0
$\begingroup$

Here below is the SDR for the O-QPSK demodulation for the Aqua satellite.

enter image description here

How are we sure that the I component we get in the Costas Loop output is not the Q component of the original signal?

Because if we don't know, then how can we build the correctly-oriented constellation diagrams?

$\endgroup$
1
  • 1
    $\begingroup$ The way for a receiver to distinguish I from Q is to get synchronized with the transmitter. I had asked a question regarding frame sync markers before and I still am curious as to how best to define a special word that, when received, will tell the receiver that a block of data is beginning and to align the receiver quadrature phase with that of the transmitter. It still seems to me that 0011001100110011 and 0110011001100110 (which spin the I/Q vector at maximum speed) is a useful concept. $\endgroup$ Commented Sep 29, 2023 at 18:21

1 Answer 1

1
$\begingroup$

An IQ constellation with random data is rotationally invariant so unless we have prior information of what is in the constellation, there would be know way to know.

That said, information is used to distinguish I from Q. I am not familiar with the Aqua satellite signal specifications specifically, but this is often done with the header that includes other apriori information for timing, synchronization and channel equalization. For example, part of the (known) header could be transmitted completely with a BPSK pattern on I only which would serve to resolve I from Q and have better performance for acquisition under lower SNR conditions. A pattern would still be needed that would be able to resolve the 180° phase ambiguity that would still exist.

$\endgroup$
2
  • 1
    $\begingroup$ Seems to me that there really is a 90° phase ambiguity in OQPSK without a time reference that has to come from another source or come from the frame sync marker or preamble. You can detect when the phase is changing, either clockwise or counter-clockwise and that should be happening at 0°, 90°, 180°, or 270°. But you could rotate the square any of four ways, if you have no phase reference. And then, for data like 00110011001100110011001100110011 and 01100110011001100110011001100110, the IQ vector is just spinning at a solid rate anyway and you don't even have these 90° markers to detect. $\endgroup$ Commented Oct 1, 2023 at 17:39
  • $\begingroup$ @robertbristow-johnson yes, very good- also to mention differential encoding that could be used. $\endgroup$ Commented Oct 4, 2023 at 12:05

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.