I want to design FD (frequency dependent ) IQ (In-phase,Quadrature) distortion that changes the phase mismatch linearly like a+b*K ( where K is 1 to 64) and amplitude remains constant to test my estimation algorithm against it. I need this filter to be in time domain before FFT but changes the IQ imbalance parameters across different subcarriers(hence the name FD) .I wonder how to achieve this ? I was thinking something like H = alpha *exp(2*pi*1i*(a+b*K)) as the frequency response and just take the IFFT of this and convolve with the signal. Do you think this is correct , any help will be appreciated ?

  • 2
    \$\begingroup\$ I'm having severe IQ distortion trying to read and understand that question... Different meaning of IQ though I guess... Abbreviations don't help on here. \$\endgroup\$ – Trevor_G Aug 11 '17 at 16:06
  • 1
    \$\begingroup\$ I edited the question for that. I in IQ means In-Phase and Q means Quadrature. \$\endgroup\$ – math Aug 11 '17 at 16:09
  • \$\begingroup\$ What are the degrees-of-freedom for your signal? and how can those be mapped into other signal domains? \$\endgroup\$ – analogsystemsrf Aug 11 '17 at 17:18
  • \$\begingroup\$ @analogsystemsrf, what do you mean by degree of freedom in my signal? the input signal is OFDM signal passed through the channel. Assume it is y and I need to distort it by IQ mismatch such that phase is changing across subcarriers linearly and amplitude is just constant \$\endgroup\$ – math Aug 11 '17 at 17:23

What you describe, phase being linear to frequency, is just a linear-phase system.

We usually design FIR filters that way, because it leads to constant group delay and hence, a lack of dispersion.

Anyway, any FIR filter that has symmetrical (complex case: hermitian symmetry) in its taps does that.

The flat response in amplitude means you're looking for an all-pass filter. There's plenty literature out there on the design of FIRs.

| improve this answer | |
  • \$\begingroup\$ What I was asking is that is my original approach correct to take the inverse IFFT of H and convolve with my signal (Is H all pass filter)? Can I model an all pass filter as constantexp(1iLinear function) \$\endgroup\$ – math Aug 11 '17 at 18:52
  • \$\begingroup\$ no, because that is a complex sinusoid, and that transforms to a single impulse! \$\endgroup\$ – Marcus Müller Aug 11 '17 at 20:35
  • \$\begingroup\$ This is not a complex exponential since K is changing from 1:64 so it is group of 64 exponential function and also H is in Frequency domain the phase is a+bK so it is H= Aexp(1i*(a+b*K)). \$\endgroup\$ – math Aug 11 '17 at 20:39
  • \$\begingroup\$ @math nonsense! usually, linear function has shape \$f(x) = ax,\,a\in\mathbb R\$; and your function is \$Ae^{i f(x) } = A e^{iax}\$, and that is a complex sinusoid, nothing else. Even if you say \$f(x) = ax + b\$ , then your function becomes \$Ae^{iax + ib}\$, and that is still a complex sinusoid! This does even hold true for any \$a,b\in \mathbb C\$!! \$\endgroup\$ – Marcus Müller Aug 11 '17 at 20:48

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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