0

Lets say i have a LPF response that has some kind of droop (some attenaution in the passband, usually could be due to cic filter how to make CIC compensation filter,,,, or can also be due to large ripple in passband).

Nevertheless, let say i have some droop in the passband , like the below figure, enter image description here

So now what i need is this passsband droop compnesation using a IIR-2nd order filter ( direct-ii- transposed or is lattice-ladder better , with regard to memory, quantization etc ?).

Can someone guide me what are the steps that i need, assuming i already have the pole-zeros of the lpf either in s-domain --> impinvar--> z-domain.

So how to proceed-...

  • Something like a pole-zero compensation with RMS error calibration.

Any help would be appreciated. Newbie in DSP.

BandW
  • 105
  • 9
  • Can you provide the details of your example with the poles and zeros in the s-domain and z-domain? I have an answer for you but would like to test out the approach with your case. Also why are you restricting yourself to IIR solutions, and why did the options at the link you gave not work for you (as that would be one approach)? – Dan Boschen Jun 05 '20 at 14:32
  • Yes that would work only for sinc compensation, but my LPF transfer function can be anything (could also include some passband ripple,any order, any shape, so its poles, zeros , and number of them is not defined. ). Also i need the compensation in IIR , since 2nd order is stable, and direct-form-II transposed is canonical , so i use this.

    Fs= 200MHz, corner 30 MHz, in sdomain [poles;zeros] = [(0,207345115.136926);(1 207345115.136926)] and zdomain [(1.03672557568463,0);(1,-0.354613939478490)]

    Thankyou

     – BandW Jun 05 '20 at 15:36
  • compensation can be in FIR. 2nd order IIR is not necessarily stable. Sounds like your problem space may be applicable to least square channel equalization techniques; have you considered this? – Dan Boschen Jun 05 '20 at 15:40
  • yes, adaptive filter techniques are there, but i am not sure how to proceed (i dont want to get lost in too much theory there ). I have to do it in IIR, i understand we can do it in FIR, but i need 2nd order IIR only. – BandW Jun 05 '20 at 15:47
  • can you explain why you must do it in IIR? That may help give further insight in to the answer you would need. Is this a homework problem? – Dan Boschen Jun 05 '20 at 16:46
  • because i need to use very small number of coefficients, also IIR filters are found to be better at approximating gain curves in pass-band. If you think FIR is the way to go.. then please go ahead.. I just need to know the steps.. I think then i can adapt to IIR also. Thankyou – BandW Jun 07 '20 at 17:26

0 Answers0