feedback on input channel of USB-6363

[terry92054]

I'm new to using the NI USB-6363 but this behavior just doesn't make sense to me.

Symptoms

Connect a 2v peek to peek sign wave to AI1.

Use NI Measurement & automation tool to create a voltage input task (Create New NI-DAQmx Task)

Run the task to verify input is displaying as expected.

Add another voltage to the task from AI0.

Run the task again 

- AI0 now has a sine wave too but nothing is attached to it ...

Using a separate scope verifies that AI0 now has a sine wave

Configuration is very simple.

Signal going to pinout 4.

Ground going to pinout 3.

Chassis ground going to grounding lug on USB-6363.

NI task - continuous samples, samples:500, rate: 1k, terminal configuration: RSE

This behavior was noticed when setting up two inputs, one each on AI0 and AI1 (one with about 2v sine wave the other with about 20 milivolts of noise. The LabView program appeared to be combining them so I used the Measurement & Automation tool and found the results described above.

How can I be getting output on a dedicated input pinout?

[johnsold]

This is called crosstalk.  The input capacitance of the multiplexer can charge to voltages on adjacent channels if not driven from a low impedance source.  If you connect a low impedance signal to AI0 (even just wiring it to ground), the sine signal from AI1 should not appear.

All inputs which are used should be driven from low impedance sources and all unused inputs should be connected to ground or some other voltage source within the input common mode range.

Lynn

[terry92054]

Thank you for your reply Lynn. One detail I failed to note in my posting was that I only see the feedback (crosstalk) when I run either my labview program or the NI Measurement & Automation tool (as confirmed by my external scope.) Of course, when the software is running, the USB interface gets energized as well but I'm not sure I understand why I only see this when the software is running, acquiring data from the two input channels.

Regarding the impedance; the 1-2vac signal is from an accelerometer and the 20mv signal is from an ultrasonic transducer. Both are processed but I believe they are still low impedance. Need to confirm.

Bottom line, I can't afford to have the 20mv signal overridden by the 1-2v accelerometer input.

I'm going to try moving the accelerometer to a more distant pinout but I don't expect that is going to change anything.

Thanks again for your input!

Terry

[John_P1]

Hi Terry,

The proximity of the input pins isn't going to make a significant impact.  For more information on ghosting, see this article:

How Do I Eliminate Ghosting From My Measurements?

The 6363 is multiplexed, meaning that all of the channels in your scan list are fed sequentially into a single ADC.  The ADC itself has a settling time, as the input capacitance must have enough time to charge or discharge.  The settling time is dependant on your source impedance since the charging/discharging mostly occurs through the signal source (the input impedance is over 10 GOhms on the 6363 so not much current will flow through the ADC path).  The image from my previous link I think shows it the best:

So, if you have an open channel, the charge stored up on the capacitor doesn't have anywhere to go after the MUX makes its switch and so you'll read the same value as what was on the channel before the switch occurred.

To prevent this from affecting your measurements, you'll need to acccount for the settling time of the device which can be found in the specifications in the form of the following chart:

The error you will see is a function of:

1. step size -- you can't control this too much unless you know ahead of time what you will be measuring and you arrange your scan list in such a way to minimize voltage transitions between consecutive channels.

2. time -- you can set the convert clock rate yourself in the DAQmx driver.  The time between channels is the inverse of the convert clock rate.

3. source impedance -- you can reduce this by buffering your inputs if they are not already low impedance.  An open channel can theoretically be thought of as "infinite" source impedance.

Best Regards,

[johnsold]

Terry,

One trick which may help: Set up your task to read twice as many channels as the number of signals.  In your case you have two signals. So read four channels.  Put you signals on channels 0 and 2 and gorund the inputs to channels 1 and 3 (and the rest of the unused ones).  When you read the data, simply discard channels 1 and 3.  By doing this you allow the input capacitance to discharge on the grounded channels. 

Of course this can only work if you have enough empty channels and enough time to read the extra channels.  It seems likely that in your case it could work.

If you add buffer amplifiers as John P suggested, put some gain on the transducer signal so that both signals are about the same magnitude.  Then even if a bit of crosstalk or ghosting occurs, it may be only a small effect on a large signal.

Lynn

[terry92054]

Lynn and John,

Thank you for some great insight, suggestions, and visual aids. I have a much better understanding of what is going on now and some techniques to test.

Thanks again,

Terry

[terry92054]

Are the analog input channels read round-robin based on their position (e.g. AI0, AI1, AI2, ...) or based on the order they get defined within the software?

Thanks,

Terry

[John_P1]

The scan order is based on the order that the channels are defined within software (true for almost all multiplexed NI products, including the 6363).

Best Regards,

[terry92054]

We put one grounded channel between the accelerometer input and the sensor input and no longer see the ghosting on the sensor channel. We tested with one accelerometer and two sensors and didn't see ghosting between any of the three. While we plan to eventually have 5 or more sensors, this configuration is sufficient for us to proceed with our v1 prototype.

Thank you for your assistnace and suggestions,

Terry