A signal being contaminated by another one

[kenkar]

I have 7 analog signals to be sampled. Four of them are amplified thermocouple signals; only 3 of which need to have their waveforms sampled. I only need one point at the start of data acquisition for the fourth one for cold junction compensation (CJC). One channel is from a slotted switch giving a grounded TTL pulse of 5 V in every revolution of the drive shaft. The last 2 channels are grounded signals from hot wire anemometry electronics.

I am using BNC-2090 rack-mount BNC accessory with PCI-MIO-16E-1. To get the best noise rejection, I have configured the BNC-2090 and MIO board to DIFF mode.

All channels were defined in MAX Explorer with 0-10 V limits, i.e. no gain, except the one for CJC whose output was con
ditioned to 10 mV/?C, so I applied a linear scaling to it.

I used the intermediate VI to acquire the multiple waveforms because it allowed me to set the interchannel delay. My VI was based on the simple-buffered analog input with a write to spreadsheet file example in figure 7-8 in LabVIEW Data Acquisition Basics Manual (see file attached).

I was sampling at 5 kHz to acquire 500 samples with interchannel delay set to auto (-1). I found that one of the thermocouple channels had got an unusual step, which coincided with the switching of the slotted switch signal (from low to high and then high to low). I had tried to increase the interchannel delay, the step on the thermocouple signal did reduced in magnitude. The strange thing was that even I increased the interchannel delay to 50 microseconds (the maximum allowed without getting an error from LabVIEW), the step still existed. The MIO board was supposed to settle to full-scale step within 3 microseconds with +/- 0.012% accuracy. H
ow could my situation be explained? What should I do to solve the problem? The chart and the acquired data in an Excel spreadsheet are attached for your reference. All suggestions are welcomed. Thank you.

[mikeporter]

I don't think this is a matter of interchannel delay. It looks to me like a case of good ol' fashioned crosstalk. The signal TC2 is the most dramatic, but the signal RJT also has a clock being impressed on it from somewhere.

Mike...

[Rob Cole]

I think that Mike is right. This looks like crosstalk. Some things that you can try... Move the signals as far apart as possible (channel wise) in the I/O board and ground all of the unused lines in between. Too often, the traces on I/O boards are run in parallel, so having grounded traces between signal traces can reduce crosstalk. Filtering the signals can help as well. Reject frequencies outside the required one.

Keep us informed of what you find and how you fix it.

Rob

[kenkar]

Thanks, Rob. One thing I did not mention in my post was that voltage levels of the thermocouple signals (TC2, TC3 & TC5) were around 0.37 V whereas the SW signal was at 5 V, i.e. TC2 was much smaller. than SW. Does it mean a combination of a weak signal and a strong signal would increase the likelihood of crosstalk? If the SW signal was much weaker, would the crosstalk be completely eliminated?

[Rob Cole]

You would not completely eliminate the crosstalk. With a stronger signal (from your TC's), you increase your signal-to-noise ratio. While the noise doesn't change, it becomes a smaller percentage of the signal (and looks smaller on the graph).

Reducing the noise-causing signal has the same effect. At some point, the noise that it generates would become unmeasurable. There is also a point at which the noise being generated cannot get past the insulating material. Ground lines and ground planes can be used as insulators (the ground will absorb some of the signal). The power of the noise is also governed by the inverse square law for distance (doubling the distance would give a signal with 1/4 of the power). So increasing the distance between si
gnals will increase the signal to noise ratio.

Hmmm, this is beginning to sound like a lecture, but there it is.

Rob

[kenkar]

Thanks again, Rob. I was trying to ground all the unused lines in between as you suggested.

I am using the BNC-2090 rack mount with the I/O card. The analog inputs are numbered from 0 to 15. If I want to move the signals as far apart as possible, do I connect the SW channel (the one which is causing crosstalk) to say ch0 and have the affected channel (TC2) to ch15. I am worring that the I/O board "sees" the channels continusouly, so that the channels next to ch 15 are ch14 and ch0. What do you think?

Does the channel order in LabVIEW VI program, more specifically the order to AI Config.VI matter? Do I need to sample SW channel first and sample TC2 channel last?

[Rob Cole]

I do not know the BNC-2090, but what you are trying to do is to get physical seperation between the signals. Where they are on breakout, the board and where the wires are in the cabling are what matters. What I would do is try it.

If this is a physical crosstalk situation, then the sampling order does not matter. If you are reading thermocouples, then the reading should be close to steady state (or changing very slowly). You could take a number of readings (stored in an array) and pass them though a low-pass filter. This should get rid of the sine wave.

Rob

[kenkar]

Thanks, Rob. I think I have solved the problem. Please read the following post. http://exchange.ni.com/servlet/Redirect?id=5468250