Hi, I am working on a uncode 16-QAM OFDM simulation. The simulation works fine when used in AWGN channel. I have a couple of questions now when i am simulating performance in multipath channels. 1) What I want to understand through simulation is the effect of cyclic prefix length on performance in a multipath channel. For example consider that the channel impulse response is [0.7071 0 0 0 0 0 0 0 0 0.7071], which is a channel with impulse response length 10 ( static paths with no fading component). Now if we consider four cases 1)cp=0, 2)cp=5, 3) cp=11 4) cp =15 The first two cases, based on theory should result in ISI. My question is: when i perform the simulations i do not see much of a difference. Will the difference be seen only at high SNR's ? 2) Presently, I transmit a known OFDM symbol at the beginning of the simulation, At the receiver after FFT operation I divide the output with the known OFDM symbol values and store them as channel state information for the remaining simulation. For all subsequent transmissions after the FFT operation i divide the output with the stored channel state information ( Zero forcing in frequency domain). My question is even for cyclic prefix lengths greater than the impulse response, don't we still need to do equalization operation at the receiver ( something trivial as i am doing now). Also can anyone suggest me a reference to other frequency domain equalization approaches. Thanks
OFDM cyclic prefix/Equalization question
Started by ●August 2, 2006
Reply by ●August 3, 20062006-08-03
> >Hi, > >I am working on a uncode 16-QAM OFDM simulation. The simulation worksfine>when used in AWGN channel. I have a couple of questions now when i am >simulating performance in multipath channels. > >1) What I want to understand through simulation is the effect of cyclic >prefix length on performance in a multipath channel. > >For example consider that the channel impulse response is [0.7071 0 0 00>0 0 0 0 0.7071], which is a channel with impulse response length 10 ( >static paths with no fading component). >Now if we consider four cases >1)cp=0, >2)cp=5, >3) cp=11 >4) cp =15 > >The first two cases, based on theory should result in ISI. My question >is: >when i perform the simulations i do not see much of a difference. Will >the >difference be seen only at high SNR's ? >Having a cyclic prefix that is too small will cause one OFDM symbol to interfere with another, which causes inter-carrier interference (ICI) in the frequency domain. If the cyclic-prefix is sufficiently long, you will mitigate ICI, but within a frequency bin, you will still have ISI. Both of these effects will defocus the constellations. When you say you don't see much of a difference, what criterion are you using to compare the two scenarios?> >2) Presently, I transmit a known OFDM symbol at the beginning of the >simulation, At the receiver after FFT operation I divide the output with >the known OFDM symbol values and store them as channel state information >for the remaining simulation. >For all subsequent transmissions after the FFT operation i divide the >output with the stored channel state information ( Zero forcing in >frequency domain). >My question is even for cyclic prefix lengths greater than the impulse >response, don't we still need to do equalization operation at the >receiver >( something trivial as i am doing now). >Also can anyone suggest me a reference to other frequency domain >equalization approaches. > >Thanks > >
Reply by ●August 4, 20062006-08-04
>> >>Hi, >> >>I am working on a uncode 16-QAM OFDM simulation. The simulation works >fine >>when used in AWGN channel. I have a couple of questions now when i am >>simulating performance in multipath channels. >> >>1) What I want to understand through simulation is the effect of cyclic >>prefix length on performance in a multipath channel. >> >>For example consider that the channel impulse response is [0.7071 0 0 0 >0 >>0 0 0 0 0.7071], which is a channel with impulse response length 10 ( >>static paths with no fading component). >>Now if we consider four cases >>1)cp=0, >>2)cp=5, >>3) cp=11 >>4) cp =15 >> >>The first two cases, based on theory should result in ISI. My question >>is: >>when i perform the simulations i do not see much of a difference. Will >>the >>difference be seen only at high SNR's ? >> > >Having a cyclic prefix that is too small will cause one OFDM symbol to >interfere with another, which causes inter-carrier interference (ICI) in >the frequency domain. > >If the cyclic-prefix is sufficiently long, you will mitigate ICI, but >within a frequency bin, you will still have ISI. > >Both of these effects will defocus the constellations. > >When you say you don't see much of a difference, what criterion are you >using to compare the two scenarios? > >> >>2) Presently, I transmit a known OFDM symbol at the beginning of the >>simulation, At the receiver after FFT operation I divide the outputwith>>the known OFDM symbol values and store them as channel stateinformation>>for the remaining simulation. >>For all subsequent transmissions after the FFT operation i divide the >>output with the stored channel state information ( Zero forcing in >>frequency domain). >>My question is even for cyclic prefix lengths greater than the impulse >>response, don't we still need to do equalization operation at the >>receiver >>( something trivial as i am doing now). >>Also can anyone suggest me a reference to other frequency domain >>equalization approaches. >> >>Thanks >> >> > > >Hi Mike, I am comparing performance in terms of bit error probability vs signal to noise ratio.
Reply by ●August 4, 20062006-08-04
>> >>Hi, >> >>I am working on a uncode 16-QAM OFDM simulation. The simulation works >fine >>when used in AWGN channel. I have a couple of questions now when i am >>simulating performance in multipath channels. >> >>1) What I want to understand through simulation is the effect of cyclic >>prefix length on performance in a multipath channel. >> >>For example consider that the channel impulse response is [0.7071 0 0 0 >0 >>0 0 0 0 0.7071], which is a channel with impulse response length 10 ( >>static paths with no fading component). >>Now if we consider four cases >>1)cp=0, >>2)cp=5, >>3) cp=11 >>4) cp =15 >> >>The first two cases, based on theory should result in ISI. My question >>is: >>when i perform the simulations i do not see much of a difference. Will >>the >>difference be seen only at high SNR's ? >> > >Having a cyclic prefix that is too small will cause one OFDM symbol to >interfere with another, which causes inter-carrier interference (ICI) in >the frequency domain. > >If the cyclic-prefix is sufficiently long, you will mitigate ICI, but >within a frequency bin, you will still have ISI. > >Both of these effects will defocus the constellations. > >When you say you don't see much of a difference, what criterion are you >using to compare the two scenarios? > >> >>2) Presently, I transmit a known OFDM symbol at the beginning of the >>simulation, At the receiver after FFT operation I divide the outputwith>>the known OFDM symbol values and store them as channel stateinformation>>for the remaining simulation. >>For all subsequent transmissions after the FFT operation i divide the >>output with the stored channel state information ( Zero forcing in >>frequency domain). >>My question is even for cyclic prefix lengths greater than the impulse >>response, don't we still need to do equalization operation at the >>receiver >>( something trivial as i am doing now). >>Also can anyone suggest me a reference to other frequency domain >>equalization approaches. >> >>Thanks >> >> > > >Hi Mike, I am comparing performance in terms of bit error probability vs signal to noise ratio.
Reply by ●August 6, 20062006-08-06
Hello, In order to check the effect of the CP, you take two values of CP one less than the max delay of your muli-path channel and othe other CP is greater then this delay... When the CP is less than the max delay of the channel, ICI will occur.. when will consecueltly increase the BER at the same SNR value "compared with the other case of CP < MAx delay".. Hope that helps Thanks Hany ramsay wrote:> >> > >>Hi, > >> > >>I am working on a uncode 16-QAM OFDM simulation. The simulation works > >fine > >>when used in AWGN channel. I have a couple of questions now when i am > >>simulating performance in multipath channels. > >> > >>1) What I want to understand through simulation is the effect of cyclic > >>prefix length on performance in a multipath channel. > >> > >>For example consider that the channel impulse response is [0.7071 0 0 0 > >0 > >>0 0 0 0 0.7071], which is a channel with impulse response length 10 ( > >>static paths with no fading component). > >>Now if we consider four cases > >>1)cp=0, > >>2)cp=5, > >>3) cp=11 > >>4) cp =15 > >> > >>The first two cases, based on theory should result in ISI. My question > >>is: > >>when i perform the simulations i do not see much of a difference. Will > >>the > >>difference be seen only at high SNR's ? > >> > > > >Having a cyclic prefix that is too small will cause one OFDM symbol to > >interfere with another, which causes inter-carrier interference (ICI) in > >the frequency domain. > > > >If the cyclic-prefix is sufficiently long, you will mitigate ICI, but > >within a frequency bin, you will still have ISI. > > > >Both of these effects will defocus the constellations. > > > >When you say you don't see much of a difference, what criterion are you > >using to compare the two scenarios? > > > >> > >>2) Presently, I transmit a known OFDM symbol at the beginning of the > >>simulation, At the receiver after FFT operation I divide the output > with > >>the known OFDM symbol values and store them as channel state > information > >>for the remaining simulation. > >>For all subsequent transmissions after the FFT operation i divide the > >>output with the stored channel state information ( Zero forcing in > >>frequency domain). > >>My question is even for cyclic prefix lengths greater than the impulse > >>response, don't we still need to do equalization operation at the > >>receiver > >>( something trivial as i am doing now). > >>Also can anyone suggest me a reference to other frequency domain > >>equalization approaches. > >> > >>Thanks > >> > >> > > > > > > > > Hi Mike, > > I am comparing performance in terms of bit error probability vs signal to > noise ratio.