Optimum sampling freq. in PCI-6220 DAQ

[student_1]

The NI PCI-6220 DAQ has a fixed anti-aliasing filter (called "small signal bandwidth" in the documentation) of 700 KHz. The DAQ itself has a maximum sampling rate of 250 KHz.

Currently, I sample a given signal (which may have lots of interferences) at 256 Hz, and then digitaly filter it and decimate it to get a final sampling rate of 16 Hz. This gives me a signal of a bandwidth up to 8 Hz, which is fine. Thus, I effectively oversample by 16, and then decimate by 16.

I am wondering that, since the anti-aliasing filter is 700 KHz, then I might be better off sampling at a much higher rate, say 200 KHz, and then filter-decimate accordingly. I believe this should give a better performance since the anti-aliasing filter will work best in this case, while if I sample at 256 Hz then I am effectively not benefiting much from it.

[johnsold]

That is not an anti-aliasing filter because its bandwidth is far beyond the Nyquist frequency for the device (125 kHz).  Most NI DAQ devices do not have internal anti-aliasing filters.

To determine whether you can avoid the use of an external anti-aliasing filter you need to tell us the bandwidth or frequencies of the highest interfering frequency component.  It would also help if you tell us the amplitudes of the largest interfering signal and of the smallest desired signal.

Lynn

[student_1]

I expect the interefering signals to be as high as up to tens of kHz, with signal-to-interferirence ratio around 50-100 dB. I would want to choose the sampling rate that can work best for my case, and then I can do any additional filtering-decimation in software. I think that a higher sampling rate (than 256 Hz) would be better.

Needless to say, I do not want any external anti-aliasing filter.

[johnsold]

Your sampling rate needs to be at least twice as high as the highest frequency interfering signal, so it sounds like >=100 kHz will be needed to meet the Nyquist criterion on all the interference.  Any interfering component which is aliased in cannot be removed by filtering or other post processing. So you need to prevent the aliasing.

The 16-bit converter of the 6220 has a theoretical dynamic range of ~96 dB. So to get the best measurement of the interfering signals you want to make sure that the signal is very near the full scale of the selected range.

Even a simple RC antialiasing filter with a time constant of ~10 ms would substantially reduce the higher frequency interference components and make you signal processing much simpler.  Why do you not want an external filter?

Lynn

[student_1]

Actually I do have an RC filter (which is inside the Lock In amplifier, from where I get my signal) set to a 20 ms time constant, so there's no need for another filter. This filter is good, but still there are wires between the Lock In and the ADC, and these can pick up some interfering signals. To mitigate these interfering signals, I thought that sampling at a much higher rate than the current one (256 Hz) should be a good idea.

[johnsold]

The output of the lockin amplifier should be low impedance, so pick up on the wires should be quite small. What are the sources of the interference?  If you are getting interference due to the wiring between the lockin and the DAQ device, you should probably be looking at better wiring practices.  How long are the wires?

Lynn

[student_1]

The wires are less than 2 meters long. While this is not so long, I still would like maximum immunity of the intererences. 

[johnsold]

Perhaps the best approach might be to use the SHC68-68-EPM shielded cable to a BNC-2110. Then use coaxial cable with BNC connectors at both ends to connect the BNC-2110 to the lockin amplifier. This will provide a complete shielded path from the amplifier to the analog input. This is likely to provide the best immunity to pick up of the interfering signals. It is recommended by NI on page A-5 of the M Series User Manual.

Lynn

[student_1]

This is exactly what I am currently doing. I was thinking there might be extra measures of protection, but looks like I'm already protected (at least to a good level). Thanks.

[johnsold]

If you are doing that and still seeing significant interference, then you need to have a close look at the entire system. Find out what generates the interference. Either you have a system which has extreme measurement conditions or you have it badly configured.

Determine if it can be shut down during the measurements. Determine if a different physical layout or wiring arrangement can reduce coupling. Determine if measuring the interfering signal directly can be used for cancelation.

These are very general guidelines, but I cannot be more specific without much more information about the overall system.

Lynn