Maximum sampling rate of the myRIO-1900

[andreasbaldur]

Hi

I'm considering to use the myRIO-1900 to measure the return echo signal of a 200 kHz ultrasonic transducer. The reason is that I would like to do some real-time signal processing. My concern is that the available sampling rate is not sufficiently fast. The datasheet of the myRIO says that the "aggregate sample rate is 500 kS/s". Is there any ways that the sample rate can be increased when making proper use of the FPGA or is the 500 kS/s an absolute maximum?

Best Regards,

Andreas

[ikaiser]

Hi Andreas,

welcome to this forum!

Short answer: No.

Long answer:

• The sample rate is limited by the hardware Analog-to-Digital Converter (ADC), and it is 500k Samples per second, as you said.
• It is called "aggregated", as this specific device has only one ADC which reads all input channels one after another. In other words: The analog inputs are multiplexed towards the ADC. See page 7 of the device's manual for more information. With this, you can reads 500k S/s on one channel or e.g. 250k S/s on two channels each.
• The FPGA is connected to the AI pins through the ADC (see the block diagram on page 3 of the above-linked manual or in color here: NI myRIO-1900 Block Diagram (myRIO Module)). This makes sense, as the signal has to be converted somehow. So no matter what the speed of the FPGA is, the ADC is the limiting factor.

What is the maximum signal frequency that you want to measure? The 200 kHz itself or slight changes in the echoed signals (much higher than 200 kHz)? For just detecting the signal 500 kHz is enough, it's 2.5 times above it, so it obeys the WKS theorem / Nyquist theorem. For a good reproduction of the waveform you'd need something near 2 GHz anyways.

[SuperLearner]

Hello,

Just out of curiosity I am asking, why do we need 10k samples per 1 period of the waveform that we want to reproduce? Isn't 100 samples or 1k samples enough?

Thank you!

[Bob_Schor]

Sorry for the late response, but your question about the relationship of sampling frequency to "frequencies in the signal" is not entirely foolish.  I'm guessing that you have not had a course on Signal Theory, but here's the basic idea:

Let's say you have a signal, and you know that your signal is a "pure sine wave", with no "noise" (so all you see is an endless up-and-down of the waveform at a fixed rate).

Go ahead and sample it at the frequency of the sinusoid (let's say it is 100 Hz, and you sample it at 100 Hz).  What will you see?  What will you be able to say about the signal (remember, you can't peek at the oscilloscope that shows a 100 Hz sinusoid, you only have your samples to go on ...)?

The answer is you'll see all the samples have the same value, i.e. you'll see no signal at all!  Can you see why that is?

So now let's sample at 10 times that frequency, 1 kHz.  You should see a sinusoid that repeats every 10 points (because you are sampling 10 time faster than the sinusoid, so you get 10 samples/cycle).

Now let's add noise (boo!).  Noise typically is high-frequency stuff that interferes with your attempt to record your signal.  So what do you do?  You filter the signal and try to remove the noise.  

So to answer your question, you need to consider the frequency content of your signal.  In my (simple-minded) example, there was only 1 frequency, 100 Hz, so I only need to sample, say, 10 times the highest frequency of interest, or 1000 Hz.  If I want to remove stray noise, I can low-pass filter 5 or 10 times higher than the highest frequency of signal that I expect.

A good course on Signal Theory will say a lot more than this, and will introduce you to Jean-Baptiste Joseph Fourier and a host of other characters (like Harry Nyquist).

Bob Schor