daq-card

[Hans]

Hi,

I want to aquire data with a PCI-6023 card from NI and write them with
time stamps to a file. The problem is that the coordination of time
stamp to the acquired data must be very exact.
For the realization I used a counter which I initialized at the
beginning of the acquisition to the actual system-time. I set the
samplerate for the "AI start.vi" to 10 (per second) and incremented the
counter by 1 second after the "AI read.vi" read 10 samples.
Since the PCI-card uses a 20MHz oscillator I thought that my counter had
an accuracy of 50 ns. I tested my program for several hour and noted
that after 3 hours the difference between my counter and the system-time
had grown up to 400 ms. Since I want to aquire data over a periode of 24
hours the inaccuracy at the en
d would be more the 3 seconds........

I hope this explanation of my problem is not too circumstantial, so that
somebody could tell me, want's wrong with my considerations.

thanks,
hans



--

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Johann Gimpl
Student at Graz University of Technology
http://www.sbox.tu-graz.ac.at/home/h/hons
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

[srm@viewpointusa.com]

there is an output on the AI Start.vi called actual scan rate. Every sample you acquire should use this value as the delta T value. This is the most common cause for apparent drift. your problem on the other hand comes from the PPM accuracy of the crystal itself. The generic spec for this card is .01% which is ~8 seconds over 24 hours. While you can try to "calibrate" this difference, the crystal frequency varies with temperature and your calibration will only be good for a specific, steady temperature. If you want a deterministic time reference better that .01%, you need to use an external clock reference either from a reference standard of your own or the NI 6602 (.005%) or 6608 (.0000005%)
Stu