valid Frequency Domain Analysis ?

[ProfP]

There seems to be two issues with using the MyDAC for frequency domain analysis ( Bode Plots etc. )

1) The system does not employ  Anti-Aliasing Filters . This can create major frequency fold-back problems.

2) The analog inputs are multiplexed which leads to a channel to channel phase error .. up to 180 degrees at the Nyquist frequency.

Is there any 'behind the scenes' processing done to mitigate these issues ??

Thanks in advance

[Corby_B · Accepted solution]

Hello ProfP,

Thanks for the post! Sorry to leave you hanging for so long. I have some answers for the 2 points you asked about above.

1. You are correct, the myDAQ system does not employ Anti-Aliasing Filters. The ELVISmx Bode Analyzer artificially limits the input frequency when using the myDAQ to 20 kHz. At that frequency, you

only get 10 points per period from both the AO and AI.

Anti-alias filters behind the multiplexer would slow down settling behavior, so they were not included. You can add filters in front of the multiplexers (at the AI0 and AI1 pins). I'd recommend they be buffered, though. Charge injection on the multiplexer makes high speed scanning very sensitive to source impedance. Note the effect in the Settling Time versus Different Source Impedance figures (21 and 22) of the myDAQ manual: http://www.ni.com/pdf/manuals/373060b.pdf 

2. You are correct. Two-channel acquisitions are phase-skewed by the 1/(2*scan rate) when the scan speed is faster than 40 kSps/ch. For example, if you acquire at 200 kSps/ch, the interval between consecutive measurements on AI0 and AI1 is 1/(2 * 200 kHz) = 2.5 us. Below 40 kSps/ch, the phase skew is constant at 12.5 us. This is intended to help very slow acquisition rates appear more like simultaneous acquisitions rather than keeping 180 degree phase skew unnecessarily. These are default configurations. You can adjust the conversion interval parameter through the DAQmx API if you prefer to customize the behavior.

Let us know if this is the information that you needed and thanks for using NI myDAQ!

Cheers

Corby

Academic Product Support Engineer

[ProfP]

Hello Corby,

Thanks for the reply ...

Yes, the idea of sampling signals in time ( 2 ch ) as

12                         12                     12

rather than

1             2             1            2           1             2

does Greatly reduce the ch-ch time skew ...  I was wondering if NI did that     🙂

For aliasing  'elimination'  at lower frequencies ( audio and vibration ) and lower sample rates, an idea could be to always sample the data at 200 KHz.  Digitally Lo-Pass the data at the Nyquist freq of the desired lower sample rate and then decimate the 200KHz data to create the lower sample rate. Only signals above 100 KHz would then be a  potential fold-back issue.

Regards,

Alex P