BER performance of Viterbi decoder

Started by lanbaba January 10, 2006
Hi all,

does anyone know what BER should I expect after the Viterbi decoding if
the channel BER is 7.0e-3? Constraint length 7, Generator (177, 133),
punturing to code rate 5/6. Thanks!

Cheers, Lanbaba


lanbaba wrote:

> Hi all, > > does anyone know what BER should I expect after the Viterbi decoding if > the channel BER is 7.0e-3? Constraint length 7, Generator (177, 133), > punturing to code rate 5/6. Thanks! >
Indeed the R = 5/6 K = 7 code looks inadequate for the input BER = 7e-3, and I doubt if there is going to be any BER improvement at all. I would use a code with R ~ 1/2...1/3 for this kind of BER, expecting the improvement about a factor of 10. Vladimir Vassilevsky DSP and Mixed Signal Design Consultant http://www.abvolt.com
On Tue, 10 Jan 2006 08:08:29 -0600, "lanbaba" <lanbaba@gmx.ch> wrote:

>Hi all, > >does anyone know what BER should I expect after the Viterbi decoding if >the channel BER is 7.0e-3? Constraint length 7, Generator (177, 133), >punturing to code rate 5/6. Thanks! > >Cheers, Lanbaba
There are a lot of curves available for the (177, 133) code, since it is nearly universally used. I'm assuming you mean for antipodal (BPSK) signalling in AWGN, and that curve shouldn't be too difficult to find. You might try some of the common FEC/ECC textbooks or Lindsay and Simon. Data sheets used to be a good source for this, and perhaps somebody like AHA has it on their website. If I had one handy I'd make it available, but I don't. Eric Jacobsen Minister of Algorithms, Intel Corp. My opinions may not be Intel's opinions. http://www.ericjacobsen.org
>Indeed the R = 5/6 K = 7 code looks inadequate for the input BER = 7e-3, >and I doubt if there is going to be any BER improvement at all. I would >use a code with R ~ 1/2...1/3 for this kind of BER, expecting the >improvement about a factor of 10.
Hi Vladimir, actually I share your view. But when I saw a performance curve of the Xilinx Viterbi decoder I was a bit surprised. In Figure 29 of their data sheet in www.xilinx.com/bvdocs/ipcenter/data_sheet/viterbi.pdf a large gain is achieved for channel BER 7e-3 even with code rate 7/8! Any advice? Cheers, Zhenlan
>You might try some of the common FEC/ECC textbooks or Lindsay and >Simon. Data sheets used to be a good source for this, and perhaps >somebody like AHA has it on their website. >
Eric, Who/what is AHA? Lanbaba

lanbaba wrote:

>>Indeed the R = 5/6 K = 7 code looks inadequate for the input BER = 7e-3, >>and I doubt if there is going to be any BER improvement at all. I would >>use a code with R ~ 1/2...1/3 for this kind of BER, expecting the >>improvement about a factor of 10. > > > Hi Vladimir, > > actually I share your view. But when I saw a performance curve of the > Xilinx Viterbi decoder I was a bit surprised. In Figure 29 of their data > sheet in www.xilinx.com/bvdocs/ipcenter/data_sheet/viterbi.pdf a large > gain is achieved for channel BER 7e-3 even with code rate 7/8! Any > advice?
From fig.29 it is clear that the 7/8 code actually results in the loss in BER at the channel BER = 7e-3, and this is what is anticipated. VLV
> From fig.29 it is clear that the 7/8 code actually results in the loss >in BER at the channel BER = 7e-3, and this is what is anticipated. > >VLV >
Sorry, but why that? In Fig. 29 of the Xilinx data sheet, the coded BER is about 0.0008 with code rate 7/8 for uncoded BER = 0.007, 10 times lower, or do I see something fundamentally wrong? LBB
I wrote a small matlab program and get coded BER = 0.0110 for uncoded BER =
0.0078 for the aformentioned code with hard decision Viterbi. This verifies
Vladimir's statement. Still wondering about Xilinx's data sheet.

LBB

    

On Tue, 10 Jan 2006 16:12:16 -0600, "lanbaba" <lanbaba@gmx.ch> wrote:

> >>You might try some of the common FEC/ECC textbooks or Lindsay and >>Simon. Data sheets used to be a good source for this, and perhaps >>somebody like AHA has it on their website. >> > >Eric, > >Who/what is AHA? > >Lanbaba
www.aha.com Eric Jacobsen Minister of Algorithms, Intel Corp. My opinions may not be Intel's opinions. http://www.ericjacobsen.org
lanbaba wrote:
> > From fig.29 it is clear that the 7/8 code actually results in the loss > >in BER at the channel BER = 7e-3, and this is what is anticipated. > > > >VLV > > > > Sorry, but why that? In Fig. 29 of the Xilinx data sheet, the coded BER is > about 0.0008 with code rate 7/8 for uncoded BER = 0.007, 10 times lower, or > do I see something fundamentally wrong? > > LBB
Thats what I see also.... I see approx 2dB gain????? please explain.... Mark