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

23rd
Feb 2016
Scott Leibrand
@scottleibrand
Feb 23 2016 00:46
we still have a ways to go to catch up to Type Zero, though.
they've had hundreds of people testing the system for about a year each
Chris Hannemann
@channemann
Feb 23 2016 00:53
@scottleibrand that seems a bit high; they claim 184k clinical trial hours on their site, which would be about 21 person-years
Still many more than OpenAPS, of course
Jason Calabrese
@jasoncalabrese
Feb 23 2016 00:56
but how long does it take to get people to test a new/changed algorithm for them
tazitoo
@tazitoo
Feb 23 2016 00:56

I need some help. I've started looping at the office, and I notice two things:
1) I am impatient and openaps as I have it configured has a hard time bringing highs down. My current max_daily_basal is 0.95
2) I also seem to have a long slow steady drift when I am below 100 - openaps has my temp basal set to 0, but I think I don't have it configured to act more quickly to stop it.

It seems like I have a nonlinear ISF - a constant doesn't quite cut it. Above ~180, I become pretty resistant, but below 100 I can run on fumes for hours.

Suggestions on an approach to modify my current setup?

Scott Leibrand
@scottleibrand
Feb 23 2016 00:56
thx, I was looking for that 184k number. couldn't remember the units.
didn't think it'd be right there in size 64 font. ;-)
@jasoncalabrese yeah, we've made a lot more progress on a lot fewer patient hours.
@tazitoo which branch are you running? master? dev?
tazitoo
@tazitoo
Feb 23 2016 01:02
master I believe - just followed install instructions
Scott Leibrand
@scottleibrand
Feb 23 2016 01:02
dev has a more aggressive high-temp algorithm that deals better with highs
and also is more aggressive with low-temps when you're dropping
tazitoo
@tazitoo
Feb 23 2016 01:04
To switch to dev, its it a git pull, and set my path to run those scripts rather than /usr/local/bin?
Scott Leibrand
@scottleibrand
Feb 23 2016 01:04
probably easiest to use the quick-src script to switch from using packages to using sources
tazitoo
@tazitoo
Feb 23 2016 01:05
</me looks slightly baffled> ... locate quick-src* returns nothing.
tazitoo
@tazitoo
Feb 23 2016 01:10
Thanks! I will give that a shot.
But even with a non-agressive algo - my wife has had some big smiles with the smoothness of the time series. Thanks!
Chris Hannemann
@channemann
Feb 23 2016 01:13
@scottleibrand yeah, seems like that's their big selling point: look at all these hours! 184k hours means that people love our stoplight UI!
Dana Lewis
@danamlewis
Feb 23 2016 01:13
haha
Scott Leibrand
@scottleibrand
Feb 23 2016 01:24
Hah, they have admitted that the stoplight will not be around for the next 200k. :-)
Chris Hannemann
@channemann
Feb 23 2016 01:52
That's real progress
Scott Leibrand
@scottleibrand
Feb 23 2016 02:19
You should read the articles about the cell novo partnership. Lots of good details.
amoskane
@amoskane
Feb 23 2016 03:56
Hey guys, I’m having a major nerd win moment! I got my 522 with a button error working again. Others have done this, yes, but I took a video and am proud to share it on youtube so other crazy hackers won’t be as reticent to open it up and go for it. :) Link is https://youtu.be/nWRVSHXN5cQ
Matthias
@ceben80
Feb 23 2016 04:00
@danamlewis thanks for sharing it, (n=1)*37 that is great, yeah.
Scott Leibrand
@scottleibrand
Feb 23 2016 04:01
Sweet! Good job @amoskane ! Good video too. :-)
Jason Calabrese
@jasoncalabrese
Feb 23 2016 04:01
awesome @amoskane
Scott Leibrand
@scottleibrand
Feb 23 2016 04:02
someone should tweet that and mention @OpenAPS, so I can RT it from there. :-)
amoskane
@amoskane
Feb 23 2016 04:04
:) !!!
Jim Matheson
@jmatheson
Feb 23 2016 04:16
:thumbsup:
Dragan Maksimovic
@dm61
Feb 23 2016 04:42
@amoskane great video :smile: :+1:
Matteo Neri
@francescaneri
Feb 23 2016 13:51
@scottleibrand when I can complete it without mistakes willingly ! ;)
Jim Matheson
@jmatheson
Feb 23 2016 21:38
@bewest I just did a hot switch of my transmitter to sensors. Seems to have kept passing through BG values without a warm up period. 10-15 second swap time. I did not pay attention to timing of a BG reading but did not miss one so clearly got lucky.
Jason Calabrese
@jasoncalabrese
Feb 23 2016 21:40
what's the trick for that? just moving a running transmitter to a new sensor?
Jim Matheson
@jmatheson
Feb 23 2016 21:40
getting some big BG drop though from the new sensor. Did a 2x test calibration right now.
Scott Leibrand
@scottleibrand
Feb 23 2016 21:41
normally when we hot-swap transmitter to a new sensor the receivers go ??? for about an hour
Jim Matheson
@jmatheson
Feb 23 2016 21:41
yeah. Yesterday i had to do a 2 hour warm up and today i decided it was time to switch sensors.
Scott Leibrand
@scottleibrand
Feb 23 2016 21:42
we usually restart one of them and let the other pick up after the hour
Jim Matheson
@jmatheson
Feb 23 2016 21:42
i am just using the ios app
Scott Leibrand
@scottleibrand
Feb 23 2016 21:42
is this G4 or G5?
Jim Matheson
@jmatheson
Feb 23 2016 21:42
but now getting a <||> symbol in NS. what does that mean?
G5
Scott Leibrand
@scottleibrand
Feb 23 2016 21:42
ah, no idea how that is different
Jim Matheson
@jmatheson
Feb 23 2016 21:43
the app shows BG values fine, but over on NS its got a left/right arrow thing.
Jason Calabrese
@jasoncalabrese
Feb 23 2016 21:45
thats some direction status we get from dex, I picked some random unicode chars for them a long time ago
Scott Leibrand
@scottleibrand
Feb 23 2016 21:47
the G4 does something similar (no direction arrow) when it doesn't trust the data
Jason Calabrese
@jasoncalabrese
Feb 23 2016 21:48
also depends on the uploader, there were some differences, maybe with share2NS too
Jim Matheson
@jmatheson
Feb 23 2016 21:49
well it is doing crazy things with the BG values now
lots of movement
Jason Calabrese
@jasoncalabrese
Feb 23 2016 21:50
without raw it would be hard to know whats going on
they don't want you to know
Jim Matheson
@jmatheson
Feb 23 2016 21:50
yeah it is the none symbol.
so it must be showing bg values when it intends to show ???
i dont have the receiver going separately. otherwise would be interesting to see what that was showing.
Scott Leibrand
@scottleibrand
Feb 23 2016 21:52
given that it uses the same sensors, I'd give it at least an hour before you use those BG values for anything
Jim Matheson
@jmatheson
Feb 23 2016 21:52
@scottleibrand did you mention you guys were doing a seprate transmitter start time on different receivers and that’s how you got it from 2 to 1 hr?
Scott Leibrand
@scottleibrand
Feb 23 2016 21:52
yep
we're also "pre-soaking" the new sensor: putting it in with a dead transmitter for 24h before doing the hot-swap
Jim Matheson
@jmatheson
Feb 23 2016 21:53
seems like they have a decent reason to do a warmup after seeing these values :)
oh i see, so that would reduce this issue potentially.
Scott Leibrand
@scottleibrand
Feb 23 2016 21:54
yeah, the full 2h is probably about 2x overkill, but there are likely some cases where it's wonky that long on a brand new one. pre-soaked, though, they're good faster.
Jim Matheson
@jmatheson
Feb 23 2016 21:54
seems like doing your method but with the G5 / app it might be nearing zero
Scott Leibrand
@scottleibrand
Feb 23 2016 21:54
helps a lot with the first-day jumpiness too
Jim Matheson
@jmatheson
Feb 23 2016 21:54
will have to try that for next swap.
Scott Leibrand
@scottleibrand
Feb 23 2016 21:55
could be. our method with raw data probably would also make it useful immediately
Jim Matheson
@jmatheson
Feb 23 2016 21:55
but you have raw, right?
Scott Leibrand
@scottleibrand
Feb 23 2016 21:55
we haven't set up our openaps to upload CGM data: we're just using share2ns
Jim Matheson
@jmatheson
Feb 23 2016 21:56
i see
Ben West
@bewest
Feb 23 2016 21:58
yeah, I'd like to set up openaps to upload glucose to NS
Jim Matheson
@jmatheson
Feb 23 2016 21:58
actually i cant test presoak until my current transmitter dies :)
Ben West
@bewest
Feb 23 2016 21:58
ns-upload is there already, but need a tool to convert glucose to NS format
Jim Matheson
@jmatheson
Feb 23 2016 22:00
interesting. i thought that was all already working. i am still hoping to find time soon to get G5 data to openaps via ble and then to NS. if that ends up being more reliable that the g5 app i’d be happy to remove it from my loop (and share to NS).
Ben West
@bewest
Feb 23 2016 22:01
there's no tool to do G5 ble yet either
only g4+share
if someone can lend me g5 I'm sure @loudnate can walk me through the process
Jim Matheson
@jmatheson
Feb 23 2016 22:02
yeah. i was going to see if i could pull it off. started to review your code actually. maybe we can discuss tonight.
Ben West
@bewest
Feb 23 2016 22:02
we can give it a shot tonight, we'll see
Jim Matheson
@jmatheson
Feb 23 2016 22:03
looks like i am back to a right arrow instead of the None symbol. i think that was ~30 mins
seems to be a reasonable reading now as well
Scott Leibrand
@scottleibrand
Feb 23 2016 22:05
:+1:
Jim Matheson
@jmatheson
Feb 23 2016 22:07
@bewest i think i am about 30d into my G5 transmitter life. I am happy to give it to you when it gets a bit closer to end of life if we don’t have it solved already.
tazitoo
@tazitoo
Feb 23 2016 23:13

Switched from master to dev today, and as my BG glides ever lower, I see this:
{
"bg": 86,
"temp": "absolute",
"snoozeBG": 110,
"recieved": true,
"rate": 0.0,
"reason": "Eventual BG 110>110 but Avg. Delta -2.67 < Exp. Delta 1.1; cancel",
"eventualBG": 110,
"timestamp": "2016-02-23T18:07:49.116091",
"duration": 0,
"mealAssist": "Off: Carbs: undefined Boluses: undefined Target: 120 Deviation: -21 BGI: 0.84",
"tick": -1
}

Am I misreading this? To me, if BF continues to go lower, I would continue a 0 rate - and not cancel.

BF == BG <mitten touch typing>
tazitoo
@tazitoo
Feb 23 2016 23:23
I'm just finding my way around the code, in determine-basal.js at line 243, it starts the logic if eventual BG is above min but falling faster than expected.
Is there a test for if the temp rate is a high temp or low temp?
Scott Leibrand
@scottleibrand
Feb 23 2016 23:31
You are in bolus snooze, so the assumption is that you'll have carbs active, and we don't want to run low temps unless snoozeBG drops below target.
Looks like you maybe over-bolused? Was that a correction or meal bolus?
tazitoo
@tazitoo
Feb 23 2016 23:33
Don't think I should be in bolus snooze - lunch was at noon EST, about 6 hours ago...
Scott Leibrand
@scottleibrand
Feb 23 2016 23:33
when was last bolus?
tazitoo
@tazitoo
Feb 23 2016 23:34
12:11 23 Feb
Scott Leibrand
@scottleibrand
Feb 23 2016 23:34
And what is IOB showing? Those numbers don't seem right.
tazitoo
@tazitoo
Feb 23 2016 23:35
I don't see IOB on the status screen, I manually set a 0 temp basal for 30 min
Scott Leibrand
@scottleibrand
Feb 23 2016 23:35
Maybe paste the stderr output of determine-basal too?
tazitoo
@tazitoo
Feb 23 2016 23:35
{"delta":2,"glucose":91,"avgdelta":1.333}
{"duration":0,"rate":0,"temp":"absolute"}
{"netbasalinsulin":-0.7,"bolussnooze":0,"basaliob":-0.628,"iob":-0.628,"activity":-0.0065}
Scott Leibrand
@scottleibrand
Feb 23 2016 23:35
what status screen? NS?
tazitoo
@tazitoo
Feb 23 2016 23:36
nope - pump status screen
Scott Leibrand
@scottleibrand
Feb 23 2016 23:36
i was looking for net IOB, which you just pasted. Pump IOB is useless
Oh, I misread. EventualBG and snoozeBG are the same. You just have lots of negative IOB already.
So it's thinking that should be enough to get you up to 110
tazitoo
@tazitoo
Feb 23 2016 23:37
It just canceled my manual temp again
{
"bg": 91,
"temp": "absolute",
"snoozeBG": 123,
"recieved": true,
"rate": 0.0,
"reason": "Eventual BG 123>110 but Delta +0 < Exp. Delta 2; cancel",
"eventualBG": 123,
"timestamp": "2016-02-23T18:35:58.687456",
"duration": 0,
"mealAssist": "Off: Carbs: undefined Boluses: undefined Target: 120 Deviation: -7 BGI: 2.1",
"tick": "+0"
}
Scott Leibrand
@scottleibrand
Feb 23 2016 23:38
Do you have reason to think that negative IOB won't be sufficient?
looks like you did flatten out and shouldn't drop much more.
If you really want to override it you can temp then suspend.
tazitoo
@tazitoo
Feb 23 2016 23:39
experience makes me worry when I see it dropping below 90
Scott Leibrand
@scottleibrand
Feb 23 2016 23:40
Does your experience come from <90 with negative IOB?
Dana Lewis
@danamlewis
Feb 23 2016 23:40
@tazitoo is that true even with large mounts (i.e. more than 1u) of negative IOB? That’s something that changes for me after looping.
Makes a difference knowing how much the temps have been working, often it turns my BGs around for me.
tazitoo
@tazitoo
Feb 23 2016 23:41
I don't know - previous to the loop I never had that kind of measurement. BUt I have been on long low temp rates with the t:slim...
Dana Lewis
@danamlewis
Feb 23 2016 23:41
but that may develop over time as you have more experience with the loop doing temps for you, getting to see the negative IOB have that impact
Scott Leibrand
@scottleibrand
Feb 23 2016 23:41
also, oref0 will resume low temps if you do keep drifting down. Below ~80 it gets really aggressive with them.
What is your min_bg set to?
tazitoo
@tazitoo
Feb 23 2016 23:42
110
Scott Leibrand
@scottleibrand
Feb 23 2016 23:42
and max the same?
tazitoo
@tazitoo
Feb 23 2016 23:42
130
Scott Leibrand
@scottleibrand
Feb 23 2016 23:42
k
sounds like you might be more comfortable with min_bg a bit higher.
tazitoo
@tazitoo
Feb 23 2016 23:43
thought process was a target of 120 - and max_bg min_bg bracket that given measurment error
Scott Leibrand
@scottleibrand
Feb 23 2016 23:43
We run with them equal now. It oscillates a bit, but nothing to worry about. 120-130 would be fine...
tazitoo
@tazitoo
Feb 23 2016 23:44
ok -I'll give that a shot
Scott Leibrand
@scottleibrand
Feb 23 2016 23:44
and then you can drop it as you get more comfortable.
tazitoo
@tazitoo
Feb 23 2016 23:46
So based on my first paste with BG=86 a while back, a delta less than expected and it cancelled the temp rate, that's intentional to not have the code concerned about it being a high or low temp?
Scott Leibrand
@scottleibrand
Feb 23 2016 23:47
Right. In range = cancel temp.
The 110>110 is a cosmetic typo. Should be >=
tazitoo
@tazitoo
Feb 23 2016 23:48
And that assumes the ISF is correct?
Scott Leibrand
@scottleibrand
Feb 23 2016 23:49
yeah. In meal-assist I have some code for detecting excess sensitivity or resistance, but otherwise we assume basals and ISF are close.
tazitoo
@tazitoo
Feb 23 2016 23:50
I think below 100, I get much more sensitive, and above 180, I get insensitive