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

15th
May 2015
Scott Leibrand
@scottleibrand
May 15 2015 00:25
I love the way @bewest thinks so far ahead on so many of these things. :-)
diabeticgonewild
@diabeticgonewild
May 15 2015 00:26
Agreed :)
Jason Calabrese
@jasoncalabrese
May 15 2015 00:28
@scottleibrand let me know if you want to meetup tonight, headed home now but could go down to La Jolla later
diabeticgonewild
@diabeticgonewild
May 15 2015 02:45
Has anyone looked into wusb (wireless USB) ( http://en.m.wikipedia.org/wiki/Wireless_USB ) devices or IP over USB (with like a wifi device that does not require driver installation) to eliminate the need to carry the RPi2 (or other boards such as mobile ITX form factors or others like some of these http://www.supermicro.com/products/motherboard/Atom/ ). I mean if we could just connect the Dexcom to a USB transceiver of some sort that communicates wirelessly with the development board (same goes with the carelink stick) I think we could get away with avoiding some RF engineering challenges, plus bulk
IEEE 802.11ad is another option coming out soon
diabeticgonewild
@diabeticgonewild
May 15 2015 02:51
On a RPi2 obviously
RPi I mean (can't edit this for some reason on my phone). That post is worth a read though.
diabeticgonewild
@diabeticgonewild
May 15 2015 03:23
Well I looked around and it looks to risky
Ben West
@bewest
May 15 2015 03:30
openaps build status
diabeticgonewild
@diabeticgonewild
May 15 2015 03:32
Nice!
I found pico ITX (a form factor for computer cases) for a small development board that runs on the Intel x86 architecture they can run MATLAB but I was looking at price and ouch http://axiomtek.com/m/Default.aspx?MenuId=Products&FunctionId=ProductView&ItemId=1105
But could run offline theoretically on MATLAB For the work I am doing with a #VirtualPancreas
With a controller...
diabeticgonewild
@diabeticgonewild
May 15 2015 03:53
More details and prices... Don't know if offline is worth it or not but it works theoretically http://www.mouser.com/Search/m_Search.aspx?Ntk=P_MarCom&Ntt=113406332
diabeticgonewild
@diabeticgonewild
May 15 2015 04:35
Been sipping on opium in a diet coke...LOL.
Damn I feel good.
Ben West
@bewest
May 15 2015 07:23
@danamlewis , @scottleibrand , I've been studying our mini-hackathon from last week, I think there are at least 5 isolated "uses" in order to structure things in terms of openaps... what I'd like to do is create openaps-diyps module that "deconstructs" everything that happens into independent, safe commands, https://gist.github.com/bewest/7b679cc2169a4ec5e293
diabeticgonewild
@diabeticgonewild
May 15 2015 18:22
I think when I do chemo next Wednesday I am going to bring a printed out copy of glucose infusion rate graphs from the studies that John Walsh has cited in previous papers, but I will primarily be referencing that article that we got in an argument over when I was PMSing for Apidra, Humalog, and Novolog plus a ruler. I will interpolate the values on the x and y (orthogonal) axes and derive accurate IOB calculations for those insulins based on certain euglycemic clamp studies by fitting equations in a certain way to the curves according to that paper I referenced in the past. I know that it is merely an approximation so please don't judge it. This is not intended for mathematical modeling or #OpenAPS , but for Nightscout users. This is a mini project compared to the work I am currently doing but it will benefit several people besides myself. I don't think we have to worry about IP concerns like @scottleibrand has brought up because this was derived easily through equation fitting from pump data, but he is reasonable and I do understand the concern. Hence using studies and fitting data from particular insulins versus pumps is an alternative.
In an #OpenAPS , rate of glucose change can be used to "fit" a IOB or insulin activity curve, so this is a totally different situation. I am just trying to integrate standard data used in diabetes care (not trying to build "new features" that may or may not be useful) on the Nightscout platform nicely for people, as this can be a deal beaker for using this versus the share.
Matthias Granberry
@mgranberry
May 15 2015 18:26
@diabeticgonewild when you do it, try and use a monotone hermite spline. You can extract a passable insulin activity estimate from a monotone spline through a finite difference, and it is possible to use certain optimizations around it because they are continuous and differentiable. Also low-order fittings tend to be better-behaved in the real world. Anything more than cubic is almost guaranteed to be overfitted, especially if you try and fit the whole curve to a single equation.
diabeticgonewild
@diabeticgonewild
May 15 2015 18:26
IOB for particular insulins would be cool, on the Nightscout platform, that's for sure.
But, gone wild again so I will stop rambling.
Matthias Granberry
@mgranberry
May 15 2015 18:27
The graphs look like a 2-3 part quadratic spline would be easy, but you can match derivatives and set tangents to ensure monotonicity with a cubic spline.
diabeticgonewild
@diabeticgonewild
May 15 2015 18:30
Thank you. They are also described as "curvilinear". I used polynomial fitting in MATLAB last time and it worked very nicely without concerns about overfitting, from what I could see and what I programmed in iob.js . That is extremely helpful. Thanks again!
Matthias Granberry
@mgranberry
May 15 2015 18:31
one thing to consider is that the pediatric population is overrepresented in NightScout and I don't know of any studies considering them. They will also tend to retain some amount of residual insulin production so the tails of insulin activity may be masked.
diabeticgonewild
@diabeticgonewild
May 15 2015 18:33
But it was like 5th and 6th degree polynomials, depending on the pump curve I fitted last time around. The key is to be diligent, patient, and pay attention when fitting. It is not complicated, the fitting, at least to me, but it's more of finding the "best fit" and not overfitting. But if you find the right fitting algorithm/technique, at least in MATLAB, for whatever you are doing, you are pretty much set
I can compare it against the pump IOB algorithms and do some reading on the pediatric population.
That is true. Thank you.
Matthias Granberry
@mgranberry
May 15 2015 18:34
you lose some nice properties when you start going down that path. Does MATLAB allow fitting to splines/piece-wise functions?
diabeticgonewild
@diabeticgonewild
May 15 2015 18:35
Yep
Matthias Granberry
@mgranberry
May 15 2015 18:35
Give it a try with lower-order polynomials. You might find that 2-3 points will give you a good fit at a low order.
though I personally think tools for constructing a personal function from tagged records is probably the way to go. It would be great to have personalized abdominal/thigh IOB functions
diabeticgonewild
@diabeticgonewild
May 15 2015 18:37
See http://www.mathworks.com/help/matlab/ref/pchip.html , included monotonicity (sorry, missed your Initial post and just read it)
Will take your advice! :)
Yeah that's what I would want to do but simple is best for Nightscout, for now. Perhaps an experimental patch for what you suggested would be most realistic for the stuff you recommended with thighs and stuff
I just want to keep it familiar for now for the user. But the stuff that @bewest referenced when I was in a bad mood when PMSing looked like a good solution with respect to personalization. It was basic math pretty much. I knew what to do with it just by looking at it.
So I could build upon that. And then build upon what I intend on working on with personalization.
diabeticgonewild
@diabeticgonewild
May 15 2015 18:42
It's not hard at all. Seriously. Easy stuff compared to #VirtualPancreas work and even some #OpenAPS stuff. It's good stuff to focus on to keep productive when not feeling great. It's not taxing on my brain at all, just requires energy.
Matthias Granberry
@mgranberry
May 15 2015 18:43
A monotone cubic hermite spline is actually pretty easy to do, too. http://en.wikipedia.org/wiki/Monotone_cubic_interpolation
and I would probably fit the derivative to a spline and numerically integrate that to derive an IOB function from real data
pretty sure it would work, but I've never actually tried it on any of my own data
diabeticgonewild
@diabeticgonewild
May 15 2015 18:45
Yeah, polynomials and rate of change/derivatives is the worst of it from what I can tell. But MATLAB will fit it in that form and I just have to "check" my work thouroughly
But all of the statistical data plus visualization and "goodness of fit" will be there so it will work very nicely in MATLAB
Matthias Granberry
@mgranberry
May 15 2015 18:49
the same technique could be paired with a HRM to quantify changes in insulin sensitivity wrt to activity, but there are some nonlinearities that become more important there because the liver egp-glucose-insulin interaction gets kind of complicated
too bad CGMs didn't exist in their current form back when I was in grad school. I probably would have tinkered my way through a Ph. D
diabeticgonewild
@diabeticgonewild
May 15 2015 18:51
Yeah IOB is just (original bolus delivered within current DIA timeframe)*((integral from t=0 when the bolus was delivered to t=time elapsed from t=0 of the glucose infusion rate graph function--that is modified according to the paper I referenced when I was PMSing)/(integral from t=0 when the bolus was delivered initially to t=DIA time in minutes of the glucose infusion rate graph function--that is modified according to the paper I referenced when I was PMSing))
Remember I figured that out when I was driving...but math is like common sense to me. Crazy still but I am talented at math.
Hovorka put out a paper on physical activity with respect to exertion recently. I can send it your way but I am on a trip and don't have access to it.
Yeah man I am struggling to get through undergrad. Would have blasted through even while sick if I didn't end up getting so sick in 2011-2012...and just getting sicker... I feel your lain
*pain
Matthias Granberry
@mgranberry
May 15 2015 18:54
I don't think the research is helpful there. Nobody knows WTF goes on when T1s exercise. It's all trial and error, but I suspect the intrapatient variability is limited.
Actually I don't regret leaving, but I enjoyed research back in the day.
diabeticgonewild
@diabeticgonewild
May 15 2015 18:54
IDK they have been modeling VO2 relatively recently
So theoretically...it's coming to fruition.
Rate of perceived exertion and HR stuff can be pretty comparable unless you have autoimmune dysautonomia (autonomic neuropathy) or just dysautonomia in general
Matthias Granberry
@mgranberry
May 15 2015 18:56
sounds to me like they are having trouble and are throwing in slack terms, but I could be wrong. I don't trust maximal models in general, and they aren't likely to be terribly useful in practice. Give me a minimal model with 2-4 knobs and a robust safety system any day.
diabeticgonewild
@diabeticgonewild
May 15 2015 18:57
There are problems mathematically with minimal models
With respect to APs
That stuff is relatively outdated and they are using stochastic stuff now in lieu of thst
Matthias Granberry
@mgranberry
May 15 2015 18:58
I'm honestly just interested in doing better than I can unaided and waking me at times like last night where an infusion set wasn't working and I had to swap sites at 2am
sensors and insulin delivery still aren't good enough for perfect closed-loop control in a noisy system like a bag of meat.
diabeticgonewild
@diabeticgonewild
May 15 2015 19:00
Controllers now can "figure it out" and "tell you"
That's what controllers do
Same with detecting conclusions and sensor failures
Matthias Granberry
@mgranberry
May 15 2015 19:01
sure. That doesn't need a maximal model, though.
diabeticgonewild
@diabeticgonewild
May 15 2015 19:01
Nope
Agreed
But it's not that simple. It's built upon several things, but overall it's a nonlinear model predictive controller using stochastic differential equations as models with Bayesian prediction plus algorithms/safety features for bolus calculators, occlusion detection, glucose prediction, sensor failure detectors, error propagation and estimates, etc.
Matthias Granberry
@mgranberry
May 15 2015 19:04
and I'll grant you that a maximum likelihood Bayesian approach can tune parameters very well, even in a complicated model and even point to likely problems, even pooled insulin if one of the factors for "compartment" transitions goes outside of a range.
but the more variables you add to a system, the more likely it is that the system will optimize itself into a good fit in an insane state. Complex models are hard to work with.
The more variables you add, the more work you need to put in on your system to determine feasibility. These things get hard to reason about very quickly. Simple linear systems, if they work, tend to work predictably and well
diabeticgonewild
@diabeticgonewild
May 15 2015 19:09
Well remember that I am doing #OpenAPS first and I am going to do my work with #VirtualPancreas . There are issues with APs and minimal models. I am not an appli
Applied mathematician
I am doing things a certain way for a reason
There were issues with "fluke values" with about 10% of people if I recall in minimal models with AP work. Like really off values that did not fit at all.
I meant maximum likelihood
But I am distracted. Ordering lunch for my mom and I and we are doing work at my grandmas house
Matthias Granberry
@mgranberry
May 15 2015 19:14
have fun. It's been interesting. Hopefully Lane et al. have some luck with their APS. Looks like they got a bridge loan recently.
diabeticgonewild
@diabeticgonewild
May 15 2015 19:15
Remember that I am sick and I have serious health issues. I have the determination but I can't work on things as efficiently or as I want to. Or I would be working on the controller by now.
I mean I know, at least in my mind that the mathematical models are solvable. I wrote out how they are solvable.
The controller builds upon the mathematical models.
There is a PID controller for an AP published for MATLAB that is publicly available...but I go all out on things and I have the patience to the work I want to. I just get frustrated with my health problems.
diabeticgonewild
@diabeticgonewild
May 15 2015 19:24
J
See: http://www.mathworks.com/matlabcentral/fileexchange/48019-blood-glucose-regulation-in-a-type-i-diabetic-in-simulink . All you would have to do is interface this with the MATLAB Database Toolbox and #OpenAPS tools
Or you could do @bustavo did and just put/place a Fuzzy logic controller in Simulink in MATLAB and interface it with #OpenAPS tools. Requires no math.
diabeticgonewild
@diabeticgonewild
May 15 2015 19:30
Like it's really no big deal to me that it might take awhile to get my particular work done. I have the patience and I go all out on things and you should be glad that I want to do things a certain way.
Plus the stuff I am working on had already been proven to work in studies. I am not "playing around" with models. I am "reprogramming" a copy that can be used that is accessible. I have thought this out and I am not blindly programming stuff. I am using stuff that had been used in clinical trials.
Matthias Granberry
@mgranberry
May 15 2015 19:35
sure
diabeticgonewild
@diabeticgonewild
May 15 2015 19:41
Seriously, just reprogramming stuff already used.
Matthias Granberry
@mgranberry
May 15 2015 19:44
I coded a system of dosing adjustment decisions into a logic satisfiability system back while I was in school. It sounds a lot like what bustavo did, although with a different set of outputs.
diabeticgonewild
@diabeticgonewild
May 15 2015 19:44
That's cool. :)
Matthias Granberry
@mgranberry
May 15 2015 19:53
I need to go fuzz the tslim and see how to dose insulin. I'm 99% sure I can do it.
once they release the firmware updater it's going to get even easier
diabeticgonewild
@diabeticgonewild
May 15 2015 19:54
That will be so awesome!
Jason Calabrese
@jasoncalabrese
May 15 2015 20:00
with the firmware uploader we'd still need to get a custom firmware right? and it wouldn't work with dexcom integrated version
Matthias Granberry
@mgranberry
May 15 2015 20:01
to enable BT? Yeah. It just makes it easier to disassemble bits of it and understand what's going on. And the firmware upload command is hidden behind their manufacturer test mode, and knowing what the magic is to enable that is helpful
they have commands to issue temp basals/boluses. I think they require a low-level big-little microbolus command, but there are events in the log that I think map to the command
Jason Calabrese
@jasoncalabrese
May 15 2015 20:07
I can't think of a way to really make a 522 work for me, mostly just the range issue with carelink
Matthias Granberry
@mgranberry
May 15 2015 20:07
I think an xDrip-style BT bridge is going to have to be the answer there
Jason Calabrese
@jasoncalabrese
May 15 2015 20:08
yeah, thats what I was thinking too
the new pi @danamlewis had last night isn't big, but too much to wear all the time
Matthias Granberry
@mgranberry
May 15 2015 20:09
that or cutting an antenna trace and attaching a real antenna, but balancing one is a big task
Jason Calabrese
@jasoncalabrese
May 15 2015 20:09
with BT range it wouldn't be much of a problem
Matthias Granberry
@mgranberry
May 15 2015 20:10
do carelink sticks work w/ Android at all? I saw some code in a branch of android-uploader but it didn't look like it did anything.
Jason Calabrese
@jasoncalabrese
May 15 2015 20:10
not with lolipop
kernel issue, but does work with kitkat
I like the idea of being able to use simple scripts like the current pi setup
and it's portable enough, if the range was just better
could have it in the same room, like the normal NS rig
Matthias Granberry
@mgranberry
May 15 2015 20:13
is power a problem with the Pi?
Jason Calabrese
@jasoncalabrese
May 15 2015 20:13
not sure how much time they're getting on the battery, but looked like a big battery
wonder if we could find someone to do the antenna work
Ben West
@bewest
May 15 2015 20:14
yeah, it works
Jason Calabrese
@jasoncalabrese
May 15 2015 20:15
antenna mod seems like it could be the faster path
Ben West
@bewest
May 15 2015 20:15
I'm working on an intel edison version
just need a sleek case to bundle battery and some other things
the edison is much much smaller
1/3rd size of pi
Jason Calabrese
@jasoncalabrese
May 15 2015 20:15
but still wearing 3 devices could be an issue
Matthias Granberry
@mgranberry
May 15 2015 20:16
that might be acceptable. That's about the right size for an SPI belt
Jason Calabrese
@jasoncalabrese
May 15 2015 20:16
unless it was xdrip sized
Ben West
@bewest
May 15 2015 20:16
hopefully it's xdrip sized and goes in a bag somewhere
I got an e-ink kit, hopeful to design e-ink as part of waterproof case eventually
or maybe make it the pump holder/clip somehow
here's some of the blocks https://www.sparkfun.com/products/13045
Jason Calabrese
@jasoncalabrese
May 15 2015 20:18
if there was a way to get antenna work done for carelink that might be a fast path
Matthias Granberry
@mgranberry
May 15 2015 20:18
that connector looks like a PITA to solder
Jason Calabrese
@jasoncalabrese
May 15 2015 20:18
the pi is ok to put in a bag
Ben West
@bewest
May 15 2015 20:18
the usb stick casing pops open, exposing a board
other option is getting the cc11111 chips onto one of the edison blocks
then your using your own antenna
Matthias Granberry
@mgranberry
May 15 2015 20:19
@jasoncalabrese that's fiddly work and would probably require a scope and tuning.
Ben West
@bewest
May 15 2015 20:19
I've been reviewing at jberian's mmcommander work, looks like it's usable for lot of things we need it for
Jason Calabrese
@jasoncalabrese
May 15 2015 20:19
smaller than I thought from other pictures
Ben West
@bewest
May 15 2015 20:19
yeah, it's really small
I'm going to stack like 20 of those battery blocks on it to see what happens
Matthias Granberry
@mgranberry
May 15 2015 20:21
I think an mmcommander+dexsomething that output to i2c would work great stuck to one of those breakout blocks
Ben West
@bewest
May 15 2015 20:21
ah, was thinking we should just do pcb with it on it
Jason Calabrese
@jasoncalabrese
May 15 2015 20:21
like how jberian use the button to unfilter some commands
Ben West
@bewest
May 15 2015 20:21
make a block with mmcommander and/or wixel on the block
Oliver Schumacher
@oschumac
May 15 2015 20:22
hi im using an edison. with the smal intel breakout board. The edison is realy nice to the battery. i am using an 1000mA lipo giving me 12h runtime.
Matthias Granberry
@mgranberry
May 15 2015 20:23
an i2c interface to whatever mmcommander runs on (and a wixel) would be a quick no-manufacturing path
@oschumac did you try running decocare on it?
Matthias Granberry
@mgranberry
May 15 2015 20:25
RF work is black magic. Takes an art to doing it right
Ben West
@bewest
May 15 2015 20:25
absolutely
Matthias Granberry
@mgranberry
May 15 2015 20:26
especially when you pair it with a 1ghz computer
Ben West
@bewest
May 15 2015 20:26
I read a lot of complaints about edison in that it's hard to integrate with because of the 70 pin computer
Oliver Schumacher
@oschumac
May 15 2015 20:26
i run a wixel via serial port on edison. but needs a level shifter. the edison eorks on 1.8 v. wixel needs 3.3v.
Ben West
@bewest
May 15 2015 20:26
with dramatatically varying loads ;-)
Matthias Granberry
@mgranberry
May 15 2015 20:28
I think for a lot of this, an STM32F4 discovery board would be enough, but the driver situation would be an issue
Oliver Schumacher
@oschumac
May 15 2015 20:28
decocare runs the carelink not. whe need another interface for edison like cc1110 chip. edidon has only an old 3.10 kernel.
Matthias Granberry
@mgranberry
May 15 2015 20:31
What's the problem with the carelink driver? A device id match or something more serious?
Oliver Schumacher
@oschumac
May 15 2015 20:33
on 3.10 kernel for any reason carelink runs not. Standard usbserial driver is used.
a tried to build a new kernel for edison. but the intel kernel Patch is not working on other versions i am waiting for the next Release from intel. support said they will release a new one.
diabeticgonewild
@diabeticgonewild
May 15 2015 20:36
@bewest 123d app lets you print via photo of the device you want to print an 3D enclosure off. Check out the 123d tutorial on Lynda.com . Although it's not free you can "Google" the topics covered in the videos and there are howtos on each particular set. Like if you just google each "section" of the Lynda.com videos. You get great tutorials and you never have to subscribe unless you want to watch videos :)
Oliver Schumacher
@oschumac
May 15 2015 20:37
i think if mmcommander or other serial rf moduls will work fine. Edison is a real cool minicomputer. low power. bt wlan included.
diabeticgonewild
@diabeticgonewild
May 15 2015 20:37
Can you run Linux on Intel Edison? Cause there is a Linux driver for Linux that works well
Awesome!
Ben West
@bewest
May 15 2015 20:39
edison runs stock linux, yeah
diabeticgonewild
@diabeticgonewild
May 15 2015 20:39
You can make clips pretty easily using 3D printing apps I believe.
Matthias Granberry
@mgranberry
May 15 2015 20:39
@diabeticgonewild there are some low-temp thermoplastics that you can get at the local cosplay supply shops, too, that you heat up in warm water and form to whatever shape you want
Ben West
@bewest
May 15 2015 20:40
@oschumac you should be able to use 3.10
Matthias Granberry
@mgranberry
May 15 2015 20:40
It works great for GhettoCases
Oliver Schumacher
@oschumac
May 15 2015 20:40
@diabetesgonewild what kink of driver do you mean.
diabeticgonewild
@diabeticgonewild
May 15 2015 20:40
Cool!
Ben West
@bewest
May 15 2015 20:40
@mgranberry there are two problems
first is simple, the getting usbserial to use the needed vid/pid
Oliver Schumacher
@oschumac
May 15 2015 20:41
@bewest. i did not make it run. Maybe i should give it one more try
Ben West
@bewest
May 15 2015 20:42
it will be... flaky
diabeticgonewild
@diabeticgonewild
May 15 2015 20:42
Let me retrieve the link. I don't have the command to download it as I don't have my computer and I am on a trip but @bewest and @TC2013 likely know the command. The link is on github and it's @bewest work
Ben West
@bewest
May 15 2015 20:42
try forking this and running make and checking the results back in, for me it takes 3 - 4 unplug/replug attempts to make it work
when it doesn't work I get error -71 in logs
diabeticgonewild
@diabeticgonewild
May 15 2015 20:43
@bewest you don't need to necessarily print a clip either. I was thinking of putting this on the back of my t:slim . My pump trainer does something like this http://www.theclip.com/store/Reinforced-Ratchet-Belt-Clip.html
but after 3.8 I think, it works so-so, sometimes
using 3.16, it works most of time, but still flaky
has to do with usb compat issue
Matthias Granberry
@mgranberry
May 15 2015 20:45
@diabeticgonewild you might guess that I'm probably too much a fan of "good enough", but Sugru will stick just about anything to anything for just about forever.
"usb compat" issue? I wonder if disabling xHCI would fix it.
Ben West
@bewest
May 15 2015 20:46
yes
Oliver Schumacher
@oschumac
May 15 2015 20:46
yes that is the point 3.10 an carelink is pain.
Ben West
@bewest
May 15 2015 20:47
if you can get it to use ehci or whatever, it should work reliably
xhci is definitely the issue, it impacts all "legacy usb" devices
Oliver Schumacher
@oschumac
May 15 2015 20:47
ehci ?
Matthias Granberry
@mgranberry
May 15 2015 20:48
hmm. Does it work after it starts, or will it drop out?
Ben West
@bewest
May 15 2015 20:48
hmmm? trouble is usb enumeration up front
Matthias Granberry
@mgranberry
May 15 2015 20:48
xhci is the new USB interface, ehci is the older one
Ben West
@bewest
May 15 2015 20:48
if you can convince it to enumerate all is well, generally
xhci == usb 3.0
it doesn't matter much unless you have all usb 3.0 usb hardware ports and no usb 2.0 ports on the physical bus
Matthias Granberry
@mgranberry
May 15 2015 20:49

it's a hack, but can you power cycle like ```

echo suspend > /sys/bus/usb/devices/1-1/power/level

echo auto > /sys/bus/usb/devices/1-1/power/level

```

Ben West
@bewest
May 15 2015 20:49
yeah, that's what we use in NS uploader
power isn't issue, I think, when xhci tries to enumerate, if it fails, it's like nothing is plugged in
there's no info on the bus about the device
diabeticgonewild
@diabeticgonewild
May 15 2015 20:50
@oschumac I am on a mobile device and due to discrepancies between mobile device and desktop sites I can't find it but it is one of the files on decoding-carelink on GitHub.com that @bewest created for Linux users
Matthias Granberry
@mgranberry
May 15 2015 20:50
oh, that's a pain
Ben West
@bewest
May 15 2015 20:51
it doesn't show up in lsusb or anything
Oliver Schumacher
@oschumac
May 15 2015 20:52
i can make some more test. but i have some fiends home a have to take care of at home.
Ben West
@bewest
May 15 2015 20:52
:-)
interestingly, this usb issue affects, all users of xhci (all os's, mac, windows, and linux)
but linux kernel is actually farthest along on porting all the legacy behavior into the xhci stack
Oliver Schumacher
@oschumac
May 15 2015 20:54
whe can open a thread im open aps group on this. :-)
diabeticgonewild
@diabeticgonewild
May 15 2015 20:57
Isn't the issue documented frequency problems wrt RF and interference from the power pulled from USB 3.0?
Matthias Granberry
@mgranberry
May 15 2015 20:59
it looks like the i2c block supplies 3v3 to sensors as well. It looks like a really easy way to have a cc1111 and a wixel communicate.
just need to plant the idea in the heads of those planning a dexcom bridge to medtronic devices.
diabeticgonewild
@diabeticgonewild
May 15 2015 21:00
@mgranberry that's awesome about Sugru! I can't always get internet on my phone at my grandmas with ATT even though she lives on the top of a hill like in the middle of the city near downtown like 3 miles at most from the space needle. They don't have LTE great with ATT in Seattle surprisingly
Unrelated but Verizon can receive phone calls and stuff in basements (under the "ground break") which shocked us as EE students in college!
So yeah...
I was trying to say Sugru is awesome @mgranberry but mobile data sucks and I tried to send the message multiple times but it was flagged as not sent...
diabeticgonewild
@diabeticgonewild
May 15 2015 21:32
I never really get occlusions but this is certainly good news http://helmsleytrust.org/news/bd-receives-fda-clearance-novel-infusion-set-technology-enhance-use-insulin-pumps (BD creates infusion set to prevent occlusions)
Eric
@ecc1
May 15 2015 21:57
@mgranberry Both cc1111 and cc2511 (wixel) have 2 UARTs each, so easy to connect them that way too. In fact so does the cc2541 (BLE HM-11); I was thinking of having it be the "master" and have it handle I/O to the other two.
Ben West
@bewest
May 15 2015 21:59
nice
@diabeticgonewild, no that's not the issue. the issue is usb enumeration and what happens when one physically inserts the usb stick
diabeticgonewild
@diabeticgonewild
May 15 2015 22:03
Thank you for the clarification :)
Matthias Granberry
@mgranberry
May 15 2015 22:09
I'd love one of those boards as a general purpose RF toolbox even without the diabetes connection. I wonder if seeed
Ben West
@bewest
May 15 2015 22:11
lazy web sparkfun/Fritzing_Parts#5
Matthias Granberry
@mgranberry
May 15 2015 22:23
Eric
@ecc1
May 15 2015 22:40
Interesting: the CC2541 can do both BLE and the proprietary 2.4GHz used by the Dexcom transmitter. Might be possible to make an xDrip-Lite out of just one chip if it could switch back to proprietary mode once every 5 minutes to catch the sensor, then back to BLE ...
Ben West
@bewest
May 15 2015 22:40
yeah, that could be interesting
Eric
@ecc1
May 15 2015 22:40
Has anyone opened up a Dexcom transmitter to see what chip they use? I know people have replaced dead batteries, but I can't find any details.
Ben West
@bewest
May 15 2015 22:41
think @lorelai knows
Ben West
@bewest
May 15 2015 22:48
@ecc1 that chip has a range extender thingy, too
Ben West
@bewest
May 15 2015 23:14
nice form factor