• Alex Holcombe
Iolabs device  might be better than the Cedrus box detailed below. And there is PsychoPy code available for the Iolabs device.
The button-box we have used for our "sensorimotor synchronisation" experiments is a Cedrus RB-730 response pad. It is on loan from Nenad and the IT office. It connects to the computer via a USB port, and is controlled by the computer as a serial port (or something like that). . To work, you must install a driver from the Cedrus website here:, and surprisingly, I also had to install a USB to serial port driver called keyspanUSAdrvr from Tripplite, formerly Keyspan: . I used the driver for their USA-28x product.
I had all of the little dip switches in back in the down position.
Basic programming with the box
Oh. the below was written before I saw Jon Pierce has put code supporting the Cedrus in psychopy, but i think his code is vulnerable to the same problems
To use the button-box with a python program, we put dipswitches 1 and 2 down so the box used XID (eXperimental Interface Device) communication protocol, which provides timestamps of button press times. Then you need these lines at the start of your code:
from psychopy import serial from struct import unpack
And to initialize the box:
s = serial.Serial('/dev/tty.usbserial-FT3ERKOD', 115200)
To erase all info in the box about previous button-presses:
To reset the button-box internal clock to 0:
Critically, the box records both the press and release of the button and sends a signal in exactly the same way (at least how it's set up now). To avoid this, you need to record both separately. In the attached file that shows how we tested the precision of the box relative to the computer clock, you can see how I did this. But basically, to check if a key has been pressed or released, you check this:
if s.inWaiting() > 0: ......
To get out the data about the button press or release:
response=s.read(6) formatted_response=unpack('<cBI', response) #'<' means little-endian order, 'c' means char, 'B' means unsigned char, 'I' means unsigned int time_in_msec = formatted_response #unsigned int, long?
Example code to increment a counter of presses and button-releases:
pressEventCodes = (48,80,112,144,176,208,240) #7 buttons from left to right releaseEventCodes=(32,64,96,128,160,192,224) #7 buttons from left to right lineInCodes=(209) #dunno the others eventType = formatted_response if (eventType in pressEventCodes): pressEvents += 1; print 'button PRESSED' elif (eventType in releaseEventCodes): releaseEvents += 1; print 'button RELEASED' elif (eventType in lineInCodes): lineEvents += 1; print 'line in TRIGGERED' else: print "WARNING! unexpected buttonbox event ", eventType print "formatted_response =", formatted_response, ", eventType = ",eventType, ",time=",formatted_response
Sending an external trigger to the box
s.write('a10') #switch button box to general purpose mode i hope s.write('a11') #or switch button box to reflective mode i hope s.write('_a1'); #ask what mode you're in msg= s.read(6) # get answer of what mode it's in numWaiting= s.inWaiting(); print 'numWaiting=',numWaiting msg=s.read(numWaiting) print msg
Input/output lines have to be set so that touching the wires together will register an event in the box. Check their current status:
s.write('_a4'); time.sleep(.2) #The XID device returns _a4 followed by the bit mask. #get number of chars in the receive buffer numWaiting= s.inWaiting(); print 'numWaiting=',numWaiting #should be 4 #_a4 will be first 3 bytes, so read those msg=s.read(3); print msg #last byte is the bitmask linesBitmask=s.read(1) linesStatus=unpack('B',linesBitmask) #binary linesStatus=linesStatus #dunno why but unpack returns tuple with empty second member binary = lambda i, c = (lambda i, c: i and (c(i >> 1, c) + str(i & 1)) or ): c(i, c) #one-line function to convert to binary format print (binary(linesStatus)) #the status of each line, 1 being input, 0 output
Retrieve a line triggered event from the box. We trigger a photodetector by drawing on the screen at a known time, and have the photodetector send an event into the box.
UPDATE: regarding the below problems, it's possible they're all avoided by using the new code that's since been developed at https://github.com/cedrus-opensource/pyxid Hisham Abboud of Cedrus says "the code detects button presses and can retrieve the timestamp, but we're currently setting the timestamp to zero because there is some drift in the hardware itself. We will update the code once we know exactly what the time drift is and can compensate for it. The code is also capable of handling two devices at the same time, e.g. a response pad and an SV-1 voice key."
- Any commands sent to button box occasionally needs to be done more than once before it takes. Box unreliable in this way (although apparently not a problem with code at github link above). Sometimes I have to restart code several times before box will listen
- If send too long a stream of events (e.g. line in left on), box seems to become unresponsive
- Sam T tried to do an experiment collecting multiple responses per trial, but seemed that the Cedrus RB-730 measured multiple responses inaccurately, accumulating a timing error of roughly 500ms over a 20-tap-piloting trial.
A file summarizing our tests of the precision of the timing is here media:buttonBoxPrecisionSummary.doc
For on-line, during-experiment timing verification we trigger a photodetector by drawing on the screen at a known time, and have the photodetector trigger an event in the button box.
Another way to verify timing and precision would be with this expensive device.