Coding Gain Definition
in [DSP]
|
Prev: How can the chirp-z transform be used in resampling?
Next: Code Composer Studio not simulating interrupts
From: JAlbertoDJ on 25 Feb 2007 18:15 I have a doubt. Gain code is defined as the reduction, usually expresed in decibels, in the required Eb/No to achieve a especified error probability of the coded system over an uncoded system with the same modulation and channel characteristics. But, what about Bit rate? The definition considers the same Bit Rate in both cases?????. For example: I have implemented a Viterbi Receptor(r=1/2 k=7. Soft Decision and df=10) for a BFSK modulation. Without coding, I need 5 dBs more than coding, but Bit Rate without coding is the double that coding Bit Rate with. Then, i could transmit without coding slower, at the same Bit Rate (the half) than coding. So i could win 3dBs and the different would not be 5 dBs, would be only 2 dBs. Then, if I read in a Book that Gain Coded for a code is X, i want to know if that parameter is for the same Bit Rate in both cases. Do you understand my question?
From: Vladimir Vassilevsky on 25 Feb 2007 18:37 JAlbertoDJ wrote: > I have a doubt. Gain code is defined as the reduction, usually expresed in > decibels, in the required Eb/No to achieve a especified error probability > of the coded system over an uncoded system with the same modulation and > channel characteristics. > But, what about Bit rate? The definition considers the same Bit Rate in > both cases?????. > > For example: I have implemented a Viterbi Receptor(r=1/2 k=7. Soft > Decision and df=10) for a BFSK modulation. > > Without coding, I need 5 dBs more than coding, but Bit Rate without coding > is the double that coding Bit Rate with. Then, i could transmit without > coding slower, at the same Bit Rate (the half) than coding. So i could win > 3dBs and the different would not be 5 dBs, would be only 2 dBs. Your reasoning is correct. When evaluating the performance of the coded system, one has to compare the coding gain against the energy per bit reduction. What makes it even worse are the implementation losses of the modem. The losses are increasing when the SNR is dropping, so for the peer-to-peer comparison the modem for the coded modulation performs worse then for that for the the uncoded. Thus, the coding does make sense only if the output BER has to be low. The simple codes like the one you are using will not provide for any gain at all if the BER is about ~1e-2. Vladimir Vassilevsky DSP and Mixed Signal Design Consultant http://www.abvolt.com
From: Eric Jacobsen on 26 Feb 2007 12:54
On Sun, 25 Feb 2007 17:15:01 -0600, "JAlbertoDJ" <nietorosdj(a)yahoo.es> wrote: > >I have a doubt. Gain code is defined as the reduction, usually expresed in >decibels, in the required Eb/No to achieve a especified error probability >of the coded system over an uncoded system with the same modulation and >channel characteristics. > >But, what about Bit rate? The definition considers the same Bit Rate in >both cases?????. > >For example: I have implemented a Viterbi Receptor(r=1/2 k=7. Soft >Decision and df=10) for a BFSK modulation. > >Without coding, I need 5 dBs more than coding, but Bit Rate without coding >is the double that coding Bit Rate with. Then, i could transmit without >coding slower, at the same Bit Rate (the half) than coding. So i could win >3dBs and the different would not be 5 dBs, would be only 2 dBs. > >Then, if I read in a Book that Gain Coded for a code is X, i want to know >if that parameter is for the same Bit Rate in both cases. > >Do you understand my question? I think I understand your question. One of the conveniences of using Eb/No as a metric is that it is independent of data rate or even modulation type for comparing power efficiency. It is an indicator only of the power efficiency of the transmission of a bit of information in Gaussian noise. Since power efficiency is a very important metric in wireless systems it is a very useful calculation. If, however, you are interested in spectral efficiency rather than power efficiency, i.e., how much data can you get through a given bandwidth, then Eb/No is NOT a relevant metric for that question. Coding gain is, generally, independent of bit rate. A given code will provide the same gain at 1Mbps as it does at 2Mbps. Eric Jacobsen Minister of Algorithms Abineau Communications http://www.ericjacobsen.org |