From: owpex on
Hello

I wanted to know the difference that can exist between(among) a flow of
bits that goes out of a decodificador viterbi + Reed solomon and the flow
of bits originally in a transmission of tdt cofdm

That is to say: do I obtain the same result to the exit of the corrector
Reed Solomon with a BER before Viterbi of 9-6 that with a BER before
Viterbi of 2-2?

Said otherwise:

Do I obtain the same result to the exit of the corrector Reed Solomon with
a BER after Viterbi of 9-6 that with a BER after Viterbi of 2-2?

Muchisimas thank you


From: Eric Jacobsen on
On 12/21/2009 7:01 AM, owpex wrote:
> Hello
>
> I wanted to know the difference that can exist between(among) a flow of
> bits that goes out of a decodificador viterbi + Reed solomon and the flow
> of bits originally in a transmission of tdt cofdm
>
> That is to say: do I obtain the same result to the exit of the corrector
> Reed Solomon with a BER before Viterbi of 9-6 that with a BER before
> Viterbi of 2-2?
>
> Said otherwise:
>
> Do I obtain the same result to the exit of the corrector Reed Solomon with
> a BER after Viterbi of 9-6 that with a BER after Viterbi of 2-2?
>
> Muchisimas thank you
>
>

Short answer: no.

You should be able to find via some google searches plots of BER vs
Eb/No or BER vs SNR for a typical Viterbi decoder, typical Reed-Solomon
(only) decoder, and typical concatenated CC-RS system with a Viterbi
decoder and RS decoder. In all cases the curves have a slope, so there
is a transfer function of input BER to output BER. For the Viterbi
decoder by itself the slope is not as steep as it is with a system that
uses an RS decoder, either by itself or concatenated with a Viterbi, but
there is still a slope.

The slope gets difficult to measure below error rates of 10e-9 or 10e-10
or so just because the tests take so long to run, and at some point in a
practical system the limits of the dynamic range and noise floor will be
reached.

So there should always be a difference in output error rate if there is
a difference in input error rate.

--
Eric Jacobsen
Minister of Algorithms
Abineau Communications
http://www.abineau.com
From: Vladimir Vassilevsky on


Eric Jacobsen wrote:


> The slope gets difficult to measure below error rates of 10e-9 or 10e-10
> or so just because the tests take so long to run, and at some point in a
> practical system the limits of the dynamic range and noise floor will be
> reached.

Brute force simulation isn't practical for low BERs, however it is
simple enough to derive the estimates in that case.

Vladimir Vassilevsky
DSP and Mixed Signal Design Consultant
http://www.abvolt.com
From: Eric Jacobsen on
On 12/21/2009 10:17 AM, Vladimir Vassilevsky wrote:
>
>
> Eric Jacobsen wrote:
>
>
>> The slope gets difficult to measure below error rates of 10e-9 or
>> 10e-10 or so just because the tests take so long to run, and at some
>> point in a practical system the limits of the dynamic range and noise
>> floor will be reached.
>
> Brute force simulation isn't practical for low BERs, however it is
> simple enough to derive the estimates in that case.
>
> Vladimir Vassilevsky
> DSP and Mixed Signal Design Consultant
> http://www.abvolt.com

Even with real equipment running real-time it's hard to measure 10e-9 or
below with most system.

--
Eric Jacobsen
Minister of Algorithms
Abineau Communications
http://www.abineau.com
From: Vladimir Vassilevsky on


Eric Jacobsen wrote:
> On 12/21/2009 10:17 AM, Vladimir Vassilevsky wrote:
>> Eric Jacobsen wrote:
>>
>>> The slope gets difficult to measure below error rates of 10e-9 or
>>> 10e-10 or so just because the tests take so long to run, and at some
>>> point in a practical system the limits of the dynamic range and noise
>>> floor will be reached.
>>
>>
>> Brute force simulation isn't practical for low BERs, however it is
>> simple enough to derive the estimates in that case.
>>
> Even with real equipment running real-time it's hard to measure 10e-9 or
> below with most system.

For real or simulated system, the case of marginally low BER is
tractable. So the expected BER can be estimated from noise floor and
code performance. There is no need to make long runs unless the goal is
verification of the decoder operation.

Vladimir Vassilevsky
DSP and Mixed Signal Design Consultant
http://www.abvolt.com