Joe's Blog - Working at Columbia Laboratory for Unconventional Electronics
Welcome to my blog!
I'm a rising senior and prospective electrical engineering major working with Professor Kymissis in a lab he created called CLUE (Columbia Laboratory for Unconventional Electronics) located on the University campus. Professor Kymissis specializes on working with thin-film electronics, an alternative of traditional silicon semiconductors, where sheets of electric components are placed on ordinary materials. A practical example of thin-film technology is the OLED display on newer smartphones. I was able to get this opportunity through the Partners in Science program, which pairs you with a professor of your interest and provides seminars on developing a research paper. The University is quite far - my working schedule is from 10am-5pm, but in reality, it's 7:30am-7:30pm including all the commuting.
A picture of the lab, CLUE Workspace for lowly interns like me
Another room with many photosensitive electronics One of the two bathrooms to choose from
Week 1
I was able to meet with Professor Kymissis, who I quickly realized was very passionate and laid-back. He prefers to be called John and would like to be contacted through Google Hangouts instead of email. At the same time, John has a very busy schedule and I don't see him too often throughout the day. For starters, John gave me some tasks to ease me in, perhaps to gauge what I would be capable of in the future. Though the Partners in Science program is research-based, I have not done any research yet. I also learned that I am working under an undergrad student, who is working under a grad student, who is working under John. Regardless, there is a wealth of opportunity and hands-on learning.
John teaches a class on different types of displays, which involve several demos and labs. Many of these labs involve a tedious process of soldering kits together. For now, my job is to help create printed circuit boards (PCB) of each of the labs so that it can be manufactured so that the students no longer need to set up the kits.
One of the demos is creating a cardboard version of the Oculus to demonstrate how to transform a 2D display into a 3D virtual image. Traditionally, students would have to print out a template and create the contraption by cutting cardboard with a pen knife. My very first assignment was to find a way to cut the template out using the laser printer in the lab, which would streamline the process for future students. This involves recreating the template with .dxg format using the laser printer software. In order to do so, I downloaded and learned how to use a software called DraftSight. Afterward, one of the graduate students, Chris, taught me how to use the laser printer and I learned that cutting cardboard is a delicate process - if your power and speed settings aren't optimal the cardboard catches on fire.
A pen knife cut version A laser cut version
Drawing file of the cardboard template made with DraftSight
Laser cutter carving out the template from cardboard
I also began to familiarize with CircuitMaker, which I will be using the in near future for schematic and PCB design. CircuitMaker is a simpler and free version of the industry standard Altium, but nevertheless difficult to learn. I spent nearly an entire day trying to create a simple multivibrator into its final form by following a tutorial. Being able to pick the right parts, modify many settings, dealing with routing errors on this new interface was quite frustrating. Though it really isn't much, it looks pretty pro to the average Joe.
A multivibrator schematic I created to familiarize with CircuitMaker
Routing the schematic design A 3D render of the multivibrator
Several days during lunch every week, Columbia invites undergrad interns and students like me to attend a lunch seminar where a professor introduces his research (free lunch too). I got to hear my own professor talk on Thursday about his research and he showed many sensors that he fabricated in the lab. Also, every week during Friday, the professor uses his excess liquid nitrogen to make some ice cream (different flavor every week) by pouring it into a blender filled with different ingredients like heavy cream. This week, the flavor was cookies and cream, and he used Milano cookies to create the taste.
Free lunch seminar with BME professor discussing microtubules and kinesins - I did not understand anything
To sum up, my first three days (I started Wednesday) have been intriguing and frustrating at the same time. The lab is half empty and most people, including John, are too busy to get in touch with. They kind of just threw me into the lab with an assignment without telling me the wifi password, the password to any of the computers or where anything was, like the bathroom, scissors, glue, etc. The first morning after a brief meeting I was alone in the lab and I had no ID (still no ID yet), so I couldn't leave the lab to use the bathroom or the door would lock. When the graduate students started rolling in, they were thankfully very helpful in guiding me. Still, I was able to learn a lot and get a good amount of work done. I'm excited to see what's ahead!
Week 2
The primary objective this week shifted to producing a single PCB board that would replace the Perception lab for John's class. I was able to meet up with my lab partner Jai, and we started by completing the lab as if we were students in order to gauge what components would be needed. The lab was fairly simple, composed of two parts: an RGB LED and a Light to Frequency (L2F) interface. Students would conventionally solder several resistors, a LED, and an AMS L2F chip on a breakout board to be attached to an Arduino Leonardo. The students would program the Arduino to set the LED at a "flicker fusion threshold" to demonstrate that at high frequencies, outputs appear to be a single image. Earlier multiplexed LCD screens use this property; monitors would seem to display a constant image but in reality, the pixels are flickering on and off, row by row, at a very high rate. The L2F board would help students visualize how color, brightness, and irradiance affects frequency.
An example of the Perception Lab John's previous students would make
Our job was to recreate the entire lab into one board: no Arduino, no external parts. We use MicroFab to manufacture the final board design, which calculates the price of boards by square inches as well as the electrical components used. In each of our boards, we must consider how to reduce the board size and also balance cost with functionality. These boards must be cheap because John will be purchasing one for each of his students on an annual basis.
We began by redesigning the lab by replacing the Arduino Leonardo with an Adafruit Trinket, another mini microcontroller. After wiring everything on a breadboard, we confirmed that the lab still worked with our new, condensed setup. With the help of Jamie, we then researched the schematic for the Trinket and figured out which components of the microcontroller we needed and which ones we could discard. We also looked into the different outputs of a MicroUSB port, which will be attached to the board to allow students to easily upload code. Ultimately, we chose to use an ATtiny85 microchip as the processor, which only had eight ports, and an Adafruit NeoPixel, a much smaller RGB LED. However, because of the simplicity of the lab, we were able to develop a schematic that successfully incorporated all necessary components:
A routed and condensed perception lab with LED and L2F Messy schematic drawing of board
After building and routing the schematic on CircuitMaker, our final design was only 3 square cm. It will take approximately 1 month for the final printed board to arrive, but in the meantime, Jamie offered to teach us how to do the delicate process of surface mount soldering so that we could make a prototype. It will be something to look forwards to in the near future!
A close up image of the Adafruit NeoPixel The final 3D render of the PCB board - only 3 cm long!
Meanwhile, we were given various miscellaneous tasks throughout the week. One task I found particularly interesting was helping a graduate student, Caroline, debug a new board design she created called the Plant Spike. The device is supposed to be stuck in the soil and would output data through Bluetooth regarding the condition of nearby plants.
Debugging the Plant Spike using a 50x camera Caroline's Plant Spike
The device was not operating properly and we started by measuring the if each of the power pins were outputting the correct voltage. This simple task turned out to be both difficult and tedious. For starters, the PSoC 4XXX_BLE56-QFN microchip on the Plant Spike had over 50 minuscule pins we needed to measure with multimeters. Fortunately, we were able to find a 50x camera to aid us in the process. Moreover, some of these pins were GPIO, meaning that I needed to write code to turn on the pin in order to test its output voltage. Though it was really just one line of code, the entire software for programming the Cypress PSoC Mini was very intricate, and programming a single pin involved creating a schematic on a graphical interface and then uploading HEX code onto a buggy PSoC Programmer. Ultimately we were unable to fix the Plant Spike, but we concluded that there was an issue with the micro USB connection. After double checking the routing for all five USB pins, we were still unable to find the error, but we hope to fix it sometime soon.
Extremely tedious PSoC Programmer interface
PSoC microchip schematic
Overall, this week was much more structured (but still not really). On the bright side, I'm glad to have a much clearer idea of what's ahead. In the meantime, I've gotten to know my lab partners and discovered many new places to eat. I also got certified to use Columbia Labs after attending a training session (IDs soon hopefully); I was told that I fell asleep three times during that lecture, but I was relieved that there was no exam. On Friday, John invited us to have his liquid nitrogen ice cream again and I discovered that he also watches Better Call Saul, a show I've been following for the past few years. I would personally suggest that you watch Better Call Saul as well, in order to become closer with prestigious professors.
Columbia mascot and logo laser cut into cardboard Mechanical Engineering Lab at MUDD
Week 3
This week I was assigned to work on many miscellaneous tasks for the graduate students while continuing to work on John’s labs. We were able to figure out that the issue with the Plant Spike last week was a capacitor which prevented many pins from having the proper output voltage. In the meantime, we were assigned to help Caroline work on her piezoelectric sensors by calibrating with the ADMET machine. We had limited resources so we laser cut some acrylic to attach to a screw in order to have a flat stable surface for measurement purposes. Another interesting task was programming a malfunctioning AD-200 lock that goes into the lab. Over the past two weeks, the graduate students had trouble getting into the lab with their IDs, so we were given a Pigeon phone and installed yet another new software, Schlage, that we needed to learn. After reading through many data sheets, we did a hard reset on the lock, downloaded the firmware, coupled the devices and reprogrammed the lock, giving everybody access once again. I still haven’t received my ID from the professor, so I just gave myself access to the lab with an old debit card.
In the meantime, we began to sort out the financial aspect of the perception lab. We use the website MacroFab in order to import our PCB design to get it produced and shipped over. However, for some reason MacroFab had trouble receiving our CircuitMaker export files whether it was in Gerber format or ODB format. We began to look into the text files that the export formats generated and we realized that it was because we had previously deleted the designators for all of our parts. After properly importing the PCB design, we began to replace the components with HouseParts, which reduces the cost of the overall board. HouseParts are part of the MacroFab storage, and using components outside of the storage would create an additional labor fee. In the end, we were able to cut the final board under $10.
John also gave us access to edit his Weebly website with all his labs and told us that it was currently very disorganized and aesthetically unpleasant. We basically spent the rest of the week editing the website and moving different blocks of text around so that the instructions are more clear. After editing, the website looks like this. However, we will be continuing to the rewrite the website throughout the summer as we simply the process of each of his labs.
Once again, we have not been assigned one formal, comprehensive task for the summer, but I have learned a lot about the PCB process from starting with a schematic idea to finalizing the financial aspects with a manufacturing company. I also had an opportunity to present some of the work done so far at a symposium, and I got to know the people at CLUE a little better.
Week 4
After ordering the parts in the financial office of Columbia, we discovered that our schematic had many flaws. After finding some ATtiny85 chips and other components, we wired our schematic on a breadboard to test the schematic for ourselves. It turned out there were critical components, such as the micro USB mount, that we didn’t even connect to power. In retrospect, my partner and I discovered many simple mistakes that we were unable to catch at first, due to inexperience. Once we finished wiring the schematic, we realized that our board did not have any means for programming the chip with bootloader code at all. In fact, we were unaware that chips needed to be programmed in the first place. This meant we needed to add on an additional set of headers that would allow the ATtiny85 to receive code from the AVR Programmer. The overall experience taught me how the Arduino board settings are integrated through USBtinyISP drivers. Pushing code to the chip was still a difficult process and we ended up pushing a .HEX file of a downloaded Trinket bootloader code through command prompt. This made our schematic effectively act as a makeshift Adafruit Trinket. We then attempted to upload code using a USB, but our connection was unsuccessful. When we created our schematic, we discarded different components that seemed unnecessary - we had thought that the zener diodes that attached to the D+ and D- lines on the USB were only a safety precaution for a voltage overflow. In reality, those channels required a specific 3.6V in order to properly transmit information, and the zener diodes were there to block current up to that specific voltage.
Meanwhile, we began to work on routing an OLED board, another one of John’s labs. This lab is a lot more extensive - the process of importing house parts and routing was quite tedious. With a substantial amount of work completed for John’s display class, we needed to look for new tasks to do. Furthermore, the Partners in Science program requires a research project, paper, and presentation - the tasks I completed thus far, none of which were experiments, are inadequate to showcase at the symposium. After voicing my concerns to several graduate students and John, I was lucky enough to participate a research project.
CLUE is currently in partnership with a biomedical company from Toronto called Synaptive Medical. They produce equipment that can be used for brain topography and neurosurgical practices. My work involves testing a piezoelectric (pressure sensitive) strain gauge sensor. Previously, CLUE had created a sensor that can be mounted on the human skin and other vital organs that would monitor any deformation or swelling. The current implementation has a Young's modulus of 0.3, with an elasticity similar to human skin - thus it can be comfortably mounted without any constraint on mobility.
The piezoelectric functionality comes from the chemical properties of polydimethylsiloxane (PDMS), a silicone structure that creates resistance as a result of deformation. Different piezoelectric sensors have varying gauge factors (GF). For instances, those mounted on concrete to detect structural damage on larger structures tend to have a higher gauge factor, meaning a more sensitive change in resistance for each deformation.
Through a proper combination of thin film electronic components, an ideal sensor can be created for its respective purpose. In the case of a skin-mounted sensor, CLUE used carbon black doped PDMS for the resistors and carbon nanotube doped PDMS for conductors. My recent research assignment involves testing a piezoelectric rod used to aid the process of brain surgery. It includes four piezoelectric pads, each with its own resistance, representing the x and y axis. Caroline gave us two objectives: we need to determine the ideal adhesive for adhering the silicone to stainless steel and also help reduce noise sampling through a series of tests such as baseline drift and noise vs. sampling frequency.
