I installed the Anaconda MKL extensions - which includes BLAS - but it doesn't include C header files, everything is precompiled. http://docs.continuum.io/mkl-optimizations/index.html

I received this error, is there an easy way to make the program use the faster MKL extensions / BLAS? (Apologies if this question doesn't make a ton of sense, I don't completely understand the Python/Cython/BLAS/MKL interaction/architecture here).

C:\Anaconda\lib\site-packages\deltasigma\_config.py:40: UserWarning: Numpy did not detect the BLAS library in the system
  warn("Numpy did not detect the BLAS library in the system")
C:\Anaconda\lib\site-packages\deltasigma\_config.py:55: UserWarning: Cannot find the path for 'cblas.h'. You may set it using the environment variable BLAS_H.
NOTE: You need to pass the path to the directories were the header files are, not the path to the files.
  warn("Cannot find the path for 'cblas.h'. You may set it using the environment variable "

I was wondering if you'd be willing to take pull requests moving the tests into their own files. I was trying to track down an issue on windows, and would get an error that I couldn't import * when I ran nosetests -v ./deltasigma/*.py. What I think is happening is on *nix systems the shell is globbing the *.py and replacing it with all the filenames. cmd.exe doesn't do this on Windows, and so when I try to just run nosetests it doesn't find any tests. I think that's because nosetests is generally looking for tests in places they're not located in this package: http://nose.readthedocs.org/en/latest/finding_tests.html

It would (theoretically) be fairly simple to move over the tests into their own file, but I don't know if you want that architecture change.

The long road to v02

And it needs an unit test.

DocumentNTF(NTF1,osr=8) shows blank plots
Errors are:
plotPZ(ntf, 'b', 6, showlist=False) line 85, in plotPZ
hold_status = plt.ishold()
AttributeError: module 'pylab' has no attribute 'ishold'
Please suggest an alternative to get the plot in spyder windows

It crashes. To my understanding we share the problem with the Matlab Delta Sigma Toolbox.

Hello everyone. I ran

nosetests -v deltasigma

and lots of errors came up. What could be the problem? I am running this inside a CentOS 7 container with what i think all dependencies installed.
nose.txt

.../numpy/core/include/numpy/npy_1_7_deprecated_api.h:15:2: warning: #warning "Using deprecated NumPy API, disable it by " "#defining NPY_NO_DEPRECATED_API NPY_1_7_API_VERSION" [-Wcpp]

The long road to v02

The long road to v02

The long road to v02

As it is also stated in the source.

The function realizeNTF_ct breaks for feed forward (FF) structure.

Try the test defined for realizeNTF but with FF from instead of FB form:

ntf = (np.array([1., 1.]), np.array([0., 0.]), 1)
ABCDc, tdac2 = realizeNTF_ct(ntf, 'FF')

Traceback:

ValueError                                Traceback (most recent call last)
<ipython-input-71-9373c6f6636c> in <module>()
      1 ntf = (np.array([1., 1.]), np.array([0., 0.]), 1)
----> 2 ABCDc, tdac2 = realizeNTF_ct(ntf, 'FF')
      3 print(ABCDc)
      4 print(tdac2)

anaconda/lib/python2.7/site-packages/deltasigma-0.1-py2.7.egg/deltasigma/_realizeNTF_ct.pyc in realizeNTF_ct(ntf, form, tdac, ordering, bp, ABCDc)
    214     sys_c = []
    215     for i in range(Bc.shape[1]): # number of inputs
--> 216         sys_c.append(ss2zpk(Ac, Bc, Cc, Dc, input=i))
    217     yy = pulse(sys_c, tp, 1, n_imp, 1)
    218     yy = np.squeeze(yy)

anaconda/lib/python2.7/site-packages/scipy/signal/ltisys.pyc in ss2zpk(A, B, C, D, input)
    223 
    224     """
--> 225     return tf2zpk(*ss2tf(A, B, C, D, input=input))
    226 
    227 

anaconda/lib/python2.7/site-packages/scipy/signal/ltisys.pyc in ss2tf(A, B, C, D, input)
    177 
    178     num_states = A.shape[0]
--> 179     type_test = A[:, 0] + B[:, 0] + C[0, :] + D
    180     num = numpy.zeros((nout, num_states + 1), type_test.dtype)
    181     for k in range(nout):

ValueError: operands could not be broadcast together with shapes (2) (3) 

The long road to v02

This topology is currently unsupported.

The long road to v02

In deltasigma/_realizeNTF.py there is no code to test the 'PFF' topology.

It looks simulateDSM operates correctly for nlev=2 and nlev=3 but not for nlev > 3:
for even nlev it gives same results as for nlev=2 and for odd nlev it gives same results as for nlev=3.

deltasigma is working nicely for me on a Mac 10.9 setup, but not on 10.11 and not on Ubuntu. The problems are all to do with cblas.
On Mac OS 10.11 and up /usr/include is "restricted", and opencblas puts things in a weird place (/opt/local/include/cblas.h).
On my linux 4.5.3-x86_64 I have /usr/include/cblas.h, and instead get the message in the title when I "import deltasigma"
Any help appreciated; your python version of deltasigma is much more useful than the Matlab one.

Files that have had tests moved to separate file in /tests/ folder:

• _axisLabels.py (PR #31)
• _bilogplot.py
• _bplogsmooth.py
• _bquantize.py
• _bunquantize.py
• _calculateSNR.py
• _calculateTF.py
• _cancelPZ.py
• _changeFig.py (no programmatic test to port)
• _circ_smooth.py
• _clans.py
• _config.py (no test to port)
• _constants.py (no test to port)
• _db.py
• _dbm.py
• _dbp.py
• _dbv.py
• _delay.py
• _DocumentNTF.py
• _dsclansNTF.py
• _ds_f1f2.py
• _ds_freq.py
• _ds_hann.py
• _ds_optzeros.py
• _ds_quantize.py
• _ds_synNTFobj1.py
• _evalF0.py
• _evalF1.py
• _evalMixedTF.py
• _evalRPoly.py
• _evalTF.py
• _evalTFP.py
• _figureMagic.py
• _frespF1.py (no test)
• _impL1.py
• _infnorm.py
• _l1norm.py
• _logsmooth.py
• _lollipop.py
• _mapABCD.py
• _mapCtoD.py
• _mapQtoR.py
• _mapRtoQ.py
• _mod1.py
• _mod2.py
• _nabsH.py
• _padb.py
• _padl.py
• _padr.py
• _padt.py
• _partitionABCD.py
• _peakSNR.py
• _PlotExampleSpectrum.py
• _plotPZ.py
• _plotSpectrum.py
• _predictSNR.py
• _pulse.py
• _realizeNTF.py
• _realizeNTF_ct.py
• _rms.py
• _rmsGain.py
• _scaleABCD.py
• _simulateDSM.py
• _simulateDSM_python.py
• _simulateSNR.py
• _sinc_decimate.py
• _SIunits.py
• _stuffABCD.py
• _synthesizeChebyshevNTF.py
• _synthesizeNTF.py
• _synthesizeNTF0.py
• _synthesizeNTF1.py
• _test_dsdemo.py
• _thermometer.py
• _undbm.py
• _undbp.py
• _undbv.py
• _utils.py
• _zinc.py

New LTI description have been added to scipy in version 0.16.0, effectively breaking the API for functions and classes in the ltisubpackage .

For now, users should stick with 0.15.x (and lower) versions of the library

For the future, we should decide what to do:

• Support 0.16.0 and break support with previous versions of the library,
• Attempt to support all versions

Obviously the latter being best, if custom code can be kept to a minimum.

The long road to v02

Possibly relying on six.

Maybe ship six within the package.

Sometimes a TF can be provided only with "simplified zpk form", ie (ndarray zeros, ndarray poles, scalar gain), sometimes an LTI object from scipy is accepted too, sometimes a tuple representation like (ndarray num, ndarray den) is also valid.

• Evaluate adding an helper method
• Ensure consistency across the package