PyFT8 3.7.0__tar.gz → 3.7.1__tar.gz

{pyft8-3.7.0 → pyft8-3.7.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: PyFT8
-Version: 3.7.0
+Version: 3.7.1
 Summary: FT8 Decoding and Encoding in Python with test/loopback code
 Author-email: G1OJS <g1ojs@yahoo.com>
 License-Expression: GPL-3.0-or-later
@@ -23,7 +23,7 @@ Dynamic: license-file
 # PyFT8 [![PyPI Downloads](https://static.pepy.tech/personalized-badge/pyft8?period=total&units=INTERNATIONAL_SYSTEM&left_color=BLACK&right_color=GREEN&left_text=downloads)](https://pepy.tech/projects/pyft8) [![PyPI Downloads](https://static.pepy.tech/personalized-badge/pyft8?period=weekly&units=INTERNATIONAL_SYSTEM&left_color=BLACK&right_color=GREEN&left_text=downloads/wk)](https://pepy.tech/projects/pyft8) [PyPI link](https://pypi.org/project/PyFT8/)
 # All-Python FT8 Transceiver GUI / Command Line Modem
 
-This repository contains the source code for PyFT8, an all-Python open source FT8 transceiver that you can run as a basic GUI or from the command line to receive and transmit. Decoding performance (number of decodes) is about 70% of that achieved by WSJT-x in NORM mode, but (tbc) slightly above ft8_lib. 
+This repository contains the source code for PyFT8, an all-Python open source FT8 transceiver that you can run as a basic GUI or from the command line to receive and transmit. Decoding performance (number of decodes) is very close to that of WSJT-x 2.7.0 running in FAST mode. 
 
 PyFT8 is somewhat experimental, with a focus on demonstrating FT8 written in Python, but can be used as a standalone replacement for WSJT-x and other software.
 
@@ -116,6 +116,28 @@ The image below shows the number of decodes from PyFT8, WSJT-x V2.7.0 running in
 
 <img width="900" height="900" alt="live playback vs wsjtx_v2 7 0" src="https://github.com/user-attachments/assets/853e98b2-5b9b-4241-b5f3-181fe193642e" />
 
+## Decoding approach
+The approach to decoding, as at v3.7.0, is as follows.
+
+ - As audio arrives
+    - FFT to a spectrum 4 times per symbol duration & assemble a time-frequency grid with 2 bins per tone.
+    - Collect all audio samples for later full cycle FFT
+ - Search the time frequency grid for candidates based on Costas array sync score (using middle Costas block)
+ - Send candidates for decode using symbol & tone points drawn directly from the search grid to get early decodes using limited LDPC cycles.
+
+This often decodes around three quarters of the candidates decoded by WSJT-x, with the steps below getting most of the rest:
+
+ - For candidates that haven't decoded:
+     - FFT the full cycle audio, downsample to 200 samples per second & shift the signal to f = 0
+     - Explore fine time and frequency offsets to maximise the Costas sync score
+     - Check for bad candidates (count maxima of Costas tones)
+     - Extract the LLR & wait for decoder to pick up the candidate
+ - For candidates awaiting decode from above:
+     - Try more extensive LDPC
+     - Try several 'A Priori' masks and input to more LDPC attempts, recording the output LLR on failures
+     - Try Ordered Statistics Decoding on the base LLR
+     - Try Ordered Statistics Decoding on the LLRs output from the AP+LDPC runs above
+
 ## Limitations
 PyFT8 doesn't decode / encode *all* message types. The table below shows which are handled.
 

{pyft8-3.7.0 → pyft8-3.7.1}/PyFT8/gui.py

@@ -209,7 +209,6 @@ class Gui:
         x = int(message['t0'] / self.waterfall_data['dt'] + message['their_tx_cycle'] * self.waterfall_data['pixels_per_cycle'])
         y = int(message['fHz'] / self.waterfall_data['df'])
         mb.set_props(x, y, message)
-        self.msg_boxes.append(mb)
 
     def update_message(self, display_text, update_dict):
         for mb in self.msg_boxes:
@@ -246,6 +245,7 @@ class Gui:
                 break
         if not mb:
             mb = Msg_box(self.fig, self.ax_wf, self.waterfall_data['sig_w'], self.waterfall_data['sig_h'])
+            self.msg_boxes.append(mb)
         return mb
 
     def _set_band(self, band):

{pyft8-3.7.0 → pyft8-3.7.1}/PyFT8/pyft8.py

@@ -11,7 +11,7 @@ from PyFT8.databases import ADIF
 from PyFT8.qso_manager import QSO_manager
 from PyFT8.message_broker import Broker
 
-VER = '3.7.0'
+VER = '3.7.1'
 PSKR_REFRESH_MINS = 20
 message_broker = None
 

{pyft8-3.7.0 → pyft8-3.7.1}/PyFT8.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: PyFT8
-Version: 3.7.0
+Version: 3.7.1
 Summary: FT8 Decoding and Encoding in Python with test/loopback code
 Author-email: G1OJS <g1ojs@yahoo.com>
 License-Expression: GPL-3.0-or-later
@@ -23,7 +23,7 @@ Dynamic: license-file
 # PyFT8 [![PyPI Downloads](https://static.pepy.tech/personalized-badge/pyft8?period=total&units=INTERNATIONAL_SYSTEM&left_color=BLACK&right_color=GREEN&left_text=downloads)](https://pepy.tech/projects/pyft8) [![PyPI Downloads](https://static.pepy.tech/personalized-badge/pyft8?period=weekly&units=INTERNATIONAL_SYSTEM&left_color=BLACK&right_color=GREEN&left_text=downloads/wk)](https://pepy.tech/projects/pyft8) [PyPI link](https://pypi.org/project/PyFT8/)
 # All-Python FT8 Transceiver GUI / Command Line Modem
 
-This repository contains the source code for PyFT8, an all-Python open source FT8 transceiver that you can run as a basic GUI or from the command line to receive and transmit. Decoding performance (number of decodes) is about 70% of that achieved by WSJT-x in NORM mode, but (tbc) slightly above ft8_lib. 
+This repository contains the source code for PyFT8, an all-Python open source FT8 transceiver that you can run as a basic GUI or from the command line to receive and transmit. Decoding performance (number of decodes) is very close to that of WSJT-x 2.7.0 running in FAST mode. 
 
 PyFT8 is somewhat experimental, with a focus on demonstrating FT8 written in Python, but can be used as a standalone replacement for WSJT-x and other software.
 
@@ -116,6 +116,28 @@ The image below shows the number of decodes from PyFT8, WSJT-x V2.7.0 running in
 
 <img width="900" height="900" alt="live playback vs wsjtx_v2 7 0" src="https://github.com/user-attachments/assets/853e98b2-5b9b-4241-b5f3-181fe193642e" />
 
+## Decoding approach
+The approach to decoding, as at v3.7.0, is as follows.
+
+ - As audio arrives
+    - FFT to a spectrum 4 times per symbol duration & assemble a time-frequency grid with 2 bins per tone.
+    - Collect all audio samples for later full cycle FFT
+ - Search the time frequency grid for candidates based on Costas array sync score (using middle Costas block)
+ - Send candidates for decode using symbol & tone points drawn directly from the search grid to get early decodes using limited LDPC cycles.
+
+This often decodes around three quarters of the candidates decoded by WSJT-x, with the steps below getting most of the rest:
+
+ - For candidates that haven't decoded:
+     - FFT the full cycle audio, downsample to 200 samples per second & shift the signal to f = 0
+     - Explore fine time and frequency offsets to maximise the Costas sync score
+     - Check for bad candidates (count maxima of Costas tones)
+     - Extract the LLR & wait for decoder to pick up the candidate
+ - For candidates awaiting decode from above:
+     - Try more extensive LDPC
+     - Try several 'A Priori' masks and input to more LDPC attempts, recording the output LLR on failures
+     - Try Ordered Statistics Decoding on the base LLR
+     - Try Ordered Statistics Decoding on the LLRs output from the AP+LDPC runs above
+
 ## Limitations
 PyFT8 doesn't decode / encode *all* message types. The table below shows which are handled.
 

{pyft8-3.7.0 → pyft8-3.7.1}/README.md

@@ -1,7 +1,7 @@
 # PyFT8 [![PyPI Downloads](https://static.pepy.tech/personalized-badge/pyft8?period=total&units=INTERNATIONAL_SYSTEM&left_color=BLACK&right_color=GREEN&left_text=downloads)](https://pepy.tech/projects/pyft8) [![PyPI Downloads](https://static.pepy.tech/personalized-badge/pyft8?period=weekly&units=INTERNATIONAL_SYSTEM&left_color=BLACK&right_color=GREEN&left_text=downloads/wk)](https://pepy.tech/projects/pyft8) [PyPI link](https://pypi.org/project/PyFT8/)
 # All-Python FT8 Transceiver GUI / Command Line Modem
 
-This repository contains the source code for PyFT8, an all-Python open source FT8 transceiver that you can run as a basic GUI or from the command line to receive and transmit. Decoding performance (number of decodes) is about 70% of that achieved by WSJT-x in NORM mode, but (tbc) slightly above ft8_lib. 
+This repository contains the source code for PyFT8, an all-Python open source FT8 transceiver that you can run as a basic GUI or from the command line to receive and transmit. Decoding performance (number of decodes) is very close to that of WSJT-x 2.7.0 running in FAST mode. 
 
 PyFT8 is somewhat experimental, with a focus on demonstrating FT8 written in Python, but can be used as a standalone replacement for WSJT-x and other software.
 
@@ -94,6 +94,28 @@ The image below shows the number of decodes from PyFT8, WSJT-x V2.7.0 running in
 
 <img width="900" height="900" alt="live playback vs wsjtx_v2 7 0" src="https://github.com/user-attachments/assets/853e98b2-5b9b-4241-b5f3-181fe193642e" />
 
+## Decoding approach
+The approach to decoding, as at v3.7.0, is as follows.
+
+ - As audio arrives
+    - FFT to a spectrum 4 times per symbol duration & assemble a time-frequency grid with 2 bins per tone.
+    - Collect all audio samples for later full cycle FFT
+ - Search the time frequency grid for candidates based on Costas array sync score (using middle Costas block)
+ - Send candidates for decode using symbol & tone points drawn directly from the search grid to get early decodes using limited LDPC cycles.
+
+This often decodes around three quarters of the candidates decoded by WSJT-x, with the steps below getting most of the rest:
+
+ - For candidates that haven't decoded:
+     - FFT the full cycle audio, downsample to 200 samples per second & shift the signal to f = 0
+     - Explore fine time and frequency offsets to maximise the Costas sync score
+     - Check for bad candidates (count maxima of Costas tones)
+     - Extract the LLR & wait for decoder to pick up the candidate
+ - For candidates awaiting decode from above:
+     - Try more extensive LDPC
+     - Try several 'A Priori' masks and input to more LDPC attempts, recording the output LLR on failures
+     - Try Ordered Statistics Decoding on the base LLR
+     - Try Ordered Statistics Decoding on the LLRs output from the AP+LDPC runs above
+
 ## Limitations
 PyFT8 doesn't decode / encode *all* message types. The table below shows which are handled.
 

{pyft8-3.7.0 → pyft8-3.7.1}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "PyFT8"
-version = "3.7.0"
+version = "3.7.1"
 license = "GPL-3.0-or-later"
 
 authors = [