Alias problem on 5112 scope digitizer

[krolt]

I am involved with acoustic measurements.  My company is implementing a new test measurement system for in-air and underwater acoustic (sonar) measurements and calibration.  We are using NI PCI-5112 scope cards (digitizers), and we have two of them.

When we measure ambient underwater noise in our indoor tanks, and while using a reference hydrophone (10 Hz to 400 kHz) we observe a strong tone at 195.3 kHz.   We have observed the same tone on both 5112 cards we own, and we have also observed the same when we install the cards into a different PC (a different model) elsewhere in the building.  That suggested that it was not noise from the computer (PCs), but from either the measurement or from the scope card itself.

We suspected the measurement and that we needed an analog anti-aliasing filter between the hydrophone and the 5112.  However we observed the same tone even when we disconect the BNC coax (from the hydrophone) to the back of the scope card (i.e. using no input to the scope).

The tone amplitude can be reduced if the sampling frequency fs is set from 100 MHz to (say) 97 MHz, but it never really goes away.  That suggested that the tone was an internal problem to the scope card and not to the DUT.

We used the canned Labview VI for showing the FFT spectrum in all cases, so that was how we observed it.  A Matlab code version was also used and implemented in parallel with the FFT VI  to see if the tone was still visible --- it was.   Finally, the on-board internal low-pass filter (10 MHz) was enabled and applied to the same data sample using both the Labview FFT spectrum VI as well as the home-grown Matlab version.  The results we similar: the tone was always present.

This is a real problem.  We measure products from kHz range to several MHz, so a ~200 kHz tone is more than just a nuisance.  I am wondering if there is a problem with the scope cards, especially since it affects both of the 5112 cards in the same way.  And for the record, there are very few hours use on the scope cards.  They were purchased 1 to 2 years ago but not used until the last few months.

[Matt_E.]

Hi Krolt,

The spur you are seeing is most likely from a switching power supply on the digitizer board itself.  Since the spur is falling in the frequency range you are trying to look at, you will be very limited as far as what you can do to filter it out.  Looking at a 5112 I have it looks like this spur is around -80dBFS with a 200mVpp range setting and -65dB with a 25mVpp range.  This tells means that the spur is getting into the digitizer's front end signal chain before the gain amplifier.  I am guessing that since you said the spur is strong you are probably using a fairly small range on the digitizer which in turn uses a high gain setting inside the board to amplify your signal to an appropriate level for the ADC.  One thing you could try is using an external amplifier to condition your sensor's signal to a range that uses a lower gain inside the digitizer, which in turn will not amplify the power supply spur.  A quick test of my board showed that it looks like the 1Vpp range brought this particular spur down to about -87dBFS which is below the noise floor of the FFT unless you do a lot of averaging so hopefully you wouldn't even see the spur.

Normally digitizers that focus on high performance in the kHz region will try to locate power supply switching frequencies outside the band of interest so that the noise from these supplies can easily be filtered out.  Since the 5112 is a wideband (100MHz) digitizer this isn't really practical to do, so greater attention is usually spent trying to keep the overall spectrum as clean as possible.  Some system spurs are usually visible on wideband products like this so generally speaking if you want the best performance possible you should get a digitizer that is optimized for the bandwidth you are interested in.

Hope this helps

-Matt