ai noise influenced by graphics

[Dekay]

I would like to find out if there is a configuration (PC or programming) change that will reduce noise in analog voltage measurements.

I noticed that dragging or opening windows when AI is in progress will change the measured voltage slightly.  The attached file shows a graph and notes, including the changes in voltage when a window is maximized/minimized or dragged. The voltage changes by about 20 microvolts, which is much more than the spec sensitivity of 6 microvolts for the 0.1V scale (16 bits). The noise level should be much less than this, due to extensive averaging and filtering. The measurements are recorded at 200KS/sec, 16667 points averaged per point x 20 for every point plotted. The program is also attached.  The noise level isn't too critical for this program, but is for a similar measurement. Any suggestions would be appreciated.

The DAQ board is PICe6353. Similar results with PCIe6251. The graphics card is also PCIe.

Windows 7, 64bit, 4GB memory, Labview 2012.

[johnsold]

I have seen things like that but not since the early days of CRT monitors. And one relatively modern oscilloscope from a well-known manufacturer. I have also seen equipment which had cards plugged into supposedly interchangeable slots but which would only work correctly with certain boards in particular slots. (That was not a PCI or even a modern computer system).

In both cases there was coupling between the actual electrical signals of the display and the the sensitive analog measurement.  In the case of the oscilloscope the problem was internal. The only fix was to replace the scope with a different brand. For the CRT monitor situation moving the monitor farther from the measurement system and routing some cables to be less parallel solved the problem.

For your case I would try moving things around - monitor, cables and so on. Also try (if possible) moving the cards to different PCI slots. 

The other thing I suggest is very carefully examining the ground returns for the signals you are trying to measure. Your signal behavior looks a lot like coupling through common conductors (which usually is ground).

Lynn