5
$\begingroup$

The way I understand it, is that to avoid aliasing on an ADC, you have to be in the first Nyquist zone (so half the sampling speed). What doesn't make sense to me is how the Hackrf is able to bypass this, and achieve 20MHz of bandwidth, even though the 1st Nyquist zone would be only 11MHz.

I notice from the schematics that the MAX5864 ADC is two channels, with the baseband I signal going into one channel, and the baseband Q signal going into another. Do the two channels get added together, to form an overall bandwidth of 20MHz?

Also, while I'm at it, what exactly is the purpose of using in an sdr? Is it to make modulation/demodulation easier?

Improve this question
$\endgroup$
1
  • $\begingroup$ Hello and welcome to ham.stackexchange.com! $\endgroup$
    – rclocher3
    Commented Feb 22, 2021 at 16:07

4 Answers 4

Reset to default
6
$\begingroup$

Another way to look at it is that an IQ ADC is really taking 2 independent (not added together) samples per IQ sample. Thus the rate of information gathered is double from a just scalar sampling at 22 MHz. The 90 degree offset between the 2 IQ sample components allows capturing phase information that can help a complex FFT (et.al.) differentiate between what would have been aliasing between spectrum above and below the heterodyning LO frequency.

Improve this answer
$\endgroup$
1
  • $\begingroup$ Thank you! That makes sense $\endgroup$
    – camerakid
    Commented Feb 24, 2021 at 2:33
7
$\begingroup$

The Nyquist limit is half the sampling rate because otherwise you can't distinguish a signal at a frequency $x$ from a signal at $f_s - x$ which starts 180° out of phase from the first one — they give exactly the same sequence of real samples. But quadrature sampling gives exactly the phase information needed to resolve this ambiguity. Knowing $\sin \omega t$ limits us to two of four quadrants, but knowing $\sin \omega t$ and $\cos \omega t$ at the same time limits us to one quadrant. With this distinction provided, it's possible to distinguish frequencies up to $f_s$ — or a little less, given that we still need to filter out frequencies above the sample rate, and filters have skirts.

Improve this answer
$\endgroup$
6
  • $\begingroup$ Interesting, I did not know that. So (if I understand correctly): if you have an ADC with two channels at freq x, you can use both to sample (and reconstruct afterwards) an I Q signal up to (almost) frequency x? Correct? $\endgroup$
    – Dieter Vansteenwegen
    Commented Feb 22, 2021 at 8:39
  • 1
    $\begingroup$ @DieterVansteenwegenON4DD You can double the bandwith with two ADC channels, but only if they are fed with orthogonal, i.e., separated by 90° shift, phases. $\endgroup$
    – OH2FXN
    Commented Feb 23, 2021 at 18:18
  • $\begingroup$ @OH2FXN Thanks, that makes sense! $\endgroup$
    – Dieter Vansteenwegen
    Commented Feb 24, 2021 at 9:27
  • $\begingroup$ "The Nyquist limit is half the sampling rate because otherwise you can't distinguish a signal at a frequency x from a signal at fs−x which starts 180° out of phase from the first one — they give exactly the same sequence of real samples" - is this really the reason for the Nyquist limit? I thought it had to do with not being able to distinguish higher frequencies (by higher I mean further from zero)? I thought the IQ sampling approach works because it turns the signal into 2 signals each having half the bandwidth, and each one is sampled at the usual Nyquist rate? $\endgroup$
    – John B
    Commented Feb 24, 2021 at 16:42
  • 1
    $\begingroup$ @JohnB $f_s$ is the sampling rate (double the Nyquist rate), so $f_s - x$ is a higher frequency. :) You also get aliases at further multiples of $f_s$ but that doesn't come into play here. And yes, what you said is true, but it's not like I is USB and Q is LSB, and you put them together to get both sides of the carrier. The relationship really works like I said in my answer. $\endgroup$
    – hobbs - KC2G
    Commented Feb 24, 2021 at 19:09
2
$\begingroup$

Are the two channels added together? Yes and no.

The whole point of collecting I and Q is so they can be treated as real and imaginary parts of a single complex sample.

This is one of the key methods of SDR.

This answer explains it well: https://electronics.stackexchange.com/questions/39796/can-somebody-explain-what-iq-quadrature-means-in-terms-of-sdr

Improve this answer
$\endgroup$
2
$\begingroup$

This is old I know and this may have been stated already, but when we normally think of Nyquist, we think of taking real samples of real signals. When you do that the signal had twice the bandwidth of the original signal because it created "positive" and "negative" frequencies.... If you look at the math of analytic (complex) signals, they are single sided and therefore the bandwidth is not doubled and so the sampling requirement is halved..... That sounds really good, but there is a catch! As others have said, each "sample" actually contains two orthogonal (real) samples that make up a single complex sample. So really the 22Msps is giving you 44Msps....

Improve this answer
$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .