Omnicore DIY hardware
Omnicore does not need special hardware beyond RileyLink. But if you are interested in having less to carry, or need a wateproof solution, read on for more custom solutions...
Here is how to put RileyLink (433) in a Samsung S7 battery case. It allows for use of aaps+omnicore with no hardware beyond the phone in its case. PLEASE NOTE this needs electronics / soldering skills so read thru and check you are confident to attempt it before starting.
-
Open case - remove faux velvet sticker, undo all screws (?6), remove metal plate, undo plastic clips and it separates into two.
-
Desolder the battery. Snip away & desolder the large USB power socket.
-
Prepare RileyLink - snip away and desolder white JST battery connector, solder on wires in its place.
4 b. Remove black socket covers on programming pins.
RileyLink with power wires and socket covers removed
4 c. Desolder antenna, replace it with 18cm piece of solid core wire (see below re tweaking the length). Alternatively, you are likely to get better results by using Edward’s custom antenna – see next section below.
4 d. I also used a silicone conformal coating to waterproof the RL - I was trying to get a silicone spray one but couldn't get it so used this one instead which you apply like nail varnish to both sides - https://rover.ebay.com/rover/0/0/0?mpre=https%3A%2F%2Fwww.ebay.co.uk%2Fulk%2Fitm%2F171365019664 . I made sure to mask the switch & USB connector so not to gum them up.
4 e. I put electrical tape on the back of the RL board to insulate it from the case charging circuit.
RileyLink testing for size / basic orientation in case
- Prepare case - get a small SMD switch (like this: https://rover.ebay.com/rover/0/0/0?mpre=https%3A%2F%2Fwww.ebay.co.uk%2Fulk%2Fitm%2F163792239569 ) and solder wires to it (middle/end poles). Super glue it down to case where the USB opening is. Make sure it is recessed sufficiently so the throw doesn't protrude/get caught but is also accessible with a pen. Tape down wires to stop movement.
Switch with wires glued in place
5 b. Create a protection around switch with sellotape, and make a 'back' to allow gluing of the usb aperture so it doesn't fill too much. Use epoxy to fill this - aim to create a solid window where usb aperture is, without gumming up the switch. Allow 24hrs to set.
Switch window created
5 c. Place RL into case - RL USB power socket down and facing bottom of case. Use a Dremmel with a small routing bit to cut away some case to allow this to fit. I needed to cut a little trough to allow space for RileyLink USB. Make sure switch on RL board turned on.
5 d. Run wire antenna up the left side of case (side furthest form usb aperture), gently bent round top. I needed to use a dremmel to cut away some of the notches to allow it to fit. Tape down the antenna.
Location of antenna along edge (these photos were actually taken after completion of later steps below
5 e. Solder one side of switch to +ve RileyLink power. Cover with heat shrink tube or electrical tape.
- Get a replacement battery. You need a smaller battery (around 4200mah but the physical dimensions are the most important) to put in place of the 6500mah battery removed from the case. Smaller size important so (i) RL will fit in case top-to-bottom and (ii) there is clearance for areal left to right. The 6500mah removed is dimensions 846088 (ie 8.4mm thick, 60mm left-right, 88mm top-bottom). You're looking for something like 845080 or slightly smaller. I found this difficult to source but found a perfect battery size using the battery out of this iphone battery case, which is 4200mah - https://rover.ebay.com/rover/0/0/0?mpre=https%3A%2F%2Fwww.ebay.co.uk%2Fulk%2Fitm%2F293048352985 (that case comes apart in similar way to S7 case... just get the battery out and chuck the rest!).
6 b. Secure battery down with double sided tape on the right hand side nearest the top (ie on USB aperture side).
-
Check for fit with everything. Case charger circuit goes on top of RileyLink. You may need to Dremmel more case away to get Rileylink to seat far down enough.
-
Solder power -ve side all together: battery, RL, case. Ensure switch is off then solder +ve battery, switch and case.
-
Test and adjust antenna. Turn switch on. Lightly place the metal case on top. Get Omnicore going, and use it to establish link with pod. Use 'update' status button to check db loss (between RielyLink and pod, not phone and RL) in conversations tab. Trim antenna down by 1mm and repeat 'update' button. Keep trimming till db worsens (ie gets a bigger -ve). eg If it was -55 then gets to -53 it has got better so trim another 1mm but if results on -56 it has got worse so stop trimming.
-
Shut case. Clip together, replace metal plate, replace screws, replace faux velvet. I managed to damage the clips at the bottom which meant I needed to superglue shut to stop it gaping but shouldn't be necessary if you were careful!
-
The end result... is being able to just carry your normal phone (albeit with a power case, albeit I always had one anyway), and nothing else to manage your diabetes except a pack of sweets... this is a big deal for me! It all works surprisingly well with Omnipod and AAPS+Omnicore. You can see the RL green LED through the epoxy which is a useful reassurance. When you need to charge your phone press the button on the case - this should give 1-2 charges (but don't exhaust the case battery entirely or the RileyLink won't get power).
In summary this is RL in a phone battery case, sharing the battery with phone backup battery; a wire antenna - up the side of the case (with about 12mm clearance from side of battery) and around the top; an accessible switch for RL.
On the question of transmission range - the case does not allow for as good a range as the RL in native case, but it is very much adequate. I found the RL in native case had a maximum range to pod of about 11m, although it was quite directional. This setup, with the wire antenna, has a shorter range (around 4m) but much less directional which is quite helpful for normal use. It works fine for phone in pocket during the day (regardless of pod location) and on bed headboard at night.
Please see details of this in Facebook post https://www.facebook.com/groups/TheLoopedGroup/permalink/2265309447019042/ and video https://youtu.be/zjlRRF2pwE4
Some things to keep in mind... 3D parts are not waterproof. You can think of them as a sieve and under water the pressure is very high even at relatively low depths. This is sufficient to force water into most 3D designs. First off, you can paint the surface with acetone (nail varnish remover) and this will melt it just a tiny bit but enough to seal the plastic.
It is also important to remember that radio waves and water do not play nicely (this is why submarines use sonar not radar and drag their aerials hundreds of metres behind for morse code communications). Google "signal attenuation". Basically the higher the frequency, the worse it gets.
Bluetooth uses 2.4GHz which for most CGM/phone combos gives a maximum range of about 5cm in fresh water and about 1cm in salt water. I use the Unihertz Atom (https://www.unihertz.com/atom) because it is waterproof and has an arm strap so you can strap the phone directly on top of the CGM (so long as you wear it on your arm). I tested this with a G6 and a Miaomiao/Libre with no problems in either a fresh water lake in Finland or the Atlantic ocean.
The omnipod uses 433MHz so the range "should" be 5.5x better than Bluetooth at the same number of decibels. Unfortunately the RL antenna is not very good and struggles to keep contact even in air sometimes. Radio theory tells us a few things to help: thicker aerial wires have less resistance which means more broadcast power; longer aerials work better than shorter ones.
The length of the aerial is very important (think tuning an analogue radio into a radio station, small movements of the dial - which change the aerial length - cause the station to quickly degrade into static). The aerial is normally either 1 wavelength, 1/2 wavelength or 1/4 wavelength.
Unfortunately the speed of light (radio is a kind of light) is also really important. The "optimum" aerial length in metres is (Speed of light in m/s)/(Frequency in Hz or "waves per second") so "metres/s" / "waves/s" = "metres * waves". To get the length, we make waves = 1, 0.5, 0.25 depending on the fraction of a wavelength we want...
Speed of light is 299'792'458 m/s, pod is 433 MHz which is 433'000'000 waves/s (does anyone know the real pod frequency?): 299'792'458/433'000'000 = 0.692361335 metre.waves which gives 69.24cm; 34.62cm; 17.31cm
The main issue is that these aerials are rather long so the solution used in the default RL aerial is to take a 1/4 wavelength aerial and wrap it in a tight coil. Unfortunately this coil is very directional in its signal so the risk of data loss is high if the aerial is not pointing at the pod. Bad data quality/signal strength = error 49 pod death and AAPS/SMB can have a lot of chatter with the pod.
I started experimenting using a very thin wire and a hybrid between a helix and a straight antenna.
For this project you will need the following things
- Some 1.5mm2 single core mains electrical wire; 2) a 5x90 screw or a small screwdriver wwith a 2mm diameter shaft; 3) some string; 4) a ruler; 5) wire cutters; 6) a potato peeler: 7) some tape; 8) some epoxy resin; 9) a Qi-mobile phone wireless charger; 10) some acetone; 11) a 3D printed case (see below)
Measuring wire accurately is almost impossible because it is very hard to keep straight. Take some string and measure that to 26cm in length. Tape it every few cm to the wire. Add 5 mm to the end (you will have lost this much in tension errors) and cut.
Next you need to remove the plastic insulation. The easiest method I found was to use a standard home potato peeler and cut the plastic insulation off by just slicing parallel to the wire away from your body like peeling potatoes (the plastic does not burn or melt because mains wire is fire resistant and it is too long to pull off easily).
Measure 2cm from the end of the wire and put a mark with a pen. Measure an additional 17.3cm from the mark and put another mark, wrap the the wire tightly around the thread of a 5x90 screw. You then need to squash the loops together to make them effective (I don't know why) but it also takes up less space if you do this.
Alternatively, if you have a 2mm diameter small screwdriver then you can wrap the wire around that instead. It is easier to make tight coils on a screwdriver. Coil until you reach the mark. Your coil should be 17.3cm (1/4 wavelength).
Measure the length of the wire below the coil and trim it to 1.73cm (1/40 wavelength). Measure the long end of the wire and trim it to 6.9cm (1/10 wavelenght): 1.73 + 17.3 + 6.9 = 25.93cm (like I said, a stupid length)
Press a kitchen knife gently between each loop of the helix and wiggle just a bit. Check that none of the turns in the coil actually touch each other using a bright light behind the coil (this is important or you get a bad signal again).
Remove the aerial from the RL either by melting the solder (best) or using nail clippers (if it is stuck and really won't budge).
Then solder the new aerial onto the same spot (put a small bit of flux on the copper wire to ensure it sticks before you apply the solder)
Tune the aerial with nail clippers checking the signal strength all around it is as strong as possible. You need to check along the sides and end-on. Check the signal strength at 1m, 5m, in the next room, upstairs. An oscilloscope is safest but you can try with a real pod also (but it may fail). The optimum length will be somewhere between 5.6cm-6.4cm (seems to depend on the the coil density but it may also be that the best total length is 1/3 wavelength for this design (17.3+5.7 = 23cm which is about 1/3 of 69.2cm) again, no idea why... you can bend the long wire and force it into a box
Does it work? Yes!
I put the phone/pi/hacked RL on the side of the bath, put a test pod in the bath stuck to the bottom using insulet pad under pod and started sending 2 unit bolus commands at different depths in open loop mode (via omnipy). Each tested twice with a five minute wait). Number is water depth
0 cm: obviously works 10 cm: instant response 20 cm: instant response 30 cm: delivery took 30 seconds for first attempt, adjusted RL position and second completed after 20 seconds ~40 cm: bolus not received, error 49 on second attempt (30cm ruler so exact measurement hard, this is the limit of my bath also!)
Obviously, your mileage may vary but unobstructed error free comms for at least 20cm underwater is a huge improvement.
If you look on the underside of the RL, you will see two holes which are labelled "Alt.PWR" on the top side. These can be used for direct charging of the RL battery. Note carefully the + and -.
I made a rectangular box with a lid at the narrow end in Fusion-360. It is best to use not more than a 1mm thickness of plastic or it will make the charging less effective. White plastic enables the LEDs to be visible.
The STL file can be downloaded from here: http://oakeley.com/wiki/RL_big_aerial_Swim.stl
This is what it looks like for real:
To waterproof "properly" paint all the electronics with epoxy. For a "good enough effort", paint the entire box inside and out with acetone (nail varnish remover). When it is dry put the electronics inside and epoxy the lid shut.