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

18th
Sep 2017
PopcornPam
@PopcornPam_twitter
Sep 18 2017 00:51
Also.. if after bolusing for food (if I know the carb count is correct) I am getting a lot of smb.. would this be an idicator that my carb ratio is off? autotune thinks it is ok
Scott Leibrand
@scottleibrand
Sep 18 2017 01:02
Can you show the NS graph from when it is doing too much, including both the purple lines and the OpenAPS pill?
PopcornPam
@PopcornPam_twitter
Sep 18 2017 01:04
Screen Shot 2017-09-17 at 9.03.45 PM.png
Screen Shot 2017-09-17 at 9.03.17 PM.png
Screen Shot 2017-09-17 at 9.05.18 PM.png
The pattern seems to be over-doing it on the smb so I need some carbs to keep from dropping too low, or too long with 0 basal so I go up too high
Martin Haeberli
@mhaeberli
Sep 18 2017 01:14
I have a Edison / Explorer combo where the charging light turns on, but the power light never turns on. I tried swapping (charged) batteries, and that’s not it. I don’t know whether I have ever flashed the Edison or not … Any suggestions?
For that matter, I don’t know for sure whether I have ever programmed that explorer board’s radio
daveklingler
@daveklingler
Sep 18 2017 03:13
@mhaeberli , I'd like my Edison to do what yours is doing. One of the items on my list is to get those lights to remain off when my Edison is on its battery.
Scott Leibrand
@scottleibrand
Sep 18 2017 03:22
@popcornpam I'm trying to drill down on one specific decision that you think was wrong. Looks like you captured theee different ones there? Hard to tell.
sadafakahmad
@sadafakahmad
Sep 18 2017 03:50
just started looping. Keep getting the following message:
Martin Haeberli
@mhaeberli
Sep 18 2017 03:51
@daveklingler I’d like that, too, if I could get the edison to do real work
sadafakahmad
@sadafakahmad
Sep 18 2017 03:51
wl_host_event: mismatched OUI, bailing
Please help
Pete Schwamb
@ps2
Sep 18 2017 04:27
does anyone know if decocare supports writing a pump basal profile?
katie disimone
@Kdisimone
Sep 18 2017 05:02
@sadafakahmad check out the flaky wifi section here. http://openaps.readthedocs.io/en/latest/docs/While%20You%20Wait%20For%20Gear/monitoring-OpenAPS.html#troubleshooting-using-papertrail. The only time I've seen that mismatch OUI message was when that avahi was the issue.
Scott Leibrand
@scottleibrand
Sep 18 2017 05:26
@ps2 I'm not aware of such capability, but if it's buried in there somewhere we might want to use it in openaps as well.
Cylon75
@Cylon75
Sep 18 2017 08:39
Having some problems with "eating soon mode" which doesn't work at all. I created the widget according to the example (which seems to have changed a little, also changed the values for mmol/L).
When I activate the widget, I cannot see it in NS (0.9.2 on heroku) or see the rig acting on it. Also when entering the temporary target on my NS site, it is also not recorded.
I do see some future treatments but not regarding the temporary target in NS which seem to go away the next day. I am running dev 0.5.3, I am near Amsterdam/Netherlands)
It seems I have to start with NS itself but where to start troubleshooting this?
PopcornPam
@PopcornPam_twitter
Sep 18 2017 11:41
@scottleibrand , this is a pattern that I have been seeing for the past two weeks, so I grabbed a few different instances of it. I have a sneaking suspicion that ithas something to do with DIA, since this has mainly been I have started with FIASP. Also, previously SMB's were usually limited to .1 and .2, and lately they are now .3 and 4 amounts which is different and not something I necessarily agree with. It seems that getting all that insulin up front, and then no basal for 1-2 hours is not ending well
also, it was actually 2 instances.. I included the extra screen shot because for the first instance the pill was covering the line so you couldn't see it
Also wondering if part of the issue is that the Fiasp tail is pretty weak, so even though the rig knows there is insulin working (from DIA), it is over estimating how much
Scott Leibrand
@scottleibrand
Sep 18 2017 15:04
If you'd rather take a shotgun approach instead of drilling down on one instance, you could try installing 0.6.0-dev and using the ultra-rapid exponential insulin activity curve...
garykidd
@garykidd
Sep 18 2017 15:06
Any quick items I can check to figure out why my OpenAPS pill reflects unknown....seem to be looping just fine despite
Scott Leibrand
@scottleibrand
Sep 18 2017 15:08
There is a whole troubleshooting page in the docs that covers that.
garykidd
@garykidd
Sep 18 2017 15:25
mlab size on disk = 496...sounds suspicious...ehh?
Dana Lewis
@danamlewis
Sep 18 2017 15:27
@Cylon75 is your NS on the correct time zone? What other problems are you having with NS? Let's troubleshoot those first
garykidd
@garykidd
Sep 18 2017 15:32
I'm compacting now....seems to be taking pretty long...we'll see.
garykidd
@garykidd
Sep 18 2017 15:48
uh oh...I accidentally deleted the devicestatus category altogether rather than deleted underlying entries....noticed that the file size on disk didnt decrease either....will it fix itself perhaps?
garykidd
@garykidd
Sep 18 2017 15:58
so i added an empty category for device status....no changes yet...any nudge in the right direction ==would be greatly appreciated...thx
got it up again---thanks
katie disimone
@Kdisimone
Sep 18 2017 16:45
@Cylon75 temp targets in NS must be set in mg/dl. They do not work in mmol
PopcornPam
@PopcornPam_twitter
Sep 18 2017 16:52
@scottleibrand, I am heading out to Seattle on Thursday for a quick trip, and between time change and what will probably be a nutty weekend with late nights, alcohol, and tons of carbs, It will be hard to see patterns since there will be a ton of other external things messing with BG.. what do you think? drill down or shotgun?
Scott Leibrand
@scottleibrand
Sep 18 2017 16:57
no reason we can't do both I suppose. :-)
PopcornPam
@PopcornPam_twitter
Sep 18 2017 16:58
true fact! And just to double check.. if Seattle is just 4 days, not worth changing time zones? (Also, not planning on bringing my laptop with me). I'll try installing dev this evening. Changing branch in the middle of the day on the first day of hte semester seems like a poor plan
Scott Leibrand
@scottleibrand
Sep 18 2017 16:59
yeah, if it's a short trip and you're going to be totally disrupting your schedule anyway, it might be easiest to leave the timezones as-is so there's one less change to keep track of...
you have two rigs so you can change just one of them and make sure it's working well?
Phil
@drnoname82_twitter
Sep 18 2017 17:01
Today we had a somewhat dangerous incident due to a faulty G5 sensor. Actual BG was 75, all of a sudden Xdrip values went up to 400 along with extreme noise levels. OpenAPS saw an increase in BG and gave insulin. As this was during the day, we noticed it and stopped the sensor immediately, but this could have caused harm had it gone unnoticed for a while.
Would it make sense to be able to configure OpenAPS not to loop on G5 values with high noise levels? That would have prevented the incident (And of course I was glad we always stuck to conservative safety settings)
PopcornPam
@PopcornPam_twitter
Sep 18 2017 17:01
Yes
Scott Leibrand
@scottleibrand
Sep 18 2017 17:03
@drnoname82_twitter oref0 is built to work with data direct from Dexcom, which in that case stops reporting SGVs and only reports "raw" values. oref0 knows not to loop on raw values if they're above a certain threshold (perhaps ~160?). it sounds like xDrip is failing to filter out too-noisy values from the sensor like it should. if possible, it'd be best to fix that within xDrip, as oref0 doesn't (AFAIK) have direct access to noise levels
a normal flat BG reading from a plugged-in G4 receiver configured for g4-upload looks like this:
  {
    "trend_arrow": "FLAT",
    "display_time": "2017-09-18T10:04:03-07:00",
    "system_time": "2017-09-18T18:06:50-07:00",
    "raw": 120,
    "direction": "Flat",
    "sgv": 116,
    "dateString": "2017-09-18T10:04:03-07:00",
    "device": "openaps://pi2-erf/cgm",
    "unfiltered": 139648,
    "rssi": 175,
    "date": 1505754243000,
    "filtered": 135424,
    "type": "sgv",
    "glucose": 116
  }
if the receiver starts showing ??? due to high noise levels, the "sgv" and "glucose" values disappear, leaving just "raw"
what does a glucose stanza from xDrip look like?
Phil
@drnoname82_twitter
Sep 18 2017 17:08
That is what I thought. @jamorham suggested to filter the values in the AP systems. I would prefer not to see any values outside of XDrip+ (Nightscout, OpenAPS etc.) which most likely are unreliable.
Tim Street
@tim2000s
Sep 18 2017 17:09
@PopcornPam it's not clear what your settings are. That may help. But also, the algorithm's prediction data suggests it is doing the right thing, which rather suggests that there's something else out of kilter.
Phil
@drnoname82_twitter
Sep 18 2017 17:11
{
    "_id": {
        "$oid": "59bff9b4995a3d4497227e3b"
    },
    "device": "xDrip-DexcomG5",
    "date": 1505753522192,
    "dateString": "2017-09-18T18:52:02.192+0200",
    "sgv": 137,
    "direction": "Flat",
    "type": "sgv",
    "filtered": 170002.25946,
    "unfiltered": 176478.96141000002,
    "rssi": 100,
    "noise": 1,
    "delta": 4.307,
    "sysTime": "2017-09-18T18:52:02.192+0200"
}
This is how an XDrip entry looks. I am just noticing that noise is always recorded as "1" no matter what the noise level reported by XDrip is.
GreatEscape
@GreatEscape
Sep 18 2017 17:13
Hi, I'm wondering if someone might be willing to take a quick look at my son's Nightscout site from today? This is our fourth round resetting autotune. Things start off okay but after 4-5 days it seems to adjust itself to the point that when AMA turns on, it always predicts BG will plummet. This causes 0 temping. With no basal, the BG climbs high and won't come down until COB is 0 and high temping starts. My son's school nurse is suggesting we kick OpenAPS to the curb as she is frustrated seeing him go so high so often. I have previously reduced the autotune_isf_adjustmentFraction to 0.3. I am thinking of reducing it lower, turning off AMA altogether and resetting autotune again.
katie disimone
@Kdisimone
Sep 18 2017 17:16
@GreatEscape why not just disable autotune and run your loop without it?
Dana Lewis
@danamlewis
Sep 18 2017 17:16
@GreatEscape I think you're confusing AMA and autotune impacts. (on a call, will hop back on more later, but wanted to flag that for any follow up discussions)
GreatEscape
@GreatEscape
Sep 18 2017 17:21
I'm getting to that point @Kdisimone but I have found that - outside of meals - we seem to have more success with Autotune on. I guess my thought is if I can learn more and correct the issue, I'd like to try that rather than forego helpful oref0 features that maybe just need to be tuned or understood better.
Scott Leibrand
@scottleibrand
Sep 18 2017 17:21
@drnoname82_twitter @jamorham regardless of where we eventually decide to filter out the bad (noisy) values, sounds like we'll need some changes in xDrip, as it's not passing along any information that oref0 could use to decide whether to filter. As I mentioned before, we probably want xDrip to behave similarly to how a Dexcom G4 receiver works: only display the "sgv" record if noise isn't high, and replace that with a "raw" record if the noise is high or for some other reason (new sensor, etc.) the data is untrustworthy.
@GreatEscape can you paste the top of your autotune recommendations report (the ISF/CR section) so we can see how far it is from your pump settings?
Tim Street
@tim2000s
Sep 18 2017 17:25
XDrip interprets the data differently anyway. The value you get from it is the raw data (at least if using G5)
GreatEscape
@GreatEscape
Sep 18 2017 17:25
@scottleibrand
ISF | 216.000 | 257.529
CSF | n/a | 9.076
Carb Ratio | 23.000 | 28.375
I'm in mmol; ISF is always reported in mg/dl though
Phil
@drnoname82_twitter
Sep 18 2017 17:29
@scottleibrand I will take your suggestion to the xDrip-plus room, thanks a lot ;)
@tim2000s But would it hurt to either a) report values as sgv (low noise) or raw (high noise) or b) to be able to configure whether high noise values should be uploaded/broadcast at all
Scott Leibrand
@scottleibrand
Sep 18 2017 17:31
@GreatEscape sounds like a lower autotune_isf_adjustmentFraction (and/or more aggressive pump ISF/CR settings, if you think those might be too high) might help somewhat for now. I'm also working on a new autotune algorithm that autotunes CR directly (rather than via ISF and CSF), which might help somewhat. I'll post here later when I'm ready to collect NS URLs from people who've been having trouble with autotune so I can check whether the new algorithm is an improvement in those cases.
Tim Street
@tim2000s
Sep 18 2017 17:31
The issue is that the filtered data is about 10-15 mins behind the raw, which is why it isn't used.
Scott Leibrand
@scottleibrand
Sep 18 2017 17:32
I don't think filtered vs. unfiltered is really a concern here. what matters isn't how SGV is derived (from filtered or unfiltered) but rather whether it's reported in high-noise situations
katie disimone
@Kdisimone
Sep 18 2017 17:38
@GreatEscape also, your ISF is being autotuned above 255, so make sure you've implemented the fix for ISFs above 255, or things don't work properly. http://openaps.readthedocs.io/en/latest/docs/Build%20Your%20Rig/OpenAPS-install.html#temp-basals-6-3-isf-255-or-carb-ratio-25-with-a-x23-or-x54
Scott Leibrand
@scottleibrand
Sep 18 2017 17:42
@Kdisimone I believe those bugs only affect you if pump ISF and pump CR are above 255 / 25: autotuned values wouldn't be affected as they don't have to be read or written to the pump
but still good to run the latest decocare just in case you have to set higher ones on the pump
katie disimone
@Kdisimone
Sep 18 2017 17:43
Ahhh, good to know. Thought I'd seen someone I was troubleshooting have problems when it autosensed over 255....maybe autotune vs autosens?
GreatEscape
@GreatEscape
Sep 18 2017 17:58
@Kdisimone and @scottleibrand - even after I rant the "decocare fix" I found that if my carb ratio was 25, autotune would report it as 0.25 and CSF would be messed up. I've wanted to try running the script again to see if I have the same result on a carb ration 25 or over. Perhaps it is a bug with the fix script and mmol. Or maybe I just messed something up. But, considering that autotune is adjusting carb ratio above 25, i wondered if this might be impacting my results. @scottleibrand confirms my suspicions that it shouldn't matter and is likely not the case because autotuned values wouldn't be impacted.
I thought that would be the code to post the image to gitter. In any case, the above link is a screenshot to today. The purple lines (AMA lines can I call them?) seem to always go to 0 even if he is quite high. As he's been rising, it has been 0 temping, causing a faster rise. My question is: would it be the CSF that is resulting in this prediction (since it is happening while the 3 prediction lines are active?)
katie disimone
@Kdisimone
Sep 18 2017 18:18
@GreatEscape if you still had .25 being reported, I would make sure you do the fix over again. I've tested the fix on several rigs and pumps and it's a solid fix. Perhaps something went wrong in the way you entered it? Like I wrote previously, I was pretty sure I saw autosens not working if the adjustment took it outside the range, and the fix was needed. Are you running autosens too?
GreatEscape
@GreatEscape
Sep 18 2017 18:20
You can see in the image, it starts high temping after COB goes to 0. But then he had his second nutrition break of the day and the 0 temping started again when more carbs came on board.
@Kdisimone Yes, I am running autosens too. I will apply the fix again before midnight tonight and put the CR up to 25. I will look at autotune after midnight to see if the 25 is reported as 25 or .25
GreatEscape
@GreatEscape
Sep 18 2017 18:27
Question on resetting autotune. I've seen the recommendation before that if one wants to reset autotune they can delete/rename the autotune directory. I've found that I also have to delete/rename the autotune.json file in settings for openaps to stop relying on old numbers. Does that sound right? I've wondered why deleting autotune.json in settings isn't mentioned along with deleting the autotune directory.
Scott Leibrand
@scottleibrand
Sep 18 2017 18:29
@GreatEscape what were the IOB and COB at the time of that screenshot? You may want to screenshot the OpenAPS pill at that same time.
@GreatEscape I believe you are correct. if you don't remove the one in settings as well, I think it will overwrite it at the next nightly autotune run, but doing so manually accelerates that, so it'd be good to include that in the docs too
GreatEscape
@GreatEscape
Sep 18 2017 18:38
@scottleibrand , here is an update that includes the openaps pill. He's been holding steady at 257 (14.3 mmol) for a while, but it still thinks that he's going to nosedive.
I'm wondering if it is CSF or ISF that is resulting in this prediction
Darryl Schick
@darrylschick
Sep 18 2017 19:10
This message was deleted
Untitled.jpg
ISF shows at 100 but the pump is set to 25. Is this just a display issue or is the pump setting not being read properly?
Dana Lewis
@danamlewis
Sep 18 2017 19:12
@darrylschick - check what you have sensitivity set to in the Nightscout profile
Darryl Schick
@darrylschick
Sep 18 2017 19:14
Do you mean in the Heroku settings? I don't see it there.
Dana Lewis
@danamlewis
Sep 18 2017 19:15
no, in Nightscout profile
Darryl Schick
@darrylschick
Sep 18 2017 19:15
Where is that?
Dana Lewis
@danamlewis
Sep 18 2017 19:16
top right corner of your NS page, click the three hamburger lines
and click on profile editor in the top
Darryl Schick
@darrylschick
Sep 18 2017 19:17
Oh I see it now. It is set to 100 there. Does that matter for the APS insulin computations, or is it just for display purposes? In other words does the APS use the pump setting or the Nightscout profile setting for ISF?
Dana Lewis
@danamlewis
Sep 18 2017 19:17
nothing in OpenAPS ever uses anything NS profile
so that's why I wanted to flag that - it's just display. don't look at anything in the basal pill dropdown. Just the OpenAPS pill.
Darryl Schick
@darrylschick
Sep 18 2017 19:19
OK I understand now, thanks!
Scott Leibrand
@scottleibrand
Sep 18 2017 19:27
@GreatEscape so you can do the math manually: ISF of 14.5 (mmol/L/U) * 18 (mg/dL) / (mmol/L) * IOB 1.47 U = projected drop of 384 mg/dL from IOB. CSF 9.076 mg/dL/g * COB 15 (g) = 136 mg/dL rise expected from remaining COB. net 384 - 136 = a drop of 248 mg/dL or 13.8 mmol/L
cameronrenwick
@cameronrenwick
Sep 18 2017 19:32
@darrylschick I'll tag in here, late to the party... but do yourself a favour and update the profile to match what your openAPS values are. It does take a bit more work to stay on top of things but that way it's consistent. At some point you may be showing an endo your interactive screen and to say "oh, that value doesn't matter" (which is technically true if you haven't kept it up to date..) it just can lead to problems IMO....
GreatEscape
@GreatEscape
Sep 18 2017 19:56
thank you @scottleibrand , really appreciate that. I have saved this to have handy. It will help me understand how much of an impact the changing ISF has on the prediction. Perhaps what @Kdisimone suggests is correct and autosens doesn't work properly without the Decocare fix applied. I thought I had applied the fix but I will try again.
Darryl Schick
@darrylschick
Sep 18 2017 20:34
@cameronrenwick I understand... She's hasn't actually seen an endo since diagnosed 10 years ago so she and I are the only people who see the screen. Mostly we just wanted to understand if that setting figured into the calculations.
sadafakahmad
@sadafakahmad
Sep 18 2017 21:26
@Kdisimone
dan lightner
@danimaniac
Sep 18 2017 21:28
FINALLY had my first post-openAPS endo appointment this morning. I sent the office some OpenAPS related info and some xDrip+ 90 day, 60 day, and 30 day graphs as well as statistics ahead of the appointment. You can read the email I sent them if you scroll up. Anyway, it went quite well. To paraphrase, "It looks like you guys have built quite a smart system. You won't see me endorsing it on television of course, but keep doing what you have been doing. What do you need from me?"
sadafakahmad
@sadafakahmad
Sep 18 2017 21:29
@Kdisimone: regarding the "wl_host_event: mismatched OUI, bailing" message I'm getting, you mentioned Avahi. I'm using Edison preflashed with Jubilinux along with Heroku and Nightscout. How does Avahi fit into all that?
Dana Lewis
@danamlewis
Sep 18 2017 21:34
@danimaniac <3 glad it went well!
PopcornPam
@PopcornPam_twitter
Sep 18 2017 21:39
Hey, I am getting stuck trying to install dev... Updating 0cbfb35..392f032
error: Your local changes to the following files would be overwritten by merge:
bin/oref0_nightscout_check.py
Please, commit your changes or stash them before you can merge.
Aborting
Couldn't pull latest oref0
git stash does not work, and I can't figure out where to find bin/oref0_nightscout_check.py
Scott Leibrand
@scottleibrand
Sep 18 2017 22:03
"does not work" how?
PopcornPam
@PopcornPam_twitter
Sep 18 2017 22:04
fatal: Not a git repository (or any of the parent directories): .git
Scott Leibrand
@scottleibrand
Sep 18 2017 22:04
you need to cd ~/src/oref0 first
umich10
@umich10
Sep 18 2017 22:06
using autotune and the ISF is setting itself higher and higher and my son's been running really high BGs. i feel like i've missed setting a "cap" somewhere - or maybe i need to just run it as a one-off?
PopcornPam
@PopcornPam_twitter
Sep 18 2017 22:06
ah.. thanks!
Dana Lewis
@danamlewis
Sep 18 2017 22:06
@umich10 assuming you haven't changed the min and max multipliers; it will cap at 20% higher than what's on the pump (so don't change what's on the pump if you don't agree with it going higher)
if you haven't changed it, and want to further limit it, you can change http://openaps.readthedocs.io/en/latest/docs/While%20You%20Wait%20For%20Gear/preferences-and-safety-settings.html#autotune-isf-adjustmentfraction autotune-isf-adjustment-fraction down from 0.5 to somewhere between that and 0 (or 0) if you want to allow carb ratio and basals, but not as much leeway for ISF
Scott Leibrand
@scottleibrand
Sep 18 2017 22:10
in oref0 master and dev up to 0.5.x (i.e. anything other than bleeding-edge 0.6.0-dev) the CR is autotuned based off the ISF, so lowering autotune-isf-adjustment-fraction will also constrain the CR as well. in 0.6.0-dev there is a new method for directly autotuning CR that isn't susceptible to the ISF inflation that autotune-isf-adjustment-fraction works around.
PopcornPam
@PopcornPam_twitter
Sep 18 2017 22:11
Where can I find the insulin profile?
umich10
@umich10
Sep 18 2017 22:13
excellent. thanks. i haven't changed the pump so i'll look at 0.6.0-dev b/c i really like the CR tuning. is that version also working in multiple CRs?
Scott Leibrand
@scottleibrand
Sep 18 2017 22:19
nope, no multiple CR support yet, but this might make it easier if we find someone with a physiological need for it (still haven't yet)
katie disimone
@Kdisimone
Sep 18 2017 22:30
Cross-posting in Loop too…181 days of THANK YOU to both systems this past year!!! What a banner year
Book1.jpg
Dana Lewis
@danamlewis
Sep 18 2017 22:30
:green_heart: :confetti_ball: :clap:
PopcornPam
@PopcornPam_twitter
Sep 18 2017 22:36
Where can I go to understand what is in, and how to use features in dev? I don't really understand the navigation in github
Scott Leibrand
@scottleibrand
Sep 18 2017 22:36
there's almost nothing in dev yet
Dana Lewis
@danamlewis
Sep 18 2017 22:36
until we merge 0.6.0-dev in, there's not much changed in there
Scott Leibrand
@scottleibrand
Sep 18 2017 22:37
https://github.com/openaps/oref0/pulls is a good place to see what's pending
Dana Lewis
@danamlewis
Sep 18 2017 22:37
@tim2000s started some notes that will be there soon, which you can see here if you're trying out 0.6.0-dev: https://github.com/openaps/oref0/pull/592#issuecomment-330027634 (thanks Tim!)
PopcornPam
@PopcornPam_twitter
Sep 18 2017 22:37
Looking for ultra-rapid exponential insulin activity curve... trying the shotgun approach