P16071: MRI-Compatible EEG Cap
/public/

Integrated System Build & Test with Customer Demo

Table of Contents

Team Vision for System Level Demo with Customer

Integrated System Build & Test Documents Directory

Integrated System Build & Test Documents Code Directory

During Phase 4, the team planned to assemble the remainder of the PCB and thoroughly test it, build the laser controller PCB, order new modulators with low insertion loss and test them, troubleshoot bias drift, finalize the shielded box design, finalize the cap design, and write 75% of the paper.

Three channels of the PCB were built up in hopes that a 3-electrode system could be demoed. The PCB continued to be plagued with problems throughout this phase, including the inability of the computer to recognize the PCB as a COM port, incorrect data return from the microcontroller, and the failure of the PCB to sample quickly enough. While these problems were solved and preliminary PCB tests were successful, the integration of the photodiode and optical front-end with the system caused unexpected additional problems.

The new modulators were characterized extensively but continued to perform poorly. The team obtained customer feedback, which encouraged us to instead pursue a demo of a 1-channel EEG system with greatly amplified voltages as an initial proof-of-concept. Because this project is expected to be a multi-year project, the creation of a shielded box was no longer placed at high priority, as it it more desirable to have prototyping components accessible for future groups to work with easily. New materials were ordered and assembled for the cap design, and the modulator attachment method continued to be explored to find the optimal solution. The laser controller PCB was assembled and a cable was built to correctly connect it to the laser mount. Integration of the laser will occur after the team more thoroughly tests the controller to ensure that it will not harm the laser with high currents or temperatures. Approximately 75% of the paper was completed.

System Architecture

System Architecture

System Architecture

The system architecture document can be viewed here.

Test Results Summary

Cap Design and Testing

EEG Cap Prototype & Testing

The EEG Cap Prototype document can be viewed here.

EEG Cap Prototype

EEG Cap Prototype

ENGINEERING REQUIREMENT: Weight of cap components

Marginal Value: <4 lbs
Ideal Value: <1 lb
Performance: 162g = 0.36 lb < 1 lb
Cap components:
Silicone Tubing - 10 Foot Piece (High Temp Hose - 500 F) (1/16" I.D. x 1/8" O.D. x 10 Foot)
3M Elastic Chin Strap
3 x Modulator Clips - (2 grams polylactic acid - melting temp 150-160 C)
2 x Clip Mount - (4 grams- polylactic acid - melting temp 150-160 C)
3 x JDSU Electro-Optic Modulator
3 X Plastics1 Ag/AgCl Electrodes + Gold Clips

ENGINEERING REQUIREMENT: Cap attachment time

Marginal Value: &#8804;20 minutes
Ideal Value: &#8804;5 minutes
Performance:
Steps 1-3: Placing cap onto head and adjusting for optimal fit: &#8804;5 minutes
Steps 4-7: Placing modulator, electrode paste and electrode onto head: 1 minute per electrode
Total time for cap and 3-system electrode placement: 8-10 minutes
Procedure:
  1. Place cap onto head
  2. Clip chin strap into place
  3. Tighten cap strings for a snug fit
  4. Clip modulator onto cap in approximate desired position
  5. Place electrode paste onto electrode
  6. Attach electrode to head
  7. Move modulator as needed

ENGINEERING REQUIREMENT: Cap removal time

Marginal Value: &#8804;20 minutes
Ideal Value: &#8804;5 minutes
Performance: &#8804;5 minutes

Procedure:

  1. Remove all electrodes from the head and wipe paste off
  2. Unclip all modulators and set them aside
  3. Unclip chin strap and remove cap from head

The Updated Modulator Clip Dimensions document can be viewed here.

Clip CAD model

Clip CAD model

PCB Testing

Amplifier & Digital Receiving PCB with 3 Input Channels Placed

Amplifier & Digital Receiving PCB with 3 Input Channels Placed

This board includes the attached photodiodes, which provide a current input to the transimpedance amplifier, and second order low-pass filter (top right of board). This is sent to the ADC (middle of left half of board), which provides the data to the MSP430 (middle of bottom half) as it calls it for data. The MSP430 is in turn told to process data through USB by an attached PC, by communicating through a UART-USB converter (right of MSP430), which is passed through the attached USB cable (bottom right corner).

ADC and Data Receiving System Minimum Required Operation

The ADC and Data Receiving System Minimum Required Operation testing document can be viewed here.

Purpose:

The purpose of this experiment was to test and prove the minimum required operation of the ADC and Data Receiving system, which includes the ADC, MSP430, UART-USB converter, and the computer receiving program. The operation is defined in minimum resolvable voltage step size, and correct receiving of expected maximum frequency signals.

Procedure:

The computer receiving program, and the MSP430 were both set to use a USB baud rate of 115200, theoretically allowing a sampling frequency of 1.92 kHz. Using Channel 1 of the ADC as an input, an 8 mV p-p, 100 Hz sine wave was applied, and the results given by the receiving program were exported to MATLAB for plotting. The 100 Hz frequency is the maximum expected frequency of signals that would be received from an EEG. The 8 mV is the previously predicted minimum voltage step size that would be expected at the input to the ADC, which is after amplification. This test was run without the amplifier as part of the system.

Results:

The minimum voltage step size can be seen in Figure 1 below, which is some of the data read by the computer receiving program. This does prove that the system is overall capable of receiving this signal level, which is a good deal smaller than what has now been observed from the optical system output. However, upon zooming out from Figure 1, Figure 2 is seen, where seemingly random spikes in the data are visible. These spikes are theorized to possibly be errors in the bit stream, or a signal somehow read as negative, which in processing, would loop back around to a bit below 10 V.

Also seen in Figure 1 is the time scale on the signal. This signal can be seen to be 100 Hz, and is read in as a decently recreated sine wave, proving that the system can successfully read the maximum required frequency.

Figure 1: Received Data Proof of Minimum Step Size, and Maximum Frequency

Figure 1: Received Data Proof of Minimum Step Size, and Maximum Frequency

Figure 2: Example of Erroneous Data in Signal Receiving

Figure 2: Example of Erroneous Data in Signal Receiving

Modulator Testing

Modulator Bias Testing

The Modulator Bias Testing document can be viewed here.

Purpose:

In the prior design phase, the poor power transmission of our new modulators was determined to be the root cause of the system sensitivity problems. As a result, LightWaveStore agreed to send us their three modulators with the lowest insertion loss specifications. These loss specifications were 1.4 dB, 3.4 dB, and 3.5 dB. The purpose of this experiment to was characterize each modulator’s insertion loss and compare the sine-squared bias voltage curve to that of the old modulator. This would allow us to assess the best possible candidate for testing against the old modulator’s performance.

Procedure:

Each modulator was hooked up to the optical system, and their power outputs were monitored as the polarization and bias voltage was changed to find the maximum power transmission through each. This was compared to the input laser power and used to characterize each modulator’s insertion loss.
The bias voltage curve of the modulator with the most favorable insertion loss was then calculated by finding the voltages at which the peaks and troughs of the sine-squared curve occur.

Results:

The modulators with 1.4 dB and 3.4 dB insertion losses were calculated to have actual losses of 3.4 dB. Oddly, the modulator with a 3.4 dB insertion loss was calculated to have an approximate insertion loss of 3.2 dB. This compares well with the old modulator, with an insertion loss of 3.1 dB. This, in combination with a slightly lower bias Vpi, makes the slope of the linear region of the sine-squared curve of this modulator highly comparable to that of the old modulator.
New vs. Old Modulator

New vs. Old Modulator

Conclusion:

The modulator shown in the figure above should have comparable performance to the old modulator based on our current knowledge of electro-optic physics. Unfortunately, this was not achieved during performance testing, as described in the subsequent minimum resolvable voltage and step size test.

Minimum Resolvable Voltage and Minimum Step Size

The Minimum Resolvable Voltage and Minimum Step Size document can be viewed here.

Purpose:

The purpose of this test was to determine the system’s minimum resolvable voltage and step size to estimate the team’s current progress towards its test plan target.

Procedure:

The minimum resolvable voltage was determined by decreasing the signal until the SNR threshold of 5 was reached. The minimum step size was determined by varying a larger-amplitude signal slightly and analyzing the average and standard deviation. When the standard deviation windows of two signals did not overlap, the minimum step size was determined to be reached.

Results:

The minimum resolvable voltage was determined to be approximately 250 uV (an SNR of approximately 7 was obtained when the system was functioning optimally, though this value is variable and sometimes dropped below the threshold of 5).
The minimum step size was also determined to be approximately 250 uV. A 2250 uVPP signal had an amplitude of 0.4291 mV +/- 0.0152 mV, while a 2500 uVPP signal had an amplitude of 0.4819 mV +/- 0.0189 mV. This is also shown visually in the figure below, with the 2250 uVPP signal followed by the 2500 uVPP signal.
Minimum Resolvable Voltage

Minimum Resolvable Voltage

Laser Assembly and Testing

Laser Mount and Temperature Controller

The Laser Mount and Temperature Controller document can be viewed here.

The laser controller PCB was assembled with the temperature controller, and a custom 9-pin Dsub cable was built to link the controller to the mount properly. The system is shown in the figure below.

Laser controller PCB (left) linked to the laser mount (right)

Laser controller PCB (left) linked to the laser mount (right)

Laser integration needs to be handled with care and performed only after the controller has been thoroughly tested. Exceeding the maximum current inputs or temperatures could destroy the laser.

The following steps need to be implemented prior to interfacing the laser with this custom controller:

1. Cut down pins to allow laser to fit in the mount.

2. Test temperature controller for function:

Connect temperature controller to 9, 10, and 11 kOhm resistors. The controller is targeting 10 kOhm as this corresponds to 25 C in the temperature controller, so this resistor should have no current flowing across it. The other two should, as the temperature controller should adjust for the varying resistances.

3.Test laser controller for function:

The input current to the laser should be 140 mA. Unfortunately, the impedance of the laser is unknown, so the proper test protocol still needs to be developed.

EEG Lab Filter Testing

The EEG lab Filter Testing document can be viewed here.

Purpose:

The purpose of this test was to confirm that the recommended EEGLab program for our prototype will be capable of adequately filtering data. Its bandpass filter was compared against that used in LabChart software.

Procedure:

Raw alpha wave EEG data was filtered in LabChart with a bandpass filter from 7-14 Hz. The same raw data was loaded into EEGLab and filtered with its bandpass filter and the same pass band. The frequency- and time-domain data were each overlaid for comparison.

Results:

The raw data, data filtered in LabChart, and data filtered in EEGLab, are shown in Figures 1, 2, and 3, respectively. Visually, it appears that the EEGLab filter worked as well as that used in LabChart in terms of filtering out noise and allowing us to see the subject’s alpha wave activity.
Figure 1: Raw Alpha EEG Data. Eye closure occurs at the artifact just after 15 seconds

Figure 1: Raw Alpha EEG Data. Eye closure occurs at the artifact just after 15 seconds

Figure 2: Alpha wave EEG data filtered in LabChart. Alpha wave content is clearly visible after 15 seconds

Figure 2: Alpha wave EEG data filtered in LabChart. Alpha wave content is clearly visible after 15 seconds

Figure 3: Alpha wave EEG data filtered in EEGLab. Alpha wave content is still visible after 15 seconds

Figure 3: Alpha wave EEG data filtered in EEGLab. Alpha wave content is still visible after 15 seconds

The overlaid power vs. frequency spectra are shown for each filtering method in Figure 4. Both plots seem very similar. LabChart was slightly more effective at filtering out noise in the 7-9 Hz and 12-15 Hz ranges; however, it also slightly diminished the power at the ~10 Hz frequency of interest.

Figure 4: Overlaid frequency content of the LabChart filtered (blue) and EEGLab filtered (red) data

Figure 4: Overlaid frequency content of the LabChart filtered (blue) and EEGLab filtered (red) data

Conclusion:

Bandpass filtering in EEGLab will work comparably to that used in LabChart.

Misc Testing

Signal Accuracy

The Signal Accuracy Preliminary testing document can be viewed here.

Purpose:

The purpose of this test was to determine the ability of the optical system’s output to accurately represent the input signal’s amplitude and frequency.

Procedure:

A 10 Hz, large amplitude signal of 2500 uV and a 10 Hz, small amplitude signal near the resolution threshold (250 uV) were both inputted into the modulator. 15 seconds of data was collected for each condition. A fast fourier transform was taken of the data, and the peak spectral width and center were recorded to assess deviation from the 10 Hz input frequency. Additionally, the standard deviation of the amplitude was recorded to assess its deviation from average.

Results:

The fourier transform of the large amplitude signal is shown below. The peak occurs at approximately 10.021 Hz, and ranges from approximately 9.865 – 10.140 Hz. It is likely that the true value is indeed 10 Hz, as the measurement is only limited by the resolution of the fourier transform (10.021 Hz was the closest value to 10 Hz, and it had the peak amplitude).
These values translate to approximately a 1.6% frequency deviation on either side of the 10 Hz signal in the fourier transform.
2500uV Signal

2500uV Signal

The next figure below shows the signal overlaid with the peak height measurements. The average peak height was found to be 0.4287 mV with a standard deviation of 0.0249 mV. This translates to a 5.8% amplitude deviation.

2500uV Signal Overlaid with the Peak Height

2500uV Signal Overlaid with the Peak Height

The fourier transform of the small amplitude signal is shown below. The peak again occurs at 10.021 Hz, and it ranges from approximately 9.889 – 10.124 Hz. These values translate to approximately a 1.2% frequency deviation on either side of the 10 Hz signal in the fourier transform.

250uV Signal

250uV Signal

Finally, shown below is the amplitude deviation in response to the small voltage signal. The average peak height was found to be 0.0668 mV with a standard deviation of 0.0146 mV. This translates to a 21% amplitude deviation.

250uV Signal Overlaid with the Peak Height

250uV Signal Overlaid with the Peak Height

Modulator Fatigue

The Modulator Fatigue testing document can be viewed here.

Purpose:

A successful MRI-compatible EEG system must be able to withstand stresses associated with attaching, removing, and transporting its contents. The purpose of this test was to monitor any performance degradation in a fiber optic cable attached to an electro-optic modulator when it is placed under bending stresses.

Procedure:

Unfortunately, RIT lacks a fatigue test fixture capable of subjecting an object to bending forces. Instead, a metal cylinder was fixed onto the optics table for manual cyclical testing. The polarization-maintaining input fiber of the bias-free modulator was wound several times around the cylinder and then unwound (see figure below). This action constituted “one cycle.”
The modulator’s baseline performance was characterized on the day before that on which the remainder of the testing was performed. In 50 cycle increments, light output through the modulator was measured and compared to the input power. An increase in power loss through the modulator would indicate that bending stresses have detrimental effects on either the system’s insertion loss or the capabilities of the stress rods to maintain polarization within the fiber. This pattern of cyclically stressing the fiber and measuring its output was continued for a total of 500 cycles.
Testing Setup

Testing Setup

Results:

Results are displayed in the figure below. With the exception of the first measurement, insertion loss remained relatively constant around 8.3 dB. The t=0 measurement of 6 dB, however, suggests that the initial bending may have affected the modulator performance. However, because this test was performed on a different day than the rest, and the data point is an outlier compared to the rest, there is some evidence to suggest that this data point could be disregarded. If the laser or other components in the system were functioning differently on the previous day, results may have varied. For example, if the laser temperature varied on the previous day, a different wavelength through the interferometer would have resulted in a different modulator power output, thereby affecting the insertion loss measurement.
To confirm the suspicion that this first data point should be excluded from analysis, a second modulator was tested before and after 50 cycles. The large increase in insertion loss was not replicated in this second test.
Fatigue Testing

Fatigue Testing

Conclusion:

There is little evidence to suggest that bending stresses affect the performance of the modulator or the fiber; however, some future investigations may be warranted.

Engineering Requirements

The Engineering Requirements document can be viewed here.

Test Plan

The Test Plan document can be viewed here.

Risk Assessment and Problem Tracker

The Risk Assessment document can be viewed here.

The Problem Tracker document can be viewed here.

Technical Paper

The Technical Paper can be viewed here.

Plans for next phase

MSDII Schedule

MSDII Schedule


Home | Planning & Execution | Imagine RIT

Problem Definition | Systems Design | Subsystem Design | Preliminary Detailed Design | Detailed Design

Build & Test Prep | Subsystem Build & Test | Integrated System Build & Test | Integrated System Build & Test with Customer Demo | Customer Handoff & Final Project Documentation