[pskmail] Re: PskR - 4 level quantizer
- From: John Douyere <vk2eta@xxxxxxxxx>
- To: pskmail@xxxxxxxxxxxxx
- Date: Wed, 1 Feb 2012 17:37:34 +1100
Stefano,
Correction to my previous email:
The values for the Viterbi decoder are: "Sure 1" is 255, "Sure 0" is 0 and
127 is "neither 1 nor 0".
Sorry for the confusion.
73, John
On Wed, Feb 1, 2012 at 3:10 PM, John Douyere <vk2eta@xxxxxxxxx> wrote:
> Hi Stefano,
>
> I can comment on the code.
>
> The Viterbi decoder is capable of handling "soft bits" meaning that we can
> feed it with variable values that represent how certain we are that they
> represent a 1 or a 0.
>
> The coding is done such that +127 represents a "very sure 1", -128 a "very
> sure zero" and 0 (or 1) a "not sure if it is a 1 or a 0".
>
> That helps the decoder make soft decisions as to which value (binary 0 or
> 1) is more likely to be the correct one.
>
> So here I use two criteria:
>
> 1. The angle is first: BPSK modulation is coded as 0 or 180 degree angle
> phases and therefore I use the angle as the first criteria: 0 or 180 degree
> decoded angle = "very sure 0 or 1" and 90 degrees is "either 0 or 1", then
> for anything in between the certainty is proportional to the deviation from
> 90 degrees.
>
> 2. The amplitude: the concept being that is the received signal is faint
> due to selective fading for example then the amplitude will be smaller than
> the average signal and therefore we should apply some "discounting" to the
> credibility of the phase information.
>
> In my tests, using the soft decoding algorithm improves the robustness of
> the signal to white noise by around 1 dB which may not be much but adds to
> the whole coding gain.
>
> Do you have a project in mind?
>
> Hope this helps,
>
> 73, John
>
>
> On 01/02/2012 7:24 AM, "Stefano" <stefano.banti@xxxxxxxx> wrote:
>
>> Hi all,
>>
>> I'm a sporadic user of pskmail, software developer and ham radio op as
>> well; I'm looking at the source code of fldigi because I'm interested in
>> the pskr software modem implementation.
>> Can someone explain (shortly) the method used for the 4-level softbit
>> quantizer ? Or give a reference to find mathematics involved ?
>>
>> ...
>> averageamp = decayavg(averageamp, sigamp, SQLDECAY);
>> if (sigamp > 0 && averageamp > 0) {
>> softamp = clamp( averageamp / sigamp, 1.0, 1e6);
>> } else {
>> softamp = 1; // arbritary number (50% impact)
>> }
>> // Compute values between -128 and +127 for phase value only
>> if (phase > M_PI) {
>> softangle = (127 - (((2 * M_PI - phase) / M_PI) * (double)
>> 255));
>> } else {
>> softangle = (127 - ((phase / M_PI) * (double) 255));
>> }
>> // Then apply impact of amplitude. Finally, re-centre on 127-128
>> // as the decoder needs values between 0-255
>> softbit = (unsigned char) ((softangle / (1 + softamp)) - 127);
>> ...
>>
>> many thanks in advance
>> Stefano ik2yxt.
>>
>>
>>
Other related posts:
