ISRN UTH-INGUTB-EX-E-2019/009-SE
Examensarbete 15 hp Juni 2019
Analysis of Accuracy for Engine and Gearbox Sensors
Daniel Johnsson Erkan Dogantimur
Teknisk- naturvetenskaplig fakultet UTH-enheten
Besöksadress:
Ångströmlaboratoriet Lägerhyddsvägen 1 Hus 4, Plan 0
Postadress:
Box 536 751 21 Uppsala
Telefon:
018 – 471 30 03
Telefax:
018 – 471 30 00
Hemsida:
http://www.teknat.uu.se/student
Abstract
Analysis of Accuracy for Engine and Gearbox Sensors
Daniel Johnsson, Erkan Dogantimur
This thesis provides a standardized method to measure accuracy for engine and gearbox sensors. Accuracy is defined by ISO 5725, which states that trueness and precision need to be known to provide a metric for accuracy. However, obtaining and processing the data required for this is not straight forward. In this thesis, a method is presented that consists of two main parts: data acquisition and data analysis. The data acquisition part shows how to connect all of the equipment used and how to sample and store all the raw data from the sensors. The data analysis part shows how to process that raw data into statistical data, such as trueness, repeatability and
reproducibility for the sensors. Once repeatability and reproducibility are known, the total precision can be determined. Accuracy can then be obtained by using
information from trueness and precision. Besides, this thesis shows that measurement error can be separated into error caused by the sensors and error caused by the measurand. This is useful information, because it can be used to assess which type of error is the greatest, whether or not it can be compensated for, and if it is
economically viable to compensate for such error. The results are then shown, where it is possible to gain information about the sensors’ performance from various graphs. Between Hall and inductive sensors, there were no superior winner, since they both have their strengths and weaknesses. The thesis ends by making recommendations on how to compensate for some of the errors, and how to improve upon the method to make it more automatic in the future.
Handledare: Andrey Gromov
Populärvetenskaplig sammanfattning
I dagens moderna fordonsindustri tillämpas sensorer i många ställen i fordonen. Sensorer ger möjligheten att mäta viktiga parametrar i fordonen, som t.ex. temperatur, varvtal, sammansättningen av kemikalier i avgas m.m. Att veta dessa parametrar möjliggör bl.a.
effektivisering av bränsleförbrukning, men även bekvämligheter som nästan omedelbar motorstart vid tändning. Eftersom att sensorer i fordonen är som ögon och öron för människor, är det viktigt att de fungerar väl. Därför behöver det säkerställas att de fungerar som de ska, men hur kan det säkerställas? Det räcker inte med att bara säkerställa att de fungerar, utan hur väl de fungerar och hur miljön och omständigheter påverkar sensorernas funktion, behöver också tas i beaktning.
I detta arbete har just den frågan besvarats. Genom att utveckla en mätningsmetod som simulerar fordonens kontrollenhet, kan sensorernas prestanda kartläggas. Efter en utförd mätning i laboratoriemiljö, fås statistiska beräkningar som används till att rita upp grafer som beskriver hur noggranna sensorerna är. Varje enskild sensor analyseras enligt denna metod, och jämförs sedan mot sensorer av samma och andra sorter. Just i detta arbete har tre olika sensortyper analyserats. För att mäta varvtal har Hallsensorer och induktiva sensorer analyserats och jämförts, medan differentiella tryckgivare analyserats för att mäta tryckskillnader i gas.
Resultaten visar att Hallsensorer vanligtvis har bättre noggrannhet i och med att mätningar gjorda med dessa ligger närmare det sanna värdet. Induktiva sensorer har däremot något bättre sammanhållning av mätningar utförda med dessa - alltså att även om de visar fel från referensvärdet, visar de ungefär samma fel varje gång och varierar således inte lika mycket som Hallsensorernas mätningar. Hallsensorerna presterar även bättre på lite längre avstånd än induktiva sensorerna. När det gäller differentiella tryckgivarna syns att de presterar bäst i höga temperaturer, vilket också är rimligt eftersom att avgaser i ett fordon har hög temperatur.
Med detta arbete kan alltså en helt ny sensormodell, t.o.m. från en annan tillverkare, sättas på prov genom denna metod. Därefter kan resultaten enkelt jämföras mot sensorer som redan används och är beprövade. Detta kan hjälpa fordonstillverkaren att avgöra bl.a. om det är lönsamt att investera i en ny sensormodell, eller om den nya sensorn klarar de specificerade kraven som finns på sensorns prestanda och för att säkerställa att sensortillverkaren faktiskt håller vad de lovar i sina specifikationer.
1. Introduction 6
1.1. Background 6
1.2. Purpose and goals 6
1.3. Tasks and scope 6
1.4. Outline 7
2. Technical background 8
2.1. Terminology - Error, Accuracy and measurement uncertainty. 8
2.2. Types of Errors 9
2.2.1. Systematic Error 9
2.2.2. Random Error 9
2.2.3. Total Error 9
2.3. Performance Characteristics 9
2.3.1. Trueness 10
2.3.2. Precision 10
2.4. Quantitative Expression of Performance Characteristics 11
2.4.1. Bias 11
2.4.2. Standard Deviation 11
2.5. Summary of Accuracy 12
3. Speed sensors and pressure sensors 12
3.1. Speed sensor 12
3.1.1. Hall Sensor 12
3.1.2. Inductive Sensor 14
3.1.3. Absolute Rotary Encoder 15
3.2. Pressure Sensor 15
3.2.1. Types of pressure measurements 15
3.2.1.1. Absolute Pressure 15
3.2.1.2. Differential Pressure 16
3.2.1.3. Gauge Pressure 16
3.2.1.4. Static Pressure and Velocity Pressure 16
3.2.2. Pressure sensors 16
3.2.2.1. Capacitive pressure sensors 16
3.2.2.2. Optical pressure sensors 16
3.2.2.3. Piezoresistive pressure sensors 17
3.3. Common Sensor Errors 17
3.3.1. Sensor Bias 17
3.3.2. Hysteresis 18
3.3.3. Non-linearity 18
3.3.4. Noise 18
4. Measurement methods 18
4.1. Rotational Speed Measurements 18
4.1.1. Measurement Requirements 19
4.1.2. Equipment 19
4.1.2.1. Custom Built Rotation Rig 19
4.1.2.2. Motor Control Unit 20
4.1.2.3. Absolute rotary encoder 20
4.1.2.4. PC for Rig Control and Data Sampling 21
4.1.2.5. XOR Logic Circuit 21
4.1.2.6. Power Supply 22
4.1.2.7. Measurand A 22
4.1.2.8. Measurand B 22
4.1.3. Methodology - Measurand A 23
4.1.4. Methodology - Measurand B 28
4.2. Differential Pressure Sensor Measurement 29
4.2.1. Measurement Requirements 29
4.2.2. Equipment 29
4.2.2.1. Pressure Controller Rig 29
4.2.2.2. PC Rig 29
5. Results and discussion 34
5.1. Speed Sensor - Measurand A 34
5.2. Speed Sensor - Measurand B 40
5.3. Differential Pressure Sensor Measurement 42
5.4. Custom Built Rotation Rig 48
5.5. XOR Circuit 48
6. Conclusions and future work 48
References 49
Appendix A: Matlab Code 50
Program 1A: Vinkel_Rigg_Hall.m 50
Program 1B: Vinkel_Rigg_Inductive.m 52
Program 2: Vinkel_stat.m 55
Program 3: Result.m 58
Program 4A: Vinkel_Rigg_Hall_cam.m 61
Program 4B: Vinkel_Rigg_Inductive_cam.m 64
Program 5: normdistrep.m 68
Program 6: CenterOffset.m 70
Program 7: DiffPres_raw.m 70
Program 8: Statistics.m 72
Program 9: Result.m 73
Appendix B: Labview Code 75
1. Introduction
1.1. Background
One way to decide which components to use in an electrical design, is to determine which functionality is needed from it. One can then proceed to specify the requirements various parts and components of the design need to meet in order to achieve the functionality of the design.
Important information relevant to the use case of these components is needed. Some of the more common pieces of information are; the temperature range they can operate in, the accuracy and tolerances of electrical values and electromagnetic compatibility etc. One way to obtain this information is to look in the datasheet supplied by the manufacturer. However, datasheets are not always ideal and it is arguably a risk to trust a supplied datasheet wholeheartedly - especially if one is building mission critical equipment that need to perform reliably in various environments. It would be beneficial to test the specifications claimed by the manufacturer, and evaluate them by comparing them to the test results.
1.2. Purpose and goals
The main goal of this thesis was to introduce a way of testing sensors against the manufacturer’s claim of accuracy, and to differentiate between different types of errors. The authors test two different kind of sensors: speed sensors and differential pressure sensors.
For the speed sensors, two different technologies were tested and compared, but for the pressure sensors there was only one. The report starts by introducing the definition of performance and by giving an explanation to common error sources. Different sensor technologies and their theoretical explanations are also briefly covered by the report.
In this thesis, the components in question are primarily rotational speed sensors and differential pressure sensors. These sensors can be supplied by a range of different manufacturers, whom can change over time. Also, the manufacturing process itself can change over time for any manufacturer. This means there are potentially a lot of variables that can play a role in the function of an electrical design. Thus, there is a need to develop a standardized way to easily and efficiently measure and evaluate the sensors, which is the primary goal of this thesis project. This goal includes both the practical measurement process as well as the statistical calculations required to judge the sensors’ performance.
1.3. Tasks and scope
The main tasks to be conducted are the following:
● Find a way to practically position sensors during measurement
● Understand how to utilize the reference sensors
● Sample signals from the reference sensor and the sensors
● Export all sampled signals for processing in Matlab
● Process the sampled signals in Matlab for statistics
● Present the results and compare Hall and inductive sensors
● Separate measurement error into systematic and random error and identify the source of error
1.4. Outline
The report follows the timeline of the actual tasks being performed and starts by giving the reader some understanding into the terminology used and also to the theoretical background of the measurements performed. The first measurement made was with the Hall speed sensor and most of the preparations for the following measurements were done using test data from this sensor. The report explains the methodology and process for each measurand in the order they were done and the results and some recommendations are presented at the end of the report. The Hall, inductive and differential pressure sensors will be explained under the theory section.
In this thesis, Daniel Johnsson is responsible for:
● Abstract
● Introduction
● Theory: Speed Sensor
● Theory: Differential Pressure Sensor
● Theory: Common Sensor Errors
● Equations
● Method: Differential Pressure Measurements
● Results and discussion: Differential Pressure Measurements
Erkan Dogantimur is responsible for:
● Background
● Theory: Accuracy - Terminology
● Method: Rotational Speed Measurements
● Results and discussion: Speed Sensor - Measurand A
● Results and discussion: Speed Sensor - Measurand B
● Results and discussion: Custom Built Rotation Rig
● Results and discussion: XOR Circuit
● Conclusions and future work
The parts not listed above are co-written by both authors.
2. Technical background
2.1. Terminology - Error, Accuracy and measurement uncertainty.
One thing that is often talked about in the context of measurement, is accuracy. The term accuracy can mean different things from one person to another, as well as in different contexts. There is however a well defined description of it in the ISO 5725 standard [1], which will be described in this section. Figure 1.1 shows how accuracy is defined by this standard.
Figure 1.1: Error, accuracy and measurement uncertainty.
A structured and clear way of interpreting Figure 1.1, is to start from the left column that is entitled “Types of errors”. These errors - systematic and random - refer to the exact errors that affect the measurement in different ways, which is explained in 2.2. The middle column entitled “Performance characteristics”, represents the systematic and random error as estimates instead of exact values, and refers to them as “trueness” and “precision”
respectively. “Performance characteristics” is further explained in 2.3. Finally, the right column entitled “Quantitative expression of performance characteristics”, shows the actual
mathematical terms used when calculating trueness and precision, and is further explained in 2.4.
2.2. Types of Errors
2.2.1. Systematic Error
Starting with systematic errors, these are such that when performing a measurement, there is a deviation from some reference value. Naturally, an “instinct” when measuring something, is to make multiple measurements. One can then take the average of these measurements to consolidate all the information into one meaningful value that is easy to interpret.
However, taking the average only reduces the random error while the systematic error remains. A few examples of common systematic errors are; an offset voltage such as a DC-offset on a sensor’s output voltage, a broken ruler which starts measuring from 1 cm instead of 0, a light intensity meter with a dirty sensor which blocks some of the light being measured, etc. These errors are deterministic by their nature, meaning a ruler with an offset of a few length units will always have that offset, no matter where on the scale one measures.
2.2.2. Random Error
Random errors however, are of the non-deterministic nature. This means they cannot be known in advance. For example, noise in a picture taken by a camera, or in an audio recording, is largely caused by the random motion of electrons in the electrical components.
The nature of this error is that it fluctuates between different values within some interval.
This error will then be superpositioned on the measured value. The most common way of reducing this type of error is simply by averaging. The average of something that is roughly equal in magnitude in both the positive and negative direction, is roughly zero.
2.2.3. Total Error
The total error is the sum of the systematic error and random error. For example, consider a known perfect reference voltage of 1.0V is measured several times. If the average of the measurements shows 1.1V, with all of the measurements being within the interval 1.0V to 1.2V, it can be estimated that the systematic error is +0.1V, while the random error is +/- 0.1V. The total error would then be +0.1V, +/- 0.1V.
In reality however, the error and the true value are closely linked together in the sense that knowing one of them exactly, is to know the other exactly. If the exact amount of error is known, it can be subtracted from the measured value to obtain the true value. However, all measurements contain some errors. Since reference values themselves are measured values and contain error, they are not true values.
2.3. Performance Characteristics
Since true values cannot be known exactly, there is a need for another way of describing the quality of a measurement. The second column of Figure 1.1, shows this quality as performance characteristics. These are used to describe the errors as estimates instead of the elusive exact values which cannot be known. The keywords in this column are “trueness”
and “precision”.
Figure 2.1 describes trueness and precision in more detail.
Figure 2.1: Accuracy in terms of trueness and precision.
2.3.1. Trueness
Trueness is an estimation of the systematic error. To find a value for trueness, the true value no longer needs to be known. Instead, a reference value can be chosen and agreed upon.
This could for example be when measuring the temperature in a room. A well calibrated temperature measurement device can for example show that the temperature in the room is 25.0°C. This value is then accepted as the reference value for the measurement. All other thermometers in the room would then be compared to this reference value, which would then be used to judge their performance.
2.3.2. Precision
Precision is an estimation of the random error. Using a finite number of measurements, it is possible to express an estimation of the random error, whereas an infinite number of measurements would be required to truly know the random error. This is because true randomness will not follow a preset behaviour, meaning each time the random error appears, the magnitude and sign of it will be unknown until measured. Hence, the need to
measure an infinite number of times to measure all of the possible outcomes of the randomness.
In the ISO 5725 standard, precision refers to the total random error - a combination of repeatability and reproducibility. The total random error is calculated in the following manner
σT ot = √σrpt2 + σrpd2 (2.1)
where σrpt is repeatability and σrpd is reproducibility.
2.4. Quantitative Expression of Performance Characteristics
When doing calculations of accuracy, the two terms used are “bias” and “standard deviation”.
2.4.1. Bias
Bias, which is also known as offset, is the distance between the mean value of the measurement data set and the reference value. If one calculates the distance between those two values, the bias is obtained. The mean is expressed as
μ = N1 ∑N
i=1
xi (2.2)
where is the number of samples and is the sample value.N xi
2.4.2. Standard Deviation
In the same example above, if one had a number of temperature measurements taken at the same temperature several times, they would form an interval - a range of values. The wider the interval, the greater the standard deviation. Greater standard deviation means worse precision. The standard deviation is expressed as
σ = √i=1∑N (x − μ)i N 2 (2.3)
where N is the number of all samples belonging to the population, xi is the sample value and is the population mean value.μ
2.5. Summary of Accuracy
The true value, and the true error cannot be known due to measurement imperfections. No matter how sophisticated equipment one has, there will always be imperfections causing uncertainty in the measurement. By introducing the terms “trueness” and “precision”, it is easier to talk about accuracy, which encompasses both of these terms.
Trueness is described by the term bias, which describes the distance between the measurements mean value and the reference value. Precision is described by the term standard deviation, which describes how spread out the measurement values are around the measurement’s mean value.
3. Speed sensors and pressure sensors
3.1. Speed sensor
3.1.1. Hall Sensor
The Hall sensor is used to detect the presence of magnetic fields and works on the principle of the Hall effect, which can be described as the difference in potential that arises due to the force induced in moving charge carriers in the presence of a magnetic field [2]. Electric current through a conductor is the movement of charge carriers that collide and exchange momentum. When looking at the overall sum of the current movement it follows a straight path. In 1879 the physicist Edwin Hall discovered that when a magnetic field is introduced to a moving current, a voltage could be measured at the sides perpendicular to both of them.
This is what is known as the Hall effect.
The Lorentz force law defines the resulting force from an electromagnetic field that is acting upon a point charge. When the moving charge carriers are affected by this force, their distribution is disturbed and the current path is curved. The resulting separation of charges gives rise to a potential difference between the sides of the conductor. This potential difference is called the Hall voltage [2].
The Hall-voltage is given by Equation (3.1) and is directly proportional to the magnetic field strength. The applicational use of the Hall-effect is implemented in a number of devices including one of the sensors tested in this thesis project.
Vh = neDIB (3.1)
where I is the current through the conductor, B is the magnetic field strength, n is the density of mobile charges, is the electron charge and is the conductor height.e D
The sensor casing usually contains an integrated permanent magnet whose field is amplified with the presence of ferromagnetic materials. The voltage generated from the Hall effect is in the order of 30µV when affected by one gauss magnetic field and needs signal conditioning including amplification and temperature compensation in order to be of any practical use [2].
The basic hall effect sensor is an analog device that outputs a voltage proportional to the proximity to a constant magnetic field. Many hall sensors, such as the ones used in this thesis project include electronics such as differential amplifiers and schmitt triggers to give a digital output. These sensors only have two output states - either a high or low voltage.
Figure 3.1: Hall sensor.
The sensors using Hall effect technology tested in this project are active components and need a supply voltage to operate. The sensor supply pins are connected to a power supply Vcc and ground and the signal is read between the signal pin and ground. When these sensors are placed next to a gear wheel, the output of the sensor follows the structure of the gear teeth and goes high for every tooth and low for every gap in between.
3.1.2. Inductive Sensor
Inductive sensors work on the principle of Faraday's law which states that the induced voltage in a loop of wire is proportional to the rate of change in a magnetic field [3], which is given in the following manner
Ei = NdΦdt (3.2)
where Ei is the induced voltage, is the number of coil turns and is the magnetic flux.N Φ
The voltage is generated no matter how the magnetic field is changed - be it a change in magnetic field strength, direction or angle [4].
Figure 3.2: Inductive sensor.
When the inductive sensor passes the gear teeth of the wheel, the magnetic field from the permanent magnet inside the sensor, is focused with the proximity to the teeth. As the gear teeth get closer, more of the magnetic field is focused and amplified. The voltage induced by the coil results in an alternating voltage that is directly proportional to the change in the magnetic field. Since the sensor induces its own voltage, no supply is needed. One drawback of this is that the inductive sensor only gives a measurable output voltage once the field is changing rapidly enough e.g. when the wheel rotation is fast enough. It is therefore not possible to measure anything stationary with this type of sensor.
3.1.3. Absolute Rotary Encoder
As a reference to the sensor measurements a high resolution absolute optical rotary encoder was used. The absolute rotary encoder indicates the current position of the shaft - meaning that it shows the angular position of the shaft in reference to a specific starting point. This was an important feature for the reference sensor as the measurements made with the speed sensors was done on a gear wheel and statistics was to be performed on specific teeth for different sensors. The absolute encoder had a 13 wire output corresponding to a 13-bit resolution or 213 (8192) edges of pulses for one full revolution. This resolution is obtained by detecting both the rising and falling edges of the pulses. However, due to limitations in sampling rate only the falling edges were used and a resolution of 4096 was obtained.
3.2. Pressure Sensor
Pressure sensors are utilized whenever pressure needs to be known. Solid objects exert pressure on whatever they are resting on, while fluids can exert pressure not only on what they are resting on but also on what is adjacent to them. Gases exert pressure in all directions and for the purpose of this thesis project, gas pressure, especially differential pressure, is what will be focused upon.
Figure 3.3: Picture of differential pressure sensor.
3.2.1. Types of pressure measurements
3.2.1.1. Absolute Pressure
Absolute pressure measures pressure against zero pressure, or in other words, against vacuum. For instance, if absolute pressure in a flat bicycle tire is measured, the
pressure on Earth [5]. If one was to inflate the same tire, its absolute pressure would be a few times that of the atmospheric pressure.
3.2.1.2. Differential Pressure
Instead of always measuring pressure against vacuum, any arbitrary non-vacuum pressure can be used as reference. This type of measurement is called differential pressure measurement. For example, a factory manufacturing pressurized cans might have one can to use as the reference when pressurizing other cans. In that case, differential pressure is used in the measurement process.
3.2.1.3. Gauge Pressure
Similar to differential pressure, gauge pressure is the pressure between a non-vacuum reference and the measurand. However, the reference in gauge pressure is the atmosphere.
Generally, most pressure readings done in everyday life are gauge pressure readings. For example the pressure in; tires, spray cans, diver tubes etc.
3.2.1.4. Static Pressure and Velocity Pressure
If there is no wind outside and one measures pressure - the measurement would show one atmospheric pressure. If the wind suddenly starts blowing and one measures against the wind, the measured pressure could be larger due to the added pressure from the moving air particles pushing against the measuring device. This is called velocity pressure, which simply is the pressure caused by moving particles.
Total pressure refers to the sum of static pressure and velocity pressure.
3.2.2. Pressure sensors
3.2.2.1. Capacitive pressure sensors
Capacitive pressure sensors make use of the relationship of capacitance between two plates and their distance. The sensor works by measuring the change in capacitance between two membranes that - when inflicted by the force of pressure, deflect and move closer to each other. The change of capacitance is typically in the magnitude of 50-100 pF and some signal conditioning is required [6][7].
3.2.2.2. Optical pressure sensors
Optical pressure sensors use optical fibers as a sensing element that transmits light produced by a source. The light is reflected against a diaphragm and is received with a certain time delay by receiving fibers. When pressure is exerted on the diaphragm, it deflects and the reflected light is delayed or propagated by the change in distance. This change in time delay is measured and converted into a change in pressure [8][9].
3.2.2.3. Piezoresistive pressure sensors
Piezoresistive materials are materials that change their resistivity when they are bent or compressed. This deformity is the result of a force being exerted on the material. The relationship between force and resistivity is usually a linear one and the effect is applicable for most materials, even metal to some level. A piezoresistive sensor makes use of this relationship and contains strings of semiconductor silicon that when deformed by the force of pressure, change their resistance and thereby lowering the amount of current through them.
The change in resistance is small and the sensor piezo element is usually connected to a measurement bridge like a wheatstone bridge. This technology is the one used in the differential pressure sensors in this thesis [6][7].
Figure 3.4: Illustration of a differential pressure sensor. Note: the plus signs on the right side are not electrical charges - they merely represent positive pressure relative
to reference pressure (left side).
3.3. Common Sensor Errors
3.3.1. Sensor Bias
The bias or offset value can change over time due to lifetime characteristics, material deformation, and more. All electrical sensors will have some zero offset, e.g. they will have some output difference from a 0V input. This can be due to the tolerance in components used in manufacturing. For example, inbuilt resistors or other components can differ in value from sensor to sensor resulting in some difference in output. Some sensors will develop this error overtime due to heavy use while other sensors will have a settling in period and offset can decrease with time. This is called zero drift and is unique for every sensor.
3.3.2. Hysteresis
During continuous measurements with sensors, the output signal can be different depending on if the measured value is increasing or decreasing. Hysteresis is the retardation of a force acting on an object after the force changes direction. For example, when reading the output from a temperature sensor encased in a plastic body. If the temperature suddenly drops outside the casing, the output signal will continue to read the same value as before the drop until the sensor body and sensor element have reached the same temperature as the air outside. Hysteresis can be thought of as some form of latency, not necessarily in time but in the same coordinate as the measurand, which is temperature in the previous example.
3.3.3. Non-linearity
The maximum deviation from an ideal linear output and the characteristics from a real output is the non-linearity error. It is preferable for any sensor to have a linear output response. This makes for easier prediction of abnormal behaviour and excludes the need for a lookup table.
When minimizing the error from non-linear sensor output, critical data points can be selected and interpolated. This technique can drastically lower the size of a lookup table and still yield measurements with the desired accuracy.
3.3.4. Noise
A common cause for random error is noise. Noise can be in many types of measurements, whether they be audio recordings, temperature readings or force measurements etc. As mentioned previously, one type of noise can be in the power supply, causing the unit it is supplying to exhibit a random error component in its output.
Noise can also come from an external source, such as from electromagnetic induction from other electrical equipment, or even natural phenomena such as a solar flare.
4. Measurement methods
4.1. Rotational Speed Measurements
In this section, the measurement requirements, equipment and methodology will be explained. The rotational speed measurement consists of two types of measurements, where one is to measure on measurand A, Figure. 4.3, and the other is to measure on measurand B, Figure 4.4. Measurand A will be measured with 12 Hall and 12 inductive sensors, while measurand B will only be measured with one Hall and one inductive sensor.
This is because measurand A will be used to obtain reproducibility data while measurand B will be used to obtain repeatability data.
4.1.1. Measurement Requirements
These requirements are for both measurand A and B:
● 100 Rotations
● 100 RPM
● 1 and 2 mm Distance
● Hall and Inductive Sensor(s)
● High Resolution Reference
● Fast Sampling Rate
In order to collect enough data for statistical analysis, at least 100 rotations of the measurands (the wheels) were needed to be sampled. The rotation speed was initially decided to be 500 and 1000 RPM to more closely resemble real world scenarios, but had to be lowered to 100 RPM due to a sample rate limitation of 200 kS/s on the DEWESoft DAQ.
Otherwise, the number of samples to describe the fastest changing bit from the rotary encoder was not sufficiently high, which would result in inaccuracies when detecting the edges of the signal’s pulses.
Furthermore, the measurements were required to take place at 1 and 2 mm distance from the wheels, to gain statistics of the sensors’ performance in regards to distance. Both Hall and inductive sensors were to be tested for comparison against each other, as well as against themselves as a group.
To do this, two different measurements had to be used. One focused on reproducibility while the other one focused on repeatability. This was possible by using two different wheels as measurands. Figures 4.3 and 4.4 show these two measurands and explain why they were used.
4.1.2. Equipment
In this section, the more important bits of equipment that were used are listed and explained.
● Custom built rotation rig
● Motor control unit
● Absolute rotary encoder
● PC for Rig Control and Data Sampling
● XOR logic circuit
● Power supply
● Measurand A
● Measurand B
4.1.2.1. Custom Built Rotation Rig
In order to perform measurements on a rotating wheel, rotation had to be generated. This was done through a rig which contained; a motor control unit, a motor, and an axle that
rotated. The purpose of the rig was to provide a stable rotation velocity. The rig is shown in Figure 4.1.
Figure 4.1: Custom built rotation rig.
4.1.2.2. Motor Control Unit
The motor control unit was a Leadshine AC Easy Servo Drive, model ES-DH2306.
4.1.2.3. Absolute rotary encoder
On the right side of the axle visible in Figure 4.1, an absolute rotary encoder was pre-mounted. The encoder was a Leine & Linde RHA 507.
The absolute rotary encoder had 13 output bits, which are indexed from Bit 0 up to Bit 12 in this report. The output was in the form of square waves. Bit 12 corresponds to the MSB (Most Significant Bit) while Bit 0 corresponds to the LSB (Least Significant Bit). One period from Bit 12 equals a rotation of 360°, while each subsequent bit period equals half the degrees of the previous bit. Equation (4.1) shows a general way to describe degrees as a function of bit index.
° θn = 360
2(12−n) (4.1)
where n is the bit number, increasing from the LSB to the MSB.
4.1.2.4. PC for Rig Control and Data Sampling
The rig control PC ran LabView, which controlled the rotation rig. It also ran DEWESoft X3 SP3 for data acquisition. The PC was an HP EliteDesk, with an Intel i5-7500 3.4 GHz CPU, with 8 GB of DDR4 RAM. The operating system was Windows 10 Enterprise 64-bit.
4.1.2.5. XOR Logic Circuit
Before the measurements could begin, it was decided that a Gray to binary converter circuit was needed to simplify the measurement process. This is because the output from the rotary encoder was in the form of Gray code instead of binary code. Since binary code is easier to work with, a total of 12 XOR gates were used to build a Gray to binary converter. This is a very simple build, consisting of three quad XOR gates - in this case a DM7486N and two SN7486N. The only reason for not using all three quad XOR gates of the same type is because they were not available.
Figure 4.2 shows the logical connections of the XOR circuit.
Figure 4.2: XOR circuit connection.
4.1.2.6. Power Supply
The power supply was an AIM-TTi CPX200D. Only output 2 was used, to supply 5V to the XOR logic circuit and the Hall sensor that was connected at each measurement.
4.1.2.7. Measurand A
Figure 4.3 shows measurand A, which is a metallic wheel with 38 evenly distributed teeth.
When sampled by a sensor, the sensor’s output will correspond to the teeth it detects on the wheel. Thus, one period on the sensor’s output will ideally correspond to 1/38 of 360 degrees, which is approximately 9.47°. This wheel is mainly used to gather statistics about reproducibility, by using 12 Hall sensors and 12 inductive sensors.
Figure 4.3: Measurand A.
4.1.2.8. Measurand B
Measurand B, shown partly in Figure 4.4, is a different metallic wheel with a tooth period of 4 degrees on the small teeth. A large tooth, which can be seen in the image, has a tooth period of 8 degrees. This wheel is patented and can thus not be shown in full in this report.
Because of the presence of large teeth in this wheel, the magnetic field is not repeating in the same way from tooth to tooth, as in the case of measurand A. The “disturbance” in the magnetic field caused by the large teeth, is an interesting attribute to study regarding repeatability. When measuring this wheel, only one Hall and one inductive sensor is used.
Figure 4.4: Part of wheel B.
4.1.3. Methodology - Measurand A
To begin with, the XOR circuit was mounted on a breadboard, and the outputs from the rotary encoder were connected to the inputs of the XOR circuit as shown in Figure 4.2.
Voltage and ground from the power supply were connected to the circuit. The outputs Bit0, Bit9 and Bit12 from the XOR circuit were connected to the DEWESoft DAQ using coaxial
cables. The coaxial cables transferred the output signals along with the ground reference to the DAQ. DEWESoft DAQ was connected to the PC with a USB 2.0 cable.
Measurand A was mounted on the rotation rig chuck. To make sure the wheel was not wobbling noticeably in the axial direction, a displacement measurement tool was used. This allowed for measuring how much the wheel wobbled with a resolution of 0.1 mm. The total wobble was noted to be 0.2 mm, as measured by the device.
The motor control program in LabView was opened on the PC. From there, rig rotation speed was set to a low value - approximately 10 RPM - for the purpose of positioning the sensor close to the wheel. The low rotation speed ensured there was no risk of damaging the sensor or the person mounting the sensor.
DEWESoft X3 SP3 was started on the PC. In there, sampling rate was set as high as possible, which was 200 kS/s. Four input channels were activated for sampling. Three of those channels were from the rotary encoder; the MSB, LSB and Bit 9.
The first sensor type to be measured was the Hall sensor, but there was no preference over which sensor type to start with. The sensor was held in place by the tool as shown in Figure 4.5, which could couple itself magnetically to the ferromagnetic table on which the entire measurement took place. A 10 mm thick aluminum plate was mounted vertically with two screws to the tool. The position of the plate could be finely adjusted in the horizontal plane, which allowed for positioning the sensor precisely. The sensor was in turn mounted through a hole in the plate with a screw.
Figure 4.5: Tool that holds the sensor.
The sensor connector was attached to a cable with the corresponding connector type on one end, and three banana plug connectors on the other end. Two of those banana connectors were connected to a coaxial cable, which was connected to the DAQ, and one of those two was also connected to the power supply’s ground reference, while the third one was connected to the voltage supplied by the power supply. The connection is shown in Figure 4.6.
Figure 4.6: Hall sensor measurement system connection.
All moving parts of the mounting system were locked with locking screws to ensure the sensor did not move during the measurements.
The sensor was placed in the axial direction towards the wheel, and was brought progressively closer until it made contact with the wheel. At this point, a slight scraping noise was heard. This point was assumed to be 0 mm from the wheel. The sensor was then adjusted to 1 mm away from the wheel, by using the position adjustment knobs and the measurement scale built in to the tool that holds the sensor.
The speed of the rotation rig was increased to 100 rpm and the power supply was set to 5 volts to supply the XOR logic circuit and the Hall sensor. The sensor output was checked within DEWESoft, to make sure no pulses were lost due to bad sensor placement. Figure 4.7 shows the three bits from the absolute encoder and the hall sensor output.
Figure 4.7: First: 360° pulses (MSB). Second: 45° pulses(BIT 9). Third: 0.088° pulses (LSB).
Fourth: Hall sensor pulses.
The sensor had to be placed at a specific distance along the radius of the wheel, in order to correctly register the teeth. Otherwise, some of the teeth were not registered by the Hall
sensor. Figure 4.8 shows how the sensor was positioned along the radius of the wheel.
Figure 4.8: Hall sensor radial positioning.
The record function was then used in DEWESoft, together with the option to automatically stop the recording after 63 seconds. This ensured that about 105 revolutions were sampled at a wheel rotation speed of 100 RPM. After one measurement was completed, the distance between the sensor and the wheel was increased to 2 mm, and the sampling process was repeated. When the sensor had been sampled at 1 and 2 mm distance, the rotation rig was stopped and the sensor’s screw was loosened to remove the sensor. The next sensor was then mounted to the rig and was brought to a distance of 1 mm to the wheel. DEWESoft X3 SP3 was once again set to record 63 seconds. This process was repeated for all sensors until all of them had been sampled at 1 and 2 mm distance.
Once all 12 Hall sensors were sampled, they were put aside and the same measurement process was repeated for the 12 inductive sensors. The setup for the inductive sensors were almost the same as for the Hall sensors, with one exception: they did not need any supplied voltage in order to function. Figure 4.9 shows the connection setup during the inductive sensor measurement.
Figure 4.9: Inductive sensor measurement system connection.
4.1.4. Methodology - Measurand B
For measurand B, shown partly in Figure 4.4, the process was nearly identical to measurand A. The only exception was that only one of each sensor type were used, meaning one Hall and one inductive sensor. The setup and procedure was otherwise the same, with the same requirements regarding number of rotations to sample, the rotation speed, measuring at two distances etc.
After all of the measurements were completed, for both measurands, the raw data files from DEWESoft X3 SP3 were copied to a USB drive and transferred to a computer that had both Matlab and DEWESoft X3 SP3 installed. The rig computer that was used did not have Matlab installed, which is a requirement to be able to export the data to a .mat file.
Therefore, a separate computer had to be used for exporting. From there, the files were transferred over the network to a work computer, where they were put into a folder which is read by Matlab. The Matlab script was written in such a way that once all the raw data .mat files were placed in a specific folder, Matlab imported every file in that folder and processed them, detecting edges, converting the measured data from time to degrees using the reference and storing the processed data in matrices. Those matrices were then exported, so they could be imported in another Matlab script for statistical calculations. The statistical calculation results were then exported to be used in another Matlab script, where they could easily be presented and plotted in a way that was relevant to the goal of this thesis. The Matlab script can be found in the appendix of this report.
4.2. Differential Pressure Sensor Measurement
4.2.1. Measurement Requirements
● Gauge pressure measurement from atmospheric to 400 mbar
● High resolution reference
● 9 measurement levels from min to max pressure
● Minimal leakage from pressure controller
● Incremental and decremental measurements
● Five different temperature measurements
The measurement criteria were specified to get sufficiently precise data for statistical analysis. The twelve sensors were tested in groups of six where they were pressurized using a highly calibrated reference controller. The setup was leak tested using the inbuilt leak test function of the controller and by lowering the sensors into water to look for air bubbles. To look for hysteresis effects in the sensors they were both pressurized and depressurized in steps of 50 mbar. The measurements were done in five temperature increments of 25 °C in the range of 25 to 125 °C. Since the temperature in the control chamber was distributed using a fan it was important to have the unconnected reference port of the pressure sensors facing away from the outblow of the fan. This was done to prevent any velocity pressure error.
4.2.2. Equipment
● Fluke 6270A - Pressure controller/calibrator
● Fluke PM200-G2M gauge measurement module
● Fluke PM200-A100K absolute measurement module
● Agilent U2351A data acquisition device
● Vötchtechnik VT4011 temperature test chamber
● ADLINK DIN-68S-01 terminal board
● TTi CPX200 Dual 35V 10A Power supply unit
● Pressure hose with connectors
● Netgear R6220 router with ethernet cables
● Computer - Windows 10 Home Intel core m3, 8 GB RAM
● Yokogawa DL750 Oscilloscope
4.2.2.1. Pressure Controller Rig
To generate a stable reference pressure the FLUKE 6270A pressure controller was used.
The controller was equipped with two calibration modules. The PM200-A100K absolute pressure measurement module and the PM200-G2M gauge pressure measurement module.
4.2.2.2. PC Rig
The pressure controller was remotely controlled with a computer connected to a local network through a Netgear R6220 router. The computer was an ASUS Zenbook UX305 with
