Technical Report from India

So in February / March this year, a group of lab members (PI, 3 postdocs and 1 PhD student) went off to Kolkata to record EEGs, and conduct behavioural data and speech production experiments on Bengali speakers. It's something you don't get to do that often, and as it's our first time we tried to prepare as much as we possibly could for the trip. One thing I learned from this experience is that preparation only gets you so far.

We'd switched to a new EEG acquisition system (the Biosemi ActivTwo), mainly because I wanted a system that would be relatively quick to apply but still reduce the interference from the 50Hz noise from mains electricity supplies. This had arrived in good time and we'd tested the kit once or twice in our labs. One problem that was thrown up was making our whole test rig portable. So desktop PCs had to be replaced by laptops.

The main problem with modern laptops from a psychological experimenter's point of view is the lack of connectivity. A laptop will nowadays come with USB, ethernet, video out, and *if you're lucky* an expresscard slot. If you want to input keypresses for recording something like reaction times, or output trigger information you can't really use USB (or indeed the internal keyboard or trackpad) as these connections are buffered and not a priority for the OS. That means you can get delays along the lines of (depending who you talk to) 5 - 30 ms. This delay means nothing for regular PC work, but for an experiment that could be bigger than the effect you're looking for. There are USB devices that can be purchased to get around this limitation (National Instruments do one for example), but these are costly and don't speed up the connection, they just buffer in an external memory with an accurate time stamp, so are ok for reaction time recording but pointless for trigger outputs.

Expresscard slots are more useful. You can purchase expresscard parallel adaptors, and if you purchase a good brand these will use PCIe rather than USB bus connectivity, and are much faster and give you pretty reliable data. At least, as reliable as a PCI parallel card in the back of a desktop PC. We managed to purchase one of these for our presentation PC, so that solved the issue of getting trigger data to the EEG amplifier. Unfortunately this didn't have 2 parallel sockets onboard so couldn't be used for input / output.

I toyed with the idea of creating an adaptor to do in and out over the one parallel socket, but this would have been a bit of work, and I wasn't confident that I would be able to avoid conflicts on the port. Thankfully the biosemi acquisition box had a 37 way connector designed for external triggers that I could connect into using a response box.

The response box we usually use in the lab has a serial connector, and wasn't really suitable for this (it's actually a serial mouse, they work pretty well as response pads). I also didn't want to cut it apart and break the thing. I was running out of time to find a solution (this was 1 day before leaving for India), but the electronics staff in the department of Experimental Psychology helped me out by giving me an old response box that was about to be put in the electrical waste bin. I modified the internal electronics slightly by adding some capacitors and an IC (NOT gates), and stuck in a new cable.

I don't think I was being paranoid when I packed my suitcase and included a soldering iron, solder, spare wires, capacitors, diodes and a couple of d-sub connectors. I was a little worried it would be opened up at the security desk, but they were mixed up with my spare phone, laptop charger etc.

The first job I had after getting to Calcutta was to check my response box. I hadn't made a total hash of it, but needed to resolder some connections for the IC and attach a d-sub. That was a little harder than I'd thought, as the screen from the cable needed attaching to the outer case of the d-sub, something that you normally use acid solder for, and something I'd left in Oxford. I managed to rough up the casing enough with a file to get regular solder to stick. I tested the box with the EEG kit, and after removing a couple of solder bridges on the pins it was ready to go.

 

Apparently anything that can go wrong will, and that night I was trying to get our Bengali font to work on the laptop. Eventually a copy was emailed from the UK for us. The next day started out as an experimental day, but most of the time was spent troubleshooting mismatching and overlapping characters in our visual targets, a peculiarity of the conversion of a Mac to a PC font.

The next major disaster occurred a few days later - just at the start of a new EEG recording, all of our electrodes disconnected. In my experience of the Biosemi system, this can mean that the CMS or DRL electrodes are not attached, but this proved not to be the case. Fortunately we had a backup set of electrodes to use. When the same electrode set failed another 2 times with the same fault, I figured there was something up. This type of fault, that presents itself with the "CM in range" light flashing, can be caused by a short, that triggers the safety circuitry inside the amplifier. I checked the ribbon connectors, and just to be sure I cleaned them first with distilled water, then ethanol, then spent a good while drying them under the airconditioner. The problem still occurred, so I dunked the electrodes in a bucket of saltwater to see if any cable was obviously damaged, but couldn't find any faults. I figured it was probably a broken wire inside one of the cables. Biosemi were very good and offered to replace the set under warranty, but as time was of the essence, we also ordered a new set of electrodes to be brought with our PhD student on the next flight.

We started getting a couple of weird responses from our button box. Specifically after one button was pressed, the other would appear to have been pressed shortly afterwards as well. Nothing that would screw up the data, but still annoying. I took the response box apart and realised something I should have thought about earlier - the response box was made from an EMF shielded case, so it had been coated on the inside with a conductive paint to ground any signals radiating out. This wasn't great for the exposed wires I'd put in there, but easy to fix, I covered the whole insides with electrical tape.

For the speech production experiment, we tried to keep connections as simple as possible, but in reality there's only so much that can be done. Briefly, the laptop presented audio, which went through a y-splitter to both one channel on a voice recorder and the participant's headset. The microphone on the headset fed into the second channel on the voice recorder. I'd ordered a headset before leaving for India, it hadn't arrived so the PhD student was to bring it. When she arrived in India I realised that the headset had an "easyconnect" plug instead of regular 3.5mm jacks, so was pretty useless. I formulated a plan to send the headset back after the trip, especially as this headset cost around £90. We went to the local shopping mall and bought a new headset with the appropriate adaptor.

The first problem with the recording setup was noise when the y-splitter moved. This was caused by the splitter jack not being quite long enough for the mac laptop headphones out socket. This was replaced. There were still noise issues, so the headset and cabling received a liberal coating of electrical tape.

Strange noise problems persisted, and this time noise was traced to the headphones themselves, pressing on the outside of one of the earphones created electrical noise, possibly from the wiring or the speaker assembly inside the earphone, that propagated back along the wire, through the y-splitter and into the voice recorder. Perhaps this could have been fixed by creating a new y-splitter with diodes in the lines, but I wasn't so confident. Instead I decided to butcher the expensive headset, cut the connector off the end and solder on some new 3.5 mm jacks. This worked surprisingly well, although required a little more electrical tape.

The speech production setup was put into place in a classroom, no air-conditioning, only ceiling fans, the low background wind noise meaning that the fans could only be turned on between blocks of trials. This being Kolkata, by this point the average daytime temperature was around 38 degrees C. Now a new noise presented itself - a very high amplitude buzz on the voice recorder, that only presented itself during the breaks. After a lot of diagnosing, it turned out that this noise was related to the fan controller - when the fan got up to speed it would start, and stop when the fan was switched off! The fan controller was creating electrical noise on the ring main, which fed the wall sockets supplying power to both the laptop and the voice recorder. Fortunately by this time we were able to finish the EEG experiments and move the setup into a room with air-conditioning, as running any experiment in 38 degree heat is not pleasant!

But, considering all of these problems, nothing was insurmountable. We managed to record all the data we needed, and preliminary analyses are looking promising. I think we have learned from our experience, but not in a way where I would change much of our setup or preparation.

The take-home message from this is if you're planning on running experimental fieldwork, of course you should plan ahead as much as possible and prepare for the worst (for example, by taking spares of everything you can think of). But it is just as important to be flexible, as something is bound to go wrong in a new and unexpected way.