Hi, I am trying to simulate a Reed-Solomon simulation over AWGN channel. When I plot the BER vs SNR the line stops every time in SNR 7db with BER almost 5*10^-5. I want the BER to be smaller than 10^-6. I tried to run for more codewords (e.x 10000) but the line stopped again at 5*10^-5. Any suggestions? Thanks.
BER in matlab
Started by ●August 16, 2008
Reply by ●August 16, 20082008-08-16
On Aug 16, 9:34�am, "magouliana" <h...@in.gr> wrote:> Hi, > I am trying to simulate a Reed-Solomon simulation over AWGN channel. > When I plot the BER vs SNR the line stops every time in SNR 7db with BER > almost 5*10^-5. I want the BER to be smaller than 10^-6. > I tried to run for more codewords (e.x 10000) but the line stopped again > at 5*10^-5. > Any suggestions? > Thanks.Without any information on how you are doing your simulation, it's hard to suggest anything. What do you mean by "the line stops". Are you saying that your simulation stops at SNR = 7 dB, or what you are not able to estimate results showing BER below 5e-5?
Reply by ●August 16, 20082008-08-16
julius <juliusk@gmail.com> wrote in news:b87e92d9-710d-48ba-90ce- 9a0f33f25e50@w24g2000prd.googlegroups.com:> On Aug 16, 9:34�am, "magouliana" <h...@in.gr> wrote: >> Hi, >> I am trying to simulate a Reed-Solomon simulation over AWGN channel. >> When I plot the BER vs SNR the line stops every time in SNR 7db with BER >> almost 5*10^-5. I want the BER to be smaller than 10^-6. >> I tried to run for more codewords (e.x 10000) but the line stopped again >> at 5*10^-5. >> Any suggestions? >> Thanks. > > Without any information on how you are doing your simulation, > it's hard to suggest anything. > > What do you mean by "the line stops". Are you saying that your > simulation stops at SNR = 7 dB, or what you are not able to estimate > results showing BER below 5e-5?It's possible the OP is running into problems with the accuracy of the AWGN. Most generators don't have very good tails, falling to zero at only a few sigma away from the mean. Regards, Allan
Reply by ●August 16, 20082008-08-16
On Aug 16, 12:27 pm, Allan Herriman <allanherri...@hotmail.com> wrote:> julius <juli...@gmail.com> wrote in news:b87e92d9-710d-48ba-90ce- > 9a0f33f25...@w24g2000prd.googlegroups.com: > > > > > On Aug 16, 9:34 am, "magouliana" <h...@in.gr> wrote: > >> Hi, > >> I am trying to simulate a Reed-Solomon simulation over AWGN channel. > >> When I plot the BER vs SNR the line stops every time in SNR 7db with BER > >> almost 5*10^-5. I want the BER to be smaller than 10^-6. > >> I tried to run for more codewords (e.x 10000) but the line stopped again > >> at 5*10^-5. > >> Any suggestions? > >> Thanks. > > > Without any information on how you are doing your simulation, > > it's hard to suggest anything. > > > What do you mean by "the line stops". Are you saying that your > > simulation stops at SNR = 7 dB, or what you are not able to estimate > > results showing BER below 5e-5? > > It's possible the OP is running into problems with the accuracy of the > AWGN. Most generators don't have very good tails, falling to zero at only > a few sigma away from the mean. > > Regards, > AllanDoes that create an error floor? John
Reply by ●August 16, 20082008-08-16
>On Aug 16, 9:34=A0am, "magouliana" <h...@in.gr> wrote: >> Hi, >> I am trying to simulate a Reed-Solomon simulation over AWGN channel. >> When I plot the BER vs SNR the line stops every time in SNR 7db withBER>> almost 5*10^-5. I want the BER to be smaller than 10^-6. >> I tried to run for more codewords (e.x 10000) but the line stoppedagain>> at 5*10^-5. >> Any suggestions? >> Thanks. > >Without any information on how you are doing your simulation, >it's hard to suggest anything. > >What do you mean by "the line stops". Are you saying that your >simulation stops at SNR =3D 7 dB, or what you are not able to estimate >results showing BER below 5e-5? >Thanks for the reply, Here is the output of the simulation, >> 1 : SNR = 0.0dB, BER = 1.44e-001 >> 2 : SNR = 1.0dB, BER = 1.12e-001 >> 3 : SNR = 2.0dB, BER = 7.60e-002 >> 4 : SNR = 3.0dB, BER = 4.57e-002 >> 5 : SNR = 4.0dB, BER = 1.63e-002 >> 6 : SNR = 5.0dB, BER = 4.25e-003 >> 7 : SNR = 6.0dB, BER = 6.02e-004 >> 8 : SNR = 7.0dB, BER = 0.00e+000 >> 9 : SNR = 8.0dB, BER = 0.00e+000 .. As you can see the BER becomes zero at SNR=7dB as a result the draw line stops at SNR=6dB, I have to represent a figure that BER will be smaller than 10^-6. I tried to run it for more codewords 10000 for exapmle (this example is with 1000) to produce more errors but the result is still the same.
Reply by ●August 16, 20082008-08-16
On Aug 16, 7:15�pm, "magouliana" <h...@in.gr> wrote:> >On Aug 16, 9:34=A0am, "magouliana" <h...@in.gr> wrote: > >> Hi, > >> I am trying to simulate a Reed-Solomon simulation over AWGN channel. > >> When I plot the BER vs SNR the line stops every time in SNR 7db with > BER > >> almost 5*10^-5. I want the BER to be smaller than 10^-6. > >> I tried to run for more codewords (e.x 10000) but the line stopped > again > >> at 5*10^-5. > >> Any suggestions? > >> Thanks. > > >Without any information on how you are doing your simulation, > >it's hard to suggest anything. > > >What do you mean by "the line stops". �Are you saying that your > >simulation stops at SNR =3D 7 dB, or what you are not able to estimate > >results showing BER below 5e-5? > > Thanks for the reply, > Here is the output of the simulation, > > �>> 1 � � : SNR = 0.0dB, �BER = 1.44e-001 > �>> 2 � � : SNR = 1.0dB, �BER = 1.12e-001 > �>> 3 � � : SNR = 2.0dB, �BER = 7.60e-002 > �>> 4 � � : SNR = 3.0dB, �BER = 4.57e-002 > �>> 5 � � : SNR = 4.0dB, �BER = 1.63e-002 > �>> 6 � � : SNR = 5.0dB, �BER = 4.25e-003 > �>> 7 � � : SNR = 6.0dB, �BER = 6.02e-004 > �>> 8 � � : SNR = 7.0dB, �BER = 0.00e+000 > �>> 9 � � : SNR = 8.0dB, �BER = 0.00e+000 > .. > As you can see the BER becomes zero at SNR=7dB as a result the draw line > stops at SNR=6dB, I have to represent a figure that BER will be smaller > than 10^-6. I �tried to run it for more codewords 10000 for exapmle (this > example is with 1000) to produce more errors �but the result is still the > same.It looks like your results are just fine. You can take a look at the application note by AHA regarding Reed- Solomon codes to evaluate the expected results (link: http://www.aha.com/show_pub.php?id=39). The BER at the output of the RS decoder is very low (depending on the exact code you are using) as the input BER gets better. For this code it is relatively easy to evaluate the code performance analytically instead of obtaining it using simulation. I do not think it is worth while running very long simulations to evaluate the code performance. Moti
Reply by ●August 16, 20082008-08-16
On Sat, 16 Aug 2008 12:15:19 -0500, "magouliana" <helo@in.gr> wrote:>>On Aug 16, 9:34=A0am, "magouliana" <h...@in.gr> wrote: >>> Hi, >>> I am trying to simulate a Reed-Solomon simulation over AWGN channel. >>> When I plot the BER vs SNR the line stops every time in SNR 7db with >BER >>> almost 5*10^-5. I want the BER to be smaller than 10^-6. >>> I tried to run for more codewords (e.x 10000) but the line stopped >again >>> at 5*10^-5. >>> Any suggestions? >>> Thanks. >> >>Without any information on how you are doing your simulation, >>it's hard to suggest anything. >> >>What do you mean by "the line stops". Are you saying that your >>simulation stops at SNR =3D 7 dB, or what you are not able to estimate >>results showing BER below 5e-5? >> > >Thanks for the reply, >Here is the output of the simulation, > > >> 1 : SNR = 0.0dB, BER = 1.44e-001 > >> 2 : SNR = 1.0dB, BER = 1.12e-001 > >> 3 : SNR = 2.0dB, BER = 7.60e-002 > >> 4 : SNR = 3.0dB, BER = 4.57e-002 > >> 5 : SNR = 4.0dB, BER = 1.63e-002 > >> 6 : SNR = 5.0dB, BER = 4.25e-003 > >> 7 : SNR = 6.0dB, BER = 6.02e-004 > >> 8 : SNR = 7.0dB, BER = 0.00e+000 > >> 9 : SNR = 8.0dB, BER = 0.00e+000 >.. >As you can see the BER becomes zero at SNR=7dB as a result the draw line >stops at SNR=6dB, I have to represent a figure that BER will be smaller >than 10^-6. I tried to run it for more codewords 10000 for exapmle (this >example is with 1000) to produce more errors but the result is still the >same.The BER curves for Reed-Solomon codes or concatenated code with outer Reed-Solomon codes get quite steep in their operating range. In other words, at some point increasing the SNR a little bit improves the BER a lot. When you're on that part of the curve and the error rate starts dropping to the 10e-6 to 10e-9 region, it takes a long time to accumulate enough errors in simulation to verify performance. I've not plotted the points you listed, but doing so on a log scale should allow you to graphically extrapolate a reasonable expected error rate for SNRs of 7dB an better. You may need to run a simulation a long time to get that error rate, and at some point everyone gives up and just plots enough points to show the expected trend. People who claim to have simulated, or even measured, error rates in the 10e-9 or better region are often met with justifiable suspiscion since the tests are often just not practical. I know of one case where a technology company claimed to have tested operating equipment in the lab to 10e-10 (or something like that, I don't remember the exact figure). I pointed out to some potential investors that at the claimed data rates and error metric those tests would have taken much longer than the the company had been in existence. At some point it just doesn't matter, so testing and simulation really only needs to be done to the point where there's confidence that things are as they should be. Eric Jacobsen Minister of Algorithms Abineau Communications http://www.ericjacobsen.org Blog: http://www.dsprelated.com/blogs-1/hf/Eric_Jacobsen.php
Reply by ●August 16, 20082008-08-16
On Aug 16, 12:15�pm, "magouliana" <h...@in.gr> wrote:> >On Aug 16, 9:34=A0am, "magouliana" <h...@in.gr> wrote: > >> Hi, > >> I am trying to simulate a Reed-Solomon simulation over AWGN channel. > >> When I plot the BER vs SNR the line stops every time in SNR 7db with > BER > >> almost 5*10^-5. I want the BER to be smaller than 10^-6. > >> I tried to run for more codewords (e.x 10000) but the line stopped > again > >> at 5*10^-5. > >> Any suggestions? > >> Thanks. > > >Without any information on how you are doing your simulation, > >it's hard to suggest anything. > > >What do you mean by "the line stops". �Are you saying that your > >simulation stops at SNR =3D 7 dB, or what you are not able to estimate > >results showing BER below 5e-5? > > Thanks for the reply, > Here is the output of the simulation, > > �>> 1 � � : SNR = 0.0dB, �BER = 1.44e-001 > �>> 2 � � : SNR = 1.0dB, �BER = 1.12e-001 > �>> 3 � � : SNR = 2.0dB, �BER = 7.60e-002 > �>> 4 � � : SNR = 3.0dB, �BER = 4.57e-002 > �>> 5 � � : SNR = 4.0dB, �BER = 1.63e-002 > �>> 6 � � : SNR = 5.0dB, �BER = 4.25e-003 > �>> 7 � � : SNR = 6.0dB, �BER = 6.02e-004 > �>> 8 � � : SNR = 7.0dB, �BER = 0.00e+000 > �>> 9 � � : SNR = 8.0dB, �BER = 0.00e+000 > .. > As you can see the BER becomes zero at SNR=7dB as a result the draw line > stops at SNR=6dB, I have to represent a figure that BER will be smaller > than 10^-6. I �tried to run it for more codewords 10000 for exapmle (this > example is with 1000) to produce more errors �but the result is still the > same.Um. If you are trying to reliably characterize an event that happens once every 10^5 times, you'll need to test in the order of several orders of magnitude higher than 10^5, right? Are you re-setting the random number generator's seed every once in a while? Sounds to me that you just have to be more patient and run more simulations.
