0

I have a matlab code for OFDM system, in this code the auther added a zero padded subcarriers before the IFFT stage, in the middle of the raws, and added zero padded guard interval after IFFT, my question is that why i need to add zero padded subcarriers and what is the relation between the number of null subcarriers and sampling frequency??

user24907
  • 313
  • 1
  • 3
  • 12
  • Similar question with answer https://dsp.stackexchange.com/questions/40335/the-necessity-of-null-subcarrier-in-ofdm?rq=1 – AlexTP May 04 '17 at 19:44
  • Another one https://dsp.stackexchange.com/questions/17801/how-to-calculate-bandwidth-require-for-ofdm?noredirect=1&lq=1 – AlexTP May 04 '17 at 19:45
  • these links did't answer my question, i have a code some thing like this
    %Data generator (A) rand('state',0); a=-1+2round(rand(M,1)).'+i(-1+2*round(rand(M,1))).'; A=length(a); info=zeros(FS,1); info(1: (A/2)) = [ a(1: (A/2)).']; %Zero padding info((FS-((A/2)-1)) :FS) = [ a(((A/2)+1) :A).']; i want to why all of this padded subcarriers is added before IFFT ??     – user24907 May 04 '17 at 20:51
  • you can think simply that Fs is FFT size which is power of 2 thus there must be zero padding. However this reason is minor. The main reason is in my two links. – AlexTP May 04 '17 at 22:31
  • as i know the FFT size is the number of subcarriers N, not sampling frequency – user24907 May 05 '17 at 06:14
  • Nyquist sampling frequency equals to FFT size times subcarrier spacing. You can think FFT size, i.e. Fs in your code, is number of sc if you use all sc. But in this case I don't understand your question any more. Would the answer simply be because your data length A is smaller than Fs and you need to map 1 data sample to 1 sc? Is it not obvious? – AlexTP May 05 '17 at 07:02
  • in the code i work on , fs=192khz , the number of subcarriers=512, no. of symbols per subcarriers=20, fc= 38khz, BW=2560 HZ, OFDM symbol duration= 200 msec, so all of 512 subcarriers carry data, the auther inserted zero padded subcarriers equal to (no of samples in ofdm symbol- no of subcarriers) (38400-512) – user24907 May 05 '17 at 07:57
  • what is "no. of symbols per subcarriers=20" ? fc=38kHz is carrier frequency ? BW = 2560Hz is when you use 512 subcarriers ? Thus sc spacing $\Delta f$ must be 5Hz ? If you dont use null sc, then no zero padding, Nyquist fs = 5125=2560Hz < fs=192kHz, so it is ok. You can use another FFT size such as 2048, so you need 2048-512=1536 null sc, Nyquist fs = 20485=10.24kHz still < 192 kHz, still ok. Your fs=192kHz can theoretically handle the DFT size up to 38400. I think you should update your question for more details. – AlexTP May 05 '17 at 08:14
  • number of symbols per subcarrier=20, fc is the carrier frequency, yes 2560 is the bandwidth when 512 subcarriers are used, and yes the subcarrier spacing =5 Hz, in the code a very large number of zero padding subcarriers is inserted before IFFT step i don't know why?? this is the link of MSC which has the code at its end you can have a look https://brage.bibsys.no/xmlui/bitstream/handle/11250/2369065/347472_FULLTEXT01.pdf?sequence=2&isAllowed=y – user24907 May 05 '17 at 08:41
  • I think my answer is the best I have. There will be someone else with greater knowledge can give you a better answer. Regards. – AlexTP May 05 '17 at 13:55

2 Answers2

1

Zero padded OFDM (i.e.) adding zeros after IFFT is used to combat Inter symbol interference, suppose that there is multi path in the system that spans over L samples then atleast L zero samples are prefixed at the transmitter after IFFT so that these can be thrown away at the receiver and then recover ISI free IFFT frame / OFDM symbol.

The advantage of zero padding also has to do with power saving at the transmitter, the transmitter is simply switched off during the transmission of zero padded L samples duration. This has the advantage of saving power when compared to cyclic prefix OFDM which should then also transmit the cyclic prefix.

Zero padded OFDM though can leave to noise boosting and this is a disadvantage when compared to cyclic prefix OFDM. The equalization also would be non linear in case noise boosting is to be avoided. There is a lot of literature on zero padded OFDM you could refer to that as well.

Regarding the null sub carriers at the ends, these are implemented to reduce the ACLR and make the design of anti aliasing filters easier ( the cut off can be less sharp) hence save cost

Dsp guy sam
  • 2,610
  • 5
  • 18
0

The zero padding is there to minimize the effect of multipath.

Short explanation: over the air signal gets convoluted with channel response, this blurs the signal and redistribute its energy through a wider frequency range than the original. Ideally, if the channel response is shorter than the zero padding, the received signal can be accurately equalized using that information from higher frequencies by a simple deconvolution (in OFDM this is performed by multiplying in the frequency domain at the receiver). After a quick search through the link you posted, I cannot find the equalization step, so I guess this feature of OFDM is not implemented there.

Please, see if this makes sense to you and you can put all pieces together, right now I don't have time to further elaborate the answer.

Fusho
  • 421
  • 3
  • 9