Re : 9233 Sampling rates

[das.rana]

Hi I am a Newbie to NI Cards
The NI USB 9233  datasheet says the following
   The USB-9233 uses a method of A/D conversion known as delta-
sigma modulation. If the data rate is 25 kS/s, each ADC actually samples
its input signal at 3.2 MS/s (128 times the data rate) and produces
samples that are applied to a digital filter. This filter then expands the
data to 24 bits, rejects signal components greater than 12.5 kHz (the
Nyquist frequency), and then digitally resamples the data at the chosen
data rate of 25 kS/s.

 
Data rate (fs)
  Minimum ......................................... 2 kS/s
  Maximum ........................................ 50 kS/s


My doubt is if the input data rate is 25ks/s ie.the max frequency is 25Khz  then how can we reject frequency components greater than 12.5 Khz.
Also is the mentioned datarate  min and max of 2Ks/s and 50Ks/s  mean that the input frequency range of each channel is from 2Khz to 50Khz.
Expecting a quick reply from someone

[PelleS]

Hi,

The 9233 contains dynamic anti-aliasing filters which automatically adjusts its cut-off frequency according to the sample rate you specify. That means that if you set the sample rate to 25 kS/s then the filters adjust so that no frequencies in the signal higher than 12,5 kHz will be acquired. The low-pass filter is in place to avoid aliasing.

Since the maximum sample rate is 50 kS/s frequency components above 25 kHz will be refused. 

Thanks

[Paul_C.]

Hi das.rana,

Pelle is correct that the filters are adjusted based on the sampling rate that you specify.  I would also like to add that you are correct in stating that the sampling rate on the 9233 goes from 2KS/s to 50KS/s.  The input frequency should be half of the sampling frequency to account for the Nyquist rate (max 25KHz).  The specific data rates that can be specified by the 9233 are shown on page 16 of the manual which can be found here.  I would also like to add that there is a great article that talks more about the sigma conversion method.  

I hope this helps,
Paul C.

[das.rana]

Thank you Very much for the Quick response.

[Rafo]

According to the NI9233 manual and this forum the sampling rates could be between 2 KHz to 50 KHz. Now I have a question which is the recommendable sampling rate in order to avoid noise. I am just interested in signals with low frequencies but there is not a problem to acquire data with 50 KHz of sampling rate and then decimate to the frequency of interest, so is the same to acquire data with 2 KHz and with 50 KHz? Thanks for your answer

[Matt_A]

Hello Rafo,

I'm not sure that I completely understand your question, but let me try to provide some information regarding the sample rates of the NI 9233. As you noted, the NI 9233 has valid data rates from 2 kHz to 50 kHz. So, what exactly do you mean by "is the same to acquire data with 2 KHz and with 50 KHz?"

One difference between data rates is described briefly above by Pelle S; the cutoff frequency of the antialiasing filter will be determined by the data rate you specify. So, if you specify a data rate of 2 kHz, the cutoff frequency of the filter would be 1 kHz due to the nyquist theorem. However, if you increase your data rate to 50 kHz, the cutoff frequency of the filter would increase to 25 kHz. This idea is demonstrated by Figure 5 on page 11 of the NI 9233 Operating Instructions and Specifications, here. I have copied the image below for your reference:

The cutoff frequency is important because, depending on the frequency of the noise you are trying to reject, it may be better to sample at a lower rate so that the filter can eliminate the high frequency components. For example, if you are trying to sample a 1 kHz signal and there is noise at 10 kHz, then it would be better to sample at a frequency below 20 kHz so that the filter will eliminate the noise component at 10 kHz. I hope this information addresses your question, let us know if you have any other questions.

Message Edited by Matt A on 01-07-2008 11:39 AM

[Rafo]

Dear Matt,

I am acquiring data of accelerations of structures in order to identify the main dynamics characteristics of them. This structures have tipically very low frequencies (in a range of 2Hz to 100Hz). I am acquiring data with the NI 9233  at a sampling rate of 2 kHz and then decimating the result until 100Hz. My question is if maybe I can improve  my results (I have e lot of noise in the signals) increasing the sampling rate for example until  20 kHz or more. Do the noise (could be electricity noise or whatever) is reduced with this process?

Another point in the example you sent  you say that if I am trying to sample at 1kHz and there is a noise at 10KHz it will be better to sample over 20kHz. Is it not better to sample at 2kHz because with the filter I will not consider signals with higher frequencies than 1kHz?

Thank you very much for your answer once again

[Wilfried_71]

The 9233 has AC coupling inputs, so it is not make for low freq signals because usually, sounds and vibrations are dynamic signals ...

[Matt_A]

Hello Rafo,

You could use a technique called dithering to oversample your signal and then average out white noise components over your required period. This topic is covered in a tutorial online that covers general analog input topics here. While dithering generally occurs in hardware, you could apply the same theory to perform an average on an oversamples signal in software. The Field Wiring and Noise Considerations tutorial here also discusses problems with signal noise and some processing steps to avoid them.

Perhaps I was unclear in my earlier post, but I think we are saying the same think. From my post: "then it would be better to sample at a frequency below 20 kHz so that the filter will eliminate the noise component at 10 kHz." Sampling a 1 kHz signal at 2 kS/sec would allow the filter to block any frequency components over 1 kHz, but it would also give you a very ugly signal (basically a triangle wave where the peaks are equal to the point in the sine where the sampling occurs). Though nyquist dictates that you must sample at 2x your frequency of interest to obtain the frequency information, it is generally recommended to sample at 10x your frequency of interest to get a better picture of what the signal looks like. So, going back to the previous example if we have a 1 kHz signal and 10 kHz noise and sample at 2 kS/sec, we will get 2 sample points per period of our signal and block the noise signal with the anti-aliasing filter. However, if we sample a little faster than that (say 10 kS/sec) we will still block the 10 kHz noise because the filter cutoff will be at 5 kHz, but we will also get more samples per period. In this case we'd be getting about 15 sample points per period of our signal rather than the 2 we were getting at 2 kS/sec.

I also wanted to address Wilfried's point about the AC coupling. Though the NI 9233 is designed for acquiring dynamic signals, it does have relatively good low frequency performance. Page 14 and 15 of the NI USB-9233 User Guide and Specifications provides the following information about AC cutoff frequencies:

AC cutoff frequency
–3 dB ...............................................0.5 Hz typ
–0.1 dB ............................................4.2 Hz max



So, depending on the level of attenuation that is acceptable in your application, you should still be able to use this device.

Message Edited by Matt A on 01-07-2008 04:02 PM

[Rafo]

Thank you very much for your answer I am checking what you told. Just one two more questions about another board the SCXI 1531.The system can adjust automatically to the cut off frequency? and what about the AC coupling to acquire signals with low frequencies?. 

Thank you very much