Page 38 - ITU Journal Future and evolving technologies Volume 2 (2021), Issue 7 – Terahertz communications
P. 38

ITU Journal on Future and Evolving Technologies, Volume 2 (2021), Issue 7






























          Fig. 3 – Reverse waterfall plot for M‑QAM links, M = 4, 16, over distances from 0 to 20 km with various SNRs. SNR values of 0, 10, 20, 30, 40, and 50 dB
          are designated by markers 5 k, 5, 3 6  1 respectively, and in inite SNR is denoted by a thick, solid line with no marker. Atmospheric conditions  ,  ,  , and
          are the same as in Fig. 2. The fact that the 40 dB, 50 dB, and in inite SNR curves are almost indistinguishable because beyond 40 dB, dispersion is the
          dominant source of errors, rather than noise.


          (according to point ‘c’ on Fig. 3). However, increasing the   this  is  how  future  terahertz  links  should  operate;
         modulation order to 64 offers improvement according to   indeed,  variable  bandwidth  channels  would  be  both  an
         Fig. 4, reducing the error rate to 4.7 × 10 −5  ( point ‘b’ on the   engineering  and a regulatory challenge, and probably are
         plot).  In  this  case,  the  improvement  again  relies  on  the   not  appropriate  for  most  circumstances.  Rather,  we
         increased  spectral  ef iciency  of  the  higher  order   chose  to  present  our  data  in  this  manner  because
         modulation scheme, with the stipulation that the SNR be   allowing  bandwidth  to  vary  with  modulation  order
         40  dB  or  greater.  This  exacting  constraint  on  SNR  arises   makes  the  counter‑intuitive  behavior  of  the  spectrally‑
         from  the  small  spacing  between  symbol  decision   ef icient,  low  bandwidth  links  (that  is,  256‑QAM)  the
         boundaries for the higher‑order link, made even stricter   most  clear  and  explicit.   This  does  not  change  our
         by  the  fact  that  dispersion,  though  reduced,  has  still   simulation  results;  it  is  just  a  data‑ presentation  choice.
         shifted  some  of  the  received  symbols  closer  to  the   Varying  the  bandwidth  of  the  link  was  the  best  way  to
         decision boundaries.                                  show that decreasing the bandwidth does  not  necessarily
                                                               decrease  the  error  rate,  and  that  modulation  type
                                                               becomes an important factor due to atmospheric GVD.
          Thirdly, notice from Fig. 2 that, for the 60 Gb/s case pre‑
          sented here, there are some modulation schemes that, in
          general,  constitute  poor  choices  for  a  link  without  dis‑
                                                               Finally, all discussion up until this point has been focused
          persion  compensation.  Namely,  a  256‑QAM  scheme  has
                                                               on  communication  links  in  which  dispersion  is
          a  worse  error  rate  at  all  distances  and  SNRs  than  ei‑
                                                               uncompen‑  sated.  Though  GVD  has  not  historically
          ther  16‑QAM  or  64‑QAM  (with  the  exception  of  a  slight
                                                               been  a  concern  for  wireless  communication  systems
          and  insigni icant  region  around  14  km  in  the  noise‑free
                                                               (owing  to  the  com‑  paratively  narrow  bandwidth  of
          case,  where  it  has  performance  marginally  better  than
                                                               legacy  microwave  com‑  munication  links),  it  has  been
          the 16‑QAM scheme).  While BPSK and 4‑QAM do under‑
                                                               extensively   investigated   in   iber   optics,   where
          perform  256‑QAM  over  long  distances  with  higher  SNR,
                                                               dispersion‑compensating   technology  is   relatively
          there are also two other modulation types (16‑QAM and
                                                               mature.  Additionally,  there  are  other  forms of temporal
          64‑QAM)  that  outperform  256‑QAM  in  nearly  all  situa‑
                                                               dispersion  that  have  been  identi ied,  studied,  and
          tions,  so  while  256‑QAM  is  an  improvement  over  some
                                                               compensated  in  existing  wireless  links,  which  often
          modulation schemes, it is never the best choice (and this
                                                               arise from multipath propagation.  Although atmospheric
          holds true for all higher SNRs as well).
                                                               GVD  is  a  new  phenomena  for  wireless  links,  dispersion
                                                               in general is not.  This may lead some to think that since
         At  this  point  in  the  discussion,  there  are  a  few  assump‑   dispersion  can  and  has  been  compensated  by  both  pho‑
         tions  that  need  to  be  addressed.  One  point  of  concern   tonic [27] and electronic [28, 29] means, then these tech‑
         may  be  that  in  most  applications,  the  bandwidth  of  a   nologies would be readily adapted for use in a terahertz
         wireless  link  is   ixed  and  the  data  rate  varies  with   wireless communication system.  Speci ically,  it might be
         modulation  order,  while  in  the  results  we  present,  the   assumed  digital  signal  processing   ilters,  also  known  as
         band‑width  varies  with  modulation  order  while  data   equalizers, will be able to compensate GVD and thus ren‑
         rate  is  held  constant.  However,  we  are  not  proposing  der the problem of GVD irrelevant.
          26                                 © International Telecommunication Union, 2021
   33   34   35   36   37   38   39   40   41   42   43