Trying to understand the BB60C block diagram

[rs20]

The BB60C can stream IQ samples at 80 (40 I + 40 Q) MSPS, with a 27 MHz bandwidth. I was looking at the block diagram they provide trying to understand how they achieve this, and TBH it's raised more questions than answers (mostly because I'm a complete n00b when it comes to spectrum analyzers.)



The ADC runs at 80 MSPS, but the SAW filters just preceding it are centered at 140 MHz. This would appear at first glance to violate the Nyquist criterion, 140 MHz is after the 3rd Nyquist "foldback", but of course as long as the signal is sufficiently filtered the original signal can be deduced. Indeed, I notice that the ADC used (AD9255) advertises that it can operate well beyond its Nyquist frequency (indeed, up to 300 MHz), but I'm wondering what specific advantage one might hope to gain by having the 2nd IF at a frequency at such a high frequency? Is it mostly to ease the IF filter designs, as I guess below?

Is the idea that the 2nd IF filters would have a bandwidth of approximately 27 MHz (passband ~126 MHz through 154 MHz), with steep enough rolloffs that corresponding Nyquist images (86 thru 114, 166 thru 194) are heavily suppressed? And that the first IF filters benefit from having a relatively relaxed spec thanks to the high second IF frequency; since the first IF filters don't have to worry about the rolloff for nearly a full 2*140 = 280 MHz from their 1.26 GHz and 2.42 GHz center frequencies? It seems telling that the distance from centre to heavy suppression point is 10-20% of the centre frequencies themselves in both cases; but if the 2nd IF was lower the 2nd IF filters would have a very easy time whereas the 1st IF would have very difficult-to-meet specs.

Am I understanding it right, or am I way off?

[ataradov]

That is called undersampling. They use heavy band pass filters, so signal bandwidth is limited, and Nyquist only cares about bandwidth, not the actual frequencies.

Wikipedia is helpful, as usual - https://en.wikipedia.org/wiki/Undersampling

And why exactly they did it this way is hard to tell. It may be because of available mixers and LOs, or some other considerations, which would be hard to tell from the block diagram alone.

Undersampling is not easy to get right, so there must be a very good reason to do it that way. It is not something RF people will take lightly.

[w2aew]

That is called undersampling. They use heavy band pass filters, so signal bandwidth is limited, and Nyquist only cares about bandwidth, not the actual frequencies.

Wikipedia is helpful, as usual - https://en.wikipedia.org/wiki/Undersampling

And why exactly they did it this way is hard to tell. It may be because of available mixers and LOs, or some other considerations, which would be hard to tell from the block diagram alone.

Undersampling is not easy to get right, so there must be a very good reason to do it that way. It is not something RF people will take lightly.

^^ this

Nyquist says that the sample rate must be >2x the *bandwidth* of the signal.  So, even if the signal is centered on an IF that is higher than the sample rate, as long as the *bandwidth* of that signal is less than half the sample rate, you will satisfy the nyquist criteria for that signal. 

[rs20]

Cool, thanks -- but my question was more asking whether my guess was correct about the reasonings for the high 2nd IF frequency selection? Specifically, that choosing a high second IF frequency greatly relaxes the requirements for a high Q on the first IF filters?