can someone explain what a difference is between phase and frequency shift/offset in estimation algorithm?

Honestly I dont understand this difference. If i estimate , i will do it for phase estimation.

EDIT 1 What is a difference in an implementation of FLL and PLL?


The difference in implementation between a FLL and PLL is completely dictated by what device is used as the error detector: if the error detector produces an error that is directly proportional to phase (phase detector) that is then passed into the loop filter, then the implementation will be a PLL and we would treat the VCO (Or NCO) as a phase integrator (since that device producing a frequency in proportion to its input, so produces a phase in proportion to the integral of the input). If the error detector is directly proportional to frequency (such as a frequency discriminator) that is then passed into the loop filter, then that device would be an FLL and we would treat the VCO or NCO as a proportional device producing a frequency in direct proportion to its input. Note that we can have a phase detector that we then take the derivative of its output which would then be a frequency discriminator (and this approach can be used for a combined FLL/PLL) so the actual block diagram and functionality needs to be carefully reviewed for each case.

Frequency is simply the derivative of phase with respect to time. So in a phase estimation algorithm you estimate absolute phase (relative to a reference which is typically the phase of a symbol location on the complex IQ plane), while in a frequency estimation algorithm you estimate the rate of change of phase with respect to frequency which is independent of any offset.

Picture a wheel that is spinning at a constant frequency and we use a control loop to stop the spin; if we measure error using a frequency estimation algorithm to then reverse and ultimately stop the spin, it will eventually stop but at an arbitrary offset like we would typically expect from a roulette wheel. If we use a phase estimation algorithm to measure error which is then used to stop the spin (and assuming we can keep up with and overcome the rate of change which implies sufficient loop order) it will stop at the exact place you desire (and you’ll be a winner every time!)

  • $\begingroup$ Can I use one algorithm to estimate phase and frequency or should i have two, one is for phase, second one for frequency? $\endgroup$
    – Jang Lee
    Apr 9 at 11:14
  • 1
    $\begingroup$ It is common to have an FLL for acquisition and then PLL to lock final phase, for systems where absolute phase is important. But yes this can and often is done concurrently where both frequency and phase are contributing to the loop error (as the sum of both) and the loop drives the resulting error to zero resolving both frequency and phase offset. $\endgroup$ Apr 9 at 11:35
  • 1
    $\begingroup$ @DanBoschen The spinning wheel is a great analogy. $\endgroup$
    – MBaz
    Apr 9 at 12:39
  • $\begingroup$ what is the difference in implementation? If i have a test signal rx, I will send it to FLL for f- correction and to PLL for phi-correction. At the end , I will have new f and phi and update the rx...smth like this? $\endgroup$
    – Jang Lee
    Apr 9 at 13:22
  • $\begingroup$ @JangLee The difference is all in the discriminator; I will update my answer $\endgroup$ Apr 9 at 19:53

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.