2013 Seventh International Conference on Sensing Technology Thick Film Flow Sensor for Respirator Applications Steven Frenzer, Will English IV, Ryan Rodebush, and Michael J. Haji-Sheikh Department of Electrical Engineering Northern Illinois University DeKalb, Illinois 60115 mhsheikh@niu.edu Abstract—A new approach to monitoring respiration activity for first responders is proposed. This approach emphasizes low cost and portability. Current commercial methods cost upwards of $100 and aren’t reusable or portable. The design goal was to achieve a portable stand alone unit that measures and displays both breaths per minute as well as inhalation/exhalation using simple low cost commercial components. The prototype device is portable and battery operated as well as accurate. This approach has benefits in the field for emergency medical technicians as well as other medical professionals. Keywords; Respirator; Arduino; Thick-Film; air flow; portable; INTRODUCTION Recently, several approaches to monitor respiration have been demonstrated [1,2]. These approaches rely on humidity sensor arrays and wearable textile sensors. The optimal sensor for this type of situation is a thermal mass flow sensor. These types of sensors can detect small flow rates and the direction of flow. The approach used in this project was to adapt a thermal mass flow sensor into a common respiration unit, used by paramedics to monitor an injured person's respiration. The sensor is thick film sensor designed at the Microelectronic Research and Development Laboratory and manufactured in the Northern Illinois University clean room. This sensor functions in a method similar to the microelectronic machines and systems (MEMS) device invented and manufactured by Honeywell's Sensing an Control division [3]. The sensor uses a Wheatstone bridge design excited by 4-9 V DC and 4 thermally sensitive resistors. Two of the resistors are on one side of a heating element and the other two across in the direction of flow. As air passes over the first set of resistors it is cooled by the heating element and the second set is warmed by the flow. The change in air temperature creates a change in voltage across the bridge. As flow is reversed, like the difference of exhaling compared to inhaling, the temperature changes and then voltage difference is reversed. The sensor also has an acrylic coating on the surface to prevent moisture from affecting the results. Currently the main methods of measuring respiration (spirometers) are bulky or expensive. In the research we did, none of the products were marketed as portable. A wireless version of this system could become a vital tool in monitoring patients, as well as other markets, due to its portable inexpensive design. 978-1-4673-5221-5/13/$31.00 ©2013 IEEE Fig. 1. Printed Mass flow Sensor showing the heater and thermistor bridge. DESIGN DESCRIPTION The finished flow sensor is shown in Fig. 1. The fabrication process for this sensor uses a four layer screen printing process. The screen pattern is made using a photoimageable mylar film with a pattern. This film is then placed on top of the screen (the screen is coated with a negative acting photoimageable polymer) which is then exposed to ultra-violet light. The photo-material then hardens resulting in a negative imprint of the pattern. The unexposed regions are then washed off with water. These screens are then used with an AMI/HMI Presco 485 printer. The 485 printer is an industrial printer which has the ability to set various parameters that range from squeegee sweep rat to screen substrate and planerity. Beginning with an alumina (Al2O3) ceramic substrate for the sensor, the first layer of gold (used for the long interconnects) is screen printed. After each layer is printed, the sensor is fired to 850°C. The second layer is an oxide which is meant to insulate so leads can be placed over the interconnects. This is then dried 150ºC and fired at 850ºC and forms a hard glassy coating. Thermistor paste material is added, dried and fired, and the last two layers are gold. Each layer use a screen with a different pattern and the totality of the process produces the final device. The design parameter that is the most crucial is that the closer the thermistors are to the heater the more sensitive the thermistors 540 2013 Seventh International Conference on Sensing Technology become. This close proximity is desired since this allows the device power to be reduced. However, due to the inaccurate nature of screen printing, there is a chance of shorting out the sensor, which is a drawback to the close proximity. To lessen the chance of the sensor shorting, the squeegee pressure has to be turned up high (called a dry squeegee). This additional pressure causes the material to thin out more making the material resistivity to increase to about 20K ohms per square instead of 5K ohms per square. Unfortunately the high resistance of the sensor causes a high susceptibility to noise. Leads must be kept short or be shielded. Fig. 2 shows the flow sensor assembly. The sensor itself has a high change in voltage for relatively low temperature change due to the high TCR (Temperature Coefficient of Resistance) of the sensing material. The equation for the TCR is as follows, input level of 0-5 volts. Meaning every bit corresponds to around 4 mV, which corresponds to around ± 0.01 V of absolute accuracy. The initial measurements from the sensor showed a 40 mv peak to peak swing between sides when breathing normally. This means the signal from the sensor should be amplified if shallower breaths are to be detected. The first stage of amplification was a difference amplifier which is shown in the schematic as shown in Fig. 3. The initial amplifier design was proposed to use a LM series instrument amp but it was decided to change op-amps from the suggested LM to an OP07 due to availability and better noise/offset characteristics. The amplifier (schematic capture shown in Fig. 4) was modeled in PSpice. A second low gain amplifier stage was then incorporated which amplified the signal to a usable level and further suppressed high frequency noise. After each step preliminary measurements were taken and further calculations were based off these numbers. 2 R=R 0 (1+α∗T +β∗T ) (1) Where R0 is the resistance at 0ºC, and α is the first order coefficient and β is the second order coefficient and temperature in degrees Celsius. The thermistor material used in this design has a high β (1600 ppm/C2) term which allows for the sensor to have a reasonably high sensitivity. Fig. 3. Simple difference amplifier used to amplify bridge signal. Signal conditioning is done on board the Arduino Nano. Fig. 2. Mass flow sensor assembly. The construction is thermistors on alumina with printed gold leads. The voltage change from both sides of the bridge is then compared using a unity gain differential amplifier. There is a second stage of amplification prior to the analog to digital (A/D) converter on the Arduino. Both stages of amplification uses a OP07 amplifier (low noise low drift low offset) [4]. from there the Arduino is coded to measure period of a breath and display the result. The entire design was modeled in Solidworks and Altium Designer 10 to verify everything physically. All the circuit elements can eventually fit on the same ceramic as the sensor is built on. The Arduino development board used includes an Atmel ATmega 328 microcontroller [5]. The ATmega 328 features include an Analog to Digital (A/D) converter which is used to convert the analog signal from the op-amp stages to digital values for further processing in the microcontroller. This A/D converter has ± 2 LSB (least significant bit) absolute accuracy [6]. The chip offers 1023 steps of resolution with an 978-1-4673-5221-5/13/$31.00 ©2013 IEEE Fig 4. PSpice Model used to design the difference amplifier used to amplify bridge signal. For this design, it was necessary to control 16 light emitting diodes (LED). These 16 LEDs were 7 segment displays plus 541 2013 Seventh International Conference on Sensing Technology the two decimal points adjoining each of the 7 segments. The Arduino development board only has 14 pins of digital I/O so our design needed to incorporate multiplexing to control the 16 LEDs. The multiplexing software switches which LED segment is illuminated by selecting which common cathode to drive via a NPN transistor. The NPN transistors are necessary because the Arduino can only sync/source 40 mA per pin. With multiple segments illuminated the design would exceed 40 mA. A 2N4401 switching transistor was readily available and met the specifications necessary. The base resistor was chosen to be 1.2K ohm to supply the current [7]. The resistors chosen for the display were chosen based on the data sheet of the display. The LED’s were designed to be driven with less than 35 mA at 50% duty cycle [8]. The voltage drop across the display meant that choosing a 270 ohm resistor would achieve ~10 mA which is well within the specifications for adequate brightness. The full schematic design was done in Altium Designer 10 shown in Fig. 5. voltage change can be calculated by knowing the velocity of the air across the sensor as well as the properties of the resistor material, substrate and heating element. Physically, the sensor was measured using a digital multimeter (DMM) and an oscilloscope. The results of the Pspice simulation is shown in Fig. 6 and the un-amplified results of the initial sensor test are shown in Fig. 7. The final amplified results are shown in Fig. 8. The results from physical testing gave the voltages necessary for the amplifier calculations. We were able to experimentally confirm our accuracy and results. By video recording operation and comparing the time coding of the video to the displayed number the results were verified Fig. 4. The video was recorded in 1080p 23.976 (NTSC format) Frames per second on a Nikon D5100. By counting frames and watching the video it is verified that the display goes to “ - - “ after 30 seconds of inactivity as an alert. The same technique was used to verify accuracy of breaths per minute. The inhale LED and exhale LED were verified via audio track. Fig 5. Altium Designer 10 model of the entire structure. Fig 6. PSpice output of the first stage amplifier. Gain of 6. The completed design was then soldered to the board and then programmed. Programming on the Nano then takes place using Arduino’s software development environment and was written in “Arduino C”. The program included debouncing, sampling the digital measurement values from the output of the integrated A/D converter, as well as code for multiplexing the display. The program works by using the microcontroller to measure the period of a breath. The internal RC oscillator on the ATmega328 only has an accuracy of ±10% which significantly impacts the accuracy of measurements. As suggested by Atmel, a more accurate external oscillator was used to drive the microcontroller. The 16 MHz oscillator which is available on the Arduino board was chosen. The program also illuminates the decimal places on the segments to visually display inhaling vs. exhaling. MEASUREMENT METHODS Measurement of the sensor was done both mathematically as well as in normal operation. Because the sensor was being built in house there was no data sheet for it, but the sensitivity can be measured directly. The expected 978-1-4673-5221-5/13/$31.00 ©2013 IEEE Fig 7. Measurement results of the un-amplified sensor. The signal range is 41 mV peak to peak. 542 2013 Seventh International Conference on Sensing Technology conclusion, this novel design allowed the measure breaths per minute accurately and inexpensively. The size and cost would be reduced if a microcontroller was used instead of using a development board like the Arduino. The Arduino was by far the largest and most expensive part in the design. In the future, it would be nice to see the device integrated with a better display as well as variable set points to trigger alarms. Further improvements can be made by using a application specific integrated circuit (ASIC). The original idea was to use an instrumentation style differential amplifier with a higher gain. After finding out the sensor was unbalanced, the only way to balance the offset was not generally not practical for this prototype. Design modifications are needed to the sensor to allow the offset be trimmed to near zero. The sensor had issues with high frequency which will require that the impedance be lowered. This noise was caused by the high resistance in the bridge legs. The noise was overcome in the design phase by using shorter wires and adding capacitors. The capacitors however added a time delay. This time delay was acceptable because the frequency of human breathing in normally less than 1 Hz. Additionally, a voltage regulator will be added to help control the signal to battery life. Fig 8. Signal after first stage amplification and filtering. Actual gain was only 3. This is due to the lack of a PSpice model of the OP07 op-amp. For this application this design error was acceptable. ACKNOWLEDGMENT The Authors would like to thank their families and Mr. Gregg Westberg for their support and help. I would additionally like to thank Misty Haji-Sheikh for her editing help over the years. REFERENCES [1] Andre, N.; Druart, S.; Dupuis, P.; Rue, B.; Gerard, P.; Flandre, D.; Raskin, J.-P.; Francis, L.A., "Dew-Based Wireless Mini Module for Respiratory Rate Monitoring," Sensors Journal, IEEE , vol.12, no.3, pp.699,706, March 2012 doi: 10.1109/JSEN.2011.2161668 [2] Merritt, C.R.; Nagle, H.T.; Grant, E., "Textile-Based Capacitive Sensors for Respiration Monitoring," Sensors Journal, IEEE , vol.9, no.1, pp.71,78, Jan. 2009 doi: 10.1109/JSEN.2008.2010356 [3] Bob Higashi, Semiconductor device microstructure, US patent 4,696,188 [4] Analog Devices. (2011). 150 μV Maximum Offset Voltage Op Amp. Norwood: Analog Devices, Inc. [5] Arduino. (2009). Arduino; Schematic. Retrieved January 5, 2013, from http://arduino.cc/en/uploads/Main/ArduinoNano30Schematic.pdf Arduino. (2012, November 23). Arduino Board Nano. Retrieved February 1, 2013, from http://arduino.cc/en/Main/ArduinoBoardNano [6] Atmel. (2009, November). ATmega48PA/88PA/186PA/328P. Retrieved Febuary 1, 2013, from http://www.atmel.com/Images/doc8161.pdf [7] Storr, W. (2013, April 28). Transistor as a Switch: Electonics-Tutorial. Retrieved 2013, from http://www.electronicstutorials.ws/transistor/tran_4.html [8] LITE-ON ELECTRONICS, INC. (n.d.). LITE ON. Retrieved from P_100D5623AG. Fig 9. Image capture of the functioning prototype. This device can be easily miniaturized. DISCUSSION AND CONCLUSIONS. The simulated results were more generous than the measured results for both the sensor and amplifier operation. The amplifier simulation in PSpice did not use an OP07 directly so this could be the reason for the discrepancy. In 978-1-4673-5221-5/13/$31.00 ©2013 IEEE 543