Team Vision for Detailed Design Phase
During the final detailed design phase, the team first and foremost made a final decision about which design to pursue. After much analysis and discussion, the team will no longer be moving forward with the conductive ink design will instead be solely focusing on the optical design. After three weeks of conducting different tests, a more finalized Bill of Materials was completed, taking into account new decisions and findings based on this testing.
Specifically, the wavelength sweep resulted in an optimal laser wavelength value, which allowed for research and final decisions to be made on purchasing a laser source for the final design. Testing on the current photodiode, however, has not been quite as successful. Although this could be explained in a variety of ways, such as high precision active area or sources of noise, a decision to purchase a new photodiode was made.
After modeling, the projected minimum resolvable voltage for the purchased modulator was 8 uV, however, the modulator has not quite performed as expected. This has been largely associated with the poor insertion loss fusion splicing due to the modulator’s unconventional fiber. This will be resolved by sending the modulator our for professional fusion splicing. Measurements of the depackaged modulator were taken and a CAD model was created for a new MRI-compatible packaging.
A deeper look into the phase plates in Dr. Preble’s lab resulted in an idea to also create a CAD model and produce a much less expensive version of the phase plate. It was also determined that polarization-maintaining fibers would greatly simplify the setup and reduce the overall bulk of the system.
The EEG signal receiving schematic was fully developed during this phase and all prototyping parts were purchased. The prototyping was started by utilizing resources from the Surface Mount Lab at RIT. Code for the microcontroller to convert signal protocols was planned out during this phase.
In summary, this phase was largely focused on the decision to move forward with the optical design and nail down decisions that needed to be made in order to move forward in MSD II. The design largely satisfies all customer and engineering requirements.
The directory of documents can be viewed within the Final Detailed Design Documents folder.
System Architectures and Engineering Analysis
An interactive version of the System Architecture can be viewed here
The Signal Processing Plan can be viewed here.
For the processing of the signal once it reaches an external computer, the MATLAB toolbox, EEGLAB, will be used. EEGLAB is a free, open source toolbox, specialized for EEG data signal processing, and is capable of not only implementing simple filters as MATLAB can easily do otherwise, but also can implement current techniques used for advanced EEG data filtering. It also provides a relevant GUI to work with, and different methods of plotting the data. A psuedo code has been planned out for interfacing with both the incoming signal from the receiver, and for interfacing with EEGLAB. Here, the MATLAB serial command is also used, to receive the data from one of the computer's USB COM ports.
For receiving the serial communication:
- serial command is called and constantly runs in a loop, storing data into a text file
- in the loop, three 16 bit words will be expected, the first of the words being the data received from the ADC, and the second two words being a timestamp generated from the MSP430
- with the data stored in a text file, with proper delimiters between data and timestamps, the data can then be later accessed to be processed
For processing the data:
- The data can be read in using the EEGLAB GUI, and then will be first filtered using the FIR filter tools, and creating a band stop filter at 60 Hz, to eliminate mains power noise. Afterwards, a different band pass filter will need to be implemented depending on the brain signal in question. These two filters create an amount of filtering similar to that used by the receiving box that has been used for testing thus far.
- Past that, independent component analysis (or ICA) removal can be used to observe the sources of certain components of the data deemed as noise. This method can learn what components of a signal are not required data, and then filter those components out of the data afterwards. EEGLAB fully implements two different ICA methods, of which, the basic ‘runica’ function will be used. This will mainly be used to filter out BCG noise caused by the body on the EEG data, and possibly consistent pulses caused by the MRI.
Electrode and Modulator Architecture
The Modulator Wavelength Dependence document can be viewed here.
The Modulator Case Design document can be viewed here.
The crystal will be placed within the middle slot of the device, and the connected fibers will be able to run out of the exit holes on each side. Two input holes were added to wire the signal and bias voltage to the device. Ultimately, a second rectangular piece would cap the device, as well as secure the exit fiber holes. This drawing is based on the depackaged modulator’s dimensions, which is not the modulator that we’ll ultimately use, so the drawing above is simply a rough draft and the final design will depend on any changes that will need to be made based on the other depackaged modulator’s dimensions. We are not able to establish those dimensions at this time. The newer modulator casing design will only have one electrical input hole as it does not require a bias voltage.
One important component of our project is to remove the electro-optic modulators from their metal casings so that the MRI footprint of the system may be minimized. To this end, we began deconstruction of a spare modulator provided by Dr. Preble before attempting the process on one of our own devices.
We found that milling the casing off of the modulator is trivially easy with the help of a machine shop technician. Since system dimensions were required before a plastic replacement case could be printed, only the ends of the modulator case were milled for the time being (as illustrated in Figure 1 below).
After exposing the internal components of the modulator, we sent the device to a naval research lab for imaging. This permitted for the determination of precise component dimensions, as well as a visualization of the fiber/modulator interface (which will need to be preserved during the transfer). Representative images from this analysis are presented in Figure 2.
One concern that was raised during modulator analysis was that the crystalline core appears to be bonded to the metal substrate. After contacting one of the design engineers working for JDSU, it was determined that MEK may be used to loosen the epoxy binding. This will need to be applied with high precision (using a micropipette) in order to avoid damaging the connection between the optical fiber and the LiNbO3 substrate. Once the seal has been weakened, the device will be lifted out of the modulator case using two clamps held by a micro-manipulator. The modulator will then be moved via the manipulator and set down into its new casing. Once in place, the plastic case will be closed and the modulator may be re-tested and then incorporated into our final production device.
Testing Data and Analysis
The photodiode interfacing document can be viewed here.
During the detailed design phase, the team was unable to successfully establish a permanent interface between the fiber optic and the 100 um active area of the photodiode.
To better test the photodiode, we assembled the fiber optic output cable onto micromanipulators that were capable of finely controlling the fiber’s position relative to the photodiode on all three axes. The board with the mounted photodiode was fastened securely onto a separate stand.
Unfortunately, testing of the photodiode was unsuccessful, as the team was only able to resolve 40 uV, whereas the photodetector used previously was capable of a 15 uV resolution. The team believes that the test was unsuccessful for several reasons:
- The current micromanipulator setup introduces a large lever arm into the system that makes it far more vulnerable to vibrational noise. Given that we’re trying to record very small amplitude signals, the system is likely to be highly susceptible to these noise sources.
- The team found that even small movements (on the order of microns) of the micromanipulators was enough to dramatically change the signal-to-noise ratio of the photodiode output. This setup may lack the control to precisely interface the fiber with the photodiode.
- The cables and solder used to connect the photodiode to the amplifier may be too susceptible to electrical noise.
Future plans: The team plans to brainstorm other ways of reducing noise in our current photodiode setup, but we would also like to order the following pre-interfaced photodiode for $50:
The data sheet for the photodiode can be viewed here.
The advantage of this photodiode is that it is already interfaced with an FC connector, which is the needed connector type to directly interface with our fiber. It is difficult to compare other aspects of this photodiode’s performance to the ones we’re using currently, as this manufacturer does not list noise equivalent power. However, SRICO, who supposedly used a similar type of electro-optic system for measuring EEG potentials, used the FD150 version of this photodiode, making this a promising option for testing. Unfortunately, using these photodiode will increase our project cost and put us at risk for exceeding our budget.
Proof Of Concepts
The Optical proof of concept document can be found here.
The concept for not choosing to continue pursuing the ink design can be found here.
As demonstrated during the detailed design review, we have been able to achieve a minimum resolvable voltage of 15uV using preliminary prototyping equipment in Dr. Prebble’s lab. The data collected during those trials was used to fit and validate a mathematical model which may be used to predict how performance will change in response to new and improved component pieces which we intend to employ in our final design. It is worth noting, however, that even at a minimum resolvable voltage of 15uV it still technically ought to be possible to detect alpha waves based on experimental data taken with commercial EEG equipment.
The bias free modulators we intend to use in our final product (and have begun to purchase) have a Vpi value of 3.7V. This compares favorably to that of the modulator currently used in our testing, which has a Vpi value of 8.4V. The model predicts that switching modulators will thus drive our minimum resolvable voltage down to 6.6uV. Our preliminary testing with the modulator did not see performance at this level, but we attribute this to a non-standard fiber diameter causing our in-house optical cable splicing to fail. We are in contact with a company who provides professional grade splicing services and who guarantees a low insertion loss for their devices.
Switching to our proposed 20mW laser and accounting for the increased insertion loss which accompanies splitting the laser beam into 8 channels, our model predicts a minimum resolvable voltage of 8.0uV. This does not account, however, for the expected decrease in noise accompanying the transition to an integrated detector/amplifier system. It also does not account for the lower noise equivalent power ratings of our proposed detectors, nor the high performance ratings of our proposed laser diode (incorporation of these factors into the model will require empirical test data). Even ignoring the potential improvements in performance which our system is expected to achieve, an 8uV resolution is certainly sufficient to detect alpha waves, p300 evoked potentials, and a myriad of other EEG signals of potential interest to an academic researcher.
The Engineering Requirements document can be viewed here.
Bill of Materials (BOM)
The Bill of Materials document can be viewed here.
The Risk Assessment document can be viewed here.
The Test Plan document can be viewed here.
Plans for Next Phase
Plans for IntersessionOver the intersession, the team plans to work on finalizing the cost estimate of the design, mainly the design as a PM system and possible backup electro-optical modulators. The REMCOM software will also be utilized to finalize the Specific Absorption Rate (SAR) calculations to ensure the safety of the device. The CAD model for the modulator and phase plate will also be finalized and ready to print as soon as the team returns for MSD II.
The MSP430 code will be written and tested over the intersession as well as the amplifier, filter, and MSP430 prototyping testing. The MATLAB signal processing code will also be completed and work on the PCB design will continue.