Frequency content of signals acquired using Random Interleaved Sampling

[Alias77]

I am using an NI PXI-5114 o-scope, which has a maximum real-time sampling frequency of 250MHz and a maximum Random Interleaved Sampling (RIS) frequency of 5GHz.  I am aware of how RIS works, but I am confused when I observe the frequency content of my signals when using RIS.  Attached is the frequency content of a 15MHz signal sampled with an RIS frequency of 2GHz.  The frequency content repeats at integer multiples of the real-time sampling frequency, and thus there appears to be valid content only up to 125MHz.  However, when I incorporate RIS averaging, the frequency content above 125MHz is modifed.  For example, I see a diminished amplitude in the mirror peaks of the 15MHz signal.  

Could someone tell me how to interpret these results? Does RIS not truly enhance the ability to sample at frequencies above the real-time sampling rate?  Or is there some sort of distortion?  Note that I have observed these results in both Matlab and LabVIEW, so I don't think it is related to the way in which I'm viewing the frequency content.

[Alias77]

I DO notice that the content in between the 15MHz mirrors does not exactly repeat.  So are these mirror peaks actually harmonics introduced by the RIS process?

[Beutlich]

Could you post the with and with/out RIS averaging so we can see what distortion you are viewing?

The averaging will account for the fact that the samples are no longer evenly spaced. Without averaging, you would see slight distortions associated with bumping your measurements a small Δt in random directions. Read more in the NI High-Speed Digitizers Help.

[Alias77]

Hi Eric,

Thanks for the response, and sorry for the delay.  Attached is the same signal with same conditions as the previous except with 256 averages.  

Clay

[Henrik_Volkers]

Looks like some mirror or mixing effects  [250, 500, 750] +-15  MHz.

My first test would be an analog Filter somewhere in the 20 to 120MHz region between the generator and the scope, and have have a look what that will do.

What generator are you using? (AWG with 250MHz ??)

But [2,4,6]*125 +-15 MHz  looks like a strong indicator to some RIS effects... If you use 4Ghz RIS will the peaks change?

[Alias77]

Yeah, the mirrors appear at integer multiples of the real-time sampling rate, and this is independent of the RIS sampling freqency that I choose.  I can of course filter these mirrors out, but the reason I'm using RIS in the first place is so that I can acquire frequency content above 125MHz.  I'm using a Rigol DG1012 AWG, but I have observed this mirroring effect in other signals from other sources.  

[Beutlich · Accepted solution]

This looks like the harmonics due to analog acquisition, which we spec out as Spurious-Free Dynamic Range, described on page 5 of the NI PXI/PCI-5114 Specifications. The card has a specification of 58 dBc. When you account for the Δt through averaging, that is exactly the range you see (there is a little more when you don't account for the Δt, as in the image for your original post, only about 4, since you introduced additional noise from phase offset error). This is more or less impossible to avoid for any aquisition method, but you can read more about our Spurious-Free Dynamic Range specification in the article What is Spurious-Free Dynamic Range?

[Alias77]

Thank you for your response on this Eric. I appreciate it.