> Could you recommend which book talks about good codes for QAM?

Look for "Pragmatic-TCM".

Thank you. As far as I know, MLSE is good, but it is too complex for a
highly selective channel.

Could you recommend which book talks about good codes for QAM? I have
digital communication written by Proakis, Error control codes written
by Shu Lin. I have still no clue about good codes for QAM. My limited
coding knowledge is limited to inner convolutional code plus outer RS
code, such as in DVB, 802.16 standards.

fl wrote:

> I have to add that I use frequency domain as feedforward equalizer and
> time domain as feedback equalizer. The criteria is MMSE. The simple
> frequency domain MMSE equalizer was also implemented. The results of
> both equalizer are similar.

Equalizer + symbol-by-symbol demodulation is very weak method for 
processing the signal distorted by the selective fading channel. MLSE 
algorithm should be used.

Go to the library and get a book on digital communications by Proakis.

> 
> I have read one paper uses rate 1/4, K=4 Generator matrix:13,15,15,17,
> dfree=13 convolutional code for BPSK.

So what?

> For QAM64, are there some
> powerful convolutional code besides (133,171) code?

There are great many good codes + constellations designed specifically 
for QAM. Don't reinvent the wheel, go to the library.


Vladimir Vassilevsky

DSP and Mixed Signal Design Consultant

http://www.abvolt.com

I have to add that I use frequency domain as feedforward equalizer and
time domain as feedback equalizer. The criteria is MMSE. The simple
frequency domain MMSE equalizer was also implemented. The results of
both equalizer are similar.

I have read one paper uses rate 1/4, K=4 Generator matrix:13,15,15,17,
dfree=13 convolutional code for BPSK. For QAM64, are there some
powerful convolutional code besides (133,171) code?

Thank you for your concern

First of all, the direct encoding of the data with a conventional 
convolution code is not going to fit the QAM64 modulation. There is a 
lot of information about the codes designed specifically to be used with 
QAM. The key word is "set partitioning", you can also refer to the 
classic book of Proakis on the digital communication.

Secondly, the codes you are mentioning are designed for good channels. 
The design goal was to accomplish the very low output BER. This is what 
the RS outer code is for. If the packet transmission fails it is often 
better to retransmit the whole packet rather then to carry the heavy FEC 
overhead.


Vladimir Vassilevsky

DSP and Mixed Signal Design Consultant

http://www.abvolt.com



fl wrote:

> Hello,
> I am doing some simulation about SC-FDE in a typical Hiperlan2 channel
> B environment. For each burst, the channel can be thought of as static.
> I use channel coding similar to 802.11 standard  and interested most in
> QAM64 modulation. I have a question about the simulation result.
> 
> When the channel is a mild frequency selective fading one, channel
> coding is really to improve performance. On the contrary, when channel
> is worse, channel coding result is even worse than that of no coding. I
> have seen some published simulation results about 802.11 or 802.16 use
> a lot of channel realizations to get an average performance curve.
> 
> And, I am not sure whether it is highly possible the result of using
> channel coding (convolutional coding) in a worse channel cannot improve
> performance? What's that possibility? Besides SC-FDE, can this happen
> in OFDM? My concern is that the system would change the modulation
> format to a low level modualtion in case of worse channel because in
> these application channel is about static. As for as I know, the
> channel coding defined for QAM64 in 802.11 is not a powerful one. To
> evaluate an equalizer of a high level modulation in a WLAN environment,
> is it fair to include a lot of bad channels? Or, is it useful to get
> result in such a bundle of channel realizations?
> 
> I have used 1/2 convolutional coding without puncturing. The result on
> bad channel is not good either. In 802.16, besides inner convolutional
> coding, there is outer RS code. I suspect if the inner convolutional
> coding cannot do the job, the outer RS can improve performance. Both
> 802.11 and 802.16 use the typical 1/2 (133,171), K=7 code. Can you tell
> me the effecitveness of this code in the standard defined statistical
> channel?
> 
> I have checked scatter mapping, interleaving and channel coding. It
> seems there is no problem for my coding. I have two typical channel
> realizations from one paper, one is bad one is good. My equalizer has
> some problem?
> 
> Could you shed some lights on this topic?
> 
> Thanks in advance
> 

Hello,
I am doing some simulation about SC-FDE in a typical Hiperlan2 channel
B environment. For each burst, the channel can be thought of as static.
I use channel coding similar to 802.11 standard  and interested most in
QAM64 modulation. I have a question about the simulation result.

When the channel is a mild frequency selective fading one, channel
coding is really to improve performance. On the contrary, when channel
is worse, channel coding result is even worse than that of no coding. I
have seen some published simulation results about 802.11 or 802.16 use
a lot of channel realizations to get an average performance curve.

And, I am not sure whether it is highly possible the result of using
channel coding (convolutional coding) in a worse channel cannot improve
performance? What's that possibility? Besides SC-FDE, can this happen
in OFDM? My concern is that the system would change the modulation
format to a low level modualtion in case of worse channel because in
these application channel is about static. As for as I know, the
channel coding defined for QAM64 in 802.11 is not a powerful one. To
evaluate an equalizer of a high level modulation in a WLAN environment,
is it fair to include a lot of bad channels? Or, is it useful to get
result in such a bundle of channel realizations?

I have used 1/2 convolutional coding without puncturing. The result on
bad channel is not good either. In 802.16, besides inner convolutional
coding, there is outer RS code. I suspect if the inner convolutional
coding cannot do the job, the outer RS can improve performance. Both
802.11 and 802.16 use the typical 1/2 (133,171), K=7 code. Can you tell
me the effecitveness of this code in the standard defined statistical
channel?

I have checked scatter mapping, interleaving and channel coding. It
seems there is no problem for my coding. I have two typical channel
realizations from one paper, one is bad one is good. My equalizer has
some problem?

Could you shed some lights on this topic?

Thanks in advance