Evaluating RF ADC architectures - Sub-Sampling or Nyquist ?

Started by krishk24 4 years ago4 replieslatest reply 4 years ago811 views


I am trying to decide on the most optimal ADC architecture for an high bandwidth receiver application.

The signal bandwidth of interest is 2.5 GHz. 

There are a few RF ADC like this one from TI http://www.ti.com/lit/ds/symlink/adc12dj2700.pdf

here is my question:

with sampling rates of 2700 MSPS (dual channel) and 5400 MSPS (single channel), how is the max frequency BW close to 8 GHz ?

Is this ADC using some sub-sampling strategy instead of Nyquist sampling ? 

I also see this part has internal DDC that effectively reduces the sample rate by decimation and probably does digital downconversion before handing the sampled data to the SERDES.

But I am confused about the actual mechanism of sampling used by this converter. Please advise.

[ - ]
Reply by timburnettApril 21, 2020

Having the analog bandwidth much higher than the maximum sample rate allows the sampling converter to undersample high frequency signals. 

For example, lets say you are using it in 5GSPS single channel mode, and your band of interest stretched from 5.5GHz to 6.5GHz. This band is within the ADC's analog bandwidth, so it would be received just fine. It's in the 3rd Nyquist zone, so it would alias on top of 0.5GHz to 1.5GHz (1st nyquist zone) and 3.5GHz to 4.5GHz (2nd nyquist zone). For this reason you would probably want to use a high pass antialiasing filter with a stopband covering 4.5GHz and below. 

As far as strategies go, it has a very narrow sampling aperture time, which allows it to properly convert samples from signals at those high frequencies. 

[ - ]
Reply by josefseppApril 21, 2020


with the following picture

I will try to answer you. For a channel with a sampling frequency of 5.4 GHz, a bandwidth of BW = 2.7 GHz in the vicinity of 8 GHz can be used. This area is then correctly aliased in the first Nyquist interval (NZ = 1).


[ - ]
Reply by SlartibartfastApril 21, 2020

The data sheet indicates that the 3dB rolloff is at 8GHz, and rolls off from there to 10GHz.  So I think the blue channel that is highlighted would be the least usable of those shown below 10GHz.   Channels 1, 2, and 3 could be used with less distortion.

But, yes, that's a good illustration of how the 8GHz input bandwidth is useful for a converter that samples at less than that.

[ - ]
Reply by MarcinsteinApril 21, 2020


With dual channel mode you have max BW = fs. After proper signal filtration you can capture band from 5,4 to 8,1 GHz or from 8,1 GHz to 10,8. With single channel I guess its the same but you have just different processing chain. Depends on signal you are sampling.