Hand Hygiene

I am beginning to look into the possibilities of design using the epic base. We have some RMB-2000′s for evaluation, which are the Archrock commercialization of the epic base. This particular chip has the radio and the 4GB flash both attached to serial bus 0 (usart0) of the msp430 processor, where as the RMB-2001 has the flash moved to usart1 (there are only the two on the MSP430).

The initial thrust of investigation is downloading data from the flash to a computer via USB. The idea is to create a dongle which will connect via one of the usarts of the microprocessor to allow telos-like access while still having the target system in the field be of minimal size and power consumption. USB to serial is the method employed in the telos and has been employed by Berkley for this epic platform in various incarnations.

Another idea: Can we easily and inexpensively add additional circuitry to the dongle (where real estate is not so much of a premium) such that the mote will look like a flash drive when connected to a computer, allowing any researcher to download the flash contents via USB to any computer without needing software and without utilizing the mote’s processor. Essentially we would be adding the circuitry of a flash drive to the dongle, except that the actual chip would reside on the target system. A likely scenario would be that when connected to the computer, dongle and mote would “look” like two devices – a flash drive and a mote. The mote is your target for programming, the drive is where your data may be retrieved.

We’ve got a working prototype of the puck circuit with a large flexible resistor as the sensing mechanism, feeding into a comparator.  The other input into the comparator is the output from a digital potentiometer.  The circuit consumes 333 microAmps, which means that, not counting the mote, with 2 AA batteries, we expect the batteries could last about 202 days.

Here are the diagrams:

A brief record of the projects I’ve worked on this last semester:

LED lights for plants – I was only involved in milling and soldering these, not their design.

Door Minder transmitter box – This design can be found on the desktop of the computer in the SW corner of the lab.  The folder is called NEW Transmitter.  Open Transmitter.asm for the latest assembly and parts.

Hand Hygiene anchor – Also on found on the desktop of the SW corner computer, the folder is called Swivel Puck.  The latest assembly is swivel_puck_split.asm.  This assembly also lists all the latest parts.  Within this folder is a second folder called STL 2.0.  This folder has all of the .stl files that were sent for prototyping.

Hope this is a decent guide should anyone need access to these designs.  Also, feel free to e-mail me with any questions: david-k-jones@uiowa.edu.

Thanks, have a great summer in the lab.

David

The current model for the door minder includes a potentiometer (pot) that changes the brightness of the infrared LED (the higher the potentiometer, the dimmer the LED). If the red LED on the mote is on, it means the door minder is not receiving the reflected light from the infrared LED (i.e. the beam is broken). The door minder must be calibrated so that the infrared LED is as dim as possible with the door minder still able to recognize the beam as unbroken. This is because the beam will reflect off of anything that is too close to the door minder, resulting in the door minder failing to detect people going through the door if they walk too close to it. The following is an algorithm for calibrating the door minder.

1. with the pot set to a reasonable value (perhaps 1/4 of its total), set up the door minder and reflector on either side of the door, so that the beam is unbroken (the mote’s red LED is off)

2. Turn the pot up until the red LED comes on. This means the infrared LED is dim enough so that it is not being reflected enough for the mote to recognise the beam.

3. Turn the pot down to the point where the door minder just starts to recognise the beam as unbroken again (i.e. the red LED turns off). After this step, you may see some “false breaks” occur. This is when the red LED flashes when it is not supposed to. If this is the case, you will have to turn the pot down until the false breaks stop occurring.

In the future, we may be able to use a digital potentiometer in the door minder design, and set it up so that you merely have to press a button to calibrate it.

I did three different tests to check how robust the measurements are from the current mote set-up.

Variance Test One: Motes at 1.5m and 2m were rotated to different orientation (45 degrees)

Variance Test Two: Motes were jiggled while signals were sent

Variance Test Three: Mote at 2.5m was lower than the rest of the motes

Here’s the graphs collected:

Variance Test One:

This data looked pretty decent – there was a separation and the box plots look decent (except for a few extreme angles & 3m). It seems to me that changing the orientation of motes at 1.5m and 2m did not have a great effect on the data, because all the motes showed variation compared to the previously collected data under “perfect” conditions.

Variance Test Two:

In variance test two, there was a difference in the CDF separations (but this separation was also seen in variance test 3). There is a weird spot on the boxplot graph at .5m, high angles.

Variance Test Three:

The CDF separations look definitely less distinct in variance test three. There also seems to be more outliers in this data set. As for changing the height of the mote, it didn’t seem to change the data as much compared to the other trials.

In all three data sets, the data seems to be a lot messier than the original experiment with no small variations. (http://groklab.org/handhygiene/2010/05/04/power-31-experiments/) I think something else is going on with these experiments though because in variance tests one and three, only one or two motes were changed, and all motes showed messier data.

Data Summary/Statistics:

From this table, we see a definite change in mean RSSI throughout the different experiments. I’m not sure why the other motes were affected when motes at 1.5m and 2m were changed in variance test 1. Also, when the height was changed for the mote at 2.5m, the stats for the other motes also varied a lot. I see a lot of variation in our testing- but I’m not sure where this variation is coming from.

We’re planning to start development on a new platform, the EPIC system in tinyos.  We’re organizing the electronic priorities now.  Some possibilities and things to look at:

MSP430 Ultra-Low Power MCU’s A tiny platform that does not include a radio, but has a watch model (which is currently out of stock).  Currently about $50.

EPIC RMB-2000 Series from ArchRock – a new device from Berkley.  About 1″x1″ with integrated radio.  No power, USB or antenna.

We’re having trouble getting a consistent estimate of distances around 2-3 meters.  I spent a lot of time looking for antennas that will transmit more power in the needed region.  I’m currently looking at patch antennas, which should double our distances in one direction only, at the sacrifice of a more elliptical-hemisphere broadcast pattern, compared to the current spherical broadcast pattern.  I went ahead and ordered some for ~$17.  Looked at plans to build them, but that would take a few more days and we should keep moving.

We found out that the highest radio power possible is 31. So, I did the power experiments again with the whip antenna & the F antenna so we could choose which one was better.

Whip Antenna:

Except for the strange bumps in the CDF for distances .5 and 1.5, the whip antenna seems to be pretty good. I’m not sure what could have caused these bumps.

F antenna:

The F antenna CDF looks smoother than the whip antenna, but the box plot data is very messy – many many outliers. It’s also not showing a great separation between 1 and 1.5 meters.

Here’s the comparison data, whip antenna and F antenna at both powers 20 and 31:

Again here – the whip antenna data looks pretty good. One thing strange to note about the F antenna is that the average RSSI drops from power 20 to power 31 at distance .5, while everything else increases.

With everything put together, I’m leaning toward the whip antenna as the better choice.