ColdRunner - A Temperature Feedback Running Band

High Level Design    top

Overview

This project was developed by using our knowledge of microcontrollers and hardware design. We ran our device using an Atmega1284p device and attached devices such as a voltage regulator to control voltage from a battery pack, Peltier plate and vibration motor controlled by PWM and amplifier circuits and an accelerometer to measure rate of movement. We came up with the idea because it was a good way to monitor exercise while intuitively providing feedback to the user through temperature and haptic feedback.

Logical structures

Our device operates first by taking in information from an accelerometer. This data provides information on the current acceleration of our device in two directions. Using this data, the device determines when there has been a massive change in acceleration, which we can say happens roughly when the user takes a step. Using this information, we estimate the user's current steps-per-interval (interval chosen to allow for adequate time to make an estimation). We compare this estimated steps-per-interval to one of two thresholds (walk or run) that is set by the user using a switch on the device. Either a red or green LED will light up depending on whether the device is in run (red) or walk (green) mode. The thresholds are set in the beginning during a 15 second delay period that also acts as a calibration period. The user's threshold is set during this time. If the steps-per-interval is lower than the threshold, we turn on a Peltier plate that cools the side that is in contact with the user.Once the user has increased the steps-per-interval to above the threshold, the Peltier plate will be turned off. If the user's steps-per-interval stays beneath the threshold for too long, a vibration motor will activate for a small period of time to give further feedback. If the steps-per-interval is higher than the threshold, we send three quick bursts from the vibration motor.

Standards

Though we did not find any official standard for the temperature for which extended temperature contact is safe, we found a reasonable metric from a NASA paper. The paper said that an extended period of -18°C(0°F) Celsius should be the cold threshold for unlimited contact. Our cold plate definitely does not reach this temperature. Though the warm side of the Peltier plate should not be touching the user's skin, it is still important to keep it on a temperature that would not harm the user if the user accidentally touches it. 45°C (113°F) is painful to touch for any period of time. We do not believe our Peltier plate hot side should reach these temperatures.

Relevant Patents and Copyrights

While working on this project, we found a couple of patents that could be similar to our design. These are Tactile feedback in an electronic device and Temperature feedback motion controller. The patent, Temperature feedback motion controller. is especially similar. However, this patent is used for controllers, especially video games, whereas our project is used for running or walking. The other patent, Tactile feedback in an electronic device could be applied to our project, but because it is worded so vaguely, we do not think our project infringes upon any intellectual property.

Hardware and Software Tradeoffs

The first major hardware and software tradeoff we made in our design was to use an analog accelerometer rather than an SPI protocol driven accelerometer. The advantage of this method was that we did not need to worry about following the SPI protocol and adhering to the restrictions it may provide. To use the analog accelerometer, we just needed to use the ATmega1284's built in analog to digital converter to get values from the signal received from the accelerometer. The downside to this tradeoff is that a SPI protocol controlled accelerometer may have given us more precise and structured results, but the design of our device did not require such precision so this tradeoff was acceptable.

A second tradeoff that we made for the device was to put the noise control of our device in the software part of the design. This meant that we put logic in our code to take out noise that may occur from a sudden change in acceleration that might cause the device to count extra steps that were not there. To make the software less complicated, we could have attempted to put the signal through a low pass filter of some kind, but we believed that a software solution would be easier to implement.

High Level Block Diagram

User Interface

The most simplistic human control feature was the on/off switch. It simply controls whether to turn the system on or off.

A second human interface feature was a SPDT switch on the device that allows the user to set the steps-per-interval threshold to be a lower value for the walk mode or a higher value for the run mode. This allows the user to decide between two different options for a workout.

The third and most major human interface with the device is the accelerometer and haptic feedback system. The accelerometer reads the rate of movement from the human user and adjusts the vibration motor's rate of vibration and the Peltier plate's temperature to give feedback to the user to inform him that he is not running or walking fast enough.

The Accelerometer

The steps-per-interval value that we use to determine the rate of motion of the user is found through the acceleromter. The accelerometer gives three analog signals, one for each of the three accelerations in the x,y and z direction. However, we found that we only needed two of these directions to draw accurate results, so we chose the two directions that are easiest to read from when we mount the accelerometer on the user's arm (x and y).

The Peltier Plate

The Peltier plate is a plate that gets hot on one side and cold on the other depending on the direction and amount of current that we put into it. We had to make sure that the plate got cold enough such that the user can feel the cool side easily. However, when going to temperatures that low on one side, the other side gets equally hot, so we had to ensure that the heat from the hot side did not conduct to the cold side. To do this, we had to put a heat sink on the Peltier plate. This heat sink had to be attached with thermal adhesive that would not become loose when temperatures change.

The Peltier plate gets its cooling/heating function through a phenomenon named the Peltier effect, a type of thermoelectric effect. A thermoelectric effect is one that converts voltage provided into a difference of temperature. This effect was perfect for the device we wanted to create since it does not require any outside substance for a cooling effect other than the Peltier plate, which can be easily hooked up into our circuit. The polarity of the voltage or the flow of current is what decides which side of the Peltier plate gets hot and which side gets cold. A Peltier plate is constructed out of two conductors connected by a junction that removes or adds heat to one of the conductors and does the opposite to the other.

Furthermore, to control the temperature of the Peltier plate, we needed a PWM controlled current source. To do this, we used an optoisolator to keep the Peltier plate circuit separated from the microcontroller unit so that we would not damage the mcu if something went wrong with the Peltier plate. We also used a BJT to control the current going into the Peltier plate based upon the PWM.

Because this device is supposed to attach to your arm and the Peltier plate draws a lot of current, we needed to use a separate battery pack to power this part of the system that does not power the microcontroller. We used three AAA batteries and used a voltage regulator to get the desired voltage needed for the system.

The Peltier plate was given a max of about 0.7 Amps in our circuit. With a resistance of about 1 Ohm (given im the data sheet), we expect about 0.7 V to be across the Peltier plate during use. Because of the circuit it was connected to, we observed the Volts across the Peltier plate to be more around 0.5-0.6 Volts using a multimeter. This corresponds to about a 30 degree Celsius temperature differential. From what we observed, this was roughly the case with the colder side being around 10 degree Celsius and the warm side being around 40 degree Celsius. This adheres to the safety standard since the cold side is definitely not in the danger zone and the warm side is heat sinked so prolonged contact should not hurt since we want it to be below 45 degrees Celsius, which it is.

Hardware    top

Hardware Overview

This project requires hardware to be built to control the external components such as the Peltier plate, vibration motor, etc. The Peltier plate control circuit is composed of a BJT, optoisolator and the Peltier plate. Furthermore, a voltage regulator is important to ensure the rail voltage for the Peltier plate is properly set. The vibration motor control circuit is important to allow a PWM to control it. The LED and switch component is also necessary to allow for the switching of modes. Finally, the accelerometer itself is attached to resistors to connect to the microcontroller ports.

Peltier Plate Control Component

The Peltier plate control component is run by a PWM. In our device, we send the controlling PWM from port B4 for the Peltier control. This PWM is sent to a 4N35 optoisolator, connected to a 330 ohm resistor. An optoisolator was chosen to protect the MCU from anything that the Peltier plate may do to damage it. The output of the optoisolator will then go into a TIP31 BJT. This TIP31 BJT acts as the voltage controlled current source that provides the necessary current to turn on the Peltier plate. A TIP31 was chosen because it was better rated to allow for this type of current control. To ensure that the TIP31 does not get too hot because it might be dangerous to place on somebody, we lowered the input resistance to the TIP31's base to be roughly 100 ohm. We also used a 1 ohm 2W rated resistor to connect the TIP31's emitter to ground to allow the collector current to be as high as necessary to run the Peltier plate. We determined that a current of roughly 0.68 A was good enough to allow the Peltier to be noticeably cold. A schematic can be seen in Appendix B.

The voltage rail for the voltage controlled current source needed to be 2.5 V. However, using just AA batteries, we could not get this voltage. To address this, we used a voltage regulator. For this purpose, we used a PQ30RV31 voltage regulator. This regulator takes in a voltage of 4.5 volts (3 AA batteries) and outputs a voltage of 2.58 V. A schematic can be seen in Appendix B.

Vibration Motor Control Component

The circuit for the vibration motor control component has two main parts, the BJT and the motor. We referenced a circuit that allowed us to use a PWM to control the vibration motor. This circuit is shown in Appendix B. The base resistor to the BJT was chosen to be 1 kOhm. To protect against possible voltage spikes that might occur from the motor, there is a 1N4001 diode placed across the motor as well as a small resistor in series with the motor.

The vibration motor and Peltier plate were both mounted on the inside of the armband. This can be seen below.

Threshold Switch and LED Component

We wanted to allow the user to choose between a walking mode and a running mode. To do this, we used a SPDT switch. Since we have our microcontroller looking for an active low, we hooked the middle pin of the switch to ground and the two side pins of the switch to micrcontrollers pin D3 and D4. We then poll these pins frequently to determine which mode the users want to be in. To allow the user to see which mode he is in, we also have two LEDs (one green and one red) connected to ports C0 and C1, each through a 330 ohm resistor to control the amount of current flowing through. If the red LED lights up, the user is in run mode. When the green LED lights up, the user is in walk mode. A schematic can be seen in Appendix B.

Hardware Difficulties

One difficulty that we ran into when we were building the circuit was that the amount of current needed to be drawn for the Peltier plate would make the power dissipated across some of the components too much, and the components would become dangerously hot. For example, the reason we chose the TIP31 BJT was because some of the other transistors we used could not output enough current without requiring a much higher rail voltage. The TIP31 was able to give us a good ratio, but not without also getting quite hot. To fix this, we had to adjust the resistor values that we used. We found that the 1 ohm resistor at the TIP31 emitter and the 100 ohm resistor at the base worked well enough for us to run the Peltier plate well. The 1 ohm resistor also had to be a higher wattage resistor (we used a 2 W resistor) to make sure that it would not overheat as well, which we found would occasionally occur with other resistors that we tried to use.

In a related problem, the PQ30RV31 voltage regulator would also become very hot. After fixing the problem of the TIP31 getting hot, the voltage regulator would also become hot enough to be very uncomfortable to touch if drawing a large amount of current over more than twenty seconds. To help alleviate this issue, we attached a heat sink to the voltage regulator. This heat sink was useful in keeping the voltage regulator a good enough temperature to be safely mounted on somebody.

Another alternate design that we had was for the switch method for user control. Instead of using a switch, we considered using interrupt driven push buttons. The buttons would allow the user to increase the threshold manually. This created two problems. The first was that it would be difficult to transparently inform the user on what his current setting was without using another component such as a LCD display or multiple seven segment displays, which would add to the bulkiness of our physical design. The second problem was that to debounce the interrupt driven buttons, we needed to perhaps use an active high design in conjunction with low pass filters. It was much more intuitive for the user and for us to use the SPDT switch.

Another problem that we had to overcome was that the Peltier's hot side would start to affect the cold side if the temperature differential were too different. The heat would start to dissipate through the middle material separating the two plates. To fix this, we had to also attach a heat sink on the hot side. That way, the plate would be able to cool down without heat bleeding through from the other side. Otherwise, the user may not be able to feel the temperature feedback that the device is trying to give him.

The accelerometer was sensitive but mostly only in the specific directions that it is designed to measure accelerations in. Therefore, we had to be careful in mounting the accelerometer and making sure that it is always pointed in the correct direction. This is important in deciding the information the device receives.

One of the difficulties arises when it comes to the actual design of the full circuit mounted on the arm bands that we use to hold the device onto a person. The circuit itself cannot be too large, so we had to solder the components close together. Furthermore, we needed two separate battery sources to be mounted alongside the circuit, one to power the microcontroller and one to power the Peltier plate control circuit.

Software    top

Code

ColdRunner.c

This file acts as the main of our program. It runs off of the Tiny Real Time (TRT) kernel, hosting multiple tasks necessary for our system. There are four main tasks: modeSelect, Accelerometer Read, PeltierSignal, and Vibe. These tasks will be discussed below. At startup, our code intializes the TRT kernel with a call to uart_init(). This function sets up the TRT kernel, initializes various tasks, semaphores, timers, and ports, and enables interrupts and puts the microcontroller to sleep. After this the TRT tasks begin to execute.

modeSelect

This task has the relatively simple job of determining what mode the system is in. It polls PIND to determine if the switch is set to a run or walk mode. After determining this, it sets a threshold (stepthresh) used later on for measuring steps, and turns on an LED. There are two LEDs which correspond to either a walk/run mode. The green LED represents the walking mode, and the red LED represents the run mode. This task has very short release and deadline times, so that any change in the switch can be seen quickly.

AccelerometerRead

This task was created to read in accelerometer values and measure if a step has been taken. The accelerometer that we used has an analog output, which we capture with the microcontrollers onboard analog to digital converter (ADC). These values are captured from the inputs PINA0 and PINA1. In order to capture these ADC inputs, we had to set up the ADC. This was done in the uart_init() function. This task reads in the values by first ensuring that the ADC Start Conversion (ADSC) bit and the ADC Control and Status Register A (ADSCRA) are equal. This is done to ensure that the ADC operation has finished. After doing this, we store the ADC value in an integer (either Acx or Acy). This is done by writing the integer with the value located in ADC Data Register(ADCH). After doing this, a new bit is written to the ADC Mux (ADMUX). This is done so that we can collect data on another channel of the ADC. After setting the ADMUX bit, we write a 1 to the ADSC bit. This starts another conversion. We do this process twice in our task so that the task will collect both the X and Y values. After collecting the X and Y acceleration values, two different operations occur. In the "run" mode the values are added together and stored in a variable, avg. Then avg is compared to a past value of avg, prevavg. We compare the |avg-prevavg| to a value to determine if a valid step occurred. If the value of |avg-prevavg| is within the range, we update a variable stepcount. This absolute value operation is done to measure acceleration changes relative to either forward or backward movement. After doing this, we impose a 100ms delay, so that we do not read in false steps. A similar operation occurs during the walk mode, but instead of combining the Acx and Acy value to form avg, we only set avg = Acx. This was done during testing our system, because we noticed higher amounts of false steps occurring when avg is a sum of Acx and Acy. Apart from this change, the rest of the stepcount operations are the same as the run mode. The task is updated every 5 ms, so that we are able to measure when acceleration changes occur. This was done because it produced the most accurate stepcount during testing.

PeltierSignal

This task sends a PWM signal to the Peltier plate. In order to do this, the task starts off by having an initial delay of 15 seconds. This initial delay is created so that the user can get up to their desired speed after turning the system on. The delay is also used to set the user's individual threshold. This is done so that the thresholds for the user is tailored to their individual running paces. This is done by averaging accelerometer reads after the delay period is over. After the delay, the task will compare a step count, stepcount to a threshold, stepthresh. stepcount is determined every 5 seconds. It is updated during this time from the task AccelerometerRead. stepthresh is set from the task modeControl. If stepcount is above the stepthresh, then the output OCR0B is set to 255. This will not trigger the PWM. However, if stepcount is less than stepthresh, a new variable is set to 1 (peltthresh). If peltthresh is 1, a PWM will be outputted. OCR0B is set to 0, so that the PWM will turn on the Peltier plate. After this occurs, a variable peltcount is added to. This variable was initialized to 0. However, if peltcount goes above 3, the OCR0B is set back to 255. This is done to prevent heat flowing back onto the Peltier plate. However, because the Peltier plate is turned off, the system still needs a way to alert the user that they have been running too slow. This is done by triggering the task Vibe.

Vibe

This is the last task in the ColdRunner.c file. This task checks to see if a variable vibe has been set. If it has been set when the operator is below their low threshold, it sends a PWM to OCR0A at a value of 0. This task turns on the vibration motor for .7 seconds, so that user has enough time to feel the vibration, without it being too annoying. If the variable vibe has been set when the operator is above their high threshold, a 0 PWM signal is sent to OCR0A. This is sent in 3 short bursts, so that the user has time to differentiate between a high and low threshold.

trtkernel_1284.c

This is used to run the TRT kernel. This sets up the kernel and contains functions for all the trt* methods.

trtSettings.h

This is a header file used to define various things in TRT, such as max number of tasks or semaphores. It is also used to specify the MCU's clock and UART baud rate.

Software Difficulties

Throughout this project we had some difficulties on the software side. These difficulties usually revolved around timing and reading in variables.

One of our major challenges was with the timing for the tasks. These were difficult to determine at first, because we were unsure how much data we needed before we could conclude that we had an accurate measurement from the accelerometer. This involved trying different times for release time in the AccelerometerRead task. We had started with .1 seconds, but eventually realised that the threshold needed to be lowered to about 5 ms. This was due in part to the fact that we would often miss the acceleration changes with a time of .1 seconds. After lowering the acceleration timing down to 5ms, we also ran into the issue that we would get too many acceleration values, which led to our 100ms delay after a high read.

Another timing related issue occurred with our PeltierSignal task. This was originally set to a very high release time value, so that we could compare the number of steps taken with a threshold. We believed that setting this to a large time would allow us to accurately see steps/min. However, this led to long delays in sending the PWM to the Peltier, and we wanted more control to do this. This led to us lowering the release time of this task to 5 seconds. This allowed us to achieve the control of the Peltier plate that we desired, along with fairly accurate step/min counts.

---