The last week, like the week before it, was a combination of manufacturing and app development. It’s pretty clear now that getting the mobile apps finished and in the Apple App Store/Google Play Store will be tasks driving the schedule. Unfortunately the Mooshimeter is not useful without a mobile app to talk to it.
Why not release the API and crowdsource the apps?
In short, because the app development process is teaching us a lot about the different ways iOS and Android implement Bluetooth Low Energy (BLE), and that drives changes to the API. BLE is designed for very low bandwidths, usually a few bytes per second. To quickly stream across graphs like you see in our videos takes some “protocol abuse”. This is where the real differences in the iOS and Android implementations of the protocol are exposed.
After establishing core functionality I hope that users will expand upon/write their own apps.
Racks on Racks on Racks
In the last update I showed you the calibration and testing stands. We’ve run a few hundred meters through the process and done final assembly on almost 200 of them.
On new meters, we’re seeing an average uncalibrated accuracy of about 1% with 0-50mV offset at the ADC. This is good. By design the uncalibrated tolerance should be about 2%. Presently any meter that shows an uncalibrated accuracy worse than 2% off is flagged and set aside. We might accept them later, but I want to get a better look at them and make sure nothing insidious is going on with them. FYI only about 5% of the meters so far fall outside of 2% uncalibrated threshold. Immediately after calibration the meters are well within 0.1%, I’ll check in on the calibrated meters again regularly to see how that drifts with time and temperature.
“I’m a Mooshimeter! Connect to me!”
The Mooshimeter is always on. When a smartphone is not actively connected, it sleeps to save power and wakes up every few seconds to send out a radio packet advertising its existence. It’s just a little “Hey, I’m here! You can connect to me!” message to anybody who might be listening.
When you have 200 Mooshimeters all next to each other, that little “Hey, I’m here!” message once every few seconds becomes an overwhelming torrent to any radio listening on the right channel. There are other offices in my building that work on Bluetooth Low Energy devices, and now they must filter out a constant torrent of Mooshimeter broadcasts. This might be a bigger issue down the road, but for right now the kind folks at Sheepdog Sciences are just being nice about it. Thanks guys.
My main focus now is getting all the core functionality in the apps working. In the next update I’ll try to put together some screenshots. Final assembly and calibration are happening simultaneously, but I am less worried about them because the time required for those tasks is easily characterizable. Estimating the time required for software development and debugging tasks is much harder. If you’re interested in the topic of why estimating software development time is hard, I encourage you to read this excellent essay..
As always, feel free to email or comment below. Thanks for reading,