Testing Gripper
This topic is for ONLY for GripAlyzeR.
Bimanual force coordination (GripAlyzeR & Gripper) - Hardware/Software Testing Protocol
This section provides procedures to test the gripper device using GripAlyzeR. The results obtained when performed at NeuroScript are also provided for comparison.
Introduction
The tests listed below need to be conducted by users:
- To familiarize and characterize how the GripAlyzeR measures
- To characterize the sensor performances
- To verify that the Gripper functions within the specifications and with accuracy
These tests also have to be conducted:
- When receiving a new Gripper device and updating the device settings
- Before starting a new study
- After completing part of a study to verify the Gripper was stable
- After an extraordinary event that may have compromised the Gripper
Contents
I. Determine Baseline
II. Test Force Sensors
1. Static Measurements
1.1. Calibration
1.1.1. Bottom grip-force sensor
1.1.2. Top grip-force sensor
1.1.3. Load sensor1.2. Random noise
1.3. Frequency spectrum
2. Dynamic Measurements
2.1. Relative linearity
2.2. Calibration of sampling rate and time
2.3. Sudden force changes
2.4. Inertial forces
2.4.1. Internal sensor inertia
2.4.2. Inertial forces due to the magnet2.5. Timing of the peak of the load force relative to the magnetic polarity change
III. Test Magnet properties
4. Friction
I. Determine Baseline
The baseline or reference level per session is the signal level determined during the first n baseline samples (set as 10 samples). Ideally, this value should be kept as close to 0 level as possible. The following procedure ensures this:
~ Place the top and bottom units horizontal on the table at rest.
~ Start GripAlyzeR > Run one trial for the grip-force experiment. If required, stop after first trial.
~ Trial > Chart processed data, the reference/starting level should be close to zero or zero. If not, restart the application and repeat the test.
Perform the test each the application is started.
2. Effect of magnet weight on the load sensor baseline
~ Place the top and bottom units horizontal on the table at rest.
~ Start GripAlyzeR > Run one trial for the grip-force experiment.
~ For the second trial, place the lower unit vertically on the table with the magnet facing up first and run the trial with units at rest.
~ Second Trial > Chart processed data > Z value vs. time, shows the difference in the baseline value after having included the effect of the magnet (since the second trial is with the magnet facing upwards)
~ The effect of the weight of the magnet on the magnet sensors baseline has shown to be about 0.5 N during our tests. Hence it has to be accounted during other tests involving the pulling force on the magnet. ie., there is already a pushing force of about 0.5 N on the load sensor, when placed vertically.Eg.,
We can notice from the above chart, there is a difference of about 0.5 N in the baseline values.
II. Force Sensors
1. Static Measurements
Equipment: The gripper consists of 3 force sensors (Entran's ELPM-T2E-03-25L, S/N 02102H31-B02, B03 and B04). The gripper is connected to the PC through the National Instruments DAQpad 6020E (12 bits).
Testing Environment/Settings in GripAlyzeR:
The tests were performed in GripAlyzeR 2.8/2.85 and will also apply to future versions of GripAlyzeR.
Input device settings (GripAlyzeR menu > Settings > DAQPad settings):
~ Set Scan rate per channel = 100 Hz (3 channels were sampled each time at a rate of 10000 Hz).
~ Set Number of samples = 1000 (10 s) next to the scan rate. (Recording time (s) = number of samples/scan rate). Baseline samples = 10. (After the application is started, the input value during the first 10 samples is set as the zero level for the entire session)
~ Choose the options 'Report in Newtons' and 'Double buffer mode' (20 buffer samples)
~ For all 3 channels, Gain = 100 mV and excitation voltage = 5 V (Specify the values as designed in the interface circuit)
~ Specify the full-scale load (25 lbs) and the calibration constant values as derived from the specification sheet of the corresponding sensors. Refer to Selecting Gripper Device topic on the help for detailed information on these terms.
Data collection/Experiment settings
Grip-force experiments, groups and subjects are created in GripAlyzeR. To collect trials for the different tests, add conditions/subjects to the experiment. Alternatively, you can split the data collection across two or more experiments.
~ Experiment > Experiment Settings > Recording > Choose 'Delay trial until Enter key pressed' and 'Accept/Redo trial', as required. Please refer help on Experiment Settings for the implications.
~ Settings > DAQPad settings > scan rate. Set the scan rate between 100 - 200 Hz. Also in Experiment > Experiment settings > input device settings > sampling rate, set the value equal to the scan rate.
Note: An experiment can be run per subject and not condition. If you want to redo only a particular condition, you need to perform trial > redo trial, per trial. The exact organization of the experiment(s) is defined by you.
Note: More tests/Enhancments to the existing tests might be added in due course. Check regularly for updates.
1.1. Calibration
Note: For the following calibration tests, the sensor units need to be placed horizontally at +/-6 degrees, so that the error is less than 0.1%, which is much smaller compared to the sensor accuracy of 1%
1.1.1. Bottom Grip-force Sensor
Purpose: To verify the GripAlyzeR system is measuring the correct force as applied to the lower grip sensor.
Procedure:
Run two trials for each of the 3 different weights as follows:
~ Before starting the recording, place the weight on the force sensor plate of the lower Gripper unit resting on a horizontal surface.
~ After the trial starts, wait for 2-3 s and then remove the weight.
~ After every trial recording ends, put the weight back on the sensor plate for the next trial during the trial-trial pause. (before pressing enter key to start the consecutive trial) and perform three trials.
The weights used in our tests were 50 g, 100 g, and 200 g. The equivalent force level is F = g * m, where g = 9.81 m/s2 or N/kg. Thus the equivalent forces were 0.4903, 0.9807, or 1.9614 N, respectively.
Analysis:
View raw data: View the raw data for a trial (Trial > View raw data) from the GripAlyzer consists of X, Y and Z values. The bottom sensor output is the X-axis data.
Derive & transform relevant data:
~ For each trial, input the numerical values of the raw (Unfiltered) data file in to MS Excel sheet.
~ Remove the data containing the transition corresponding to the weight removal, indicated by a change in the X-values. The stable data with weight (level1) and without the weight (level2) will be retained.
~ For example, in the below chart, the moment that the weight is being removed is visible by a sudden change (transition). The transition period displayed here is very small. Hence all the data before this transition starts is one level and after the transition ends is a second level.
~Calculate the average of the high and low levels (or vice versa) of X as AvgX1 and AvgX2. Also, calculate the standard deviation, SDX1 and SDX2. The difference in the average levels between before and after removing the weight |AvgX1-AvgX2| is an estimate of the measured force.
~In the chart above, the trial starts at 0 N level with the weight, and after the removal of the weight goes to -1.86 N level. Hence the force change is 1.86 N.
Results from our tests:
The sensitivity of the sensors from the manufacturer is approx. 2.5 mV/V at a full load of 25 N. And the amplification factor from circuitry is 504. Hence in the experiment settings, we use a calibration constant of 2.5*504 = 1260.
For two trials performed for each of the weights applied on lower grip sensor, the follows values were derived from analysis:
Table 1. Lower grip force measurements
| Weight applied (g) | Actual Force(N ) = weight applied in N | Trial 1 | Trial 2 | ||
| Measured Force = |AvgX1 - AvgX2| | % Error = (Actual - measured)/ actual x 100% |
Measured Force = |AvgX1 - AvgX2| |
% Error = Actual - measured/ actual x 100% | ||
| 50 | 0.4903 | 0.4707 | -4.0 | 0.4674 | -4.9 |
| 100 | 0.9807 | 0.9462 | -3.5 | 0.9424 | -4.0 |
| 200 | 1.9614 | 1.9030 | -2.9 | 1.8624 | -5.0 |
Conclusions:
The measured force levels are up to 5% lower than the actual force levels. That the relative departures are fairly constant implies that the sensors are fairly linear.
To obtain accurate results, the lower grip (x) calibration constant can be fine tuned by increasing it by 5%, ie, setting it to 1266.3. After changing the settings, an accuracy better than 0.5% is observed for the above measurements.
Purpose: To test whether the GripAlyzer system is measuring the correct force as applied on the upper grip sensor.
Procedure: The procedure used is the same as that used to test the bottom sensor
Analysis: Same as lower grip sensor, except that the Y-axis data (upper grip) are taken in to consideration, instead of x-axis data, to obtain, AvgY1, AvgY2, SDY1 and SDY2.
Results from our tests:
For two trials performed for each of the weights applied on upper grip sensor, the follows values were derived from analysis.
Table 2. Upper grip force measurements.| Weight applied (g) | Actual Force(N) = weight applied in N | Trial 1 | Trial 2 | ||
| Measured Force = AvgY1 - AvgY2 | % Error = (Actual - measured)/ actual x 100% | Measured Force = AvgY1 - AvgY2 | % Error = Actual - measured/ actual x 100% | ||
| 50 | 0.4903 | 0.4875 | -0.5% | 0.4899 | -0.1% |
| 100 | 0.9807 | 0.9865 | +0.6 % | 0.9680 | -1.3 % |
| 200 | 1.9614 | 1.9383 | -1.2 % | 1.9362 | -1.3% |
Conclusions:
The measured force levels are approximately 1% lower than the actual force levels. The upper grip (y) calibration constant could be left as the same 2.5, since is the error is negligible.
Purpose: To test whether the GripAlyzer system is measuring the correct force as applied on the load sensor.
Procedure: A similar procedure was used as for the lower grip sensor
Analysis: Same as lower grip sensor, except that the Z-axis data are taken in to consideration, instead of x-axis, to obtain, AvgZ1, AvgZ2, SDZ1 and SDZ2.
Results from our tests:
For two trials performed for each of the weights applied on load sensor, the follows values were derived from analysis.
Table 3. Load force measurements
| Weight applied (g) | Actual Force(N) = weight applied in N | Trial 1 | Trial 2 | ||
| Measured Force = AvgZ1 - AvgZ2 | % Error = (Actual - measured)/ actual x 100% | Measured Force = AvgZ1 - AvgZ2 | % Error = Actual - measured/ actual x 100% | ||
| 50 | 0.4903 | 0.4367 | -10.8 % | 0.4386 | -10.4 % |
| 100 | 0.9807 | 0.8578 | -12.4 % | 0.8711 | -11.1 % |
| 200 | 1.9614 | 1.7367 | -11.4 % | 1.7387 | -11.3 % |
Conclusions:
The measured force is approximately 11% lower than the actual force. To obtain more accurate results, the load sensor (z) calibration constant should be increased by 10% or so for the load sensor. ie., from 1260 to 1368.
Purpose: Estimating static equipment noise in comparison to the quantization noise.
Procedure:
Quantization Test:
Place the upper and lower units flat on the table, with the upper and lower grip force sensor plates facing upwards.
Measure 3 trials with the units at rest.
Analysis:
1. From quantization test in the section:
Right click trial > chart raw data > X and Y vs. sample number
Right click on the chart > plotting method
2. From tests in previous sections:
~ For each of the trials in lower grip (1.1.1), upper grip (1.1.2) and load sensor (1.1.3), the SD values, SDX1, SDX2, SDY1, SDY2, SDZ1, SDZ2, are considered.
~The minimum of the two trials per channel per trial reading is calculated as SDX = min (SDX1, SDX2), SDY = min(SDY1, SDY2) and SDZ = min (SDZ1, SDZ2), displayed in table 4. The minimum value is used as the best estimate, because the measured device cannot be underestimated but easily overestimated.
Results from our tests:
Table 4. Averages and static noise levels
| Lower Grip | Weights placed | Force corresponding to weights placed (N) | SDX1 (trial1) | SDX2 (trial2) | Noise in lower grip channel per weight = SDX | Noise (N) /channel = Avg of noise for different weights |
| 50 g | 0.4903 | 0.0256 | 0.0255 | 0.0255 | 0.02453 | |
| 100 g | 0.9807 | 0.0259 | 0.0237 | 0.0237 | ||
| 200 g | 1.9614 | 0.0339 | 0.0244 | 0.0244 | ||
| Upper Grip | Weights placed | Force in N for weights placed | SDY1 (trial1) | SDY1 (trial2) | Noise in lower grip channel per weight = SDY | Noise (N) /channel = Avg of noise for different weights |
| 50 g | 0.4903 | 0.0229 | 0.0254 | 0.0229 | 0.0227 | |
| 100 g | 0.9807 | 0.0389 | 0.0220 | 0.0220 | ||
| 200 g | 1.9614 | 0.0245 | 0.0232 | 0.0232 | ||
| Load | Weights placed | Force in N for weights placed | SDZ 1 (trial1) | SDZ 1 (trial2) | Noise in lower grip channel per weight = SDZ | Noise (N) /channel = Avg of noise for different weights |
| 50 g | 0.4903 | 0.0180 | 0.0217 | 0.0180 | 0.0188 | |
| 100 g | 0.9807 | 0.0193 | 0.0199 | 0.0193 | ||
| 200 g | 1.9614 | 0.0192 | 0.0241 | 0.0192 |
Conclusions:
Quantization Noise: The DAC unit in NIDAQPad converts the analog signal into a 12 bit digital value (2^12 = 4096 values). The step size for the signal will be full scale voltage/(4096 - 1) Volts. For example, if the full scale voltage is 0-5 V, then the step size (1 LSB) is (5/4095) = 1.22 mV. Hence, for each 1.22 mV change in the signal, a bit difference in the digital value is encountered. This implies that ANY signal that has an amplitude less than 1 LSB may not be detected. The corresponding Newton value for this noise is 0.0126N with 25N full scale and 5 V excitation.
Quantization noise (RMS) = 0.0126/sqrt(12) = 0.0036 N
Note from table 4.,Noise/Channel calculated from the data, is much higher than the quantization noise. Therefore, the quantization noise has negligible effect on the noise estimate.
Interestingly, the static noise level does not depend on the load force. The static noise consists of random noise plus a quantization noise. (SD^2 Noise = SD^2 Random Noise + SD^2 Quantization noise).
Purpose: To detect the presence of specific interference frequencies.
Procedure:
For each sensor (lower unit, upper unit and load), collect two trials as follows:
~ Place the units on the table horizontally (for upper and lower unit trials) and vertically (for load sensor trials).
~ In the first trial, the data is collected with the sensor unit at the rest, without any weight placed.
~ In the second trial, place a known weight after 2-3 s from start of recording.
Analysis:
~ View the trial > view processed data > Velocity Vs Frequency Spectrum plot.
Conclusions:
In the first trial (at rest) for each sensor, as in the chart below, the frequency spectrum ranges from 0 Hz to half of sampling frequency (256 Hz), ie, 128 Hz. The unfiltered frequency shows frequency harmonics at around 50 Hz, 100 Hz ...etc. with the amplitude increasing with every 50 Hz frequency. The filtered plot displays the data filtered with a filter frequency of 12 Hz.
In the second trial, there was no weight initially and then a weight of 50 g was placed gradually. Here also, the unfiltered frequency shows frequency harmonics at around 50 Hz, 100 Hz ...etc. with the amplitude increasing with every 50 Hz frequency.
Addendum:
When the trials where the units are at rest, are processed to obtain the raw data spectrum instead, the spectrum shows a triangular increase of noise with frequency suggesting the presence of white noise.
1.4 Force sensor plates: Central and off-central pressure
Purpose: To verify if there is a difference in forces measured by upper and lower grip sensors, depending on where the pressure is applied on the force-sensor plate (3.8 cm dia.).
Procedure:
Verify that the system is measuring with reference to a basline at or close to 0. (Refer section I.)
For each of upper and lower grip units,
Collect two trials as follows: place a known weight on the center of the sensor plate.
Collect two trials as follows: place a known weight on the edge of the sensor plate.
Analysis:
~ For each trial > chart processed data > X Vs. time plot for lower sensor and Y vs. time for the upper sensor.
~Note the force value measured in the above charts for each trial.
Results:
Position of the weight (100g = 0.9807 N) Average Force measured (N) Lower Sensor Upper Sensor Center 0.899 0.94 Edge 0.897 0.933
1.5 Temperature Effects on the sensor senstivity
Refer http://entran.com/elpm.htm for the sensor specifications
The manufacturer specified "Thermal Sensitivity Shift" is 0.01 FS/C, where FS is the full scale voltage.
The effects of temperature on the measurements can be roughly estimated by using the above value.
Example: If the sensors are normally placed in an temperature controlled room at 18C (65 F), and when the sensor comes in contact with hands for sometime at 37C (98.4 F), the increase in temperature is 19C.
Change in FS = 0.01 * 19 = 0.19
For a FS of 25 N, Change in N = 0.19 *25 = 0.475 N.
II. Dynamic Measurements
Purpose: To verify the sensor data vary proportionally with a variable force applied. The goal is to apply equal forces on all of the sensors and show that the values vary proportionally (as sensitivity is approximately the same for all the sensors).
Procedure:
1. Relative linearity between lower and upper grip unit sensors
Collect two trials as follows:
~Place both the gripper units on a table, facing each other.
~Move the units, such that the sensor plates of the two units touch each other.
~ Press the units together with one hand repeatedly 10 times. That is, apply equal forces on the bottom and top sensors.
2. Relative linearity between upper grip unit and load sensors
Collect two trials as follows:
~Place the bottom unit upright on the table and place the upper unit on the lower one, such that the sensor plate and the magnet surface touch each other.
~Press the upper unit over the lower unit downwards with one hand repeatedly 10 times. That is, apply equal and opposite forces on the upper grip sensor and the magnet.
3. Relative linearity between lower grip unit and load sensors
Derive from 1. and 2.
Analysis:
For each of the two combination of sensors, observe the processed data chart for each trial.
Conclusions:
1. For Relative linearity between lower and upper grip unit sensors,
~The X vs Y chart of the trial data, implies a proportional change for the force applied between the lower and upper grip sensors.
2. For Relative linearity between upper and load grip unit sensors,
~The Y vs Z chart of the trial data implies a proportional change for the force applied. The curve also shows hysteresis loops, which indicate that there is a small lag between the two forces. The width of the hysteris loop can be obtained difference in y coordinates from the chart by considering the center of the outermost elliptical loop. The average peak width of the hysteresis loop is around 1 N.
3. For Relative linearity between lower and load grip unit sensors,
~ From 1. and 2., we can see that there is no lag between X and Y and they are proportional, but there is some lag between Y and Z. Hence, we can say that there will a lag between X and Z also.
This also shows that the forces from the magnetic load sensor, when compressed may be obstructed by the cables between the sensor. In general, this effect may be as much as +/- 2 N, depending on increasing and decreasing tension.
2.2. Calibration of sampling rate and time
Purpose: To evaluate any differences in time between the instant at which change is enforced on any of the sensors and the corresponding change in the measured data. This could imply the sampling rate time and sensor response time.
Procedure:
For each sensor (lower unit, upper unit and load), collect two trials as follows:
~Place a known weight after 5 sec accurately from start of recording. The 5 s was measured using a stop watch.
Analysis:
~ View processed data: For each trial, observe, X (Lower grip) Vs. time, Y (upper grip) Vs. time and Z (Load) Vs. Time.
~ The transition time is very small and mostly ends with a small peak. To maintain consistency, the time instant corresponding to the peak is considered as the point of transition for each trial.
Results from our tests:
Table 8.
| Trial 1 | Trial 2 | |||
| Time | Error % (Deviation from 5 s) | Time | Error % (Deviation from 5 s) | |
| Lower Grip | 4.74 | (5-4.63)/5 * 100 % = 5.2 % | 4.869 | 2.6 % |
| Upper Grip | 4.5 | 10% | 4.468 | 10.64% |
| Load | 4.67 | 6.6% | 4.68 | 6.6% |
Conclusions:
It can be observed that there may be up to 10% deviation in the sensor data change time from the actual instant at which the change was forced. This could be due to the sensor response time, interface units delay, and/or sampling time.
It can be observed in the chart above that the change from high to low of the plot occurred very close to the 5 s, (up to 10% error) from the actual time.
Purpose: To verify the sensor's ability to detect sudden force changes.
Procedure:
For each sensor (lower unit, upper unit and load), collect two trials as follows:
~ A trial starts with the system at rest.
~ Take two known weights. In our tests, weights of 100 g and 200g were used.
~ After a trial has started, within a few seconds, drop the first weight on the horizontal surface of the sensor plate from a distance of about 0.5 -1 cm.
~ During the trial-trial pause (before pressing enter to start next trial), remove the weight from the sensor plate.
~ Perform the second trial with the other weight.
Analysis: For each trial,
~Open raw data file, by trial > view raw data
~ Determine the sample number (row number in raw data file) at which the transition from low to high or high to low starts.
~Determine the sample number at which the transition ends.
~The response time is the difference between these sample numbers.
Results from our tests:
Table 9.
| Sensor | Trial | Sample # at the end low/high values | Sample # at the start of high/low values | # Samples for transition = response time | Bandwidth (Hz) = Sampling rate/2*(#samples) |
| Lower | #1 (200g weight) | 163 | 169 | 6 | 21.33 |
| #2 (100g) | 168 | 177 | 9 | 14.22 | |
| Upper | #1 (200g) | 130 | 137 | 7 | 18.28 |
| #2 (100g) | 263 | 270 | 7 | 18.28 | |
| Load | #1 (200g) | 273 | 278 | 5 | 25.6 |
| #2 (100g) | 293 | 299 | 6 | 21.33 |
Conclusions:
The sensor force should ideally show a step function. The resulting curve shows an oscillation during the impact and then comes to the rest value, as show below.
The time at which the sensor was at rest to the time it takes to switch to a higher value is measured as the response time of the sensor.
In average, the response time including the oscillations was 5-10 samples. This can be visually assessed by viewing the raw data chart for the corresponding sensor after zooming in on the transition area.
2.4. Inertial forces
2.4.1. Inertial forces due to the sensor
Purpose: Estimate what load the sensor (lower grip & upper grip sensor) measures due to the sensor's mass or cover.
Procedure:
For each of the following conditions, collect trials as: Place the lower grip unit on the table with the load sensor facing upwards. Lift the lower unit before a trial starts (in the trial-trial interval or press enter key delay). After trial starts, shake it with high acceleration and frequency.
Collect 3 trials shaking in horizontal plane with the transducer cover facing towards the side. Donot touch the sensor plate with fingers Collect 3 trials shaking in horizontal direction with the transducer cover facing towards you. Collect 3 trials shaking in horizontal plane without the transducer cover (1 N magnet force) Collect 3 trials shaking in horizontal direction without the transducer cover (1 N magnet force)
Repeat the procedure using the upper grip unit.
Do not touch the grip-force sensor.
Analysis: Per condition, measure the lower grip-force sensor oscillation amplitude (peak to peak), from processed data chart, X-time graph (average of the 3 trials). Similarly, observe Y-time graph for upper grip sensor.
Results:
Old force plates (weight not available, the steps 2. and 4. done in vertical direction instead of sidewards)
| Task | Amplitude of the largest variation (N) | |
| Lower Grip sensor | Upper Grip sensor | |
| 1 (shaking in horizontal plane with the transducer cover) | 0.23 | 0.12 |
| 2 (shaking in vertical direction with the transducer cover) | 0.28 | 0.36 |
| 3 (shaking in horizontal plane without the transducer cover) | 0.13 | 0.14 |
| 4 (shaking in vertical direction without the transducer cover) | 0.14 | 0.15 |
Old force plates (weight = 10 g)
| Task | Amplitude of the largest variation (N) | |
| Lower Grip sensor | Upper Grip sensor | |
| 1 | 0.7 | 0.57 |
| 2 |
0.63 |
0.55 |
| 3 | 0.68 | 0.47 |
| 4 |
0.6 |
0.46 |
Conclusions:
For the lower grip sensor, it can be observed that having the transducer cover has a small effect on the sensor readings. Also, shaking in different directions the horizontal plane in different directions have similar effects. However, there is a slight difference in the readings between upper and lower sensors, which could be due to additional cables for the lower unit.
2.4.2. Inertial forces due to the magnet
Purpose: Estimate what load the sensor measures due to the mass of the magnet.
Test (A): Procedure
For each of the following conditions, collect trials as: Place the lower grip unit on the table with the magnet facing upwards. Before the trial starts, lift the lower unit. After the trial starts, shake it with high acceleration and frequency.
1. Collect 3 trials shaking in horizontal plane (1 N magnet force)
2. Collect 3 trials shaking in vertical direction (1 N magnet force)
3. Collect 3 trials with pulling the BOTTOM unit RAPIDLY and suddenly with the two hands holding each of upper and lower units, laying the two units put together flat on the table (1 N magnet force)
Note: Do not attempt to remove the magnet. Do not touch the load-force sensor.
Test (B): Procedure
Repeat 1. and 2., but while shaking, the thumb is placed on the lower grip sensor. This is done in addition to test (A) as when a subject is performing a pulling task, the thumb would be pressing the lower sensor and hence, is closer to the experimental scenario.
Analysis: For each trial, measure the load-force sensor oscillation amplitude (peak to peak), from processed data chart, Z-time graph. In the last 3 trials, the force includes: Residual magnetic force + suction force + inertial force.
Results:
| Condition | Test (A) | Test(B) | |
| Z amplitude (load sensor) N | Z amplitude (load sensor) N | X amplitude (Lower Grip sensor) N | |
| 1 (shaking in horizontal plane) | 0.19 | 0.2 | uneven variations |
| 2 (shaking in vertical direction) | 1.24 | 1.1 | 1.51 |
| 3 (pulling the BOTTOM unit RAPIDLY) | 0.95 | This is not done for task 3. | |
Conclusions:
In both tests (A) and (B), shaking in the vertical direction has more of an effect on the load sensor readings. This is due to the weight of the magnet mounted on the load sensor. Also, in test (B), placing the thumb on the sensor while shaking the unit in vertical direction produces an oscillation in the lower grip data (X amplitude), although pressure is not applied intentionally.
2.5.Timing of the peak of the load force relative to the magnetic polarity change
This test is for your review only, as this was not conducted using GripAlyzer, but an internal program, GripView.
Purpose: To show that the magnetic polarity change (Channel 4) during the disengagement of the upper and lower grip occurs almost simultaneously to the load force reaching a peak (Channel 3) and hence the measurement on Channel 4 can be bypassed.
Procedure:
Grip force data with time, X, Y, Z and magnetic polarity were obtained in 3 conditions for 3 trials per condition.
~ In Condition 1, trials were performed by pulling apart the upper and lower units in the sideward direction, with a fixed magnetic strength (e.g., 8 N).
~In Condition 2, the pull was on the forward direction.
~In condition 3 the pull was in the vertical direction.
Analysis: For each trial,
~ Raw data values were taken into consideration.
~ From magnetic polarity values, the point of transition of the polarity from high to low or from low to high was determined.
~ From the Z values, the point of transition to the peak value of the load was noted. It was found that the two transitions were one or two samples apart. (table 7)
Results from our tests:
Table 6. Data to indicate the transition of peak z-value and repolarization of the magnet.
| Sample # | Upper Grip- Y value | Load - Z value | Lower Grip- X Value | Magnet Polarity |
(+/-) # Samples difference |
| 203 | 41.6637 | 9.0071 | 42.4489 | 5.0635 | 2 |
| 204 | 41.5635 | 9.3417 | 42.6409 | 0.0977 (repolarization) | |
| 205 | 41.7138 | 9.5329 (max) | 42.8329 | 0.0977 | |
| 206 | 42.2651 | 7.8121 | 42.7369 | 0.0977 |
We found no difference between between the conditions. See Table 7.
Table 7. Timing of the repolarization of the magnet.
| Condition |
# samples difference between the max load force sample and the last sample nr before magnet repolarizes (For trials 1-12) |
Average difference | |||||||||||
| Sideward | 2 | 2 | 2 | 2 | 2 | 1 | 1 | 1 | 2 | 1 | 1 | 1 | 1.5 |
| Forward |
2 |
1 | 1 | 1 | 1 | 2 | 2 | 2 | 2 | 2 | 1 | 2 | 1.58 |
| Upward | 1 | 1 | 2 | 2 | 2 | 1 | 1 | 1 | 2 | 2 | 1 | 1 | 1.416 |
Conclusions:
The time corresponding to the magnetic polarity change (Point of disengagement) can be found from the peak Z value taken from the Z-value Vs time plot, while tolerating an average of 1.5 samples.
In the trial below, the magnetic polarity switches AFTER sample 203, while the load value reaches its peak AROUND sample 205, although it seems unlikely that at sample 205, the load force is on its way to decrease. It seems safer to say that the load force will have its sudden decrease AFTER its peak. Therefore, there seems to be a delay of 2 samples between the moment of repolarization of the magnet (to facilitate disengaging) and the maximum grip force.
Addendum:
When the above test was repeated for 16 N force value, the lag of the Z value load force measurement with respect to repolarization was much larger (as many as 14 samples)
II. Magnet
Purpose: To verify magnet force as specified in the GripAlyzeR condition properties. That is, set a specific magnet force and verify that the magnet disengages at the critical force.
Procedure:
~ Measure the weight of the lower grip unit on a scale (300 g ~3 N).
~ Attach a 200g weight (2 N extra) to the lower grip unit by using a scotch tape or band.
~ In the grip-force experiment in GripAlyzeR, Condition that will be used to collect trials > Condition properties > gripper > set the force to approximately equal to the force required to lift the lower gripper plus the added weight (5 N).
Collect the trial: Method I
~ Place the bottom unit on the table with the magnet face upwards with the top unit placed on it.
~ After a trial starts, lift the top unit in to the air carefully against the force. (keep the lower unit very close the surface of the table)
~If the applied force (5 N) is accurate, it should be able to hold the bottom unit (with effective weight of 5 N).
~ Reduce the weight attached by 50 g (now attached weight is 150 g ~ 4.5 N) and repeat the trial. As you go further in to trials reduce the weight by smaller units (see results table 1)
~Keep reducing the weight and repeating the trial until the lower unit is held by the upper unit.
~Note down the extra weight attached each time and tabulate as in table 1.
~Repeat the above steps for 1 more repetition.
The above test can be repeated by going from lower to higher units of weight (see results table 2).
Results:
Weight of the lower unit = 300 g ~ 3 N
Weight of the upper unit = 200 g ~ 2 N
Table 1:
Force applied by magnet (condition setting) Weight Attached (in g and N) Effective weight of the lower unit = 3 N + attached weight in N Effective pull force on the magnet = effective weight of lower unit - weight of the upper unit (push force) Lower unit attached or detached? Repetition 1 Repetition 2 5 N 200g ~ 2 N 5 N 7 N Detached Detached 5 N 150g ~ 1.5 N 4.5 N 6.5 N Detached Detached 5 N 100g ~ 1 N 4 N 6 N Detached Detached 5 N 50g ~ 0.5 N 3.5 N 5.5 N Detached Detached 5 N 40g ~ 0.4 N 3.4 N 5.4 N Detached Detached 5 N 30g ~ 0.3 N 3.3 N 5.3 N Attached Attached 5 N 20g ~ 0.2 N 3.2 N 5.2 N Attached Attached Table 2:
Force applied by magnet (condition setting) Weight Attached (in g and N) Effective weight of the lower unit = 3 N + attached weight in N Effective pull force on the magnet = effective weight of lower unit - weight of the upper unit (push force) Lower unit attached or detached? Repetition 1 Repetition 2 5 N 20g ~ 0.2 N 3.2 N 5.2 N Attached Attached 5 N 30g ~ 0.5 N 3.3 N 5.3 N Attached Attached 5 N 40g ~ 0.4 N 3.4 N 5.4 N Attached Attached 5 N 50g ~ 0.5 N 3.5 N 5.5 N Attached Attached 5 N 100g ~ 1 N 4 N 6 N Detached Detached 5 N 150g ~ 1.5 N 4.5 N 6.5 N Detached Detached 5 N 200g ~ 2 N 5 N 7 N Detached Detached
Analysis and Conclusions:
From table 1, it can be noticed that units remain detached at a pull force of 5.4 N on the magnet, and attached at 5.3 N. If the magnetic force was exact, this value should be very close to the applied force of 5 N. Hence we can see an error of 0.3 to 0.4 N.
From table 2, also similar transition is observed between 5.5 N and 6 N, and hence a difference of up to 0.5 N (10% of 5 N)
This difference could be accounted to some extent to the human hand holding the upper unit and the forces on the wires. Nevertheless
we can conclude that the magnet strength is more than 90% accurate.
Addendum:
For the trial where the lower unit starts to be detached, (from our results table 2, it can be noted that this is when 100 g is attached),
Right click the trial > chart processed data > Z value vs. time chart (load sensor)
From the chart, we notice that the effective pull force on the load sensor at the time of detachment while lifting was about 4.5 N (peak value of Z in the chart). This is about 0.5 N off the applied force of 5 N, which is accounted by the weight of the magnet on the load sensor (see section I. > test 2., where this difference is shown to be about 0.5 N)
Note: You can repeat this test for the trial with 100 g, if you have not saved or overwritten the trial.
Purpose: Measure the load force due to residual magnetism.
Procedure:
~ Create two conditions for zero Newton and non-zero Newton, say 00N and N0N.
~ Assign the condition 00N with value 0.1 N and N0N with 2 N force (Condition properties > gripper > force in N). Assign number of trials per condition as 1. Add conditions N0N and 00N to the experiment used to collect trials in the same order.
~ The conditions have to be executed in the same order as listed under the experiments, as defined by, Experiment > Experiment settings > Recording > Uncheck the option 'randomize condition and trial order'.
~The trials have to be paused after the first trial, during which the magnet has to be manually switched off. Do this by: Experiment > Experiment Settings > Recording > 'Delay trial until enter key is pressed'
~ Start recording and perform each trial as follows: Lay units flat. Pull TOP unit SLOWLY without shock. Make sure the bottom unit does not move whatsoever. When the message prompt to start next trial appears, unplug the power supply to the magnet, click ok and perform the trial similar to last one.
~ For each trial, View processed data: Z axis-time; Note down the amplitude difference of the peak to the next segmentation point.
~ Repeat the above steps for 3 trials.
~ Repeat the above steps by setting the force of condition N0N to 4N, 8N,..etc.
Analysis:
~Note the force at the point of disengagement (based on the Z-data) from the processed data chart after the magnet was turned off for each trial.
Results:
Z Peak Amplitude (point of disengagement) 20 N 0 N 10 N 0 N 2 N 0 N Repetition 1 14.56 0.429 6.95 0.32 1.05 0.29 Repetition 2 13.87 0.418 6.34 0.128 1.037 0.178 Repetition 3 13.77 0.69 6.64 0.38 1.102 0.38 Average Residual Magnetism For 20 N: 0.513 N For 10 N: 0.276 N For 2 N: 0.282 N
Conclusions:
Ideally, residual magnetism should be close to zero once the magnet is turned off. The data indicate that residual magnetism increases with increasing the force when the magnet is turned off. Increasing the time of the trial-to-trial interval may reduce residual magnetism, although that has not yet been tested.
Purpose: Estimate the magnitude of the load force that could be cause by vacuum suction when the units are pulled apart.
Procedure:
~ Use the same experiment and conditions, 00N and N0N from the Residual Magnetism test.
~ Assign the N0N with 8 N force
~ Start recording and perform each trial as follows: Lay units flat. Pull TOP unit off QUICKLY and suddenly. When the message prompt to start next trial appears, unplug the power supply to the magnet, click ok and perform the trial similar to last one.
~ For each trial, View processed data: Z axis-time; Note down the amplitude difference of the peak to the next segmentation point.
~ Repeat the above steps for 3 iterations.
Analysis:
~Note the force at the point of disengagement (based on the Z-data) after the magnet was turned off for each trial.
Results:
Z Peak Amplitude (point of disengagement) 8 N (low) 0 N (zero) - Average Z value Iteration 1 3.09 -0.5 Iteration 2 3.55 -1.25 Iteration 3 3.87 -1.25 Average suction-induce load-force amplitude For 8 N: -0.75 N
Conclusions:
Ideally, the vacuum suction should be close to zero once the magnet is turned off. The data indicate that a small negative load force exists due to the vacuum suction of pulling the units apart. The suction should remain fairly constant between trials however, if the speed of pulling the units apart remains constant. Thus, the impact of a vacuum suction is negligible.
Purpose: Estimate the magnitude of the load force that could be cause by friction of the magnet touching the walls of the magnet cylinder.
Procedure:
~ Use the same experiment and conditions, 00N and N0N from the Residual Magnetism and Vacuum Suction tests.
~ Assign the N0N with 8 N force.
~ Start recording and perform each trial as follows: Lay units flat. Lay units flat. Pull TOP unit SLOW and suddenly, but SLANTED so that the magnet rubs inside the hole. When the message prompt to start next trial appears, unplug the power supply to the magnet, click ok and perform the trial similar to last one.
~ For each trial, View processed data: Z axis-time; Note down the amplitude difference of the peak to the next segmentation point.
~ Repeat the above steps for 3 iterations
Analysis:
~Note the force at the point of disengagement (based on the Z-data) after the magnet was turned off for each trial.
Results:
Z Peak Amplitude (point of disengagement) 8 N (low) 0 N (zero) - Average Z value Iteration 1 4.8 0.87 Iteration 2 5.33 1.167 Iteration 3 5.23 0.53 Average friction-induceed load-force amplitude For 8 N: 0.85 N
Conclusions:
The results demonstrate that friction induced by rubbing the magnet against the unit increases the magnitude of load force. Therefore, participants should be instructed to use a parallel motion of pulling the units apart, without slant.
© NeuroScript LLC. All Rights Reserved.