Care will have to be taken throughout the calibration process and the recording of results. The anemometer will rotate at a fast rate, and would cause damage if it hit the user’s eye, especially as the cups are pointed. As the calibration is to be conducted in the dark room, and the length of the arm that the cups are mounted on may be underestimated, the user must take precaution.
Proposed sensor: Anemometer
I have chosen to construct an anemometer to measure wind speed. To do this I will make use of some important principles. Firstly I will construct a device to transfer the movement of the air into a rotary motion compliant with a 3v direct current motor, in the most efficient way possible to keep air resistance and friction to a minimum. The wind’s energy will be transferred to the motor, to create a voltage which I will record using a multimeter. This is in effect an alternating current generator on a very small scale and is similar to idea 1, that was proposed on page 1, except for the fact that the magnets are attached to the central axis rather than cups, enabling the magnet to be in much closer proximity, therefore instead of a small intermittent creation of current the connection of a motor enables a constant alternating current to be produced. This is because the coils of wire are always being passed through a magnetic field.
The alternating current generator, shown below, works on the principles of the motor effect. The motor effect states that if you put a direct current through a wire in a permanent magnetic field, the wire will move up or down, depending on the direction of the magnetic field. The alternating current generator works quite simply; by moving the wire through the magnetic field, a current is produced. This therefore is a way of converting kinetic energy into electrical energy.
To ensure as efficient energy transfer as possible, the design of the wind-catching device is crucial. The inclusion of three cups was not just an accident; any more than three cups and the air resistance of the cup returning against the flow of the wind is greatly increased, and less than three cups does not provide enough ‘wind catching’ area. Two cups would enable the cups to become aligned in the stream of the wind, and stop moving. Three cups offers the optimum arrangement and ensures that one cup will always be picking up the wind, without creating unnecessary air resistance. It was also necessary to make the connecting arms thin, to keep resistance to a minimum. For the same reason, the ‘cups’ were pointed to ensure they cut through the air when they were passing back through the air, against the direction of the wind.
The motor will be connected to a multimeter which will accurately measure the voltage produced from the motor. To determine whether the relationship between voltage produced, and revolutions per minute (rpm) is linear or not, I will have to calibrate the sensor. Once calibrated, I will be able to use a look-up table to convert voltage produced into wind speed.
Calibration
To calibrate the sensor, I will need to accurately record the rpm of the cups. When I have accurately recorded the rpm of the cups for certain potential differences, I will be able to produce a calibration curve for the relationship between potential difference created by the motor and wind speed.
To record the rpm of the cups I had initially intended to use light gates, and to record three interruptions per revolution. But there were too many irresolvable issues using this method. The main problem was that it was not possible to make the data harvester (the piece of hardware that translates messages from the light gate to the computer) to count continuous interruptions of the light beam. Light gates are intended to measure the time in between interruptions of two separate beams. Additionally the light gates may not have been able to detect such a small interruption, as the arm that the cup is attached to may not have been sufficient to create a large enough interruption in the beam to be registered.
To overcome this I used a stroboscope, a device which creates a constant flashing beam, whose rate can be altered. The device was able to produce 200 flashes per minute, to over 2000 flashes per minute. Necessary for my calibration was around 200 to around 500 flashes per minute. Care had to be taken as a stroboscope can trigger fits of epilepsy, as it produces high frequency flashing lights.
To record the rpm of the cups, I had to use a constant and even wind source; for this I used a fan. I then placed the anemometer with the motor attached to a multimeter into the wind stream. I then adjusted the position of the anemometer in relation to the wind source to achieve a constant potential difference created by the motor. Once a constant voltage was being produced, I could then adjust the stroboscope to produce more or less flashes per minute, until the cups appeared frozen in motion. To make this easier the practical was done in a dark room so the stroboscope was the only source of light. Once the stroboscope was set so the cups appeared frozen, I counted the number of flashes per minute produced by the stroboscope.
To do this, I used a stopwatch and counted 20, or 40 flashes depending on the rate of flashes and measured the amount of time it took for the stroboscope to produce the recorded number of flashes.
During the practical, each cup was sighted in one revolution when the cups appeared frozen. It was possible to show this as one cup had a black mark on it. This cup appeared every third ‘frozen cup’. To calculate rpm for a given potential difference the following calculation had to be followed:
Three cups were seen for every revolution; this means a flash was produced every 1/3 of a revolution. This was very convenient, as if only one flash was produced per revolution, it would be a lot harder to get the cup to appear still.
I took rpm readings for a potential difference of 40, 50, 60, 70, 80 and 90 mV, and repeated the calibration 5 times prior to the trial readings. This was to ensure that any anomalies were not influential in the final results. Initially I had set the stroboscope to produce a known number of flashes per minute, and then adjusted the anemometer’s position in the wind stream, until each cup was sighted in one revolution whilst appearing motionless. This did not prove successful as there were only a few plausible defined values on the stroboscope that I was able to use. For example, 300, 1000, 2000, and 5000 counts per minute were marked, the scale in between these was non-linear. This meant that I was only able to take a reading for 300 flashes per minute as the anemometer did not turn at more than 165 rpm, so the maximum counts per minute needed were around 495 (165*3). One rpm reading was not enough information to determine the relationship between the potential difference produced and the rpm of the anemometer. To overcome this, I amended my method as explained previously.
Calibration Results
Trial run
From these results I concluded that to increase accuracy I would count the number of flashes for a longer period of time to reduce margin for error. This part of the process was where there was margin for error on my part. So by completing 5 subsequent readings for each potential difference I was able to reduce the affect that a small error would have on the results through averaging.
There were other steps that I took to ensure fair results were recorded. When holding the motor it was imperative that I did not make a bridge between the two contacts, as the human body conducts electricity. Also the multimeter used must have its own power supply to power the LCD display, so that potential difference created by the ‘generator’ is not used in the multimeter. In addition the same motor should be used throughout the calibration and the recording of results, as different motors may produce slightly different voltages, as I am measuring 40 to 80 thousands of a volt during the calibration.
Readings 1-5 – Results for times and number of flashes in that time
Readings 1-5 – Results for calculated rpm from previous results
I was then able to create a look up able for potential difference of 40 to 80mV against rpm.
But this is still not the final result as I need to create a look-up table for potential difference against wind speed. To calculate wind speed from rpm, I needed to apply a simple equation of motion: v=s/t. Speed = distance / time. It is also important at this point that I am not calculating the velocity of the cups, as this would imply a direction. The cups’ velocity is always changing and as they travel in a fixed circle their displacement for one revolution will be 0 metres. This therefore means that in one revolution the velocity is also 0 ms-1. I can calculate the speed of the cups, to produce a speed that will be taken as the value for the speed of the wind. Whether this value is accurate will be analysed later in the report. If the values are not accurate I will suggest reasons why and suggest possible improvements.
To calculate ‘cup’ and therefore wind speed:
Circumference of a circle = pi*d (where d is the diameter)
Radius of anemometer (from cup centre to centre of axis) = 0.155m
Diameter = 0.31m
- Speed = distance / time Speed * time = distance
- Distance per minute = revolutions per minute * circumference
- Distance per second = revolutions per minute * circumference
60 (seconds in one minute)
Readings 1-5 – Results for calculated wind speed
The anemometer will be able to measure to a degree of accuracy of two decimal places, giving it a resolution of 0.01ms-1. This is because the least certain measurement of distance, the diameter of the anemometer was measured to two decimal places. This therefore restricts the accuracy of the anemometer to two decimal places when calculating wind speed. A calibration curve for potential difference against speed is now possible:
When taking my results, I may find that I encounter wind speeds of over 2.51ms-1. I predict that as the potential difference in mV increases, so will the rpm in a directly proportional manner, above 160 rpm. 80mv means an rpm of around 155 and a wind speed of 2.51ms-1. The value for the rpm is the calculated mean from the 5 results taken. I can assume that for the amount of voltage involved in this practical, the relationship between rpm and speed will remain linear. Therefore, I am going to extrapolate the calibration curve, to give me values for wind speed that relate to higher potential differences being obtained. There will be a point where the relationship between the p.d and rpm will become non-linear. But I do not think that with the wind speeds I will encounter, this limit will be reached.
An example of the kinds of rpm involved when the relationship between voltage and rpm may not be linear is when a similar type of motor is used in a motorised milk whisk, or model train. Both these motors are powered by 3v dc. This is a potential difference 75 times greater than produced by the motor turning at an average of 73 rpm. A wind speed 75 times greater than the wind speed for an average of 73 rpm is 89ms-1; this is around 200mph. This is very unrealistic, and the anemometer would not survive in such conditions. Therefore, I estimate that I will only encounter wind speeds at most, up to 18ms-1 (40mph). Up to this point I would suggest that the relationship between the potential difference and rpm would remain linear enabling me to calculate wind speed for potential differences of over 80mV.
Results
The table above shows the results obtained measuring the wind speed. I encountered several problems during the taking of the results. To obtain different potential differences, I needed to obtain different wind speeds. To do this I went to the sea front, and found different points along the sea front where the wind was blowing at different speeds. It was very difficult to accurately obtain a potential difference exactly equal to those recorded for long enough to read off the value for the datum at the same time. The wind source during the calibration was constant, and therefore fluctuations in the potential difference were minimised. As wind is not at all constant the task of recording the results was made harder.
To record the results necessary to determine the accuracy of the anemometer, I found different wind speeds along the sea front and as accurately as possible gained a potential difference as close to those stated in the table. At the same time I measured the wind speed that created this potential difference using the datum.
The results are not consistent with the findings from my calibration. Therefore I am able to analyse these results and the factors which contributed to the inaccuracy of the anemometer that I devised.
As I predicted I encountered wind speeds fast enough to produce a potential difference greater than 80 mV. Using calculated data, I would have predicted that the corresponding wind speed was 2.84ms-1, using values from Appendix 1. In fact using the datum I recorded a wind speed of 8.20ms-1. The graph shows that the expected values for wind speed, produced from the calibration are all inaccurate. This obviously relies upon the assumption that the datum is accurate, which I can do. The results show that as the potential difference created by my anemometer increases, the inaccuracy increases. Both sets of results are linear showing that the calibration was successful to a certain degree, but now using this data I can analyse and explain why the results I achieved were inaccurate. Although they are inaccurate, they are not anomalous as they follow the predicted pattern.
Analysis
There were several factors during the calibration that could have led to these inaccurate results. Firstly, when the anemometer was placed in a very narrow wind stream, the returning cups did not have to pass through the same wind which was turning the cups. This means there would have been less air resistance during the calibration than when recording results. Contextually, this means that a potential difference of 50 mV created during the taking of results may correspond to a potential difference of 60 or 70 mV created during the calibration. This agrees with the results recorded, although the degree of inaccuracy is greater than this suggested difference. In actual fact using values from Appendix 1, the potential difference required to produce a wind speed equivalent to that recorded for 50 mV using the datum (4.72 ms-1) is between 140 and 150mV.
Another factor to suggest that the absence of air resistance during the calibration had an effect can also be seen on the graph. The inaccuracy of the results increases as the potential difference increases. This can also be explained using the same idea. During the calibration, a potential difference of 50 mV was calculated to correspond to a wind speed of 1.51 ms-1; this means the returning cups were not encountering the resistance created by this corresponding wind speed. This can be compared to a potential difference of 80 mV where a corresponding value of 2.51 ms-1 was calculated. The returning cups would therefore not be encountering a greater air resistance (because of the faster wind speed), increasing the inaccuracy of the results.
The main factor that contributed towards the inaccuracy of the results, which encompasses the previous explanation is the inefficiency of the anemometer. As the wind’s energy was transferred from linear kinetic energy, to rotary kinetic energy (through the centre axle) into electrical energy (through the motor utilising the generator effect). At each stage energy will be lost, therefore making the values obtained for each rpm inaccurate.
The inefficient transfer of energy was caused by several factors; these were friction in the motor, sound created by the turning of the cups, and mainly air resistance.
The error seen in the results is a systematic error as all the results were similarly inaccurate; the speed of the cups that was calculated during calibration was slower than the actual wind speed. Obviously the size of error is greater than desirable when making a sensor, but now with this knowledge the results from the calibration could be altered to encompass this systematic error.
Another factor that affected the accuracy of the results from the calibration was the inclusion of user error when measuring the time period for x number of flashes. For example, if I measured 40 flashes in 5.50 seconds, but due to slow reactions or an error the time it took for 40 flashes to be produced was only 5.30 seconds the difference in the calculated speeds of the cups would be 0.11 ms-1. Although this is not a lot, it would still have been influential.
The success of the anemometer can be measured by analysing its effectivness in meeting the qualities desirable for a good sensor. These include a good resolution, fast response time, low systematic drift or error, appropriate sensitivity and low random variation.
The sensor had a relatively fast response time; the potential difference displayed on the multimeter was updated very quickly when a change of wind speed occurred. This made the data collection very different as the value kept fluctuating, although it showed a fast response time in respect of changing wind speed.
I was able to reduce the effect that unsystematic random error had upon the results from the calibration by taking the average of the five sets of results. Small unsystematic variations were present in all the readings that I took, but as what I was measuring (thousands of a volt) is a small quantity these small variations had a relatively large affect (this being the sensor’s sensitivity). The sensitivity of a measuring system is the ratio of change of output to change of input; this is where the sensor became inaccurate as the multimeter did not have a sufficient resolution to create a sensitive enough sensor. The sensitivity was limited, as a very small input was inaccurately converted into a large output. This is why the results were so inaccurate, as the calculations converted very small differences of potential difference containing error (systematic and random) into relatively large values for wind speed therefore amplifying any error that was present in the results.
For this same reason, the sensor’s resolution was limited. In conjunction with a high level of random variation caused by the multimeter’s insensitivity in measuring such a small amount of potential difference, the smallest degree of potential difference that I could accurately measure was 10 mV; this is ten, one thousandths of one volt. Therefore the resolution of the sensor is around 0.15 ms-1; this is roughly the wind speed calculated from the calibration results for 10mV. This is irrelevant because of the fact that results can be calculated to 2 decimal places, as I can only be sure of results to the nearest 0.15 ms-1 due to the sensor’s relatively large resolution. In comparison, the datum can accurately measure to 2 decimal places, e.g. 2.42 ms-1.
I was able to detect and explain the systematic error due to the fact that my sensor was relatively inaccurate and I had access to a much more accurate sensor designed to measure the same thing. The use of a datum enabled me to effectively analyse my results. Overall to create a more successful sensor, I would need to review the complexity of this sensor and devise a method that reduces the margin for error as the current design encompasses too many opportunities for the results to be affected.
Bibliography
– Picture of Reed switch
Advancing Physics AS – Institute of Physics
Lonsdale Science Revision guide – The essentials of OCR science double award.