TMS570LS0714: correct way to disconnect ADC channel from ADC module

The ADC inputs are connected to an analog multiplexor. The only time an ADC input is connected to the sample capacitor is during the sample time at the start of a conversion of that channel. There is always leakage current with the analog multiplexor. Why are you trying to disconnect AD1IN[0]?

I am seeing incorrect ADC readings on AD1IN[1] when I have a resistor to ground on AD1IN[0]. If I remove the resistor to ground on AD1IN[0] or lift the pin of AD1IN[0], I get clean ADC readings on AD1IN[1]. When I try to disable sampling on AD1IN[0] I still see incorrect ADC readings on AD1IN[1] (not sure if I'm doing this right in firmware). I would expect to get the same effect as lifting the pin as disabling the AD1IN[0] channel in firmware but that is not what I am observing. I would like some guidance on how to properly disable AD1IN[0] to verify. How much leakage should I expect on the AD1IN[0] pin?

Thanks,

You probably need to increase the sample time for your conversion on AS1IN[1]. 44K is a high source resistance. What is your current sample time?

currently 7 usec, what would be recommended given a 44k source impedance. Is it expected that even in disabled, there would be an effect from one ADC channel to the other ?

As a rule of thumb I use 1uS/KOhm. You have about 30pF of internal capacitance that needs to be charged. There is a default connection to channel 0 that does affect the first channel in the conversion group if the sample time is insufficient. Adding a sample cap discharge time eliminates the channel to channel dependence, but the converted value will still be off if the sample time is insufficient. I suggest you try increasing the sample time until you see the value converted match the expected value.

Bob,

is the default connection only affect the first channel in the conversion group? For example, if I do

Channel 0
Channel 2
Channel 1

Would the default connection be between 0 and 2 and not 0 and 1?

Most of the readings are accurate, but every so often, there are single point spikes in the data. Can this be explained by the high source impedance and in-adequate sampling time? I would expect a fixed offset and not drooping single point spikes that are orders of magnitude slower than the sampling period.

In the case of converting channels 0, 2, and 1 (in that order) if the sample time is insufficient, the value on channel 0, will affect the results on channel 2 and the value on channel 2 will affect the results on channel 1. If you add a sample capacitor discharge time, the result is independent of the previous channel sampled, but will be low if insufficient sample time is given. The simple explanation is that the sample capacitor is at the voltage of the previous sample taken and must charge or discharge to the new voltage. The higher the source impedance, the longer it takes for the capacitor to equalize at the new voltage.

I suspect the noise spikes on your graph of samples are caused by an induced voltage from nearby circuits. The high source impedance makes it easy for surrounding noisy circuits to offset the voltage you are trying to convert. The spikes can be reduced by decreasing the source impedance. If the signal changes slowly you can add external capacitance. Increasing sample time will only average out the noise if it is of relatively short duration. Careful layout of the analog signals is the best solution.