sampling rate vs frequency

[dphan128]

How does the sampling rate relate to the max frequency that a digitizer can measure ? If I know the frequency that I want to measure then what min sample rate should I choose from a digitizer ?

Thanks

Dan

[Travis_W]

Dan,

According to the Nyquist Theorem, you need to sample twice as fast as the highest frequency you want to measure.  For instance, if the signal you wish to measre is a 1MHz sine wave, you have to sample at **AT LEAST** 2 Mega Samples per second (MS/s) to ensure that the 1MHz frequency component of the signal is detected.

However, sampling at exactly 2 times the highest frequency is often unacceptable, especially in applications where the shape (time-domain representation) of the signal is important.  When sampling a pure sine wave, I generally make sure to sample at at least 10 times the frequency of that sine wave to make sure I can properly see the shape of the signal.

Other factors that will affect your sample rate are the bandwidth and on-board memory available to your digitizer.  If you are looking to acquire continuously, or for many seconds at a time, this will limit your maximum sample rate.  On the other hand, if you are doing short, finite acquisitions, and you have plenty of processing power, I'd say the more samples the better.

Good luck on your application, Dan.  Let me know if you have more questions.

Regards,
Travis W

[dphan128]

Thanks Travis.

Dan P.

[LiewSC]

Dear Travis,

NI PXIe-5622 with 150MS/s but the frequency is 3M to 250MHz. Why it can measure higher frequency? Can it measure lower frequency, for instance 150kHz?

-Ability to stream to disk at maximum sample rate for hours with PXI Express

Is this new technology can let us measure continuously for hours with maximum sample rate?

[Jennifer O]

There are actually two things to consider when determining if a device is capable of capturing a particular frequency.

1)Analog Bandwidth

2)Sampling Rate.

The front end of an analog sampling board such as a High Speed Digitizer can be modelled as a simple RC circuit which inherently acts as a lowpass filter.  This will limit the bandwidth and ultimately the maximum frequency that can be measured.  Frequencies which exceed this upper cutoff frequency will be significantly attenuated and cannot be measured accurately with that device. In addition, if the board is AC coupled then there will also be a highpass filter which limits the minimum frequency which can be measured.  Frequencies below this cuttoff also cannot be measured accurately.

In the case of the NI PXIe-5622 it would not be an appropriate device to measure 150kHz as this is outside the bandwidth of the device.

The Nyquist theory that Travis quoted refers to the ability to measure a signal without aliasing.  If you sample at least 2x the maximum frequency in your signal you will ensure that your signal is adequately sampled to prevent aliasing.  However if you have a bandwidth-limited signal you can employ a technique of undersampling to accurately capture signals above the nyquist frequency. They will appear at the correct amplitude, however will appear to be a lower frequency than they actually are.  If your signal is not bandwidth limited, it is possible for two frequencies to become aliased to the same frequency and it will be impossible to tell them apart.  Refer to this tutorial for more information on the Nyquist theory, aliasing and undersampling.

PXIe with a dedicated x4 communication link will provide a theoretical data transfer bandwidth of 1GB/s.  In practice data input can be streamed to disk at approximately 800MB/s.  So for a device that has a maximum sample rate of 150MS/s, taking 2 Bytes per sample, data will need to be fetched at 300MB/s to support continuous streaming.  This is within the streaming bandwidth of the PXIe technology.  Therefore, yes you can measure continuously for hours at maximum sample rate.  The limitation will generally be the ability to process or view the data in realtime.  Generally this data is streamed to disk instead for later post-processing.  You will need a large capacity storage method that is capable of keeping up with those rates such as the NI HDD- 8264 RAID Array.  In this case you can stream the data until you run out of disk space.

Hope this helps.

Jennifer O.

Message Edited by Jennifer O on 01-02-2009 09:45 AM

[LiewSC]

Dear Jennifer,

Thanks for fast response. It does help. 🙂

[nhmh]

Hi Everyone,

I need some help to verify whether I am in the right path to choose the correct digitizer for my client. 

He has an Ultrasonic Pulser/Receiver (UPR) having specifications Bandwidth 1kHz-200MHz. Currently he is connecting the UPR to a LeCroy Oscilloscope. Now I would like to replace the LeCroy Oscilloscope with NI Digitizer. Previously I proposed him to use PXI-5154, but due to budget constraint, he requested to have cheaper solution i.e low-cost PCI. In order to reduce more cost, he would appreciate if I can propose the mid or low spec of digitizer.

My question is :

1) In order for us to measure the UPR signal, we will need a digitizer with frequency range of 1kHz - (200x2 at least)MHz, correct? I;ve gone through NI Digitzer and see that the best digitizer that shall be used is the one with up to 500MHz frequency range, right?

2) if I am wrong for question no.1, what would be the minimum requirement of the digitzer? can anyone suggest me which NI PCI (digitizer) can be used?

Thankx. 

[adeel1]

Dear Jennifer

can you please tell me that how can I set the desired sampling time of my bits in labview 2009 (32 bit) and how can i put user defined wavelet filters coefficient in IDWT/DWT block.

thanks

Regards

Ahsan Adeel

[Ken3006]

Hi nhmh,

1)  If I am not wrong, I think you would need to have a digitizer that covers that range of frequencies of your device.  The higher the range (i.e. more than 200Mhz) would be great but will result in higher spec digitizer.

2) Maybe you can look at the NI-5152 series, there is a PCI version too.  Its sampling rates and freq range seems ok.

Cheers,

Ken

[tareen]

Hello Sir,

I am a research student in Electrical power electronics. I read your comments on sampling frequency under sampling frequency question. Sir, my question can the sampling frequency be equal to switching frequency. ???. I designed a grid-connected inverter, the switching frequency is 20 KHz, then can my sampling frequency be same, if yes WHY ???

I used the DSpace as controller to excute the system running. 
Thanks