Development of Miniature Dynamic Electret Pressure Sensor

Transcription

1 Author: Simo Ahtiainen (UKU) Development of Miniature Dynamic Electret Pressure Sensor SUSPOWER PROJECT Supervisors: Jens Fridh (KTH) Prof. Hannu Ollkonen (UKU) Ass. Prof. Andrew Martin (KTH)

2 2 Abstract Experiments in this SUSPOWER project were performed in KTH, Heat and Power laboratories. Overall aims in these experiments were to learn more about thin-film dynamic electret pressure sensor, confirm the manufacturing process and improve the measuring technique. The electret sensor itself is a thin-film fast response dynamic pressure sensor, which can be manufactured in different shapes and sizes. This kind of research is necessary because one of the specific targets was to miniaturize sensor size, in the order of one square millimetre while maintaining the sensor characteristic. The requirement to develop miniature sensor size technology is especially important due to the space constraints in turbines and also because of the desired high spatial resolution of the unsteady pressure distribution on certain surfaces, e.g. a turbomachine blade. Furthermore, reductions of the pick-up and background noise levels are necessary. Also a one major research topic was to find out the relationship between sensor size and the magnitude of the output signal. This information will help in future if the sensor technique will be implemented in for example wind tunnel measurements. The outcome from the experiments shows that the pick-up noise is more or less eliminated and that a reliable signal is retrieved from the electret sensor. The research study has also shown that it is possible to manufacture sensor size of 1mm 2 in a controlled way. Dynamic calibration runs for three different sensor sizes have shown that the trend of the sensor transfer characteristic is preserved regardless of the size. This facilitates an easy future calibration procedure. Acknowledgement I would like to express my gratitude to the SUSPOWER European Research Programme and the people involved in it, for making this research possible. I especially would like to thank to my supervisor, friend and work colleague Jens Fridh for great help on arranging everything and making this project interesting and instructive. Thank you JeF! I will also thank Professor Andrew Martin for administrative support and making this SUSPOWER possible. Thank you Susy Mathews for arrange my accommodation. Thank you Rolf Bornhed at the workshop for helping us to achieve good results in experiments. Also great thanks go to my supervisor Professor Hannu Olkkonen for giving me all the knowledge, support and making this possible. Dear Minna-Maria thank you for your support.

3 3 Table of contents Abstract.. (2) Acknowledgement... (2) 1. Introduction & Background.. (4) 2. Objectives... (4) 3. Electret sensor & Measuring technique.(4) 3.1. Manufacturing process......(5) 4. Experimental setup.. (7) 5. Results and Discussions... (8) 6. Future works (12) 7. Summary...(13) 8. Literature..... (13)

4 4 1. Introduction & Background Measurements of unsteady flow in turbomachines are an absolute necessary feature in engine design optimization. This is because of unsteady flow and its interaction with the internal structure may lead to situations that can be fatal for both humans and machine. Considering typical turbine unsteady flow features being of the order of 1-20 khz, unsteady surface pressure is traditionally measured by means of miniature fast-response piezoresistive transducers (e.g. Kulite) [1]. Typically many commercials sensors are expensive and the shapes of those sensors often require costly mechanical modifications of the measuring object in order to embed the sensors. Therefore it is sometimes very difficult to get a high resolution picture of the pressure distribution on certain surfaces where dimensions are small, i.e. it may be mechanically impossible to attach many single sensors on small area. The electret sensor has been recently developed and patented at the University of Kuopio (UKU) and the original applications for the sensor were in the medical and industrial fields. The electret foil and the sensor technique have been under research and development for years. The collaboration between (UKU) and (KTH) started during year 2004, in order to investigate the alternative pressure/force measurement technique that the electret senor facilitates. Pre-study tests started at KTH for this new application, in order to perform sensor characterization. One MSc study was performed in , where the goal was to understand more about the calibration procedure and the sensor technique itself. All these studies were undertaken because of the promising economical and technical viability of the electret sensor technique, especially, for cold flow dynamic pressure measurements in the area of turbomachinery. This time the application is different from previous applications of the electret sensor. The sensor must now be calibrated in order to be able to retrieve the physical value of the pressure in Pascal. Were the sensor is used there is no need to know the exact value, only the information on changes in pressure. Furthermore, an important feature of the dynamic pressure thin-film electret sensor when used in turbomachinery research is the limited obstruction in the flow channel and the straightforward way to attach the sensor in the measurement environment. 2. Objectives The overall objectives in this project were the exchange of information on the thin-film electret sensor technique and confirm the manufacturing process in order to miniaturise sensor size near one square millimetre and to confirm the calibration procedure. Also, one aim was to improve the measuring technique in order to insure the quality of the data. 3. Electret Sensor & Measuring technique The electret sensor is a thin-film fast response dynamic pressure sensor, which can be manufactured in various forms and sizes. The electret sensor has an excellent frequency response (in the range of MHz). The sensor includes an electret foil, which is a permanently polarized material with a strong permanent charge. This electret film can be manufactured of low priced commonly recognized polymers. The electret foil has long decay time, typically 100 years for 50% reduction of the corona charge, which is long enough when

5 5 considering the sensor 'life time'. Inside the electret foil there are enclosed air cavities which are not equally distributed in any specific area. Therefore it is important to make note of the direction of the polarity of the electret foil when the aim is to compare different sensors. When using this electret foil in the middle of transversally conducting layers it is possible to measure potential difference between them. Transversally conducting layers (electrodes) can be manufactured in e.g. copper where one side is coated with electrically conducting glue which contains mixed nanoscale nickel particles which are coated with silver. Thus this thin film sensor is based on the ability to produce a measurable potential difference when exposed to an external force or pressure. Raw materials in this sensor are the electret foil and thin slices of transversally conducting material on both sides of it. The wires are connected to these layers in order to connect the sensor to a measuring device. The sensor is very thin and flexible. In figure 3.1 the operating principal of the sensor is sketched. Figure 3.1: Sensor Operation. The measuring quantity is differential which should also be considered in the measuring environment. If the sensor is attached on a metal substrate it is necessary to use an insulating layer between the sensor and the substrate in order prevent capacitive connection between the two. Otherwise the measurement will no longer be differential and the charge distribution will not be equal between the transversally conducting layers. Figure 3.2 illustrates how the charges are dispersed in the metallic substrate if one of the electrodes is connected against ground, which was the case in early try out tests. Figure 3.2: Illustrates charge distribution when measuring is against ground. This explains quite clearly that in previous studies it was impossible to achieve a good shape of the transfer characteristics curves when using a small sensor size. According to figure 3.2, when decreasing the sensor size the sensor starts acting as a dipole and the output signal collapse more dramatically than when the measurement is true differential. When the aim is to miniaturize the sensor size down towards one square millimetre, a pre-amplifier is needed even when the measurement is differential. The electret sensor does not work exactly like a capacitor; however, it still has a capacitance. The equation (1) shows the dependency between capacitance and the area. C = εa. (1) d

6 6 3.1 Manufacturing process Manufacturing electret sensors by hand is easy, simple and quick. The transversally conducting material is a commercial tape from 3M (type 1181), which is cut in slices with for example a lathe. This is in order to ensure that the slices are of equally width, shown in figure 3.3. Figure 3.3: Slices of transversally conducting tape. In between these conducting slices (electrodes) is the electret foil, which is cut here to a smaller are size than the calibration pressure chamber. The electret foil must be bigger than conducting layers in the measurement point otherwise there is a risk of a closed loop. The electret sensors can be manufactured in various configurations. Figure 3.4 below shows two different ways to build the sensor. Figure 3.4: Sensors manufactured in two different ways. The only easy way to build very small sensors is as like the type shown in figure 3.4, right hand side case 'C3', where the measuring area is where these conducting layers are crossing. Furthermore it is very important to prevent capacitive connection between the sensor and any electrically conducting substrate. This can be prevented by using insulating double adhesive tape e.g. 3M (type 467). When miniaturising the sensor size it is very important to take care of the connection points because the thin electrodes are fragile and can easily brake. In figure 5.3 it can be seen that those critical parts of the sensor electrodes are fixed with structural (epoxy) glue. Naturally, these critical soldering connections can be fixed in many different ways depending on the specific application.

7 7 4. Experimental setup The experimental setup for the calibrations test runs and sensor testing is shown in picture 4.1 below. Figure 4.1: Experimental setup. The pressure pulse generator (Dyncal) gave air pulses with adjustable constant amplitude for a selected frequency (<4kHz). The air pulses are lead via a plastic tube in to the pressure chamber where two pressure sensors, one reference sensor (Kulite-XCQ-062) and the electret sensor are located in a symmetrical manner. A differential pre-amplifier (INA128) is used depending on the tested case and with a default gain of This differential pre-amplifier is of low power with a good accuracy and it is used for general instrumentations purposes. It employs a current-feedback input circuit which gives it a wide bandwidth. Signals which are coming from the sensors are connected to a fast data acquisition- and storage system (KT8000). The data files were saved and transmitted to another computer for evaluation and post processing. The data evaluations were performed using Matlab programs. Figure 4.2 illustrates the employed Faraday cage where the air pulses were transmitted and sensors were placed. The reference sensor is a piezoresistive pressure transducer. The measuring principal of the Kulite transducer is based on Whetstone Bridge and in front of every active resistor there is usually a small capacitor in order to cut all the higher harmonic components. This may explain why the Kulite raw signal is so pure and there is no need to modify the signal afterwards. This is also good when the aim is compare electret sensor vs. Kulite and find the exact Pascal value for certain amplitude.

8 8 Figure 4.2: Measuring environment. The Faraday cage was used because it shields long ground cables. Using the long ground cables helped make the manufacturing process significantly easier and also no tedious rebuilding of the system was necessary. Moreover, it is also possible to compare the results in straightforward manner. 5. Results and Discussions The initial task was to get rid off the annoying pick-up noise (basically 50Hz with multiples) by finding a proper connection between the sensor and the measuring device. Cases 'C1', 'C2' and 'C3' shown in figure 3.4 were experiments to find out the optimal connection. Also, the manufacturing procedure was investigated in order to reach a technically acceptable solution. Case 'C3' is built in a way that the shielding of the signal cable is not connected to the ground at either end. The pre-amplifier was not used and the connection was straight to the measuring device. The measuring area was approximately 10,5mm 2. In figure 5.1 is shown the raw signal and the Fourier spectrum of that case. These figures show quite clearly the size of the pick-up noise because the connections are not correct. For some unknown reason there is no peak at 100Hz (multiple of 50Hz).

9 9 Figure 5.1: Raw signal and FFT of case 'C3'. The excitation frequency was 220Hz and it is possible to see in the FFT spectrum that the signal peak of the electret sensor at 220Hz is higher than the peak at the pick-up frequency 50Hz. It is obvious that the raw signal includes the 50Hz and the pick-up that probably are possibly caused by the wrong connections. For every test that was performed the sampling frequency was set to 250 khz and the Jet pressure as 0.4 bar. The break trough test was case 3-4, which is seen in figure 5.2. In this case the measuring area is quite small, 2,25mm 2 and the signal cable shield was connected to ground at both ends. No pre-amplifier was used and the connection was straight to the measuring device (KT800). In figure 5.2 it is noted that the flicker noise or the 'background noise' is eliminated and the pick-up noise is drastically reduced. The raw signal is now clear, as it should be, but not exactly like the reference sensor Kulite as it contains all the higher components that are also affecting the resulting signal. No filtering was used. The size of the peak in the FFT spectrum reveals that the excitation frequency 220Hz is rather small, but when compared with the pick-up noise peaks it is huge. When compared to Kulite it is very small. Based on the above discussion it can be concluded that the point were it is necessary to use pre-amplification is approaching for this small sensor. In theory, both sides are at the same potential when shielding of the signal cable is in the ground at both ends. Hence, it is the correct way to connect the sensor in the measuring device, thus, configuration is also used for the remaining tests.

10 10 Figure 5.2: Raw signal and FFT of case 3-4. One specific aim was to study when it is necessary to use a pre-amplifier. Case A, B and C were cases with different sensor sizes. Dynamic calibration test runs were performed for these three different size sensors in order to confirm the shape of the transfer characteristic curves. According to the theory the shape of those curves should be the same. The case A is sensor with 1mm 2 measuring area, which is the smallest sensor that was manufactured. Also this might be the smallest size that will work reliably because of the limitations of the electret material itself. The electret material is not homogeneous and under these conducting electrodes there must be nearly the same number of charges, otherwise the output signal is no longer comparable between the cases. Figure 5.3 shows the three cases (A, B and C). In case B, the measuring area is 4mm 2 and in case C the measuring area is 9mm 2. Results from case B were eliminated in the evaluation because there was leakage at the pressure chamber during the test. Figure 5.3: Cases A, B and C. All these cases used a pre-amplifier (gain 13.82) and the measuring environment was kept similar for all cases. Repeated test runs for the sensor case A were performed in order to insure that such a small sensor works reliably. Figure 5.4 shows the case A raw signals and the Fourier spectrum, and the results seem excellent even if there is flicker noise. No 50Hz pick up is observed and the peak of the perturbation frequency 220Hz after pre-amplification is sufficiently high to get relevant information.

11 11 Figure 5.4: Case A raw signal and FFT. The frequency band of the electret sensor is larger than the Kulite transducer and no zero padding or other signal processing methods were used when evaluating data with Matlab mainly because this may affect the raw signal. This explains why the raw signal for the electret sensor does not appear as pure as the Kulite transducer signal. In figure 5.5 it is shown how the sensor size affects the output signal. Comparing at higher perturbation (excitation) frequencies is appropriate here as it is easier to understand when the output signal is so low that information would be drowned in the flicker noise. Figure 5.5: FFT for cases A and C in perturbation frequency 3500Hz. The scales for these figures are the same in order to estimate the size of the disturbance when compared to the excitation peak. Sensor size in case A is 1mm 2 and in case C it is 9mm 2. In case C there is a

12 12 measurement error of around 3325Hz but it is showed anyway in order to see the differences in the output signals. Obviously the size of peak in case C with perturbation frequency 3500Hz is bigger than in case A due to the larger sensor area. Dynamic calibration test runs were performed; however, static calibration was not done because there was a leakage in the pressure chamber. Instead the value of kpa 911 V was used for the Kulite sensor and taken from previous studies [1]. These dynamic calibration runs are necessary in order to confirm that the relevant information is captured by the electret sensor. The shapes of these curves correlate accurately with the theoretical model. Figure 5.6: Transfer characteristic curves Case A and C. 6. Future works In the future it is advisable to start with a clean table and rebuild the whole system again. Moreover, care should be taken to consider the basic rules of electronics when manufacturing the system, e.g. remove all unnecessary wires, make the ground wiring as short as possible and shield them in the same Faraday cage where the pre-amplifier is located. It is important to use precious little connections when connecting the sensor wires to the sensor. Also, confirm the sensor manufacturing process even further and figure out the way how to shield all the critical parts of the sensor in order to prevent any electrical disturbances. When using a pre-amplifier it is necessary to take the operating power from a battery placed close to the amplifier in order to reduce any unnecessary ground connections and long ground loops. The above mentioned issues should be dealt are in an appropriate manner in order to reduce the disturbances and get a pure signal. It is evident that a lot of noise that we had was coming from inappropriate connections. When these things are taken care of noise levels reduce significantly and the signal will be reliable and possibly there is no need to use any signal processing methods to get rid of noise.

13 13 When aforementioned issues are taken care of then logically the next step would be to start designing prototype sensor that would be attached into a wind tunnel, and conceive the way how to attach sensor there. Furthermore, to a make a mathematical model of the calibration curves in order to see how the sensor size and the output signal are correlated. This information will help in future calibration procedure. It is also important to study how the size of the electret foil is affecting the signal when using 1mm wide conducting strips (electrodes). Lastly, if the aim is to get high spatial resolution in the pressure distribution on a certain surface it is necessary to start investigations and make preliminary evaluations of the Electret Matrix Sensor (EMS) technology, currently patent pending. It is as simple as using a single sensor in working principle but all the necessary surrounding electronics must be built and tested prior to wind tunnel tests. 7. Summary This SUSPOWER project was an excellent way to exchange information. This project has us lead one step closer to the goal which is to use this thin-film electret sensor in real measurements in turbomachinery research. Miniaturization of sensor size was achieved and the manufacturing process is now reliable. The shape of the transfer characteristic curve reports that the relevant information is there and it goes hand in hand with the theory. Many of the disturbance factors are now known and when taking care of those, in future noise levels can be reduced a lot and signal quality improved. The measuring technique is confirmed and we are aware of the limitations, where and how the sensor can be applicable. It is now possible to manufacture miniaturized single sensor and in a real environment e.g. a wind tunnel. The next step is to go for a full matrix. The collaboration between (UKU) and (KTH) is fruitful and (UKU) are now aware of where the sensor is going to applied, the desired information and the measurement environment. To conclude this project in one sentence, it was a success 8. Literature [1] Asgeir Rygh, MSc Pre-Validation of a Fast-Response Thin-Film electret Pressure Measurement Technique, [2] Jacob F, Handbook of Modern Sensors, Physics, Design and Applications, [3] Demarest K, Engineering Electromagnetism, [4] Data sheet of Low Power Instrumentation Amplifiers (INA 128).