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

1st
Nov 2015
diabeticgonewild
@diabeticgonewild
Nov 01 2015 05:00
About 2/3 of the way done programming the ODE solver for the equations in ​Stochastic Virtual Population of Subjects with Type 1 Diabetes for the Assessment of Closed Loop Glucose Controllers​ See! https://github.com/diabeticgonewild/Artificial-Pancreas-2015-Mathematical-Model/blob/master/ODESystemSolve.m
Too tired to work on this anymore tonight :( ...and I really do want to!
diabeticgonewild
@diabeticgonewild
Nov 01 2015 07:44
Worked on it a little bit more. Can't sleep but too tired to work anymore on this tonight. See, differential equations are awesome to program...
System_of_ODEs.jpg

I don't have a ton left of work to do. Just verifying that I entered everything in correctly, running tests with test values, making corrections, and then going all out and running it.

Could legitimately be done writing the program by next week if all goes well....

Scott Leibrand
@scottleibrand
Nov 01 2015 08:22
@bewest your tz rezone/clock stuff is broken right now thanks to DST.
hopefully it's just because the 1-2am hour happens twice, so it assumes it's PST instead of PDT.
So right now clock-zoned.json is getting -0800 instead of -0700, and determine-basal is barging on the time mismatch.
Ben West
@bewest
Nov 01 2015 08:26
not sure what you mean
everything switched over?
there are two 1am's today
so if everything it's NTP updated, and all devices have the "new" time, the lapse should remain until "2am"
2am yesterday is 1am today, there are two 1am's today
Scott Leibrand
@scottleibrand
Nov 01 2015 08:28
Your tool thinks it's already PST.
Ben West
@bewest
Nov 01 2015 08:28
because you pi is NTP aware
Scott Leibrand
@scottleibrand
Nov 01 2015 08:28
it's taking 1:28am and saying "oh, that must be PST"
Ben West
@bewest
Nov 01 2015 08:29
so your pi switched
Scott Leibrand
@scottleibrand
Nov 01 2015 08:29
it's still PDT right now.
date shows PDT
bewest @bewest nodding
Ben West
@bewest
Nov 01 2015 08:30
ok, so if I understand
Scott Leibrand
@scottleibrand
Nov 01 2015 08:30
$ date
Sun 1 Nov 01:29:34 PDT 2015
Ben West
@bewest
Nov 01 2015 08:30
date shows pdt
timezone tool assumes PST, perhaps hour early?
Scott Leibrand
@scottleibrand
Nov 01 2015 08:30
$ cat monitor/clock-zoned.json
"2015-11-01T01:26:08-08:00"
ya
probably because 1:26 is ambiguous
Ben West
@bewest
Nov 01 2015 08:31
in thirty minutes, your date should report PST if you are ntp
Scott Leibrand
@scottleibrand
Nov 01 2015 08:31
yep
Ben West
@bewest
Nov 01 2015 08:31
at that point they will match
Scott Leibrand
@scottleibrand
Nov 01 2015 08:31
Here's hoping. :-)
wondering if I should switch the pump now.
Ben West
@bewest
Nov 01 2015 08:31
I think what's happening here, is python's dateutillibrary is being smart
and knows that relative to UTC, this is a done deal
locally what we do is have two 1am's
but in UTC it's done
Scott Leibrand
@scottleibrand
Nov 01 2015 08:32
it started doing this at 1am, not 7h ago
Ben West
@bewest
Nov 01 2015 08:33
right, 30 min ago
it will end in 30 min if I'm right and your pi is on ntp
Scott Leibrand
@scottleibrand
Nov 01 2015 08:34
That was my theory as well, but I'm thinking it's actually looking at the timestamp it's about to zone and figuring out whether that is during DST or not.
Ben West
@bewest
Nov 01 2015 08:34
in UTC, even translated to here, it'd past midnight
Scott Leibrand
@scottleibrand
Nov 01 2015 08:34
For 1h, it's ambiguous.
Ben West
@bewest
Nov 01 2015 08:34
yeah :-(
Scott Leibrand
@scottleibrand
Nov 01 2015 08:34
changing pump to PST
Ben West
@bewest
Nov 01 2015 08:34
well, we made august deadline, stuff like this is expected
yeah, might as well
I'm ??? right now, so doesn't matter
after second or third sensor failure today
it's already ???, when I hit the hay, no chance
Scott Leibrand
@scottleibrand
Nov 01 2015 08:36
You still haven't replaced it?
you knew sensor was dying a week ago.
Ben West
@bewest
Nov 01 2015 08:37
so, when I say I don't like needles, I really really mean it
I find these interventions completely repulsive, and if it wasn't absolutely necessary, I wouldn't do it
as-is, especially with raw, I can tell what's going on
Scott Leibrand
@scottleibrand
Nov 01 2015 08:38
:-(
Ben West
@bewest
Nov 01 2015 08:38
currently I'm 120 - 130, which is great
Scott Leibrand
@scottleibrand
Nov 01 2015 08:38
my dad is kinda that way. And he's an ER doc.
Do we need to make openaps use raw?
Ben West
@bewest
Nov 01 2015 08:39
yes, my response to dr's are much like dr's own stereotypical response
well... it's tricky
I would love to
it's obvious that one in 45 minutes drops out
or one in 30
but overall, it's still visible to human eye, what's going on
I'm clearly 120 - 130
oh, now I'm back!
dexcom says 104, within 20%, fair
really need to think of glycemia as range that it is, it's constantly moving, in different body parts it'll be different depending on how it's measured
it's not a single number
even through all that noise, it's still pretty clear what the likely range is
other thing we can do to up the anty is consider location, which is very tricky due to sensitive nature
Scott Leibrand
@scottleibrand
Nov 01 2015 08:42
Hah, can't win: $ cat monitor/clock-zoned.json
"2015-11-01T00:40:33-07:00"
Ben West
@bewest
Nov 01 2015 08:43
hmm
my prediction is it'll be tough till 2am
Scott Leibrand
@scottleibrand
Nov 01 2015 08:43
that time is before 1am, so it is properly PDT
Ben West
@bewest
Nov 01 2015 08:43
npr told me there are two 1am's tonight
Scott Leibrand
@scottleibrand
Nov 01 2015 08:43
yeah, nothing I can do for 17m
Ben West
@bewest
Nov 01 2015 08:44
my laptop reads 1:44am
Scott Leibrand
@scottleibrand
Nov 01 2015 08:44
16 now. :-)
Ben West
@bewest
Nov 01 2015 08:44
:-)
so, this is the experiment
Scott Leibrand
@scottleibrand
Nov 01 2015 08:44
wonder if I need to change Dex time too.
Ben West
@bewest
Nov 01 2015 08:44
sometimes, science asks you to stay up until 2am
yes
the logic is similar
it will grant local time
errr
oooo
you're getting from share using js
not python
hmmm
no clue
ask dexcom
lol
Scott Leibrand
@scottleibrand
Nov 01 2015 08:45
I'm doing both
Ben West
@bewest
Nov 01 2015 08:45
oh?
Scott Leibrand
@scottleibrand
Nov 01 2015 08:45
when plugged in, I get it from USB
Ben West
@bewest
Nov 01 2015 08:45
ah, ok
the bultin dexcom vendor
Scott Leibrand
@scottleibrand
Nov 01 2015 08:46
when that fails, Nightscout actually now
was using share2bridge directly in openaps-js
now just using NS API, and letting it talk to share.
Ben West
@bewest
Nov 01 2015 08:46
I'm using NS, kevin and ross did a good job making that robust
but I'm depending on old tech now
Scott Leibrand
@scottleibrand
Nov 01 2015 08:47
Main reasons I use USB are offline and quicker new-BG updates.
Ben West
@bewest
Nov 01 2015 08:47
counting on demand for g4 style features with g5 to reshape what they ship
dedicated device has certain performance benefits
nate did good job understanding g5 usb records
need a dedicated device that can promise connectivity to both (MM or accessible-controlable-pump) and (Dexcom or trustworthy sensor)
currently both dexcom and MM are EOL
Scott Leibrand
@scottleibrand
Nov 01 2015 08:54
We'll be fine with Dex G5 share.
The pump is the bigger concern. Hope the Omnipod or Accu-Chek RE bears fruit.
Ben West
@bewest
Nov 01 2015 08:56
accu-check made it quite clear to me that giving me the protocol would expose all of their patients to risk
Scott Leibrand
@scottleibrand
Nov 01 2015 08:57
Yeah. Wonder if we can figure out how to pair without cracking existing pairings.
Ben West
@bewest
Nov 01 2015 08:57
I took this to mean that (however strong or weak) their encryption is, it's a shared-key mechanism that once known for one pump applies to all
Scott Leibrand
@scottleibrand
Nov 01 2015 08:57
:-(
Ben West
@bewest
Nov 01 2015 08:57
second, the pump has two wireless modes
BT vs IR
Scott Leibrand
@scottleibrand
Nov 01 2015 08:58
meter controls pump over BT, right?
Ben West
@bewest
Nov 01 2015 08:58
while in BT, it needs to emulate ALL of the meter functions
including keeping history and all kinds of tasks
you cannot query the pump for history using BT
Scott Leibrand
@scottleibrand
Nov 01 2015 08:58
hmm
Ben West
@bewest
Nov 01 2015 08:59
if something happens locally on the pump you won't know about it
for that, you have to use IR mode
during which the pump goes into NO-OP
it won't dose any insulin during this, and it requires line of sight for IR
good job, they made it so hard, it's completely unattractive, borderline useless
I mean, I may have had some wine
Scott Leibrand
@scottleibrand
Nov 01 2015 09:00
Groundhog hour, round 2, go!
Ben West
@bewest
Nov 01 2015 09:00
oh wow
it's 1am AGAIN
no joke
lol
that's quite amusing
aw... back to ???
will probably install a new one in next few days
what good is showing it off to your parents and maintaining it if it's not working?
Ben West
@bewest
Nov 01 2015 09:06
will probably have one of them help me install it
Scott Leibrand
@scottleibrand
Nov 01 2015 09:06
We need to get you on a regular replacement schedule with someone who can assist with insertion.
Ben West
@bewest
Nov 01 2015 09:06
I actually try to maintain a social schedule around this
my folks are in town, lots of visits with borther in law (who is md/phd) and mom/dad
mom did great job helping last time
sometimes I use brandon
have used you before :-D
how are results now?
my NS is interesting, it reports 1am again, but no overlaps in dots
so wall clock was preserved so far, at least for NS-uploader-dexcom
no pump stuff yet, so we'll see
I could bolus a bit
Scott Leibrand
@scottleibrand
Nov 01 2015 09:09
Had to wait for all 5 cgm data points to be in the same timezone.
timomer
@timomer
Nov 01 2015 09:10
@ecc1 thanks for that link! @bewest @scottleibrand you can get history of bolus from the pump and therefore the meter as you can access all pump functions from it. Your really dont want to use IR mode as it shuts the pump down and is of limited use
Ben West
@bewest
Nov 01 2015 09:10
ah, need 5? wasn't sure
howdy @timomer, how are you
I see, so BT is more functional than I thought?
timomer
@timomer
Nov 01 2015 09:10
good thanks buddy, just making my morning coffee ;)
Ben West
@bewest
Nov 01 2015 09:10
very encouraging
Scott Leibrand
@scottleibrand
Nov 01 2015 09:10
I have iter_glucose 5, so apparently.
Ben West
@bewest
Nov 01 2015 09:11
gotcha
timomer
@timomer
Nov 01 2015 09:11
yeah, there is a mode on the meter where you go access all functions of the pump as if the meter is the pump - of that make sense
Ben West
@bewest
Nov 01 2015 09:11
@timomer strong, I hope, then watered down and with sweetener like us americans like it
sort of.. you can page through ... "pages" of things?
Scott Leibrand
@scottleibrand
Nov 01 2015 09:12
$ cat pump.ini
[device "pump"]
serial = <snip>
expires = 2015-11-01T02:01:38.167968
model = 522
see a problem there?
Ben West
@bewest
Nov 01 2015 09:13
well
timomer
@timomer
Nov 01 2015 09:13
@bewest but in the end, you only need temp basal access really as everyone uses the meter for carb logging and most do for bolus. I agree that there is a common shared key for all pumps
Ben West
@bewest
Nov 01 2015 09:13
it could be wrong until 2am
try wiping it out
that expires line
Scott Leibrand
@scottleibrand
Nov 01 2015 09:13
yeah, gonna delete that. Might want to bug-fix it.
Ben West
@bewest
Nov 01 2015 09:13
because ... at1:55, expires would be 2:05 for 10 minute session
not, 1:05
Scott Leibrand
@scottleibrand
Nov 01 2015 09:14
Maybe delete the expires if commands fail?
Ben West
@bewest
Nov 01 2015 09:14
at which point 1:10 would fail
Scott Leibrand
@scottleibrand
Nov 01 2015 09:14
(in decocare)
Ben West
@bewest
Nov 01 2015 09:14
nah, this is once every six months
the expires is openaps specific
Scott Leibrand
@scottleibrand
Nov 01 2015 09:14
really once a year.
Ben West
@bewest
Nov 01 2015 09:14
right now I'm fine with this as hack
Scott Leibrand
@scottleibrand
Nov 01 2015 09:14
No such problem for spring forward.
Ben West
@bewest
Nov 01 2015 09:14
it'll fix it self at 2am proper
or delete the expires line now
and it'll work
also
add
minutes = 3
:-D
and that will prevent this next time, actually
Scott Leibrand
@scottleibrand
Nov 01 2015 09:16
How?
Ben West
@bewest
Nov 01 2015 09:16
this time the problem is that the "session length" + cron was possible to span an hour
Scott Leibrand
@scottleibrand
Nov 01 2015 09:16
don't think so
Ben West
@bewest
Nov 01 2015 09:16
with session length of 3 minutes that's not possible
yes
Scott Leibrand
@scottleibrand
Nov 01 2015 09:16
at 1:55 it would set expires of 1:58
Ben West
@bewest
Nov 01 2015 09:16
why was it set to 2am?
Scott Leibrand
@scottleibrand
Nov 01 2015 09:16
so would still be screwed until 1:58
Ben West
@bewest
Nov 01 2015 09:17
no, you were screwed until 2am
Scott Leibrand
@scottleibrand
Nov 01 2015 09:17
cause it ran at 1:50?
Ben West
@bewest
Nov 01 2015 09:17
1:55
Scott Leibrand
@scottleibrand
Nov 01 2015 09:17
if it had written an expires of 1:58, that would still be there
and it wouldn't retry until 1:58 PST
Ben West
@bewest
Nov 01 2015 09:18
expirty = (1:51 - 1:59) + 10 minutes
that puts you at 2 am
that means at 1:08 (second time)
you're failing with outdated expiry
but it won't try again until 2am
Scott Leibrand
@scottleibrand
Nov 01 2015 09:19
explain how it would work if minutes = 3
Ben West
@bewest
Nov 01 2015 09:20
because 10 minutes + 1:55 = 2:05 vs 3 minutes + 1:55 = 1:58
so it doesn't cross the hour boundary
Scott Leibrand
@scottleibrand
Nov 01 2015 09:20
So it would fail until 1:58 PST
hour boundary shouldn't matter
what matters is falling back and thinking you have an extra hour of pump comma
Comms
Ben West
@bewest
Nov 01 2015 09:22
well, time keeping is a political thing more than technical
seems like twice or once a year there will be an hour at least of confusion
maybe once every less often an extra second or so
Scott Leibrand
@scottleibrand
Nov 01 2015 09:23
and looks like we'll also be screwed from the time we change the devices' time until DST actually switches.
Ben West
@bewest
Nov 01 2015 09:23
yes
it makes no sense to change device clocks until after, basically
Scott Leibrand
@scottleibrand
Nov 01 2015 09:23
if you don't the loop is down until morning.
Ben West
@bewest
Nov 01 2015 09:24
and it may make some sense to delay processing the NTP updated time until the devices have confirmed an update in time
Scott Leibrand
@scottleibrand
Nov 01 2015 09:24
It's not NTP that adjusts for DST
Ben West
@bewest
Nov 01 2015 09:24
or to figure out a way to interact/compensate
Scott Leibrand
@scottleibrand
Nov 01 2015 09:24
it's the Pi itself.
Ben West
@bewest
Nov 01 2015 09:24
the Pi gets it from NTP
otherwise it's just like microwave
Scott Leibrand
@scottleibrand
Nov 01 2015 09:24
It gets UTC from NTP
you set TZ and it does all the adjustments from there
Ben West
@bewest
Nov 01 2015 09:25
when you select your timezone, it's syncing it's next updates from NTP
there's a series of graceful degradations
Scott Leibrand
@scottleibrand
Nov 01 2015 09:25
If I shut off wifi 2h ago it still would have switched from PDT to PST
Ben West
@bewest
Nov 01 2015 09:25
but NTP gets priority when it can
right, because it's recently synced with NTP's expectation
Scott Leibrand
@scottleibrand
Nov 01 2015 09:26
knowledge of the DST schedule is built into the timekeeping packages.
Ben West
@bewest
Nov 01 2015 09:26
via NTP
via variety of packages, actually
Scott Leibrand
@scottleibrand
Nov 01 2015 09:26
if I had never used NTP and just set the time and TZ manually, it'd do the same.
Ben West
@bewest
Nov 01 2015 09:27
probably because your mirrors are up to date
any build in last six months would be up to date
Scott Leibrand
@scottleibrand
Nov 01 2015 09:27
NTP doesn't communicate DST changes. It's pre scheduled.
Ben West
@bewest
Nov 01 2015 09:28
further, likely
Scott Leibrand
@scottleibrand
Nov 01 2015 09:28
Only have to update when congress changes the schedule.
Ben West
@bewest
Nov 01 2015 09:28
there was a weird leap second thing recently
but think it was well known
Scott Leibrand
@scottleibrand
Nov 01 2015 09:28
Which has happened once in my adult life
yeah, leap seconds are diffeeeent.
That actually changes UTC.
Oref0 is happy now.
it thinks the temp I manually set last hour is running now, but that won't screw up IOB too badly.
Also got new IOB working again today by using zoned inputs and sorting tempHistory.
Keeping an eye on how closely it tracks DIYPS, but pretty decent so far.
Scott Leibrand
@scottleibrand
Nov 01 2015 09:35
'Nite
Ben West
@bewest
Nov 01 2015 09:46
nice
'night :-)
Tim Howard
@tghoward
Nov 01 2015 13:02
Finally awake && in same space as sarah to test what I've got so far from oref0. Of course, DST is intervening.
Tim Howard
@tghoward
Nov 01 2015 13:26
[
  {
    "body": "",
    "packet_size": 4,
    "name": "DateTimeChange",
    "raw": "87-c6-01-0f",
    "date": "2015-11-01T07:06:00",
    "date_type": "secSpecific",
    "_tell": 13,
    "op": "0x0c"
  },
  {
    "body": "",
    "packet_size": 4,
    "name": "DateTimeChange",
    "raw": "88-c4-21-0f",
    "date": "2015-11-01T08:04:00",
    "date_type": "secSpecific",
    "_tell": 18,
    "op": "0x0c"
  },
  {
    "packet_size": 0,
    "name": "DataEnd",
    "date": "2015-11-01T08:59:00",
    "date_type": "none",
    "_tell": 1,
    "op": "0x01"
  },  
  {
    "name": "GlucoseSensorData", 
    "date_type": "prevTimestamp",
    "_tell": 2, 
    "sgv": 154, 
    "date": "2015-11-01T08:54:00", 
    "packet_size": 0, 
    "op": 77
  }, 
  { 
    "name": "GlucoseSensorData", 
    "date_type": "prevTimestamp", 
    "_tell": 3, 
    "sgv": 154, 
    "date": "2015-11-01T08:49:00",  
    "packet_size": 0, 
    "op": 77
  },
In the above, note the time change records and the BG data. The BG data are still an hour off! BG records should be 7:54 and 7:49. Checked the pump three times: it is reading the new time. Totally a mystery to me.
Tim Howard
@tghoward
Nov 01 2015 13:38
Here's a read-clock call and then an iter-glucose call.
pi@betacb ~/bcb-s $ openaps use pump723 read_clock
"2015-11-01T08:36:11"pi@betacb ~/bcb-s $ openaps use pump723 iter_glucose 5
[
  {
    "body": "", 
    "packet_size": 4, 
    "name": "DateTimeChange", 
    "raw": "87-c6-01-0f", 
    "date": "2015-11-01T07:06:00", 
    "date_type": "secSpecific", 
    "_tell": 25, 
    "op": "0x0c"
  }, 
  {
    "body": "", 
    "packet_size": 4, 
    "name": "DateTimeChange", 
    "raw": "88-c4-21-0f", 
    "date": "2015-11-01T08:04:00", 
    "date_type": "secSpecific", 
    "_tell": 30, 
    "op": "0x0c"
  }, 
  {
    "packet_size": 0, 
    "name": "DataEnd", 
    "date": "2015-11-01T09:39:00", 
    "date_type": "none", 
    "_tell": 1, 
    "op": "0x01"
  }, 
  {
    "packet_size": 0, 
    "name": "SensorWeakSignal", 
    "date": "2015-11-01T09:34:00", 
    "date_type": "prevTimestamp", 
    "_tell": 2, 
    "op": "0x02"
  }, 
  {
    "packet_size": 0, 
    "name": "19-Something", 
    "date": "2015-11-01T09:29:00", 
    "date_type": "prevTimestamp", 
    "_tell": 3, 
    "op": "0x13"
  }, 
  {
    "packet_size": 0, 
    "name": "SensorWeakSignal", 
    "date": "2015-11-01T09:29:00", 
    "date_type": "prevTimestamp", 
    "_tell": 4, 
    "op": "0x02"
  }
Tim Howard
@tghoward
Nov 01 2015 13:50
had her give a bolus. That time is correct
pi@betacb ~/bcb-s $ openaps use pump723 read_clock
"2015-11-01T08:49:11"pi@betacb ~/bcb-s $ openaps use pump723 iter_pump 1
[
  {
    "programmed": 0.2, 
    "_type": "Bolus", 
    "_description": "Bolus 2015-11-01T08:41:42 head[8], body[0] op[0x01]", 
    "timestamp": "2015-11-01T08:41:42", 
    "_body": "", 
    "_head": "0100080008000000", 
    "amount": 0.2, 
    "unabsorbed": 0.0, 
    "duration": 0, 
    "type": "normal", 
    "_date": "aae928010f"
  },
Tim Howard
@tghoward
Nov 01 2015 14:02
Amazing. It looks like it is reporting correctly now. So I'm thinking it needed an hour (the amount the clock was set back) before switching over for cgm data.
pi@betacb ~/bcb-s $ openaps use pump723 read_clock
"2015-11-01T09:01:18"pi@betacb ~/bcb-s $ openaps use pump723 iter_glucose 5
[
  {
    "packet_size": 0, 
    "name": "DataEnd", 
    "date": "2015-11-01T09:06:00", 
    "date_type": "none", 
    "_tell": 1, 
    "op": "0x01"
  }, 
  {
    "name": "GlucoseSensorData", 
    "date_type": "prevTimestamp", 
    "_tell": 2, 
    "sgv": 168, 
    "date": "2015-11-01T09:01:00", 
    "packet_size": 0, 
    "op": 84
  }, 
  {
    "name": "GlucoseSensorData", 
    "date_type": "prevTimestamp", 
    "_tell": 3, 
    "sgv": 168, 
    "date": "2015-11-01T08:56:00", 
    "packet_size": 0, 
    "op": 84
  },
Sorry for cluttering up the line here. Perhaps this will be useful for something! :smile:
Tim Howard
@tghoward
Nov 01 2015 16:53
@bewest , another bump in the road. The MM glucose data was throwing errors and I figured out that it works with the non-cgm records removed. Good! so I can do a json filter, like this `|json -c "this.name== 'GlucoseSensorData'"
and I can make a use command that works
openaps use --format=json --output=monitor/filterglu.json cat shell monitor/glucose.json | json -c 'this.name=="GlucoseSensorData"'
that was after making cat a device
but no matter what iteration I try I can't make it work as an alias or a report. :worried:
Tim Howard
@tghoward
Nov 01 2015 17:40
no .. that use doesn't work
Nathan Racklyeft
@loudnate
Nov 01 2015 18:29
Hey @tghoward we've got a plugin for cleaning glucose entries. openapscontrib.glucosetools
Install via pip or easy_install
Tim Howard
@tghoward
Nov 01 2015 18:54
@loudnate , excellent, thanks, I'll check it out.
Tim Howard
@tghoward
Nov 01 2015 19:26
That did it! The clean tool from openapscontrib.glucosetools did the trick! I am one happy camper now! Our oref0 implementation runs and we are commencing testing.
Scott Leibrand
@scottleibrand
Nov 01 2015 19:26
:+1:
Let us know how testing goes, and when you "graduate" to running live/unattended! :-)
Tim Howard
@tghoward
Nov 01 2015 19:36
Absolutely, will do. :smile:
Dana Lewis
@danamlewis
Nov 01 2015 19:44
:)
More reasons for people to setup Nightscout as part of their openAPS implementation, even if they don't need/want to share data with others - best quick glance visualization of what the loop has done/is doing on top of your BGs: https://twitter.com/danamlewis/status/660903318256926720
diabeticgonewild
@diabeticgonewild
Nov 01 2015 20:16
The Dexcom G5 is not capable of extracting RAW data, right? I need to make a decision between the G4 and G5 in the next couple of weeks because I am out of warranty and realistically I will need RAW data for optimizing the algorithm for my controller. However, this is not imperative but is very, very desired and is a deal breaker for me.
???
Matthias Granberry
@mgranberry
Nov 01 2015 20:20
@diabeticgonewild No raw. I just now decided to go with a G4 transmitter because of that and a lack of a workable Android option right now.
diabeticgonewild
@diabeticgonewild
Nov 01 2015 20:21
Thanks. Yeah I need raw for modeling the sensor error (part of the nonlinear model predictive control) and deconvoluting the sensor signal (to improve accuracy) and more....
Definitely a deal breaker.
I am going to use that really small Android phone with the xDrip case combined. I don't know what it is called, but that is ultimately what I am going to use for my online setup. For offline OpenAPS it will just be the G4 receiver...
Too bad OpenAPS tools doesn't allow extraction of the RAW data from the receiver.
Matthias Granberry
@mgranberry
Nov 01 2015 20:27
xDrip behaves differently from the G4 receiver. The BG estimation algorithm is different in several ways, less prediction/filtering and different modeling of sensor age for starters.
diabeticgonewild
@diabeticgonewild
Nov 01 2015 20:28
I'm very close to solving the stuff I have been working on for months now, and I have a copy of a nonlinear model predictive control, that I have to parameterize for the stochastic model I am using http://etd.dtu.dk/thesis/248263/ep09_46_net.pdf
All I need is the RAW data. Is xDrip capable of that?
Matthias Granberry
@mgranberry
Nov 01 2015 20:29
yes. I'm just not sure if it's the same raw. You might ask @StephenBlackWasAlreadyTaken
Scott Leibrand
@scottleibrand
Nov 01 2015 20:39
Wouldn't be too hard to add raw support to dexcom_reader if it's not somewhere in a dev branch already. Nate just added G5 support...
but yes, xDrip has raw, so that would also work.
Matthias Granberry
@mgranberry
Nov 01 2015 20:40
@scottleibrand the G5 receiver doesn't appear to fetch raw data from the transmitter as far as I know.
Scott Leibrand
@scottleibrand
Nov 01 2015 20:41
Right, two different issues.
Supporting G5 at all, vs. supporting G4 raw.
diabeticgonewild
@diabeticgonewild
Nov 01 2015 20:43
Sorry driving
diabeticgonewild
@diabeticgonewild
Nov 01 2015 20:51
Yeah touchpad driver on my laptop isn't working so had to take a break and buy a mouse to use my computer until I resolve the issue :P
It's definitely G4 for me.
They should make G5 raw data available too...
Jason Calabrese
@jasoncalabrese
Nov 01 2015 20:56
maybe we'll be lucky and someone will find a way to get raw from the transmitter
Milos Kozak
@MilosKozak
Nov 01 2015 21:07
Is there something like dexcom studio for g5?
timomer
@timomer
Nov 01 2015 21:08
@danamlewis the temp basal vis you have in NS, what version of NS is that in, beta? Is it added as a custom treatment? I plan to be syncing more data from HAPP to NS this week
diabeticgonewild
@diabeticgonewild
Nov 01 2015 21:09
@MilosKoziak I am kind of out of the loop on that but I think there is a Dexcom Integrity App that does Dexcom Studio and more on the cloud.
@MilosKozak
Milos Kozak
@MilosKozak
Nov 01 2015 21:10
Cloud cut you off from data. If something like dexcom studio still works for download from g5 receiver there could be possible to look for raw in stored data
diabeticgonewild
@diabeticgonewild
Nov 01 2015 21:11
Oh yeah! Wasn't thinking about that!
I don't see what is the big deal about turning over RAW data
Scott Leibrand
@scottleibrand
Nov 01 2015 21:18
@timomer viz is in 0.8.2 (just released)
the big deal for Dexcom with raw data is there's zero upside, and big downside risk, of releasing it.
There might be a way to get it off the G5 receiver, but no reason for them to put it into the protocol to transmit over BLE.
timomer
@timomer
Nov 01 2015 21:21
@scottleibrand :+1:
Scott Leibrand
@scottleibrand
Nov 01 2015 21:21
Tools for uploading properly formatted data are in wip/bewest/dev of oref0
diabeticgonewild
@diabeticgonewild
Nov 01 2015 21:21

I guess, but keeping it locked just sucks. It really is a deal-breaker for me.

FWIW, I am going to use xDrip in the special case for the Melrose S9 phone with a Melrose S9 phone with a Ting SIM card and service....Then I am going to use Nightscout API to retrieve the RAW data properly....but this is all eventually...

I think there is a Melrose S9 that does CDMA, so Verizon might be an option. Verizon has the best coverage overall. Overall, I have to do a cost-factor analysis...
Maybe not though...either way, it's nothing to worry about. I am trying to fix my computer, then back to programming! I'm in the home stretch with respect to finishing up "mission impossible"!