Solution
These glitches are expected behavior provided they fall withing the Glitch Energy Specification of the device. Please refer to your devices specifications for information on what is the expected glitch energy. For example, here is the glitch energy specification for two devices:
PCI-6221:
Glitch energy
Magnitude ...................................100 mV
Duration .......................................2.6 μs
PCI-6251:
Glitch energy at midscale transition, ±10 V range
Magnitude ...................................10 mV
Duration ......................................1 μs
On some DAQ devices, you may see the glitch energy specified in units of nanovolt seconds (nV*s) rather than a magnitude and duration.
Some devices, such as the NI 9264 and NI 9263 will not have detectable glitching (cannot be picked out from the noise). Others, such as the NI 9262 and NI 9269 as well as X Series multifunction DAQ cards do have detectable glitching.
Make sure to consider the glitch energy level when choosing a device for your application.
To reduce the impact of these glitches, you can try the following:
- You can build a low-pass deglitching filter to remove the high frequency components of the glitch. Depending on the frequency and nature of the output signal, the low frequency components of the glitch may not be filtered. See related links for more information on deglitching circuitry.
- You can build an external deglitching circuit. The foundation of this particular idea is to use sample and hold circuitry to hold the previous value while the DAC is settling on the current value.
- Since the greatest glitch energy occurs at the midpoint transition of the output range (see the note in Additional Information below), you could try to avoid letting the output level cross this transition. Usually, the midpoint transition will be when the signal crosses 0 V. For example, rather than output a +/- 1 V sine wave centered at 0 V, you could output a sine wave between 2 and 4 V. The amplitude would still be the same, but you are introducing a DC offset so that the waveform never goes below 0 V. This should acceptable as long you are connecting the signal to a differential or non-referenced floating input.