Spring 2016 A-TeChToP Project Manager Advice to Future A-TeChToP Groups
By: Cody Dunn (Project Manager)
Introduction
A Project Manager’s Could Have, Should Have, Would Have, Guide to a Successful A-TeChToP
The Spring 2016 A-TeChToP project was not as successful as expected for many reasons. However, instead of dwelling on mistakes and poor decisions made (there were many), I thought it would be better to provide instructions on how to make a successful A-TeChToP in future semesters. In theory, if an A-TeChToP group were to follow this guide step-by-step, it would have a solid foundation for innovation and meeting the requirements set by the customer.
Table of Contents
Central Sensor Suite
Sensors and Major Components
A-TeChToP should be thought of as six sensors, two of which are already on the phone. The four sensors not on the phone are the temperature sensor, pulse sensor, blood oximeter, and electrocardiogram (ECG). The two sensors found on the phone are the accelerometer and camera. Our group and past groups used an extra accelerometer, which became unnecessary with proper placement of the phone.
These were the most successful sensors, aside from those in the phone, for implementation based on our research and the research of past groups:
Temperature: LM34 (It is simple, but it does not require the delay of the DS18B20.)
Pulse Sensor: Pulse Sensor Amped (This provides heart rate but will not give a true ECG signal, contrary to statements in past blogs.)
Blood Oximeter: TSL235R Light-to-Frequency Sensor placed on the ear (This was the only option found by the groups.)
ECG: This sensor has given every group trouble so far. Both a custom circuit and the AD8232 have been tried, neither of which works once the subject starts moving. Furthermore, the placement of electrodes makes manufacturing difficult. Our group made the mistake of believing that noise from the signal could be filtered when the subject moved. However, filtering does not matter if the electrodes are not making proper contact, which seemed to generally be the case given the confines of the project’s budget. A quality ECG signal can be produced while stationary, which would help in the analysis of arrhythmias. Otherwise, this sensor should be replaced with something better designed for playgrounds (perhaps next semester’s innovation).
The first groups used the Arduino Lilypad as the microcontroller, but recommended against it for future iterations. Our group switched to the Arduino Pro Mini, which was easier and offered more pins. The major downside was that a separate FTDI was necessary for programming. If the future group only needs 4 analog pins (2 less than the Pro Mini), the Arduino Pro Micro is much more convenient [1].
For batteries, LiFePO4 batteries are recommended because they are comparatively safer than other options. [2]
Functional code for the sensors can be found in the CSS folders at https://www.dropbox.com/sh/9lg023uivp07bp8/AAAv6DZW-Dqmp9vylJc5koYqa?dl=0
One major area for sensor improvement in the future would be digital filtering, which is relatively non-existent at the moment. This would help clean signals and prevent sudden changes in values to improve accuracy.
Arxterra
The Control Panel is essential for displaying the A-TeChToP sensor signals. The phone’s camera and accelerometer automatically link to the Control Panel, which makes those two the easiest sensors to implement. Ask Professor Hill to upload your desired child pictures for the accelerometer and then you can determine the child’s orientation. Currently, there are a limited number of bars to which a user can send data and then display the values. However, one of the most important new implementations is the threshold flags. These should be implemented in all future versions of A-TeChToP so that parents can better understand when the child is in danger. I recommend, with the permission of Professor Hill, talking to Jeff Gomes for help with implementation.
This semester’s group established a sold firmware foundation for communication with Arxterra using an iPhone6, Arduino Pro Mini, and the HM10. This foundation can also be found on the Dropbox link above. The code from the semesters before spring 2016 rarely worked (if it even compiles). The most important thing to be aware of is that Arxterra should not be overloaded with sending data. Furthermore, it is beneficial to write code with counters instead of delays. This will not prevent the sensors from reading but will prevent too much data from being sent. The code should be further improved to only send data when sensor values change, otherwise sending data is unnecessary.
One interesting feature to add to future versions of A-TeChToP would be a power saver mode. Using a control slider on the Control Panel would allow the user to save power by determining how often sensors read and how often data is sent. The device would last longer if the sensors were turned off a larger percent of the time and data was sent less frequently.
Please reserve ample time at the end of the sensor for proper sensor code integration. This will also allow for better implementation of the Control Panel functions.
Manufacturing
Manufacturing for A-TeChToP does not seem difficult initially and is often underappreciated. However, for a device that is truly safe and sensible, the manufacturing task is somewhat difficult. There should not be wires hanging out from the harness. The central container should be waterproofed. The current strap design is quite reasonable as are the placement of the central container and phone. The major difficulty with manufacturing was determining how to conceal the wires, especially those going to the ear sensors, which was not entirely successful this iteration. It is likely Professor Hill will ask many questions concerning child safety, and it is important to ensure the design was quantitatively tested based on the requirements.
Seizure Watch
The seizure watch utilizes two sensors, an electrodermal activity sensor and accelerometer, to detect convulsive seizures. Since spring 2016 was the first semester this project was explored, a lot of the work was research and the project was not completed. The main problems stemmed from trying to make the device extremely small like a Nike Fuelband.
SAMB11/BTLC1000:
The SAMB11 is basically the BTLC1000 but with flash program memory. It is an extremely small microcontroller with Bluetooth capabilities. Unfortunately, it requires and special Xplained board to program. In theory, the board would be programmed and then soldered onto the final PCB.
The group ran into trouble using the SAMB11 because of poor documentation (at the time of this post the device had come out only a few months ago and may have been rushed to market). The difficulties ranged from connecting to Arxterra to using a simple A-to-D converter. In order to run with Arxterra, the current format is set up to read an HM10 or an HC06 Bluetooth module with ease. The SAMB11 had to emulate the HM10 to connect to Arxterra, which was challenging. The A-to-D conversion would only read one value but would not loop properly when trying to read consecutive values using the provided functions. For these reasons, if possible, it would be best to use an HM10 and an easier to use microcontroller. These better fall within the design of the course, although they are considerably larger in size. If the size was increased to that of an arm sweatband, then the device could be completed using more standard course methods while including more features such as a step counter or sleep tracker (taking advantage of the current fitness watch trend).
Fig. 1: An arm sweatband with a size large enough to use more standard class components [3].
Seizure Activity Detection
The seizure activity detection ended up relatively successful. More testing is required on true seizure patients, but as of now the preliminary testing yields accurate results. The off-the-shelf Bitalino EDA sensor proved more accurate and efficient than the homemade EDA sensor circuits tested. The one area for improvement is the seizure detection algorithm. Currently it runs off simple thresholding, but more complex algorithms would probably yield more accurate results. Studies should also be conducted to determine whether thresholds would change based on the individual or over time as the sympathetic nervous system adjusts to new challenges.
Manufacturing
The manufacturing for the watch is not too difficult except for placement of the electrodes and keeping the overall design small. Future iterations should avoid using metal on the watch as this could interfere with accurate readings of skin conductivity. Furthermore the design should be sweat resistant and waterproof. Adding a display would increase the manufacturing difficulty and improve usability. Lastly, future iterations should be sleeker and aesthetically appealing.
General Recommendations
- Put an ON/OFF switch on the final device
- Double check wiring. Ten extra minutes of checking the wiring could save time waiting for new parts to arrive because the old ones burnt out.
- Do not plan on completing the demonstration the night before. It will probably not work in front of the review board.
- Turn your assignments in on time to the Professor, Division Managers, and Project Managers. When a Project Manager must assemble a presentation at 5:00 a.m. because slides were turned in late, a lower grade will result.
- Finalize all circuits before designing and ordering a PCB. This requires planning. Also, if the board has been approved by the Division Managers, do not make any changes without their permission.
- Anticipate problems. Design and assembly are a small portion of the process. Testing is the best way to determine whether the demonstration will work.
- Assuming funds are available, buy more than one of each part and account for shipping times when making schedules.
- Analyze how many units and hours at work you will have before starting the semester. Taking 18 units with 400D as one of the classes is not a good idea! Also working a lot of hours and taking 400D will make it difficult.
Recommendations for Future Project Managers
- Push the team extremely hard at the beginning of the semester. The team must have a sense of urgency to complete the project.
- Assume the project will fail until proven otherwise. This will create the necessary urgency.
- Be wary of the Systems and Subsystems Engineers. They tend to overstate progress in order to earn a good grade and will tell you they have completed a task when they have not. Require demonstrations of progress every meeting.
- Ask for blog posts early and often. This will prevent an even larger blog post headache at the end of the semester.
- Ask team members why they are doing something and then question the methodology. Keep probing until you are truly satisfied with the final answer and design.
- Stay strict with team members concerning deadlines. You will still receive a lot of late work, but not as late as it would have been without deadlines.
- Develop solid relationships with the division managers as they can be very helpful.
- Keep the President and others in the chain of command above you in the company informed concerning the progress of the project.
Good luck and may your project meet most of the requirements by the end of the semester!