Well, it certainly sounds like crosstalk to me. On this board (and most other boards with multiplexed front-ends) there are several ways for crosstalk to occur:
1) Capacitive coupling: signals are coupled by capacitance between adjacent wires, either in the cabling or on the board itself. (This could happen on any board - not just ones with multiplexers). Since it's capacitive, it's frequency-dependent, and high frequencies couple much better than low frequencies. Since you're seeing ~55 Hz, this is probably not the mechanism behind your problem.
2) Channel-to-channel settling crosstalk: signals on one channel influence the measurement on the next channel in the scan sequence. When the multiplexer switches from one signal to the next, the signal at the A/D converter does not switch instantaneously from one to the other. Due to limited bandwidth of the amplifier and the A/D converter, and capacitance in the multiplexers, there is a settling time as the signal transitions from one level to the next. At first blush it usually has the shape of an exponential function, though once that part has settled out there's usually slower effects that aren't exponential. Those slower components of the settling curve are often due to thermal transients and dielectric absorption in capacitors. In any case, if the A/D converter samples the signal before it's through settling adequately to the new level, the new signal will carry some residual influence of the previous channel. Since the curve usually decreases monotonically, waiting longer to settle decreases the amount of influence. So one thing to try is to lower the channel-to-channel sample rate within each scan - if that decreases the noise, then your problem is likely related to settling.
3) Track-and-hold feedthrough: this is not as common as the previous effect, but it can happen in some systems. While the A/D converter is busy converting a signal that's being held steady by its track-and-hold, oftentimes the multiplexer switches to the next channel to allow the amplifier and other circuitry time to settle before the next sample is taken. The resulting transient can couple a little of the signal from one channel into the measurement of the -previous- channel since the A/D converter is still working away on the previous channel. One way to test for this effect would be to tie the same signal to two (consecutive) channels instead of one, and scan three channels instead of two. Since the voltage at the A/D converter would be remain the same when the multiplexer switches to the second copy of the signal, there should be no feedthrough on the first conversion of the signal (but there still might be settling from the channel before that). I don't think this is the problem in your case, but it's worth investigating.
Hope this helps,
EBL
