1,-1,-1,1],"lastCarbTime":1516383521040,"bwFound":false}
@live4sw one thing that sometimes causes COB to reappear is if the bolus that caused it to detect most of the absorption ages off and stops being considered
live4sw
@live4sw
Jan 20 2018 00:11
Could you explain? I don't understand.
happy to try to PR it later, I really just want to make sure that there isn't something wrong
Scott Leibrand
@scottleibrand
Jan 20 2018 00:27
COB is calculated by figuring out what insulin activity should have been, and then adding up all the deviations from that, and attributing that to carb absorption. If the rig “forgets” about an old bolus, it will find very minimal deviations, and calculate very little absorption.
The question then becomes whether the boluses from the hour or so before the carb entry were being properly considered when calculating COB.
live4sw
@live4sw
Jan 20 2018 00:30
Is it conceivable that it would cause the entire prior COB, that had entirely decayed, to pop back up in full? Since the previous carb entry had been 30.
Scott Leibrand
@scottleibrand
Jan 20 2018 00:31
That depends on your min_5m_carbimpact, but generally not.
live4sw
@live4sw
Jan 20 2018 00:31
I don't know if it is at all relevant that almost immediately after my lunch I noticed that my dexcom had stopped receiving readings, so I changed my sensor.
live4sw
@live4sw
Jan 20 2018 00:44
So the rig was not receiving readings for about 4 hours, and during that time the COB decayed to zero, and then after 3 or so new readings had come in from the new sensor, carbhistory was refreshed and COB popped back up to 30
Maybe there's some explanation here: Jan 19 16:28:57 [rig name] pump-loop.log: SMB enabled for COB of 30 Jan 19 16:28:57 [rig name] pump-loop.log: Last carbs 230 minutes ago; remainingCATime: 8.8 hours; 40% carbs absorbed Jan 19 16:28:57 [rig name] pump-loop.log: Carb Impact: 2.8 mg/dL per 5m; CI Duration: 8.8 hours; remaining CI (~2h peak): 0.3 mg/dL per 5m
So it sounds like it thought that the carbs were absorbing over a very long period of time. Not sure why, but at least that seems to explain how there could be a non-zero COB number despite having not had any carbs for hours.
Dana Lewis
@danamlewis
Jan 20 2018 00:58
Ah. CGM down time is contributing there
live4sw
@live4sw
Jan 20 2018 01:02
how so? makes sense but can't quite put it together
Scott Leibrand
@scottleibrand
Jan 20 2018 01:41
Oref0 doesn’t interpolate quite right over big CGM gaps.
So after CGM data comes back, it may treat carbs entered during the gap as if they’ve had no decay, because there was no CGM data and therefore no deviations to attribute to absorption.
If someone wants to set up a test rig and do some debugging there, it could probably use some more love. I got interpolation mostly working, but apparently not quite.
Ebgineer
@Ebgineer
Jan 20 2018 01:54
Its been about a week and haven't been able to reestablish BGs to nightscout. I made a second NS site and attempted to connect to dexcom share2 only (not to my rig) via the bridge settings, but its not getting BGs either. MongoDB shows 0 documents in Entries, where the BG records should be. Follow on my phone is getting BGs so the dexcom username and password are verified correct. Is there a way to see NS calls to the bridge, or stores to MongoDB?
Scott Leibrand
@scottleibrand
Jan 20 2018 02:28
Follow doesn’t use the same username as Share, does it?
Ebgineer
@Ebgineer
Jan 20 2018 03:50
Follow works off an invitation email, but displays the username from Share2
mkericson
@mkericson
Jan 20 2018 14:33
Do we have instructions in the openaps docs on how to reflash the firmware on the expoler board?
alimhassam
@alimhassam
Jan 20 2018 14:41
otherwise there are instruction to reflash the edison somewhere here: http://openaps.readthedocs.io/en/latest/docs/Resources/Edison-Flashing/all-computers-flash.html
mkericson
@mkericson
Jan 20 2018 17:11
@alimhassam Thanks. I have a new explorer board that is failing to communicate - they told me to try to reflash the firmware but having trouble figuring out how to do that.
I reset what was my working rig after removing log files that filled the disk space. I reran setup and trying to get it back up and running. Its not looping yet - I have the following note on the pill that seems to indicate a clock issue. It says this, "BG data is too old(It's probably this), or clock set incorrectly. The last BG data was read at Mon Nov 20 2017 21:29:18 GMT-0500 (EST) but your system time currently is Sat Jan 20 2018 09:32:43 GMT-0500 (EST) IOB: 0U"
Dana Lewis
@danamlewis
Jan 20 2018 17:16
Double check for typo's in NS URL in your setup, that means it's not getting BGs to the rig
mkericson
@mkericson
Jan 20 2018 17:16
Will do - Thanks Dana.
@danamlewis NS site is good. copied the link straight from "cd ~/myopenaps && cat oref0-runagain.sh" into a web browser and its good.
Dana Lewis
@danamlewis
Jan 20 2018 17:23
K. Is rig online?
mkericson
@mkericson
Jan 20 2018 17:24
I removed several files in "./log" to clear our space. But reran setup thinking that would but everything back that was needed.
Yes rig is online. have IP address when I type "wifi"
Brian Cadre
@bcadre
Jan 20 2018 18:26
Having many issues with sensor BG staying exactly the same (i.e. no variation) for hours at a time, despite actual (blood) glucose changing normally. Likely problem with sensor or transmitter (Medtronic), correct? Pump reflects what sensor says. Additionally, getting a “git fetch failed” message at the end of my pump-loop logs, despite successful looping. Any insight, anyone? Thanks in advance.
Dana Lewis
@danamlewis
Jan 20 2018 18:44
Git fetch is just checking version, nbd. And yes that's a medtronic sensor issue, have you tried switching to another sensor or is this a recurring problem across multiple sensors?
Brian Cadre
@bcadre
Jan 20 2018 18:53
Thanks, Dana. We’ve changed 4 sensors in the last few days and I’ve called Mdt about the transmitter. No obvious defects determined. Since we’ve been looping, she gets “Sensor Error” whenever she inserts a new sensor. She simply enters the “New Sensor” command a second time and it usually works. But, the first day, data is wonky. However, if she calibrates and the true BG is way off, she gets “Cal(ibrate) Error.” 1 or 2 of those errors and she gets a “Bad Sensor” error message and she’s forced to insert a new sensor.
Dana Lewis
@danamlewis
Jan 20 2018 19:02
Bleh :worried: that's why many of us stopped using.
Brian Cadre
@bcadre
Jan 20 2018 19:13
Just now we had to stop looping. This has been going on for 4-5 days now and I’ve been racking my brains trying to figure it out. She has a volleyball tourney and can’t risk too much insulin on board. She’s already overdosed due to SMB (appropriately) treating bad sensor data… Poor kid’s fingers look horrible from all the fingersticks… We’ll try again another day… Have a great weekend and thanks for the help/advice. I have a little time off next week and will work on some docs for you.
Dana Lewis
@danamlewis
Jan 20 2018 19:14
:( Sadly don't have a suggestion re sensor fixing of medtronic, but see twitter dm
Chris L
@woodduckie
Jan 20 2018 20:35
Anyone know how UAM prediction curves could be less than COB (frequently)?
Tim Street
@tim2000s
Jan 20 2018 20:43
@bcadre Presume you’re using MDT CGM with a loopable pump?
Scott Leibrand
@scottleibrand
Jan 20 2018 21:44
@woodduckie the UAM and COB curves are completely independent. there’s no reason one should always be higher than the other
Brian Cadre
@bcadre
Jan 20 2018 22:27
@tim2000s We are. Update: 4th call to Medtronic; transmitter dead/defective. Medtronic no longer shipping old transmitter d/t 670 Guardian replacing old format (per rep). Gotta find a transmitter that will work with her pump...
Maybe Dexcom...
Tim Howard
@tghoward
Jan 20 2018 23:58
howdy all, the flashing the edison page on readTheDocs suggests jubi 0.3.0 does not work and I should flash 0.2.0. I don't see any reference to problems here on intend-to-bolus. Is it worth trying 0.3.0 or do not go there?