locked How is the SNR internally measured? #WSPR #reception


Nils Schiffhauer
 

Hi - so far, I didn't find an answer to a seemingly simple question:

* How -exactly- is WSPR measuring the SNR?

I know that the SNR represents dB against background noise@2500Hz.

What I want to know:
* does "SNR" shows the maximum or average/median value withing the 110 second's period? At how many samples?
* is "noise floor" represented by the lowest signal within the 2500Hz channel over 110 seconds? Or how is it measured otherwise?

Background:
I am comparing WSPR SNR data against fading from broadcasters from the same direction. Maximum time resolution of these data is usually one second but can be raised up to 1 millisecond. From these raw data, I can calculate "chunks" of, say 110 seconds. For a comparison, I need to know how WSPR handles fading within the 110 seconds. Standard deviation for fading is usually around plusminus 4dB, at least.

Thanks and 73: Nils, DK8OK
--
dk8ok.org


Joe
 

On Mon, Apr 4, 2022 at 10:33 AM, Nils Schiffhauer DK8OK wrote:

Hi - so far, I didn't find an answer to a seemingly simple question:
* How -exactly- is WSPR measuring the SNR?
To see the exact algorithm you will need to read and understand the source code.

I know that the SNR represents dB against background noise@2500Hz.

What I want to know:
* does "SNR" shows the maximum or average/median value withing the 110
second's period?
The program measures the average signal power over the full transmission. It measures the average noise power (per Hz of bandwidth) in whatever noise baseline segments (no signal present) are available. It then scales the noise power up to what would be contained in 2500 Hz bandwidth, and calculates SNR = (average signal power)/(average noise power).

-- 73, Joe, K1JT

At how many samples?
* is "noise floor" represented by the lowest signal within the 2500Hz channel
over 110 seconds? Or how is it measured otherwise?

Background:
I am comparing WSPR SNR data against fading from broadcasters from the same
direction. Maximum time resolution of these data is usually one second but can
be raised up to 1 millisecond. From these raw data, I can calculate "chunks"
of, say 110 seconds. For a comparison, I need to know how WSPR handles fading
within the 110 seconds. Standard deviation for fading is usually around
plusminus 4dB, at least.

Thanks and 73: Nils, DK8OK
--
dk8ok.org


Williams, G (af8c) <af8c@...>
 

If I may ask Joe Taylor:

WSJT-X  2.2.2  scaled S/N ratio one way that resulted in the lowest observed S/N being -21.   In 2.5.4 the lowest is now -24 or -26 (I can't remember).  What changed?

--73, Glenn, AF8C

On 4/5/2022 10:25 AM, Joe via groups.io wrote:
On Mon, Apr 4, 2022 at 10:33 AM, Nils Schiffhauer DK8OK wrote:

Hi - so far, I didn't find an answer to a seemingly simple question:
* How -exactly- is WSPR measuring the SNR?
To see the exact algorithm you will need to read and understand the source code.

I know that the SNR represents dB against background noise@2500Hz.

What I want to know:
* does "SNR" shows the maximum or average/median value withing the 110
second's period?
The program measures the average signal power over the full transmission. It measures the average noise power (per Hz of bandwidth) in whatever noise baseline segments (no signal present) are available. It then scales the noise power up to what would be contained in 2500 Hz bandwidth, and calculates SNR = (average signal power)/(average noise power).

-- 73, Joe, K1JT

At how many samples?
* is "noise floor" represented by the lowest signal within the 2500Hz channel
over 110 seconds? Or how is it measured otherwise?

Background:
I am comparing WSPR SNR data against fading from broadcasters from the same
direction. Maximum time resolution of these data is usually one second but can
be raised up to 1 millisecond. From these raw data, I can calculate "chunks"
of, say 110 seconds. For a comparison, I need to know how WSPR handles fading
within the 110 seconds. Standard deviation for fading is usually around
plusminus 4dB, at least.

Thanks and 73: Nils, DK8OK
--
dk8ok.org

--
This email has been checked for viruses by Avast antivirus software.
https://www.avast.com/antivirus


Sam Birnbaum
 

HI Glenn,
Based on the information that I believe is correct here are the two functions that I wrote for my program for conversion to/from SNT to RST. 
Hope these help.


Function WSJTX_SNR_2_RST(Const SNR : integer) : Integer;begin
   Case SNR of   -999..-19 : Result := 529;    -18..-13 : Result := 539;    -12..-07 : Result := 549;    -06..-01 : Result := 559;      0..5   : Result := 569;      6..11  : Result := 579;     12..17  : Result := 589;     Else  Result := 599;  End;
end;
Function WSJTX_RST_2_SNR(Const RST : integer) : Integer;begin     Result := -0;     if RST < 100 then     begin          Case RST of            52 : Result := -24;            53 : Result := -18;            54 : Result := -12;            55 : Result := -06;            56 : Result := 0;            57 : Result := 6;            58 : Result := 12;            59 : Result := 18;          else Begin                    if Rst < 52 then Result := -30                    else Result := 30               End;          end;          Exit;     end;
    Case RST of      529 : Result := -24;      539 : Result := -18;      549 : Result := -12;      559 : Result := -06;      569 : Result := 0;      579 : Result := 6;      589 : Result := 12;      599 : Result := 18;      else Begin                if Rst < 529 then Result := -30                else Result := 30           End;    end;
end;

73,

Sam W2JDB

-----Original Message-----
From: Williams, G (af8c) via groups.io <af8c@...>
To: main@WSJTX.groups.io
Sent: Tue, Apr 5, 2022 11:16 am
Subject: Re: [WSJTX] How is the SNR internally measured? #WSPR #reception

If I may ask Joe Taylor:

WSJT-X  2.2.2  scaled S/N ratio one way that resulted in the lowest
observed S/N being -21.   In 2.5.4 the lowest is now -24 or -26 (I can't
remember).  What changed?

--73, Glenn, AF8C

On 4/5/2022 10:25 AM, Joe via groups.io wrote:
On Mon, Apr  4, 2022 at 10:33 AM, Nils Schiffhauer DK8OK wrote:

Hi - so far, I didn't find an answer to a seemingly simple question:
* How -exactly- is WSPR measuring the SNR?
To see the exact algorithm you will need to read and understand the source code.

I know that the SNR represents dB against background noise@2500Hz.

What I want to know:
* does "SNR" shows the maximum or average/median value withing the 110
second's period?
The program measures the average signal power over the full transmission.  It measures the average noise power (per Hz of bandwidth) in whatever noise baseline segments (no signal present) are available.  It then scales the noise power up to what would be contained in 2500 Hz bandwidth, and calculates SNR = (average signal power)/(average noise power).

    -- 73, Joe, K1JT

At how many samples?
* is "noise floor" represented by the lowest signal within the 2500Hz channel
over 110 seconds? Or how is it measured otherwise?

Background:
I am comparing WSPR SNR data against fading from broadcasters from the same
direction. Maximum time resolution of these data is usually one second but can
be raised up to 1 millisecond. From these raw data, I can calculate "chunks"
of, say 110 seconds. For a comparison, I need to know how WSPR handles fading
within the 110 seconds. Standard deviation for fading is usually around
plusminus 4dB, at least.

Thanks and 73: Nils, DK8OK
--
dk8ok.org



--
This email has been checked for viruses by Avast antivirus software.
https://www.avast.com/antivirus


Alan G4ZFQ
 

WSJT-X  2.2.2  scaled S/N ratio one way that resulted in the lowest observed S/N being -21.   In 2.5.4 the lowest is now -24 or -26 (I can't remember).  What changed?
Glenn,

The possible minimum depends on the mode.
Sometimes an improvement to the decoders is announced. https://physics.princeton.edu//pulsar/K1JT/Release_Notes.txt

73 Alan G4ZFQ


Nils Schiffhauer
 

... thanks, Joe. 73 Nils, DK8OK

Am 05.04.2022 um 14:25 schrieb Joe:
On Mon, Apr 4, 2022 at 10:33 AM, Nils Schiffhauer DK8OK wrote:

Hi - so far, I didn't find an answer to a seemingly simple question:
* How -exactly- is WSPR measuring the SNR?
To see the exact algorithm you will need to read and understand the source code.

I know that the SNR represents dB against background noise@2500Hz.

What I want to know:
* does "SNR" shows the maximum or average/median value withing the 110
second's period?
The program measures the average signal power over the full transmission. It measures the average noise power (per Hz of bandwidth) in whatever noise baseline segments (no signal present) are available. It then scales the noise power up to what would be contained in 2500 Hz bandwidth, and calculates SNR = (average signal power)/(average noise power).

-- 73, Joe, K1JT

At how many samples?
* is "noise floor" represented by the lowest signal within the 2500Hz channel
over 110 seconds? Or how is it measured otherwise?

Background:
I am comparing WSPR SNR data against fading from broadcasters from the same
direction. Maximum time resolution of these data is usually one second but can
be raised up to 1 millisecond. From these raw data, I can calculate "chunks"
of, say 110 seconds. For a comparison, I need to know how WSPR handles fading
within the 110 seconds. Standard deviation for fading is usually around
plusminus 4dB, at least.

Thanks and 73: Nils, DK8OK
--
dk8ok.org


--
dk8ok.org