0

What are the techniques which can be used to perform digital IQ amplitude and phase correction in a wideband receiver system when input signals can be of unknown different frequencies and occupy different bandwidths?

malik12
  • 488
  • 4
  • 24

1 Answers1

1

The same techniques when that isn't the case can be used, but the question is if the correction is closed loop or open loop and if the offset and quadrature error is the same or similar enough across the entire bandwidth. For high accuracy correction under conditions where phase and amplitude offsets can change with time, temperature or from unit to unit, a measurement is made of the resulting carrier offset and sideband suppression with known signal conditions and the correction is applied using the techniques referenced here in a nulling loop. There are cases and conditions where an open loop correction would be sufficient as well and this comes out of the requirements for DC offset (carrier feedthrough from origin offset) and sideband leakage (from quadrature and amplitude error), and an assessment of how good the analog hardware is in terms of initial conditions and maintaining that over time and temperature.

Other than that the correction approach itself begins with structures that I have detailed in this other post. The additional complication to consider is that the correction under wideband conditions would likely be frequency selective, and would have multiple coefficients resulting in a "complex equalizer" rather than the "single tap equalizer" that the referenced post indicates. I detail the implementation of a complex equalizer at this post.

Dan Boschen
  • 50,942
  • 2
  • 57
  • 135
  • 1
    nice! The old-school RF-daughterboard-carrying Ettus USRPs implement static IQ correction through observing crosstalk from PRNG – which statistically should have as much power in I as in Q. If that's not the case -> correction factor. – Marcus Müller Mar 24 '22 at 18:44
  • 1
    @MarcusMüller yes very nice! So could do DC offset as well similarly assuming equiprobable distributions. – Dan Boschen Mar 24 '22 at 18:57
  • My memory betrayed me, they use leakage, but a tone. RX DC offset cancellation is done live with a IIR;TX DC offset calibration again throug tone leakage. – Marcus Müller Mar 24 '22 at 19:04
  • 1
    @MarcusMüller and you "single tone" comment clarified for me that it would likely be an equalizer that does the correction with all the same techniques (wideband sounding signal with full complex FIR-- what I show in the link is essentially a 1 tap complex FIR which is typically what we do for relatively narrow band solutions where the error applies to the whole bandwidth). An equalizer is essentially a multi-tap nulling loop :) – Dan Boschen Mar 24 '22 at 19:11
  • @DanBoschen I have gone through your answer in the other post so basically for wideband correction I need to calculate the frequency of the input signal first before applying the coefficients of the "complex equalizer" accordingly, there is no way to do it other than that? – malik12 Mar 24 '22 at 19:22
  • Once you have multiple taps, then I think traditional equalizer approaches would be well suited for this. (If you really can't do it with a single tap). Ultimately all the solutions are the same, it's just once it is 1 tap it is simpler to not explain it in the same terms and process as an equalizer. I will add a link I have here that details how to construct the equalizer solution. (But definitely investigate first if you can't do a single tap correction--- will the error corrected by a tone at one frequency suffice at your other frequencies? that is what you want to answer for yourself). – Dan Boschen Mar 24 '22 at 19:26
  • Ok, I'll check that out, thank you. – malik12 Mar 24 '22 at 19:35