Hi, My simulations show that having IIR low pass filter to simulate baseband model of a communication systems will introduce inter symbol interference (ISI). regardless the filter type (Butterworth, Cheby, Elliptic,...) all gives a significant ISI. The simulation shows it gives a significant BER degradation, it can be around 3 to 5 dB. Is that a realistic value compared to the real implementation ? Anybody would like to share his experience, how many dB of degradation due to the ISI introduce by IIR filter ? Is there such away to mitigate this ? Thanks. Berry
IIR Filter for Baseband comms
Started by ●October 5, 2005
Reply by ●October 5, 20052005-10-05
>>>>> "Mas" == Mas Bas <buzzberry@gmail.com> writes:Mas> The simulation shows it gives a significant BER degradation, it can be Mas> around 3 to 5 dB. Is that a realistic value compared to the real Mas> implementation ? Probably. I think it really depends on how much ISI there is and what your receiver does. Mas> Is there such away to mitigate this ? Use a filter that doesn't produce ISI? Add an equalizer to remove the ISI? Ray
Reply by ●October 5, 20052005-10-05
Raymond Toy wrote:>>>>>>"Mas" == Mas Bas <buzzberry@gmail.com> writes: >Thanks for the reply.> Mas> The simulation shows it gives a significant BER degradation, it can be > Mas> around 3 to 5 dB. Is that a realistic value compared to the real > Mas> implementation ? > > Probably. I think it really depends on how much ISI there is and what > your receiver does.My simulation has a simple 16 QAM Tx and Rx with AWGN channel in between. I inserted a pair of IIR filter (one at Tx and one Rx) Rx filter has wider 3db Bandwidth. I am quite surprise as the degradation due to the filter is such that high (3-5 dB).> Mas> Is there such away to mitigate this ? > > Use a filter that doesn't produce ISI?Is there any guidelines to choose the right filter (small ISI) ?> Add an equalizer to remove the ISI? > RayBerry
Reply by ●October 5, 20052005-10-05
Mas Bas wrote:> Hi, > > My simulations show that having IIR low pass filter to simulate baseband > model of a communication systems will introduce inter symbol > interference (ISI). regardless the filter type (Butterworth, Cheby, > Elliptic,...) all gives a significant ISI. > > The simulation shows it gives a significant BER degradation, it can be > around 3 to 5 dB. Is that a realistic value compared to the real > implementation ? > > Anybody would like to share his experience, how many dB of degradation > due to the ISI introduce by IIR filter ? > > Is there such away to mitigate this ? > > Thanks. > > BerryWithout knowing the details, I would say that 3-5 dB is a reasonable degradation to expect. A modern way to mitigate ISI is to use a maximum likelihood sequence estimator (MLSE), which can be efficiently implemented using the Viterbi Algorithm. The ISI is modeled as a finite state machine (FSM), and the estimator uses knowledge of the FSM to figure out the most likely sequence of state transitions (shortest path) looking back a finite duration into the past. The FSM input bits which could produce those transitions are the recovered data. Your handy GSM cell phone uses this technique. It dynamically updates the ISI model many times per second using channel estimates. In your case the ISI model can be static. John
Reply by ●October 5, 20052005-10-05
On Wed, 05 Oct 2005 16:26:41 +0200, Mas Bas <buzzberry@gmail.com> wrote:>My simulations show that having IIR low pass filter to simulate baseband >model of a communication systems will introduce inter symbol >interference (ISI). regardless the filter type (Butterworth, Cheby, >Elliptic,...) all gives a significant ISI. > >The simulation shows it gives a significant BER degradation, it can be >around 3 to 5 dB. Is that a realistic value compared to the real >implementation ? > >Anybody would like to share his experience, how many dB of degradation >due to the ISI introduce by IIR filter ? > >Is there such away to mitigate this ? > >Thanks. > >BerryButterworth Tx filters didn't used to be that unusual before digital filters became more practical. Do you intend to have an IIR on both ends of the link or only one? I.e., both Tx and Rx filters are IIR? If not, you can make a FIR on one end that matches the response of the IIR on the other and still wind up with a zero-ISI filter. Eric Jacobsen Minister of Algorithms, Intel Corp. My opinions may not be Intel's opinions. http://www.ericjacobsen.org
