Integrated System Build & Test
Team Vision for Integrated System Build & Test Phase
During this phase, the team planned to receive, assemble, and test the PCB, troubleshoot and test the microprocessor and ADC, and test the new modulator choice. The team also planned to order enough parts to demonstrate the use of a 3-modulator system in acquiring alpha waves.
Unfortunately, this phase was plagued with a variety of problems that prevented the acquisition of alpha waves with a 3-modulator prototype. Ongoing PCB troubleshooting prevented its full assembly, as one of the eight channels was assembled first to verify functionality. The splitter and three modulators were received and tested extensively, but ultimately, the modulator performance was insufficient to resolve EEG signals. Bias control efforts were continued throughout the phase, but bias control remains a problem in our system. The new laser was received, integrated into a controller, and tested. Code was written to acquire EEG data from the system in real time. The cap design was reworked in response to customer concerns and more detailed design plans were created.
The Systems Architecture can be viewed here.
The Cap design proof of concept document can be viewed here.
EEG Cap - Proof of Concept
- Small Diameter, very flexible silicone tubing
- Comfortable chin strap
- Universally fit
- MRI-Compatible materials
- 10-20 electrode system possible
- Electrode Materials
- Ink Cap: Ag/AgCl
- Conventional: Gold (0.66 Ω/m)
- XFDTD Simulations
- RF frequency: 300 MHz
- MRI: 7T
- Input Power: 1 W
- Both Ag/AgCl and Gold
proven safe in MRI
- Silicone tubing ordered
- Needs to be fitted to average head size
- Number of channels TBD
- Chin Strap:
- Ordered and received
- Silicone Tube Connectors:
- Find smaller connectors that allow for stability and flexibility
- Possibly design on CAD & 3D print
- Silicone Tube to Chin Strap Connectors:
- Need to design on CAD & 3D print
- Contingent on final channel number
- Modulator to Silicone Tube Connectors:
- Designed on CAD
- Dimensions need to be double checked
- Need to be 3D printed
Laser Control Design
EEG Capture Software
The source code for the capture software can be viewed here.
The Laser Stability Document can be viewed here.
- The purpose of this test was to integrate our laser into a mount and controller system in Dr. Preble’s lab and compare its baseline noise stability to that used in previous testing. The secondary purpose of the test was to characterize the dependence of laser power output on the input current.
Laser stability - procedure and results:
- After the new QPhotonics laser was integrated, its noise at a power of approximately 2 mW was compared to that obtained with the previous Agilent laser. A Thorlabs power meter was used to plot laser output over time, and the RMS2 value was calculated to compare noise between the two models. Additionally, the Agilent laser was integrated with the isolator to compare the performance of that system to that of the new modulator. The following plot was obtained:
The noise of the Agilent laser was many orders of magnitude higher than that of the new QPhotonics laser. After the isolator was integrated into the Agilent system, the noise became closer to that of the QPhotonics system; however, it was still higher.
- The new QPhotonics laser performance is superior to that of our previous system.
Laser power – procedure and results:
- The power of the laser was sampled at 5 currents, including the maximum recommended current, to create a linear model of the dependence of laser power on input current. Results are shown below.
- Laser power reaches a maximum of approximately 19.7 W. Laser power has a linear relationship with current, and the equation obtained above can be used in the future to set the current according to the desired laser output power.
Laser/splitter system together – procedure and results:
- Once the 8-channel PM splitter was received, the laser was routed through it to characterize its performance. 1.7 mW were recovered from each of the eight channels, so the insertion loss performance was as expected. However, the laser was observed to be very noisy, as shown below:
It is possible that this noise is resulting from internal reflections within the splitter itself. The laser controller has also been unreliable throughout the course of Phase 3.
- Further investigation into the stability of the laser is warranted.
The Modulator Bias Control testing document can be viewed here.
- The purpose of this set of experiments was to find methods of controlling the bias voltage to help control the bias drift problems in our system.
- Various methods of bias control were tried, including a MOSFET that switched on an off to discharge the capacitor more quickly than grounding it. A small subset of the results are outlined below.
- With a 50% duty cycle in our MOSFET system, we were able to achieve fairly stable bias performance (i.e. the modulator was consistently outputting a power around 19 uW when the bias was turned on).
Unfortunately, the modulator appears to take time to “turn on” after the bias voltage is turned on, so the signal is generally not resolvable with a 50% duty cycle.
A 70% duty cycle bias was tried with a 10 ohm resistor; however, the system seemed to be highly sensitive to movement and may have exhibited some drift.
Below is an example of the modulator output with a 70% duty cycle and a 250 uVPP, 10 Hz signal:
Unfortunately, the system was also found to be highly susceptible to motion, as artifacts such as those shown below were observed when people moved:
Even if the switch is used at a 70% duty cycle (which may be too long to prevent eventual bias drift over time), data would only be collected for ~50% or less of the time.
The bias drift phenomenon was also investigated over large periods of time. Initially, the system was turned on and left to settle to a certain bias point plateau, a phenomenon which occurred in ~45 minutes. After this time, the bias voltage was relatively stable. Unfortunately, technical reasons prevented this data from being saved.
This was investigated more in detail later. When the bias voltage was set to 7.1 V, the following plot was obtained:
Here, the bias voltage failed to truly settle on any given value. The bias voltage was increased to 7.2 V after 2.5 hours to see the effect of changing the bias voltage on the performance:
Here, the bias did appear to roughly stabilize at 10 uW after 15 minutes, though it is unclear whether or not this trend would have continued.
Increasing the bias voltage to 7.6 V:
Therefore, it does appear that the bias voltage will plateau at a different power output depending on the bias applied. If this relationship is consistent, it is possible that the system would perform favorably after a warm-up period; however, as seen in these plots, the system can take large amounts of time to settle on a particular bias voltage.
Commercial bias control circuitry relies on monitoring the output of the modulator with a second photodiode that then feeds into a bias controller. Splitting the output fiber of the modulator to integrate a second photodiode is simply not practical for our project.
- Current methods have failed to control the modulator bias drift appropriately, warranting further investigation into bias control methods.
Modulator Bias Vpi Comparison
The Modulator Bias Vi Comparison testing document can be viewed here.
- The new OC-192 modulators are exhibiting decreased performance when compared with Dr. Preble’s old modulator, despite their similar technical specifications. To better understand this phenomenon, the power output of each modulator at a given bias voltage was studied to roughly characterize its bias Vpi and insertion loss.
- To characterize the approximate bias control curve of the new modulator and compare it to that of the old modulator.
- The bias voltage on each modulator was varied until both maximum and minimum values were achieved. These voltages were recorded along with the power output at maximum bias. These numbers were then used to re-create the modulators’ approximate bias sine squared curves for visualization.
- The sine-squared curves for both the new and old modulators are shown below:
In general, a smaller Vpi means that a smaller voltage is needed to give a resolvable change in light intensity. The old modulator’s bias performance curve is shown in blue above. If this same curve were adjusted to reflect the smaller Vpi of the new modulator (shown as the gray “theoretical curve), the linear region of the sine-squared curve would have a steeper slope, thereby giving us a better resolution.
Unfortunately, the actual sine-squared curve of the new modulator is shown in orange. While the Vpi was lower than that of the old modulator, the insertion loss (empirically) appears to be higher, as we get a lower maximum power output. This larger insertion loss translates to a far shallower slope in the linear region of the modulator. This helps to explain the poor resolution of our new modulator lot.
- In order to achieve a resolution suitable for EEG, we must be able to minimize insertion loss beyond typical manufacturer specifications. Empirical data suggests that the insertion loss of the old modulator is <3.2 dB, while manufacturers generally spec to <5 dB.
Risk Assessment and Risk Chart
The Risk Assessment document can be viewed here
The Problem Tracking Document can be viewed here.
The Test Plan Document can be viewed here.
The Detailed Test Plan Document can be viewed here.