Quantifying ADC Channel Crosstalk

Dont know if this is of interest to others, but it answers a question from a previous thread on interfacing to the Panda II ADC inputs, and I could not find similar data in searching the forum.

Several forum threads and the LPC 2387 datasheet point out that an over-driven ADC input, i.e. greater than 3.3 volts, can affect the conversion accuracy of adjacent channels, but the effect was not quantified. This left some uncertainty about just how big a problem this is if you have a signal that is only slightly above the max of 3.3 volts.

Since my application needed to quantify this affect I took some measurements that give some idea of what happens to adjacent channels when one is over driven. The data was taken using two batteries and voltage dividers to provide a 2.4 to 3.8 volt level on channel AD0.0 and a constant 1.6 volt level on channel AD0.1.

As the AD0.0 voltage was increased up to the 3.3 volt max there was very little error introduced on the AD0.1 channel – a slight rise, but less than +4%. However, when the voltage was increased to 3.8 volts the error in the adjacent channel AD0.1 increased to +25%. A plot is attached showing the effect.

I guess the take away is that if you keep below the 3.3 volt max – no worries if up to 3 % error on adjacent channel is OK. Going much above, like > 3.8 volts is problematic.

A caveat is that this is DC data and may be different from what would be measured with an AC signal.

@ idafez,

If you apply a true AC signal which would go negative your caveat turns into a disaster. However you may mean a sinusoidal waveform, which stays above 0volts :slight_smile:

Well detailed. Thanks

Davef, Right you are. My setup uses a 1.6 volt offset on an op amp input so that the signal presented to the ADC is always between 0 -3.2 volts.

Interesting. Thanks.