MSP430F5328: MSP430F5328 A/D Cross Talk

Hi Tom,

I would expect a very small amount of crosstalk between ADC channels, but seeing nearly the same signal on a different ADC channel would definitely be concerning.

I suspect that it's an easy fix though. For each Port 6 analog input here, have you made sure to set the CBPD.x bit to disable the output driver and input Schmitt trigger to prevent parasitic cross currents when applying analog signals? Footnote 1 under Table 6-52 in the MSP430F5328 datasheet mentions this. Try making this change and let me know if the issue goes away.

Regards,

James

Hi James,

Thank you for the suggestion. Unfortunately this did not correct the issue. Any other ideas?

Tom

I found that if I do not collect data on the first channel the cross talk on the second channel is almost gone. As soon as I include the first channel the cross talk returns.

Part Number: MSP430F5328

We have an issue with A/D crosstalk. This can not be the way the system works. It has to be something we are doing wrong but we can not seem to figure it out. 

We are measuring up to 3 different amplifier outputs.One amplifier has a DC coupled input and the other 2 have an AC coupled input, there are 2mf cap in series with the inputs to these amps. If I measure the DC coupled amp the readings are clean and stable. When we add either one of the AC coupled amps the readings on the DC coupled A/D counts change. When we measure the AC coupled amps there is about a 5-6 second charging of its output and when measuring the DC signal at the same time we can see this charging on its signal as well.

We have measured the analog output of the DC amp in all configurations and the voltage does not change or display this charging.

It was suggested we follow settings in Table 6-52 of the data sheet and we did without any change to the outcome.

Does anyone have suggestions or code they could share that measures multiple A/D channels without crosstalk.

Hi Tom

Could you measure the power supply of the MSP430 when the AC coupled?

How about increase the ADC's sample hold time by the ADC12SHT0x bits?

Could you provide the schematic of you circuit?

Best regards

Gary

That bit about being AC coupled concerns me. Especially without a schematic to look at. Opamps must always have a DC path to a power rail on their inputs or the bias currents will do things you don't like.

Hi Gary,

Below is the schematic as well as two screen shots of the signals. The scale on the left is actual A/D raw counts.

The supply to the MSP is from a 3 volt battery. The voltage to the processor is the same (meassured) The voltage from the 2.5 volt regulator is confirmed at 2.5 volts. The regulator is also the REF voltage to the processor. It turns on to stable in about 200 microseconds when any of the amps are turned on. The last two screen shots show the two AC coupled amp outputs, the bottom signal is the DC coupled output. You  can see the DC signal moving up as the AC amps are charging. When meassuring just the DC amp (last Screen shot) the base line is different from the baseline when including the AC amps. You will also see that the DC signal when meassured alone come upto a stable level very quickly in the last screen shot.

Thank you for looking into this for us.

Tom

Hi David,

I posted the schematic and some comments. 

Thank you for your help.

Tom

Call me old fashioned but I like the standard opamp symbol for circuits rather than a package outline. The outline makes it much harder to see what is going on. But I have questions:

1) Do the sensors have a ground reference? They must as every bridge sensor I have seen is a four wire device. Also in the category of things unseen is the ground system. I see only one symbol used but the ADC front end should be on AVss and AVcc. Some filters to isolate those from the digital side would also be nice.

2) When you say that only the DC circuit was measured, exactly what does that mean. Was only the ADC configuration changed? Was power to the AC coupled circuits enabled?

The design of the ADC input mux makes it very unlikely that the mux is the cause of your trouble. More likely is a fault in the analiog front end. The time constant of the problem appears to match the time constant of the voltage reference circuit (why not the ref included in the MAX part?) so look there.

Hi David,

Thank you for the reply.

1) Yes the connections to the amp is output of a full bridge. The power and ground to the bridge comes from the voltage regulator for all sections, bridge, amps, amp references and processor REF. The grounds are all connected on an internal ground plane on the board. The board itself is only about 1/2 inch in diameter.

2) When measuring the output of the DC amp through the A/D. The other amps are not turned on and the A/D configuration is changed accordingly. When we measure the analog voltage directly from the DC amp we do not see any of the charging using an oscilloscope. We only see the charging in the A/D readings from the DC amp when we turn on the AC amps and read them through the A/D.

We measured the output of the voltage reference (voltage regulator) and this is very stable within about 200 microseconds. The charging effect on the readings is about 5 seconds.

We noticed today that if we move the DC amp to the first channel of the A/D from the second the cross talk is diminished considerably but does not disappear.

Tom

The reason you don't see the voltage ramp when the AC coupled amps are not powered is because the AMP_BIAS also gets its power from that switched output. This is the long time constant signal that I am referring to.

Because you have RC low pass filters between your opamps and the ADC input you do have to worry about things that don't happen with nicely buffered signals. This works well at low sample rates where the capacitor voltage has plenty of time to recover from the charge transfer between samples. You don't lose a lot of charge per sample to that 20pF sample and hold capacitor but it can add up at high sample rates. The amount of charge transferred is effected by any residual charge on the sample and hold capacitor.

Hi David,

I am sorry for not mentioning that in the test circuit we removed the caps, C24,C22,C27 and C30. The only cap that is still in the circuit is C26. We did this to take this out of the equation on the DC amp.

The sample rate we are using is 250 samples/sec for the most part.

The strange thing is that the same time constant of charging occurs in the DC measurement but only when the AC measurements are taken. This is why I think the cross talk is the problem in the A/D part. We do not see this charging at any time when we measure the analog voltage at the output of the DC amplifier with a scope.

As I mention before when we changed the order signals of the A/D channels the crosstalk gets diminished but not gone, another reason we think it has to do with the A/D section.

We are taking 2 A/D readings per meassurment and averaging, do you think this could have anything to do with the issue?

Tom

More things you left out.

I don't see C14, C18, or C30 on that list and those are the caps on the ADC inputs. I still suspect an interaction between those capacitors and the sample and hold capacitance. Short out the R in those RC filters to verify. If the crosstalk goes away you have to do something to buffer your signals.

People mean a lot of different things when they say they are averaging two samples. Uniformly sampled (equal times between samples) is always the best way to go. But it will not cause your particular trouble.

Hi David,

Yes again sorry C30 is also out of the circuit. 

So you are thinking that when measuring the AC amps the C14 and C18 are inserted into the equation and might be the cause of this.

I will remove the C14 and C18 as well as short out the R's to see if that does the trick.

Tom

Hi David,

Removing C14 and C18 did not do much.

There is virtually no cross talk to speak of any longer after we moved the A/D channels around so that the DC amp is measured first in all cases. 

Now the only problem is that when we measure the DC amp by itself the base line is 1338 counts and when we add one of the AC amps we get counts of 1370 for the DC signal and finally when we add the second AC amp we get 1388 counts for the DC signal. The counts are changing depending on how many channels we include.

We are turning on the DC amp for about 200 microseconds every 2msec and taking a measurement about 140 microseconds after the amp is turned on. I can see when the A/D is sampling because we have an output pulse generated on an output pin from the processor when it is triggered. This pulse does not change its position at all on the output signal of the DC amp when measuring both the pulse and signal with an oscilloscope. The point of measurement based on this pulse is the same no matter how many amps are turned on.

Any Ideas?

Tom

By the way we do not turn the AC amp on and off like the DC amp. These amps are turned on at 100% duty cycle when they used, they use very little current compared to the DC circuit which is why we duty cycle the DC amps to save battery power.

I am really puzzled by this cycling of power to the DC amp. It has a RC low pass filter on its output that has a time constant much longer than the time that you are enabling power. How does that work?