How to remove ghosting totally? NI USB-6363

[Rex0725]

Thank you in advance. Right now, i am using NI6363-BNC to measure three phase voltage signals. I meet the problem of ghosting effect. I already read your page about ghosting effect. I wonder how to decide whether i remove ghosting effect or not. Even the voltage error is 1mV, it will lead to a wrong result. I try to use AI0(input 5V signal) alone, and it gives me 5.005V when i use AI0(input 5V signal) and AI1(input 5V signal) at the same time. It gives me 5.001V. If 5.005V is correct, does it mean ghosting effect still exist? 

[Henrik_Volkers · Accepted solution]

The only way to remove ghosting is to use simultanious sampling 🙂

Low impdance driver for each input and the lowest possible samplerate are the way with multiplexed DAQs.

Have a look at the spec sheet at the settle error chart.

[LV_Warrior]

The story with the time constant of the multiplexer capacitance / source impedance combination is only part of the story: https://knowledge.ni.com/KnowledgeArticleDetails?id=kA00Z0000019KzzSAE

What is missing explicitely in the datasheet of the USB 6363, is that it probably has a sigma delta AD-converter. This type of needs more time for higher accuracy. The datasheet does present a chart showing a relationship which may include the effects of the sigma delta topology but, the source impedances hint to the multiplexer story. It shows accuracy improving with about 60dB/Decade.

The chart starts at 1us. If however you were trying to sample two channels at 1MSPS, there is only 500ns for each channel. Which is left of the chart.

The huge 2,5% error that you get than, can no longer be expained by the multiplexer time constant.