Forums

Fair comparision: Reed Solomon code and convolutional code

Started by cpshah99 July 15, 2008
Hello People

I want to compare Reed Solomon code [255,239] and rate 1/2 convolution
code with constraint length 5, i.e. generator polynomials are [23 35]_8;

Now, my channel is frequency selective, doppler effect and noise.

Also the way I am scaling the noise is 

sigma=sqrt(0.5*Eb/(Rc*snr_lin))

where for RS, the Rc=239/255 and for convolution encoder Rc=1/2;


Is this fair comparision?

What should I expect in terms of performance? Which one will perform
better?

Your help will be greatly appreciated.

Chintan
>Hello People > >I want to compare Reed Solomon code [255,239] and rate 1/2 convolution >code with constraint length 5, i.e. generator polynomials are [23 35]_8; > >Now, my channel is frequency selective, doppler effect and noise. > >Also the way I am scaling the noise is > >sigma=sqrt(0.5*Eb/(Rc*snr_lin)) > >where for RS, the Rc=239/255 and for convolution encoder Rc=1/2; > > >Is this fair comparision? > >What should I expect in terms of performance? Which one will perform >better? > >Your help will be greatly appreciated. > >Chintan >
This doesn't look like a fair comparison as their coding rate difference is too significant. Also, if you add random noise, convolution code is sure to win. Michael
>>Hello People >> >>I want to compare Reed Solomon code [255,239] and rate 1/2 convolution >>code with constraint length 5, i.e. generator polynomials are [23
35]_8;
>> >>Now, my channel is frequency selective, doppler effect and noise. >> >>Also the way I am scaling the noise is >> >>sigma=sqrt(0.5*Eb/(Rc*snr_lin)) >> >>where for RS, the Rc=239/255 and for convolution encoder Rc=1/2; >> >> >>Is this fair comparision? >> >>What should I expect in terms of performance? Which one will perform >>better? >> >>Your help will be greatly appreciated. >> >>Chintan >> > >This doesn't look like a fair comparison as their coding rate difference >is too significant. > >Also, if you add random noise, convolution code is sure to win. > >Michael >
%%%% Hi Thanks for your reply. I have not much experience of RS codes, I am jsut doing it because it is the requirement. Anyways, if you forget doppler for now, just frequency selective channel and noise (random but scaled properly), convolution code performs better than RS (same set up). What should I do to make it fair? or is this comparision valid at all? Thanks chintan
On Jul 15, 3:03 pm, "cpshah99" <cpsha...@rediffmail.com> wrote:

> What should I do to make it fair? or is this comparision valid at all?
Which is better for chiseling wood, a screwdriver or a ball-peen hammer? Neither is very good for the task, but you could do a little bit, maybe, with the screwdriver and hardly anything with the hammer. So, "clearly" the screwdriver is the "better" tool. But, ah, when you have to drive nails, the hammer is far far better than the screwdriver (think of CDs and DVDs), isn't it? And when you need to drive screws, the screwdriver is the best tool, though you might be able to do a little with a chisel if you have a flat-head screw (forget Phillips heads!) and nothing with the hammer.
>On Jul 15, 3:03 pm, "cpshah99" <cpsha...@rediffmail.com> wrote: > >> What should I do to make it fair? or is this comparision valid at all? > > >Which is better for chiseling wood, a screwdriver >or a ball-peen hammer? Neither is very good for >the task, but you could do a little bit, maybe, with >the screwdriver and hardly anything with the >hammer. So, "clearly" the screwdriver is the >"better" tool. But, ah, when you have to drive >nails, the hammer is far far better than the >screwdriver (think of CDs and DVDs), isn't it? >And when you need to drive screws, the >screwdriver is the best tool, though you might >be able to do a little with a chisel if you have a >flat-head screw (forget Phillips heads!) and >nothing with the hammer. >
%%%%% Hi Thanks for this carpentry stuff which I already know!!!! chintan
On Jul 16, 2:10 am, "cpshah99" <cpsha...@rediffmail.com> wrote:
> >On Jul 15, 3:03 pm, "cpshah99" <cpsha...@rediffmail.com> wrote: > > >> What should I do to make it fair? or is this comparision valid at all? > > >Which is better for chiseling wood, a screwdriver > >or a ball-peen hammer? Neither is very good for > >the task, but you could do a little bit, maybe, with > >the screwdriver and hardly anything with the > >hammer. So, "clearly" the screwdriver is the > >"better" tool. But, ah, when you have to drive > >nails, the hammer is far far better than the > >screwdriver (think of CDs and DVDs), isn't it? > >And when you need to drive screws, the > >screwdriver is the best tool, though you might > >be able to do a little with a chisel if you have a > >flat-head screw (forget Phillips heads!) and > >nothing with the hammer. > > %%%%% > > Hi > > Thanks for this carpentry stuff which I already know!!!! > > chintan
Good. Now tell your boss that the convolutional code is going to be better in almost any application and the Reed-Solomon code is useless (unless he has a CD or DVD (or a concatenated coding scheme) in mind.) Reed-Solomon codes are not very good as channel codes, but they make great straw men to be knocked down while touting the merits of a coding scheme that you really ought to be using in the first place.
>On Jul 16, 2:10 am, "cpshah99" <cpsha...@rediffmail.com> wrote: >> >On Jul 15, 3:03 pm, "cpshah99" <cpsha...@rediffmail.com> wrote: >> >> >> What should I do to make it fair? or is this comparision valid at
all?
>> >> >Which is better for chiseling wood, a screwdriver >> >or a ball-peen hammer? Neither is very good for >> >the task, but you could do a little bit, maybe, with >> >the screwdriver and hardly anything with the >> >hammer. So, "clearly" the screwdriver is the >> >"better" tool. But, ah, when you have to drive >> >nails, the hammer is far far better than the >> >screwdriver (think of CDs and DVDs), isn't it? >> >And when you need to drive screws, the >> >screwdriver is the best tool, though you might >> >be able to do a little with a chisel if you have a >> >flat-head screw (forget Phillips heads!) and >> >nothing with the hammer. >> >> %%%%% >> >> Hi >> >> Thanks for this carpentry stuff which I already know!!!! >> >> chintan > >Good. Now tell your boss that the convolutional code >is going to be better in almost any application and the >Reed-Solomon code is useless (unless he has a CD or >DVD (or a concatenated coding scheme) in mind.) >Reed-Solomon codes are not very good as channel >codes, but they make great straw men to be knocked >down while touting the merits of a coding scheme that >you really ought to be using in the first place. >
%%%% Hi Thanks again. I had expected convolutional code to do better and it is doing the same. But another advantage of RS code is that it corrects burst error, which is problem in fading channel, and because of burst errors DFE (Decision Feedback Equaliser) will not work properly. But if you use RS, this problem is solved. Now to solve the same problem with convolution encoder, u will have to use 'interleaver', otherwise forget that even BCJR will decode the burst errors. So my system with convolution encoder and interleaver outperforms RS system. Everything is same when I am comparing both the systems, channel, doppler, noise, adaptive algorithm for DFE. But I was wondering that if I can claim that given the channel perticular channel conditions, convolution code will outperform RS. BTW, nobody is boss. Chintan
cpshah99 wrote:

   ...

> Thanks for this carpentry stuff which I already know!!!!
Some of us think that knowledge from one field can carry over to a related one, and go so far as believing that analogy can have a useful place in explanation. Others feel that the only way to deal with a subject is in its own terms, with no reference to parallel disciplines. You apparently belong to the second group. I'll try to keep that in mind when answering your future questions. Jerry -- Engineering is the art of making what you want from things you can get. &#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;
On Wed, 16 Jul 2008 08:53:36 -0500, "cpshah99"
<cpshah99@rediffmail.com> wrote:

>>On Jul 16, 2:10 am, "cpshah99" <cpsha...@rediffmail.com> wrote: >>> >On Jul 15, 3:03 pm, "cpshah99" <cpsha...@rediffmail.com> wrote: >>> >>> >> What should I do to make it fair? or is this comparision valid at >all? >>> >>> >Which is better for chiseling wood, a screwdriver >>> >or a ball-peen hammer? Neither is very good for >>> >the task, but you could do a little bit, maybe, with >>> >the screwdriver and hardly anything with the >>> >hammer. So, "clearly" the screwdriver is the >>> >"better" tool. But, ah, when you have to drive >>> >nails, the hammer is far far better than the >>> >screwdriver (think of CDs and DVDs), isn't it? >>> >And when you need to drive screws, the >>> >screwdriver is the best tool, though you might >>> >be able to do a little with a chisel if you have a >>> >flat-head screw (forget Phillips heads!) and >>> >nothing with the hammer. >>> >>> %%%%% >>> >>> Hi >>> >>> Thanks for this carpentry stuff which I already know!!!! >>> >>> chintan >> >>Good. Now tell your boss that the convolutional code >>is going to be better in almost any application and the >>Reed-Solomon code is useless (unless he has a CD or >>DVD (or a concatenated coding scheme) in mind.) >>Reed-Solomon codes are not very good as channel >>codes, but they make great straw men to be knocked >>down while touting the merits of a coding scheme that >>you really ought to be using in the first place. >> >%%%% > >Hi > >Thanks again. I had expected convolutional code to do better and it is >doing the same. > >But another advantage of RS code is that it corrects burst error, which is >problem in fading channel, and because of burst errors DFE (Decision >Feedback Equaliser) will not work properly. But if you use RS, this problem >is solved. > >Now to solve the same problem with convolution encoder, u will have to use >'interleaver', otherwise forget that even BCJR will decode the burst >errors. > >So my system with convolution encoder and interleaver outperforms RS >system. > >Everything is same when I am comparing both the systems, channel, doppler, >noise, adaptive algorithm for DFE. > >But I was wondering that if I can claim that given the channel perticular >channel conditions, convolution code will outperform RS. > >BTW, nobody is boss. > >Chintan
The bottom line is that care has to be taken to match codes to channels. The assumptions about the error distributions in the channel can make a big difference in how the code is applied or architected into the system. While on the surface the fact that RS codes can correct "burst errors" is true, the nature of the "burst error" can make a big difference. Often the length of the error burst can overwhelm the correcting capability of the RS code. The performance curve (in an error probability vs SNR sense) of a system with an RS code is pretty steep. What often happens is that the region from performing well to performing poorly is small, and RS codes exhibit a sometimes undesirable characteristic of aliasing codewords when in the "poor" region. i.e., many of the codewords coming out of the decoder will be marked as corrected, but they'll be incorrect. At least with a convolution code and a Viterbi decoder you can tell when it's unlocked and spewing garbage. So the nature of the error bursts matters, and the sensitivity of the system to output errors matters. Interleavers are often used to reduce the length of error bursts so that the channel code works better. There are all kinds of considerations, and I think Dr. Sarwate's initial response hints at the complexities of picking the appropriate tool depending on the situation. Eric Jacobsen Minister of Algorithms Abineau Communications http://www.ericjacobsen.org Blog: http://www.dsprelated.com/blogs-1/hf/Eric_Jacobsen.php
>cpshah99 wrote: > > ... > >> Thanks for this carpentry stuff which I already know!!!! > >Some of us think that knowledge from one field can carry over to a >related one, and go so far as believing that analogy can have a useful >place in explanation. Others feel that the only way to deal with a >subject is in its own terms, with no reference to parallel disciplines. >You apparently belong to the second group. I'll try to keep that in mind
>when answering your future questions. > >Jerry >-- >Engineering is the art of making what you want from things you can get. >&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533;&#65533; >
%%%% Hi Jerry I am really sorry but you misunderstood me. But I am also from first group, because when I explain something to some stranger or to my friends what I am doing (who are not from EE filed) I use real life examples, because this is what makes it simple isn't it? But sometimes you are in such a state that you dont know what to do. So I just replied like that. I am asking abt this comparision because I dont want to cliam something which is not true on paper. And please use these sort of examples when u answer my posts next time. Sorry for any misunderstanding. :-) chintan