Convertion to INT Crashes pyfdax
Tarik-T opened this issue · 7 comments
pyFDA is a remarkably beautiful and powerful tool. Thank you for making it available to the public!
Although it worked well for many design steps and filter evaluation, at the moment it crashes when I ask for integer coefficients, as follows:
Traceback (most recent call last):
File "C:\Python310\lib\site-packages\pyfda\input_widgets\input_coeffs.py", line 840, in fx_base2qdict
self.qdict2ui()
File "C:\Python310\lib\site-packages\pyfda\input_widgets\input_coeffs.py", line 795, in qdict2ui
self.refresh_table()
File "C:\Python310\lib\site-packages\pyfda\input_widgets\input_coeffs.py", line 630, in refresh_table
self._refresh_table_item(row, col)
File "C:\Python310\lib\site-packages\pyfda\input_widgets\input_coeffs.py", line 552, in _refresh_table_item
if self.ba_q[col + 2][row] > 0:
TypeError: 'int' object is not subscriptable
Steps:
load filter from config file
update all fields (Like Fbp, Fsb, Fc, etc. by clicking into it, as the fields are not updated by loading the filter data (A minor bug in my mind)
Set Filter design class , as this is not part of saved / loaded data
Design Filter
Switching to tab "b, a"
Selecting in the box from "float" to "DEC" -> Crash
I can't rule out the possibility that I'm doing something wrong, but it would be nice if a misoperation ended with an error message rather than a crash.
Thanks for the warm words! If the software crashes, there is obviously something wrong with the software itself :-)
I have two ideas about the cause of the crash (although I cannot reproduce it):
- Are you using v0.7.0? There was a related bug that I fixed in 0.7.1
- Something seems to be wrong with your filter file - with a sampling frequency of 1 Hz it shouldn't be possible to use a frequency of e.g. f_c = 500 Hz. I agree that the filter import routine is buggy (I'm afraid of touching the related part of the code, it is a big dark mess) and this might also cause the mismatch of sampling and corner frequency.
Could you please check your pyfda version, attach your filter file (also for debugging the filter import routines) and retry designing your filter with integer coefficients without importing a filter file? Thanks a lot for your help!
I'm sorry - version is 0.7.1
Remaining version information (just to be on the safe side), below.
I agree that a sampling frequency of 1 Hz and a filter frequency of 300 doesn't match very well.
I think I forgot to update sample freq (It is 19200) after importing the filter design file:
Butter375.npz.zip
(Suffix is seems to be zip, but remove that. It is npz).
I believe there is a special function hidden on the "b,a" tab that is randomly (I have not found the trigger yet, but keep trying :-) .) activated when someone wants integer coefficients.
I have another one at that stage. I hope it makes it easier for you in figuring out the issue.
Traceback (most recent call last):
File "C:\Python310\lib\site-packages\pyfda\input_widgets\input_coeffs.py", line 325, in process_sig_rx
self.load_dict()
File "C:\Python310\lib\site-packages\pyfda\input_widgets\input_coeffs.py", line 661, in load_dict
self.qdict2ui() # set quantization UI from dictionary
File "C:\Python310\lib\site-packages\pyfda\input_widgets\input_coeffs.py", line 795, in qdict2ui
self.refresh_table()
File "C:\Python310\lib\site-packages\pyfda\input_widgets\input_coeffs.py", line 630, in refresh_table
self._refresh_table_item(row, col)
File "C:\Python310\lib\site-packages\pyfda\input_widgets\input_coeffs.py", line 552, in _refresh_table_item
if self.ba_q[col + 2][row] > 0:
TypeError: 'int' object is not subscriptable
Nevertheless, thanks to this nice tool, I am making progress in understanding and developing and digital filters.
Thanks!
Tarik
C:\Python310\lib\site-packages\pyfda
User Module Dir None
Home Dir C:\Users\Tarik
Temp Dir C:\Users\Tarik\AppData\Local\Temp
Config Dir C:\Users\Tarik.pyfda
pyFDA Config C:\Users\Tarik.pyfda\pyfda.conf
Log. Config C:\Users\Tarik.pyfda\pyfda_log.conf
Logfile C:\Users\Tarik\AppData\Local\Temp.pyfda\pyfda.log
External modules and libraries
Module Version Licence Purpose
Python 3.10.2 (64 Bit) PSF
numpy 1.23.3 BSD Fast array numerics
scipy 1.9.3 (no mkl) BSD Library for scientific computing
numexpr 2.8.3 MIT Fast numerical array expression
matplotlib 3.6.1 PSF-based Plotting library
Qt5 5.15.2 LPGLv3 Widget library (UI etc.)
PyQt 5.15.7 GPLv3 Python bindings for Qt5
Markdown 3.4.1 BSD Markdown implementation
docutils 0.19 GPLv3 a.o. Plain text -> markup formats
mplcursors 0.5.2 MIT Interactive cursors (needs Matplotlib >= 3.1)
YOSYS not found ISC Framework for Verilog RTL synthesis
Thanks, I can reproduce the crash!
I had to rewrite the part for reading and writing filters as numpy arrays with irregular shapes ("ragged arrays") are discouraged / deprecated (https://numpy.org/devdocs/release/1.24.0-notes.html) and have become difficult to handle. Therefore, pyfda v0.8.0 no longer crashes but it also cannot read the old filter format anymore. Sorry aboout that!
Good day! I have the same problem. Version: 0.8.2 (Windows). When I click on the save icon the program closes.
Terminal message: ValueError: "setting an array element with a sequence. The requested array has an inhomogeneous shape after 1 dimensions. The detected shape was (2,) + inhomogeneous part."
If you save Float then there is no problem. The problem is only when saving Fixpoint.
hilbert_pyfdax.zip
Thanks for reporting, probably you're right and the bug is in the same region of code. I'll take a look!
Fixed in latest, please test! Version 0.8.3 will be released today or tomorrow.