I had a little unannounced demo today. I had bits working, and it seemed to make a relatively good impression.
As it turned out, the alternate blinking is indeed problematic, as Windows on the computer I got has severe scheduling problems. I wish I hadn’t worked on such a tight schedule, so I could re-install it (with Linux, of course).
I got an other computer today: it hasa Quadro 600 GPU, some reasonable amount of RAM and CPU. However, I found a bug in the display driver: when you select an ‘unsupported’ resolution, the display refuses to work after a reboot. The resolution I set was supported by my monitor and worked – until I had to reboot. The only way to restore display is to boot in safe mode and uninstall the driver. I don’t know where the per-display settings were stored (as opposed to /etc/X11/xorg.conf or …/xorg.cond.d/*), so I got one more shot.
Thankfully, using the latest, driver solves this problem.
There is a 64-bit version of Psychtoolbox, and I am installing it right now. As of this moment, I am in a dependency hell because of that. But not for long!
There is an old Hungarian saying: ‘A puding próbája az evés’. That is, the only way to see if your pudding (as in dish) is any good, is to eat it. Well, we had our very long, very detailed pudding (as in dessert) today, and it was great! And believe me, we went down to the bottom!
Finally got the gadget to work. Tested against the results I managed to get earlier, and it seems to hold.
Since I am just about to finish this project (well, at least, time is up), it’s worth taking a look back: After fixing some experimental design issues, I got some wacky results. “Now that’s strange” I thought, and remained quite sceptical. It seemed so unrealistic, as the hypothesis was so straightforward that I was convincing that I will get the boring results everyone expected.
The results were nothing like it when I make the processing routines of course. What do you do in that case? You immediately assume that you have made a mistake at some point, and you have a list of suspected things you think might be held responsible. However, the results held against my every attempt to prove them inconsistent/wrong, and believe me, Jenny’s ‘vicious reviewer’ comments and my critical thinking made us to examine things thoroughly and systematically eliminate every factor that might have explained what we have seen or measured.
Today, I can say, that this three-week-long full-blown crusade against errors, mistakes and misinterpretations has failed. Going through thousands of measurements, quite a few display devices and even creating and using a purpose-built visual device (I couldn’t find an other name for it) have failed to do any harm to the credibility of the results.
end of line
After some fiddling with the code, the circuit works. Delays seem to be fine. Results seem to be fine. Found out how to control the serial port. I should upload some shell scripts to make life easier when using a campus pc. I managed to cover up the effect of the earth loops, so the circuit is stabilised properly.
Now, it’s on to matlab, and will need to spend time thinking about the implementation.
Things now slow down a bit. I expected much more rapid progress with the micro. There were some obstacles that should not have been obstacles: The default reset value for an output pin is analogue input (took a good 3-4 hours to find out). My new USB/TTL UART doesn’t have the CTS/RTS pins wired internally to the power supply, so I can’t bypass flow control that easily (2-3 hours). The battery lasts around 5 hours in the oscilloscope (2×1 hour break). The poor thing sometimes re-sets because of interference. This will be solved as soon as I get rid of all the earth loops caused by the programmer, the UART and the oscilloscope and power supply.
Good news is that current-mode PWM powered LEDs are very closely linear. I managed to find my components in the morning (well, the components found me, actually…)
The moral of today: One measurement is not a measurement. Practically the only two things I can do is to build the stimulator circuit and get an other photometer to make sure it’s not a calibration artifact. Then, all I can do is wash my hands how Pontius Pilate did. Then, I have locked out all possibility to have a misconducted measurement, and then, I can say is that my measurement is indeed correct no matter how wacky and unrealistic it seems. So the question is now if I have something publishable or more modest. Two weeks to go.
The weird result I am seeing is not caused by the monitor. Also, it is not caused by the different refresh rate and stabilised plate current. I am getting the same things over and over again, and the test show that there is a significant deviation to what I was expecting. I can’t think of any other thing I might have missed, and the results are pretty consistent. I currently have 10 matlab figures open and can see the same pattern over and over again. Okay, the coefficients are different, but the other monitor is brighter, smaller and behaves differently.
To do an other verification, I will make a circuit, and will check it with LEDs as well. This tendency should vanish.
Do you know what is the answer to the ultimate question of life, the universe and everything? If you say 42, well, you are at least as much of a nerd as I am.
I think I found today the psychophysics equivalent of it: 1.4142136. I don’t want to jump into conclusions, so I am doing extra verification, replacing displays, redoing experiments, so I can be a hundred zillion percent sure that it’s not there because I forgot something. Guess I will find out in the next few days.
Also, I had my arm zapped for a good 20 minutes. Georgia was astonished how fast I found the nerve in my arm to test a reflex. Weird.
Spent pretty much all day playing with Matlab. I have my experiment done (it works, verified), I also have my initial result processing script done (which also works, I verified it). What I don’t understand is why things go haywire when I sort the final array. As usual, I sort by the first column, but this time it seems that it doesn’t keep the rows. Anyway, I could get around it eventually, but it’s strange…
I also think I found an artifact in the results. Will need to replace the monitor to see if it’s real.
Got more results today. I see people getting bored with the experiment. For instance, my arm will be zapped for two hours in exchange for an experiment. But also it includes the more generous spreading of chocolate, teas, etc.
I have a theory on how we should mathematise the findings, we’ll see if it stands…